384kHz PCM ??? [Archive] - AudioBanter.com

Jay - atldigi

November 16th 03, 09:36 AM

In article >, "Roger W. Norman"
> wrote:

> > are no 384k converters (that I know of) available at the moment, and
> > there's certainly some question as to whether it would be a good idea
> > to do so.
>
> Certainly Roger Nichols thinks so. He seems to recommend upsampling, and
> in terms of gaining finer resolution on EQ and such, it might be worth it.
> Then again, without the wherewithal to actually hear what might be a
> difference or not, it certainly doesn't make sense, does it?

Upsampling is another issue. The point was whether 384 kHz A to D
conversion would be a worthwhile endeavor. Upsampling for processing can
be useful, though not universally. Dynamics processing may actually
benefit more than EQ. However, performing A to D at 384 likely offers no
benefit and according to some may actually have some unintended negative
consequences. We certainly don't have the data to say definitively at
the moment. This also is not the same question as oversampling which
does offer benefits and is in common use today.

As for Roger Nichols opinion, I haven't seen exactly what he has said so
I couldn't really comment specifically. When it comes to technical
issues dealing with conversion, I may be inclined to listen more closely
to Dan Lavry (Lavry engineering, formwely db technologies, widely
regarded as the finest AD/DA conversion available) than Roger Nochols.
If it was a question of production techniques, certainly Roger's advice
would hold more weight.

Dan is soon to release a white paper on the subject of >96k conversion
that is sure to cause a stir. His website is
http://www.lavryengineering.com/ and the papers show up under the
support section. The new paper isn't there yet, but keep an eye out.
When somebody as knowledgable as Dan offers opinions on conversion, it
causes one to stop and think.

--
Jay Frigoletto
Mastersuite
Los Angeles
promastering.com

Dave Collins

November 16th 03, 09:54 AM

In article >,
"Roger W. Norman" > wrote:

> Certainly Roger Nichols thinks so. He seems to recommend upsampling, and in
> terms of gaining finer resolution on EQ and such, it might be worth it.

Except linear processes like EQ are the least likely to benefit from
upsampling.... What does "finer resolution" mean, anyway? Like you can
get 1.00001kHz?

> Then again, without the wherewithal to actually hear what might be a
> difference or not, it certainly doesn't make sense, does it?

It does to Roger. And he writes an article every month.

DC

Arny Krueger

November 16th 03, 11:44 AM

"exbeatle" > wrote in message
om
> from the Pyramix website:
>
> Pyramix DSD 16 channel Features
> *2, 8, or 16 Ch DSD Record/Editing/Mixing and Mastering System.
> *DSD De-noise and De-click for Re-Mastering SACD.
> *SONY and PHILIPS SACD authoring compliance to Scarlet Book
> specification.
> *DST Encoding.
> *Multi-channel Punch-in Punch-out record capability.
> *Integrated Stereo Monitoring.
> *Superior DSD Noise Floor Technology.
> *Up/Down conversion from/to 44.1 KHz, 88.2 KHz, 96 KHz, 192 KHz, 352
> KHz, 384KHz and DSD.

> 352KHZ, 384KHZ -
> !!when did this happen, and what do you do with a 24b/384kHz PCM
> recording??

If you're coding audio, 24/384 does nothing practical, but it does waste a
lot of storage space, time and processing power. Since 24/192 has zero sonic
advantages over 24/96, and the ear is more sensitive to improvements that
are closer to its core capabilities, by logical induction it is safe to
assume that 24/384 can do nothing for sound quality.

Given that digital storage and processing power remain quite extensible at
the 24/384 point, it's hard to imagine how far this madness will go.

Arny Krueger

November 16th 03, 11:48 AM

"Dave Collins" > wrote in message

> In article >,
> "Roger W. Norman" > wrote:
>
>> Certainly Roger Nichols thinks so. He seems to recommend
>> upsampling, and in terms of gaining finer resolution on EQ and such,
>> it might be worth it.

> Except linear processes like EQ are the least likely to benefit from
> upsampling....

People need to understand that upsampling only provides advantages in
certain very specific cases. Just because it makes it so much cheaper and
easier to have good converters doesn't mean that its going to improve
*everything*.

> What does "finer resolution" mean, anyway? Like you
> can get 1.00001kHz?

Ironically, you can get 1.00001kHz-type frequency resolution with ordinary
16/44. Getting that kind of frequency resolution is just a matter of having
a long enough sample. The sample size required to get that kind of
resolution in the frequency domain is the same in both the analog and
digital domains.

Roger W. Norman

November 16th 03, 02:35 PM

"Jay - atldigi" > wrote in message
...
> Upsampling is another issue. The point was whether 384 kHz A to D
> conversion would be a worthwhile endeavor. Upsampling for processing can
> be useful, though not universally.

True, but we weren't necessarily discussing 384 kHz A/D conversion, at least
in terms of the original post. In that case it was 384 kHz downsampling
from 2.8 mHz of DSD, which, when one wants to edit, the conversion has to go
to PCM and 384 kHz is probably a nice figure that allows editing without
losing the "flavor" of the DSD. Certainly I wouldn't want to downsample it
to 96 kHz. However, going UP the scale in A/D conversion, then obviously
even at the 96 kHz rate one has only technical aspects that suggests this is
even necessary, such as a filter rate that is eminently more smooth than a
somewhat brickwall anti-aliasing filter at 22.5 kHz.

Just as obvious, some have argued that even 96 kHz sampling is simply a
waste of bandwidth and others suggest that it's possible to hear a
difference. More than likely it wouldn't make a difference to my war-torn
ears, nor do I think, unless the space is absolutely pristine for recording,
that one gains much benefit from 96 kHz other than maybe some spacial
placement clues that one doesn't actually hear but more likely perceives.

However, you are correct, and I did interject a somewhat different slant on
the conversation. You most definitely said 384 kHz converters, and I jumped
it to doing upsampling, so even with the original idea of downsampling I
still missed the boat! <g>

--

Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.

> In article >, "Roger W. Norman"
> > wrote:
>
> > > are no 384k converters (that I know of) available at the moment, and
> > > there's certainly some question as to whether it would be a good idea
> > > to do so.
> >
> > Certainly Roger Nichols thinks so. He seems to recommend upsampling,
and
> > in terms of gaining finer resolution on EQ and such, it might be worth
it.
> > Then again, without the wherewithal to actually hear what might be a
> > difference or not, it certainly doesn't make sense, does it?
>
Dynamics processing may actually
> benefit more than EQ. However, performing A to D at 384 likely offers no
> benefit and according to some may actually have some unintended negative
> consequences. We certainly don't have the data to say definitively at
> the moment. This also is not the same question as oversampling which
> does offer benefits and is in common use today.
>
> As for Roger Nichols opinion, I haven't seen exactly what he has said so
> I couldn't really comment specifically. When it comes to technical
> issues dealing with conversion, I may be inclined to listen more closely
> to Dan Lavry (Lavry engineering, formwely db technologies, widely
> regarded as the finest AD/DA conversion available) than Roger Nochols.
> If it was a question of production techniques, certainly Roger's advice
> would hold more weight.
>
> Dan is soon to release a white paper on the subject of >96k conversion
> that is sure to cause a stir. His website is
> http://www.lavryengineering.com/ and the papers show up under the
> support section. The new paper isn't there yet, but keep an eye out.
> When somebody as knowledgable as Dan offers opinions on conversion, it
> causes one to stop and think.
>
> --
> Jay Frigoletto
> Mastersuite
> Los Angeles
> promastering.com

Roger W. Norman

November 16th 03, 02:45 PM

"Dave Collins" > wrote in message
...
> Except linear processes like EQ are the least likely to benefit from
> upsampling.... What does "finer resolution" mean, anyway? Like you can
> get 1.00001kHz?

Hi, Dave. You're right, and mostly I was referring to the original post of
downsampling DSD to PCM for editing, which makes 384 kHz seem to be a
reasonable sampling rate and one that would allow some of that finer
resolution in EQ and dynamics. I interjected the upsampling statement,
disregarding Jay's 384 kHz A/D converter statement and reading it instead as
an upsampling argument, whether for or against.

Long day, no time and so I missed the boat.

As for Roger's perception, for a lot of us the difference won't be
noticeable due to rooms, equipment, condition of our hearing, etc. I don't
dispute nor argue for Roger's perceptions, since I'm obviously not
qualified. I just mentioned them because he's made the case more than once.
I still haven't made any effort whatsoever to find out on my own. I leave
that to people who have great rooms, excellent equipment and ears that
haven't been assaulted by grenade concussions! <g> Not to mention such
lumenaries as yourself who's credentials suggest a level of knowledge and
expertise most of us won't achieve.

Just chalk it up to my normal way to screw up the conversation.

--

Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.

> In article >,
> "Roger W. Norman" > wrote:
>
> > Certainly Roger Nichols thinks so. He seems to recommend upsampling,
and in
> > terms of gaining finer resolution on EQ and such, it might be worth it.
>
>
> > Then again, without the wherewithal to actually hear what might be a
> > difference or not, it certainly doesn't make sense, does it?
>
> It does to Roger. And he writes an article every month.
>
> DC

Erik de Castro Lopo

November 16th 03, 08:24 PM

Arny Krueger wrote:
>
> If you're coding audio, 24/384 does nothing practical, but it does waste a
> lot of storage space, time and processing power. Since 24/192 has zero sonic
> advantages over 24/96, and the ear is more sensitive to improvements that
> are closer to its core capabilities, by logical induction it is safe to
> assume that 24/384 can do nothing for sound quality.
>
> Given that digital storage and processing power remain quite extensible at
> the 24/384 point, it's hard to imagine how far this madness will go.

There is another wrinkle in this fabric.

The digital filters used for EQ and whatever else are far harder
to design when the targeted frequency is small in comparison to
the sample rate.

For instance, attempting to design a 4th order Butterworth high
pass filter with a cutoff of 20Hz at a sample rate of 48kHz is
difficult in 32 bit floating point but it can be done in 64
bit double precision floating point.

Once you up the sample rate to 384kHz, that same high pass filter
at 20Hz is going to be difficult the design and implement eve in
64 bit doubles.

Erik
--
+-----------------------------------------------------------+
Erik de Castro Lopo (Yes it's valid)
+-----------------------------------------------------------+
Traditional capital was stuck in a company's bank account or investments.
It could not walk away in disgust. Human capital has free will. It can
walk out the door; traditional capital cannot.

Jay - atldigi

November 16th 03, 10:15 PM

In article >, "Roger W. Norman"
> wrote:

> However, you are correct, and I did interject a somewhat different slant
> on the conversation. You most definitely said 384 kHz converters, and I
> jumped it to doing upsampling, so even with the original idea of
> downsampling I still missed the boat! <g>

No trouble - still on topic to the thread. I just wanted to clarify
since your post had me quoted but wasn't really in response to what I
said.

One last comment: I thought I might have heard that Pyramix will edit
DSD without downsampling and only needs to downsample if you want to do
signal processing (EQ, dynamics etc). I'm not really sure. It may
downsample DSD the moment it comes in and upsample it again on the way
out. Does anybody know for sure which way it works? All downsampled, or
just for processing?

--
Jay Frigoletto
Mastersuite
Los Angeles
promastering.com

Remixer

November 17th 03, 12:25 AM

Pyramix can record and playback pure DSD. You can edit and move clips
around in pure DSD but there is no cross-fade, level change, or any other
processing possible, unless you downsample to 8Fs, render, and bump back up
to DSD.

Roger W. Norman

November 17th 03, 09:27 AM

"Jay - atldigi" > wrote in message
...
> One last comment: I thought I might have heard that Pyramix will edit
> DSD without downsampling and only needs to downsample if you want to do
> signal processing (EQ, dynamics etc). I'm not really sure. It may
> downsample DSD the moment it comes in and upsample it again on the way
> out. Does anybody know for sure which way it works? All downsampled, or
> just for processing?

Simply cut and paste editing, yes. Geez, I'd hate to get down into sample
accurate edits on DSD! <g> But this type of editing for DSD should be fine
considering the type of market it seems to be garnering. I don't suppose we
need worry about an AC/DC DSD recording, but in terms of bluegrass, jazz and
classical, maybe even folk or what's today called Americana, where dynamics
are key components to a performance, it seems like a natural. As I recall,
however, that inflexibility is one of the reasons that DSD never established
itself as a real contender in the DVD-A standards fight, even though there
were some very strong proponents of the technology. I remember many a
conversation on the standards fight with Stuart Robinson back in the
Compuserve days, along with Michael Gerzon just before he passed away. Too
much math to be a fun conversation sometimes, though! <g>

--

Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.

Arny Krueger

November 17th 03, 11:21 AM

"Erik de Castro Lopo" > wrote in message

> Arny Krueger wrote:
>>
>> If you're coding audio, 24/384 does nothing practical, but it does
>> waste a lot of storage space, time and processing power. Since
>> 24/192 has zero sonic advantages over 24/96, and the ear is more
>> sensitive to improvements that are closer to its core capabilities,
>> by logical induction it is safe to assume that 24/384 can do nothing
>> for sound quality.
>>
>> Given that digital storage and processing power remain quite
>> extensible at the 24/384 point, it's hard to imagine how far this
>> madness will go.
>
> There is another wrinkle in this fabric.
>
> The digital filters used for EQ and whatever else are far harder
> to design when the targeted frequency is small in comparison to
> the sample rate.

Not really.

> For instance, attempting to design a 4th order Butterworth high
> pass filter with a cutoff of 20Hz at a sample rate of 48kHz is
> difficult in 32 bit floating point but it can be done in 64
> bit double precision floating point.

Not really. There are zillions of working counter-examples.

> Once you up the sample rate to 384kHz, that same high pass filter
> at 20Hz is going to be difficult the design and implement eve in
> 64 bit doubles.

Not at all. If this were right, programs that implement your so-called
difficult filters like Adobe Audition would be far slower and more complex
than they actually are.

Kurt Albershardt

November 18th 03, 07:38 PM

Remixer wrote:

> Pyramix can record and playback pure DSD. You can edit and move clips
> around in pure DSD but there is no cross-fade, level change, or any other
> processing possible, unless you downsample to 8Fs, render, and bump back up
> to DSD.

Said 8fs being 352.8k in the case of DSD, rather than 384k BTW.

Kurt Albershardt

November 18th 03, 07:40 PM

Jay - atldigi wrote:
>
> there
> are no 384k converters (that I know of) available at the moment, and
> there's certainly some question as to whether it would be a good idea to
> do so.

Claude mentioned to me awhile back that he was expecting dCS to offer a
converter with both DSD and 352.8/24 PCM outputs (using similarly gentle
filter slopes on the PCM side.)

Jay - atldigi

November 18th 03, 10:42 PM

In article >, Kurt Albershardt
> wrote:

> Jay - atldigi wrote:
> >
> > there are no 384k converters (that I know of) available at the moment,
> > and there's certainly some question as to whether it would be a good
> > idea to do so.
>
> Claude mentioned to me awhile back that he was expecting dCS to offer a
> converter with both DSD and 352.8/24 PCM outputs (using similarly gentle
> filter slopes on the PCM side.)

The most exciting part of this is the gentle filter slopes, not the
sample rate (though, heck, why not offer the option if you can?).
Honestly, I'm beginning to think 96k with very gentle filters that start
a gentle rolloff not excessively far above the audible band would be the
best way to go. Perhaps they'll offer a filter choice like this for the
96k rate. That would certainly make me very interested in the converters.

--
Jay Frigoletto
Mastersuite
Los Angeles
promastering.com

Kurt Albershardt

November 19th 03, 01:37 AM

Jay - atldigi wrote:
> In article >, Kurt Albershardt
> > wrote:
>
>
>> Claude mentioned to me awhile back that he was expecting dCS to offer a
>> converter with both DSD and 352.8/24 PCM outputs (using similarly gentle
>> filter slopes on the PCM side.)
>
>
> The most exciting part of this is the gentle filter slopes, not the
> sample rate (though, heck, why not offer the option if you can?).
> Honestly, I'm beginning to think 96k with very gentle filters that start
> a gentle rolloff not excessively far above the audible band would be the
> best way to go.

You're far from alone in that thought. I'd spring for a dCS or
Lavry-priced converter today if I could get 88.2k - 192k conversion with
filters that started rolling off at 21-24k. I have enough faith in
the rest of their designs but this unresolved filter issue looms over my
decision process--if it's a significant as some of us think it might be,
it would mean a lot of replacement purchases.

Erik de Castro Lopo

November 19th 03, 02:35 AM

Arny Krueger wrote:
>
> Not really.
>
> > For instance, attempting to design a 4th order Butterworth high
> > pass filter with a cutoff of 20Hz at a sample rate of 48kHz is
> > difficult in 32 bit floating point but it can be done in 64
> > bit double precision floating point.
>
> Not really. There are zillions of working counter-examples.

So your argument is that not all filters are difficult to
design therefore they must all be "not difficult".

Sorry, that is plain and simple poor logic.

> > Once you up the sample rate to 384kHz, that same high pass filter
> > at 20Hz is going to be difficult the design and implement eve in
> > 64 bit doubles.
>
> Not at all. If this were right, programs that implement your so-called
> difficult filters like Adobe Audition would be far slower and more complex
> than they actually are.

I'm curious. Do you know anything at all about digital filter
design or are you going to stick to your faultly reasoning
above?

I've seen the arguments you get into in this group and others.
Sometimes you do know what you are talking about but in this
case I doubt it.

Erik
--
+-----------------------------------------------------------+
Erik de Castro Lopo (Yes it's valid)
+-----------------------------------------------------------+
"TLC declared bankruptcy after they received less than 2 percent of the $175
million earned by their CD sales. That was about 40 times less than the
profit that was divided among their management, production and record
companies." -- Courtney Love on the REAL piracy

Arny Krueger

November 19th 03, 12:27 PM

"Erik de Castro Lopo" > wrote in message

> Arny Krueger wrote:
>>
>> Not really.

>>> For instance, attempting to design a 4th order Butterworth high
>>> pass filter with a cutoff of 20Hz at a sample rate of 48kHz is
>>> difficult in 32 bit floating point but it can be done in 64
>>> bit double precision floating point.

>> Not really. There are zillions of working counter-examples.

> So your argument is that not all filters are difficult to
> design therefore they must all be "not difficult".

Straw man argument. You named a very specific filter "4th order Butterworth
high
pass filter with a cutoff of 20Hz" and now you're trying to generalize to to
"all filters".

> Sorry, that is plain and simple poor logic.

Straw men tend to be that way.

>>> Once you up the sample rate to 384kHz, that same high pass filter
>>> at 20Hz is going to be difficult the design and implement eve in
>>> 64 bit doubles.

Why would an even higher sample rate be of benefit to the implmentation of
a 20 Hz filter, when 48,000 Hz sampling puts the Nyquist frequency more
than 10 octaves above the corner frequency of 20 Hz?

>> Not at all. If this were right, programs that implement your
>> so-called difficult filters like Adobe Audition would be far slower
>> and more complex than they actually are.

<notice that Erik doesn't respond to this very cogent point>

> I'm curious. Do you know anything at all about digital filter
> design or are you going to stick to your faultly reasoning
> above?

The faulty reasoning is something you fabricated, Erik. Please feel free to
critique it sharply, early and often.

> I've seen the arguments you get into in this group and others.

The facts are generally on my side. I've presented several and the response
is this silly straw man.

> Sometimes you do know what you are talking about but in this
> case I doubt it.

Just tell us why a sampling rate more than 10 octaves above the corner
frequency won't do the job, and why moving up an additional 3 octaves
somehow makes things better.

Erik de Castro Lopo

November 19th 03, 10:21 PM

Arny Krueger wrote:
>
> Just tell us why a sampling rate more than 10 octaves above the corner
> frequency won't do the job, and why moving up an additional 3 octaves
> somehow makes things better.

Your have missed my point completely.

The problem is already there when the sampling rate is 10
octaves above the corner freq and it becomes very much
worse if the sampling rate moves up another 3 octaves.

The problems are due to the inherent inaccuracies of floating
point arithmetic. Computer floating point arithmetic has
accuracy problems when performing arithmetic which mixes very
large numbers and very small numbers.

Filters implementing low critical frequencies (corner for
HP/LP, center for notch/peak etc) at high sampling rates
results in coefficients with a very large spread in values.
Hence the problems.

Erik
--
+-----------------------------------------------------------+
Erik de Castro Lopo (Yes it's valid)
+-----------------------------------------------------------+
"There is no reason why anyone would want a computer in their home"
Ken Olson, DEC, 1977

Arny Krueger

November 20th 03, 12:33 AM

"Erik de Castro Lopo" > wrote in message

> Arny Krueger wrote:
>>
>> Just tell us why a sampling rate more than 10 octaves above the
>> corner frequency won't do the job, and why moving up an additional 3
>> octaves somehow makes things better.
>
> Your have missed my point completely.
>
> The problem is already there when the sampling rate is 10
> octaves above the corner freq and it becomes very much
> worse if the sampling rate moves up another 3 octaves.
>
> The problems are due to the inherent inaccuracies of floating
> point arithmetic. Computer floating point arithmetic has
> accuracy problems when performing arithmetic which mixes very
> large numbers and very small numbers.
>
> Filters implementing low critical frequencies (corner for
> HP/LP, center for notch/peak etc) at high sampling rates
> results in coefficients with a very large spread in values.
> Hence the problems.

You said:

"For instance, attempting to design a 4th order Butterworth high
pass filter with a cutoff of 20Hz at a sample rate of 48kHz is
difficult in 32 bit floating point but it can be done in 64
bit double precision floating point."

So let's put your claim to the test. Cool Edit Pro has a variable frequency
high pass Butterworth filter of any reasonable order including 4. It can be
run at any sample rate up to 999,999 Hz including 48 Khz and 384 KHz. AFAIK
it is implemented in 32 bit floating point arithmetic.

What bad thing to look for when operating it? Bad frequency response?
Nonlinear distortion? Bad phase response?

dan lavry

November 20th 03, 08:17 AM

Erik de Castro Lopo > wrote in message >...
> Arny Krueger wrote:
> >
> > Just tell us why a sampling rate more than 10 octaves above the corner
> > frequency won't do the job, and why moving up an additional 3 octaves
> > somehow makes things better.

Hello,

My name is Dan Lavry, and I hope you do not mind that I barge in. I do
want to say a few words on that 192Khz. As a designer, I have long
realized that faster sampling means less accuracy. You can get 1 bit
at a GHz, 8 bits at 100MHz, 12 bits at 30Mhz…. Nearly 24 bits at a few
Hz. Well, I am not suggesting to sample at 10Hz. We need to cover the
audio bandwidth. Going too slow is bad. Going too fast is also bad.
Let's face it, you want to say charge a cap, or settle an amp, and it
does dot happen at zero time. Nothing does. Weather you get there
exponentially or you have that step with some ringing… whatever it is,
if you wait till it settled you get it closer and more accurate. These
are practical limitations. The thing that gets me is that all the EE's
I talk to make jokes about 192Khz – a common one is "my dog can not
hear it". But many in the recording and mastering side are a part of a
different crowd. There seem to be a gap between the EE designer types
and the recording types. I am not suggesting that EE's are better or
worst. There have been times when the good ear is right, and the EE
learns new things. But there should be times such as right here and
now for the recording types to be open-minded. And folks, that 192KHz
is a crock.

I have wondered about how is it possible that a whole industry can go
in the wrong direction. After all, Few musical instruments will under
special cases yield a -50dB harmonic at 50Khz, and almost nothing at
100KHz. Experiments showing that special case required special mics
and gear. Most recording mics will not pick up a thing at 60khz… The
speakers will not play it, the ear will not hear it, no matter how big
the ego is!
So the sounds are negligible at those high frequencies, the mics will
not pick them, the speaker will not play it, and you can not hear it.
So how can such a crock take place? And let me assure you the folks
that made it happen at the semiconductor houses and workstation houses
did not include the EE department... I think it was and is about
making money. I talked to many engineers that stated privately that
they are not talking because they need their jobs! Other "did not want
to make waves", "stir the pot"…

So is everyone trying to cheat? Of course not. It started with 48Khz
better than 44.1, which is true. Also, while a bit excessive, 88.2 and
96KHz are in some cases a better compromise than 44.1. That is due to
the old FIR pre shooting (pre echo) that can happen under some cases
with a lot of processing. By the time you are at 60KHz it is so far
down you will not hear it, so 70Khz is a pretty reasonable place to
be. So after we have improved some as we went higher, the expectation
was than to continue upwards. Where does it stop? 384? Why not 1Ghz?
Oh well, the will be pretty bad, will it not?

But there is more to that than that "trendy upwards worked before so
lets continue". The common sense tells you that more is better. For
example, more pixels yield more detail. More bits gives better sample
accuracy. So if we take that analog wave, made out infinite points
connecting into a line, will we not benefit from a better "tracking"
of the wave? Well here is the news, and it is pretty old stuff, though
it may not be easy to grasp with simple common sense. More pixels for
better picture- yes! More bits for better accuracy (assuming
theoretical case - no noise) – yes! More sample density (higher sample
rate) for more accuracy –NO! This is the beauty of the sampling
theory. Nyquist did not say: We take more points and we will get
closer approximation. What he said – and it is a FUNDUMENTAL THEORY,
is that once we agree to deal with a limited bandwidth (called the
Nyquist bandwidth), all we need to do is sample at a tiny amount
greater than twice that bandwidth. This will yield 100% of the
waveform information in the data stream. We may need to filter a
signal (anti alaising) to make sure we do not have energy over
Nyquist, but than we are home free. Taking 4 times as many samples
does not yield 400% of the information. You can only have 100%! How do
you retrieve the information? You use a filter and it connects the
sample points in such a way that that you get the original wave shape.
A filter does not connect the dots (sample points) with straight
lines, or parabola… It recreates the original wave! You do not need to
help things with extra point in between. It buys you nothing!

I also see a lot of confusion regarding that Nyquist, upsampling,
oversampling, gradual filters… Some folks think that a 96KHz AD will
require a sharper anti aliasing filter than say 192KHz. This is
typically very wrong. The 96KHz or 192KHz AD refers to the OUTPUT RATE
of the converter. The antialias requirement is determined by the INPUT
SAMPLING RATE which is usually way beyond 192Khz. This days, most
modern AD's are running at input sampling rates of 3-12Mhz!!! DSD is
64fs and many mulibit IC's run even faster… So even with 50Khz audio,
Nyquist is so high the a gradual 3 pole will yield 120dB at the input
rate Nyquist. The outcome of the high rate modulator (input) is than
down-sampled to whatever – 44.1, 96, 192… Of course, when the sales
guys try to stick you with it they tell you need more bandwidth, but
they also ALL I saw regarding semiconductor and gear makers alike:
specs for 192KHz device are with A weighting – which states
(indirectly?) that you do not even need to measure flat to 20Khz. So
is it a crock? It is!

Theoretically, there is "no harm" in more points, and there is "no
added good". But as I pointed out, faster is less accurate! And yes,
you double the sample rate and the processing power requirement, and
so is the file size. These are serious draw backs! Don't you say you
do not care about file size: The DVD audio has 12 to 1 compression
(Dolby AC), We do not even get near a 1 to 1, and we want to push it
higher?

I realize that with all that reasoning and science and engineering,
someone is going to tell me that they hear it and like it. In fact,
someone told me that they still hear that high frequency in the
44.1KHz CD. I will not dignify that impossibility. If you hear some
distortion you like on the 44.1K CD, you did not need to go any faster
than 44.1KHz to generate it. I am not arguing against controlled
distortions (such as tube sound and what not). If you like it is fine.
It may be artistic decision. If you feel like you need to go to 1Mhz
than down to like it, fine. I think you are letting the gear control
you instead of the other way around, but fine! Just as long as I get
you to realize that you can get those distortions with a 44Khz… And we
do not all need to double the file size and processing power, and buy
new gear that is less accurate.

192 is a crock! 382 is a super crock! 88.2/96Khz is a bit excessive,
but not too far from a good rate. I too can glue a faster IC on the
board and make more money. My 192 DA prototype is not bad, but the
96KHz bits it by a lot.
Anyone telling you that more points will give better aproximation is
lacking lacking some know how.

Thanks for your patience.

Dan Lavry
Lavry Engineering

Mike Rivers

November 20th 03, 04:05 PM

In article > writes:

Hi, Dan. Welcome to rec.audio.pro.

> I have wondered about how is it possible that a whole industry can go
> in the wrong direction.

Marketing. It's easier to sell a new format than it is to sell an
improved version of an old one.

> So the sounds are negligible at those high frequencies, the mics will
> not pick them, the speaker will not play it, and you can not hear it.
> So how can such a crock take place?

There are throries (though no valid tests yet) that suggest that
what's up there that we can't hear affects what we can hear. As I said
in another posting, I related a listening demonstration where I heard
what I considered a subjective improvement when moving from a 48 kHz
to a 95 kHz A/D/A converter chain (and they were your converters, back
when you were still selling under the db name), though I have no
reason to believe (or not believe) that a similar improvement would be
realized at 192 kHz.

> Theoretically, there is "no harm" in more points, and there is "no
> added good".

The "harm" is that more samples per cycle is easier to achieve than
more accurate samples, so we are often sold the wrong technology
because it's cheaper and it looks better than the previous technology
on the ad copy.

--
I'm really Mike Rivers - )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

KikeG

November 20th 03, 04:31 PM

Very good and enlightening post, Dan. However, there's one thing I
somewhat disagree with:

> So is everyone trying to cheat? Of course not. It started with 48Khz
> better than 44.1, which is true. Also, while a bit excessive, 88.2 and
> 96KHz are in some cases a better compromise than 44.1. That is due to
> the old FIR pre shooting (pre echo) that can happen under some cases
> with a lot of processing.

This pre-echo (I'd better call it pre-ringing) is a consequence of the
brickwall FIR filter cutoff frequency, at around 20-21 KHz. Due to
this, a) this pre-ringing has a frequency of around 20-21 KHz, which
should be hardly audible, b) it happens just with signals that have
strong content at these frequencies, c), usual lengths of this
pre-ringing are in the order of a few milliseconds (less than 3 ms
with a 256-tap FIR filter), very close to the limits of audibility,
and d) this pre-ringing is important just in case of impulsive signals
where it could really make a difference.

So, it seems that this pre-ringing should have little importance in
practice. I have yet to see someone prove it is audible by means of a
reliable double-blind listening.

Even if it was proved to have some audible consequences, I think it
could be overcome by means of simply using adequate non-symmetric,
minimum pre-echo, non-linear-phase FIR filters. Our ear is little or
no sensitive to phase distortion at high frequencies, so I think the
use of this kind of filters should have little or no audible
consequences.

dan lavry

November 20th 03, 07:15 PM

Erik de Castro Lopo > wrote in message
> The problems are due to the inherent inaccuracies of floating
> point arithmetic. Computer floating point arithmetic has
> accuracy problems when performing arithmetic which mixes very
> large numbers and very small numbers.
>
> Filters implementing low critical frequencies (corner for
> HP/LP, center for notch/peak etc) at high sampling rates
> results in coefficients with a very large spread in values.
> Hence the problems.
>
> Erik

I agree with you and will add some additional comments:

It is not really easy to explain quickly, but there is a relationship
between filter complexity (number of coefficients) and the ratio of
sample rate to filter corner frequency. For example, making a low pass
or high pass FIR at exactly Nyquist is desirable because this is a
specific case called "half band" where half the coefficinet become
zero. The further you deviate from the Nyquist/2 point, the more
coefficients are needed for a given quality. The deviation comment is
symetrical: the 1KHz is like 21.050KHz(=22.050-1KHz) the example here
is based on say 44.1KHz sampling.
You want some intition to work for you? It is tough to explain. Maybe
you can think of it as a need to contain at least a significant part
of a cycle in the FIR pipe line. After all, 100 samples near the peak
of a 20Hz sine wave are almost all at value of 1. What can you do with
that data? On the other hand 100 samples of a higher frequency audio,
say 10KHz let you look at some significant part of the wave shape...
In any case, try to do a low pass at say 500Hz, with 100dB, and it is
not a specific filter I am talking about, very general comment here,
and you get certain number of coefficients. Now double the sample
rate, and your requirnment doubled. The point is that it is about the
RATIO between sample rate to corner frequency. A 44.1KHz 100Hz fith
certain coefficients yields say Xdb with Y transition band... Double
the sample rate to say 88.2KHz, and the SAME coefficients will give
the Same X and Y performance but at 200Hz - the ratio of sample rate
to corner frequency stays the same. You still want 100Hz? Start adding
compute power and a lot of ot!

One only needs to realize that a 100Hz FIR at 192KHz is THE SAME
FILTER as 25Hz at 48KHz... Do not anyone tell me that a 25Hz FIR is
easy. It is huge, even for real crapy DSP. That is why at very low and
very high frequency, folks do IIR's, with all the phase compromises...
And that is when coefficient acuracy may realy bite you. No one does
FIR DC removal, because the frequency ratio (sampling to corner) is
way too much at 44.1KHz. It more than 4 times worse at 192Khz.

All that talk about frequncy resolution being better with higher
sampling, that is also upside down! Say you want to do some peaking at
1Khz, and you want it to be a certain bandwidth. For any given compute
power, you have better control with the lower sampling rate.

BR
Dan Lavry

dan lavry

November 20th 03, 11:30 PM

(KikeG) wrote in message >...

> This pre-echo (I'd better call it pre-ringing) is a consequence of the
> brickwall FIR filter cutoff frequency...

> So, it seems that this pre-ringing should have little importance in
> practice....

OK. I am glad you agree that preringing may not be a big deal. I was
just trying to be as carfull as posible so I brought the issue to the
discussion. You know who it is - if i didnot someone would say I
forgot some big issue...

BR

Dan Lavry

dan lavry

November 21st 03, 12:07 AM

(Mike Rivers) wrote in message news:<znr1069336652k@trad>...
> In article > writes:
>
> Hi, Dan. Welcome to rec.audio.pro.
>
> > I have wondered about how is it possible that a whole industry can go
> > in the wrong direction.
>
> Marketing. It's easier to sell a new format than it is to sell an
> improved version of an old one.

I am gald to hear someone else say it. It feels lonely :-(

> There are throries (though no valid tests yet) that suggest that
> what's up there that we can't hear affects what we can hear.... though I have no reason to believe (or not believe) that a similar improvement would be
> realized at 192 kHz....

I read the same studies, about the alfa brain waves... but even when
you take that research into account, the issue was about extending
audio to 26KHz or so. Certainly not to 96KHz bandwidth (192KHz
sampling). 88.2KHz sampling should more than take care of that extra
bandwidth...

> The "harm" is that more samples per cycle is easier to achieve than
> more accurate samples, so we are often sold the wrong technology
> because it's cheaper and it looks better than the previous technology
> on the ad copy.

You know, right after my tutorial AES presentation on AD's, when both
Dr. Rich Cabot and I were pretty vocal about that 192KHz crock, some
industry salesmen started talking about 384KHz. I think such talk is
done delibaratly so that folks will not go against 192KHz. By making
the the conversation about 384, folks may be more willing to end up
with 192K.

The conversation should be about the fact that 192KHz is a crock
suported by major semiconductor houses, huge worksation makers and so
on. It is one thing for some industry salesman to fall for a crock. It
is another thing to get out of that serious presentation start
promoting 384KHz!

Arny Krueger

November 22nd 03, 03:56 PM

"Mike Rivers" > wrote in message
news:znr1069336652k@trad

> In article >
> writes:

>> I have wondered about how is it possible that a whole industry can go
>> in the wrong direction.

> Marketing. It's easier to sell a new format than it is to sell an
> improved version of an old one.

Agreed. It's the tyranny of numbers. I find it extremely ironic that some of
the people who are the first to queue up behind larger number are people who
claim they are subjectivists.

>> So the sounds are negligible at those high frequencies, the mics will
>> not pick them, the speaker will not play it, and you can not hear it.
>> So how can such a crock take place?

Agreed. People routinely publicly enthuse about mics monitor speakers that
roll off rapidly above as little as 13 KHz or as much as 23 KHz. Needless to
say, these products do right within the nominal 20-20 KHz band. AFAIK the
mics and speakers that have anything like flat response up to 44 KHz can
probably be counted on one hand. They are also anything but flat if only a
little bit off-axis. Many of the recordings that people have been praising
the SACD and DVD-A re-releases of are made from 48 KHz digital recordings or
15 ips tapes. Response plummets like a stone above 23 KHz or so.

> There are theories (though no valid tests yet) that suggest that
> what's up there that we can't hear affects what we can hear. As I said
> in another posting, I related a listening demonstration where I heard
> what I considered a subjective improvement when moving from a 48 kHz
> to a 95 kHz A/D/A converter chain (and they were your converters, back
> when you were still selling under the db name), though I have no
> reason to believe (or not believe) that a similar improvement would be
> realized at 192 kHz.

And I'm still asking about bias controls.

>> Theoretically, there is "no harm" in more points, and there is "no
>> added good".

> The "harm" is that more samples per cycle is easier to achieve than
> more accurate samples, so we are often sold the wrong technology
> because it's cheaper and it looks better than the previous technology
> on the ad copy.

Agreed. Consider the irony of so-called "24 bit" converters whose dynamic
range underperforms earlier products that were by design and application 16
or 20 bit devices.

Tommi

November 23rd 03, 11:35 PM

"dan lavry" > wrote in message
om...
> I also see a lot of confusion regarding that Nyquist, upsampling,
> oversampling, gradual filters. Some folks think that a 96KHz AD will
> require a sharper anti aliasing filter than say 192KHz. This is
> typically very wrong. The 96KHz or 192KHz AD refers to the OUTPUT RATE
> of the converter. The antialias requirement is determined by the INPUT
> SAMPLING RATE which is usually way beyond 192Khz. This days, most
> modern AD's are running at input sampling rates of 3-12Mhz!!! DSD is
> 64fs and many mulibit IC's run even faster. So even with 50Khz audio,
> Nyquist is so high the a gradual 3 pole will yield 120dB at the input
> rate Nyquist. The outcome of the high rate modulator (input) is than
> down-sampled to whatever - 44.1, 96, 192.

Burr Brown,(a division of Texas Instruments), manufactures both one-bit and
true multibit converters. They recommend to use multi-bit converters for
"waveform synthesis applications requiring very low distortion and noise."
I'm not an expert in converter design, however, wouldn't this imply that
their finest converters are true multibit converters with an input sampling
rate of 192kHz or below? This would mean that 96kHz converters would
actually require a sharper anti-alias filter than 192kHz and so on.

November 24th 03, 02:58 AM

2 track mastering or a surround is feasible. but a full multi-track
session...well you would need some wicked sort of raid setup to have
your hard drives keep up with the data throughput.

dan lavry

November 24th 03, 03:08 AM

"Tommi" > wrote in message >...
> "dan lavry" > wrote in message
> om...
> > I also see a lot of confusion regarding that Nyquist, upsampling,
> > oversampling, gradual filters. Some folks think that a 96KHz AD will
> > require a sharper anti aliasing filter than say 192KHz. This is
> > typically very wrong. The 96KHz or 192KHz AD refers to the OUTPUT RATE
> > of the converter. The antialias requirement is determined by the INPUT
> > SAMPLING RATE which is usually way beyond 192Khz. This days, most
> > modern AD's are running at input sampling rates of 3-12Mhz!!! DSD is
> > 64fs and many mulibit IC's run even faster. So even with 50Khz audio,
> > Nyquist is so high the a gradual 3 pole will yield 120dB at the input
> > rate Nyquist. The outcome of the high rate modulator (input) is than
> > down-sampled to whatever - 44.1, 96, 192.
>
>
> Burr Brown,(a division of Texas Instruments), manufactures both one-bit and
> true multibit converters. They recommend to use multi-bit converters for
> "waveform synthesis applications requiring very low distortion and noise."
> I'm not an expert in converter design, however, wouldn't this imply that
> their finest converters are true multibit converters with an input sampling
> rate of 192kHz or below? This would mean that 96kHz converters would
> actually require a sharper anti-alias filter than 192kHz and so on.

The answer to your question will be very long. First, much of what the
IC makers has to be taken with a grain of salt. Some of it is
engineering and some is marketing, Engineering is not why .why they
sell 192KHz for audio.
But one may not realize that different applications require different
focus on specific aspects. For example, audio conversion does not
require the best accuracy in terms of DC offset or gain. An +/-10V
audio converter will not cause problems if the DC offset (error) is
1mV. A 5% gain is "no much.". Just adjust the volume by a tiny portion
of a dB. Try that 1mV and 5% for an instrumentation device. You do
want your meter to give precise readings... Than there are devices
that should match (in pairs or groups) so the offsert and gain must be
right on. For audio there is a set of requirnments (including
THD+N)....

Regarding the filter, I will say it again: The input of the AD
(modulator) of the DSD and the modern mulibit noise shaping
converters, run at at least 64fs and some are at 128fs or 256fs. So
from antialiasing stand point the filter can be prteey gardual. Say a
device is a 96K. That number is NOT what you use to figure the anti
aliasing. You look at the oversampling and find it is 64fs. That means
the input is sampled at about 6Mhz and Nyquist is 3Mhz. That is the
frequncy that you look at for antialiasing. Say you want to pass
30Khz, and the INPUT Nyquist is 3MHz. Say you want 100dB rejection at
3MHz (that is a lot because you do not expect to see much signal
there). Every pole yields 20dB per decade. You have 2 decades
(30-300K, 300K-3MHz). So a pole is worth 40dB attenuation at 3Mhz. so
3 poles yield 120dB.

In the good old days, a 44.1KHz required huge number of poles. This
days, the input is sampled much higher. That modulator high speed
output is decimated down to manufacture the 192/96/44... Just like DSD
goes from 64fs (2.8M sampling or so) to 44,1KHz CD rate, so do the
multibit high rate.

For audio, a more importent distinction should be made between noise
shaping and regular type AD's. Between single bit dhaping (DSD) and
mulibit shaping.

BR Dan Lavry

Tommi

November 24th 03, 05:58 AM

"dan lavry" > wrote in message
om...
> "Tommi" > wrote in message
>...

> > Burr Brown,(a division of Texas Instruments), manufactures both one-bit
and
> > true multibit converters. They recommend to use multi-bit converters for
> > "waveform synthesis applications requiring very low distortion and
noise."
> > I'm not an expert in converter design, however, wouldn't this imply that
> > their finest converters are true multibit converters with an input
sampling
> > rate of 192kHz or below? This would mean that 96kHz converters would
> > actually require a sharper anti-alias filter than 192kHz and so on.
>
> The answer to your question will be very long. First, much of what the
> IC makers has to be taken with a grain of salt. Some of it is
> engineering and some is marketing, Engineering is not why .why they
> sell 192KHz for audio.

> Regarding the filter, I will say it again: The input of the AD
> (modulator) of the DSD and the modern mulibit noise shaping
> converters, run at at least 64fs and some are at 128fs or 256fs.

> For audio, a more importent distinction should be made between noise
> shaping and regular type AD's. Between single bit dhaping (DSD) and
> mulibit shaping.

Dan, thanks for your informative comments. When you're talking about modern
multibit noise-shaping converters, do you actually mean 1-bit delta-sigma
converters which convert the data stream to PCM, or, say a 4-bit converter
with noise-shaping etc, or do you mean real 20-24 bit pcm converters?
I'm asking this because I don't understand why would you have to do a real
20+bit
converter with noise-shaping since the noise floor would be theoretically
over 120+dB down at all frequencies, so you wouldn't have to shift it
anywhere.

I have been under the impression that noise-shaping converters are generally
inferior
to real pcm converters without any kind of noise-shaping, one of the reasons
being that transient signals will have a poor resolution in noise-shaping
systems. If the signal doesn't endure for a long enough time, the error will
not be minimized by the noise shaper. As I said, I'm not an expert at
converters, so I'm still unsure about which type
of converter is most suitable to audio.

Scott Dorsey

November 24th 03, 03:20 PM

Tommi > wrote:
>
>Burr Brown,(a division of Texas Instruments), manufactures both one-bit and
>true multibit converters. They recommend to use multi-bit converters for
>"waveform synthesis applications requiring very low distortion and noise."
>I'm not an expert in converter design, however, wouldn't this imply that
>their finest converters are true multibit converters with an input sampling
>rate of 192kHz or below? This would mean that 96kHz converters would
>actually require a sharper anti-alias filter than 192kHz and so on.

Since when do you believe marketing propaganda?

I'm still trying to figure out why none of the current Burr-Brown offerings
have as good low level performance as the long-discontinued PCM-63 multibit
converter.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Tommi

November 24th 03, 10:11 PM

"Scott Dorsey" > wrote in message
...
> Tommi > wrote:
> >
> >Burr Brown,(a division of Texas Instruments), manufactures both one-bit
and
> >true multibit converters. They recommend to use multi-bit converters for
> >"waveform synthesis applications requiring very low distortion and
noise."
> >I'm not an expert in converter design, however, wouldn't this imply that
> >their finest converters are true multibit converters with an input
sampling
> >rate of 192kHz or below? This would mean that 96kHz converters would
> >actually require a sharper anti-alias filter than 192kHz and so on.
>
> Since when do you believe marketing propaganda?
>
> I'm still trying to figure out why none of the current Burr-Brown
offerings
> have as good low level performance as the long-discontinued PCM-63
multibit
> converter.
> --scott
> --
> "C'est un Nagra. C'est suisse, et tres, tres precis."

My point was more on the multibit/single-bit issue. One can always say that
IC manufacturers just want people to buy the most expensive circuits they
have, but it would seem logical that multi-bit converters are generally
better for A/D for example than Delta-Sigmas which convert to pcm, since
real multibit converters can cost over
6 times as much as 1-bits. This is also related to the SACD-DVD-A issue.

Scott Dorsey

November 25th 03, 03:36 PM

Tommi > wrote:
>
>My point was more on the multibit/single-bit issue. One can always say that
>IC manufacturers just want people to buy the most expensive circuits they
>have, but it would seem logical that multi-bit converters are generally
>better for A/D for example than Delta-Sigmas which convert to pcm, since
>real multibit converters can cost over
>6 times as much as 1-bits. This is also related to the SACD-DVD-A issue.

Why does expense have anything to do with quality?

Multibit converters are difficult to make linear, because of the trimming
issues. Lots of resistors that have to be very precise. So any good multibit
converter will be hand-trimmed and that is expensive.

Sigma-delta converters are easy to make linear, but it's hard to make them
quiet. Idle tones become a bit issue.

You pays your money and you takes your chance.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

dan lavry

November 25th 03, 08:50 PM

(Scott Dorsey) wrote in message >...
> Tommi > wrote:
> >
> >My point was more on the multibit/single-bit issue. One can always say that
> >IC manufacturers just want people to buy the most expensive circuits they
> >have, but it would seem logical that multi-bit converters are generally
> >better for A/D for example than Delta-Sigmas which convert to pcm, since
> >real multibit converters can cost over
> >6 times as much as 1-bits. This is also related to the SACD-DVD-A issue.
>
> Why does expense have anything to do with quality?
>
> Multibit converters are difficult to make linear, because of the trimming
> issues. Lots of resistors that have to be very precise. So any good multibit
> converter will be hand-trimmed and that is expensive.
>
> Sigma-delta converters are easy to make linear, but it's hard to make them
> quiet. Idle tones become a bit issue.
>
> You pays your money and you takes your chance.
> --scott

Hello Scott,

I agree with completely. I hope you do not mind if I use your comment
as a starting point to take it in a somewhat different direction.
Based on your comments, I suspect you will agree with me. So here I
go, with a long one:

One needs to make a distinction between a multi bit and a noise
shaping multi bit.

The classical multi bit is, as Scott points out, a more difficult
design, in terms of matching resistors or capacitors in some designs.
Most of these converters are beyond a simple R-2R ladder, with
improved architectures (such as segmented design) providing faster
settling and less resistor dependence. But still, they are precision
made devices with a lot of trimming. They provide a theoretical flat
noise floor to Nyquist. Given the high number of bits, the circuits
tend to operate slower than their noise shaping counter parts. That is
not to say that running the front end is not best to run at some
faster rate than X2 the required bandwidth, to help the anti aliasing
filter requirements.

Now, think of a 1MHz 16 bit AD. Say we decide to take the average of
every 2 adjacent samples. That will yield 500KHz, but we gain 3dB of
dynamic range. In theory we get 3dB improvement each time we divide
the clock by 2. So a 64fs which is 2*2*2*2*2*2 yield 6*3=18dB (3 bits
improvment). Of course one does not want to go 64fs for 18dB, but it
does demonstrate that speed and accuracy are a tradeoff, which is one
of the reasons why 192KHz - too fast for audio, is a crock!

The noise shaping multi bits are a different animal. After some
evolution, from a 2nd order 1 bit, to 3rd order, to 5th order 1 bit at
64fs (such as much of DSD), we stated getting into the area of
diminishing returns, in terms of increased filter order. The only way
to improve 1 bit significantly was to go faster, which is fine for an
AD maker, but too much data for DSD, which already takes a lot of
space.

But DSD offered that "perfect differential linearity". True it has a
large amount of noise rising fast at about 22KHz all the way to
1.4MHz, but it was the "linear way" to go. A 1 bit has only one
comparator and the decision is simple: Is the signal over zero or
under zero. If the threshold is not exactly zero, it is still ok (it
is just a DC offset). With more than 1 bit you have steps and they
must match extremely well (on the DA in the feedback loop of the noise
shaper). Say you have the following comparator levels -2,-1,0,1V. It
is no longer just a question of say "greater than or less than" -1V or
0V. The difference between thresholds must be the same, very
accurately so.

True, if you go for, say 3 bits noise shaper, the trimming and
precision is about 8 levels, much less than the classical (no noise
shaping) AD. But 8 levels at very high speed is no walk in the park.
Again, the faster you go, the less accurate! Keep that in mind - There
is alwayse a speed vs. accuracy tradeoff.

But some smart folks figured way to deal with it. The main one methode
is called DEM (dynamic elements matching). You basically take a lot of
elements (say resistors) and keep rotating which one to use. Given
that you do it very fast, you end up with some nice "average value".
Say one resistor is really off, but it is only 1 out of 16. It gets
used only 1/16 of the time, thus it does not "pull" the error "that
bad"…

The multi bit noise shaping AD's has less noise amplitude rise than
DSD, and with DEM, better dynamic range than the 1 bit DSD. But they
can not be used as a format. A say 3 bit noise shaping AD with 64fs is
makes 3 times more data than DSD. Now, some of the converter makers
choose 128fs or even 256fs. That is not going to go well for the
consumer 64fs 1 bit DSD format. But the amount of data, modulator
sampling rate, high frequency noise, are all just an part of an
intermediate step. That 3 bit 128fs modulator (for example) will be
made into an audio PCM by means of decimation.

Now here is an important point. Please think of a balloon, partially
filled with air – you squeeze at one place, the air goes to another
place… Given an oversampling rate and a given an N bit shaping, one
can design the modulator and decimator so that the noise will start
rising at say 22KHz (for a 44.1KHz sampling AD outcome), or 48KHz
audio (for a 96KHz outcome), or for a 96KHz audio bandwidth (for
192KHz sampling rate outcome). Guess what: The 22KHz case will yield
better audio band performance than the 48KHz and the 96Khz audio (for
192K sampling) provides much worst than the 0-22KHz performance. It is
like that balloon. You want more audio bandwidth and something has to
give. It is important to "park" the design at a good point. If you
just keep rasing the sampling rate, you may end up at 3 dB at a GHz
:-)

I can not argue with those that say we hear 22KHz. I doubt it that we
need more than 44 or 48KHz audio (88 or 96KHz sampling). But 192KHz is
a CROCK!!! 384 is a SUPER CROCK!!!

We love to talk about how to get from point a to b, by way of topics
such as bipolar vs FET's, comparing transient response or jitter of
sigma delta to the classical one… and so on. But here we are dealing
with FUNDUMENTALS. Humans can not hear 96KHz audio, most microphones
do not pick it, speakers don' play is, and there is hardly any sound…
We are at the place where the fundamental theory (Nyquist) states that
we do not get ANY benefit from adding samples, because the properly
filtered (anti aliasing) sampled signal, contains 100% of the
information – read it as "100% of the wave shape". More samples is not
more money, more pixels, more bits. More samples is more NOTHING. And
that is in THE FUNDUMENTAL THEORY OF SAMPLING. In practice, more
samples means everything is speeded up so the overall accuracy is
LESS! The old speed accuracy compromise.

So 192KHz is a CROCK!!! Just the mention of 384KHz or 192KHz is an
outrages attempt to have folks think "faster is better". Folks that
push that either lack some serious background, or are trying to figure
a way to sell more stuff. It is eiteher disrespectfull to reality or
to the customers. I wish we concentrated on improving what needs
improving, instead of screwing things up. I am doing my best to try
and stir that big ship back in the right direction. I can not do it
alone, but I am not alone. I am finding out many real good ears and
serious talent on the recording and mastering side that know what I am
saying "by ear". I do believe history will prove that science, math
and engineering will prevail over ignorance and short term marketing
plans.

BR
Dan Lavry

Rob Adelman

November 25th 03, 09:07 PM

dan lavry wrote:

> With more than 1 bit you have steps and they
> must match extremely well (on the DA in the feedback loop of the noise
> shaper). Say you have the following comparator levels -2,-1,0,1V. It
> is no longer just a question of say "greater than or less than" -1V or
> 0V. The difference between thresholds must be the same, very
> accurately so.

Dan, thanks for all of the information, though I have to admit most of
it is over my head. But somehow I am guessing that the above section is
a pretty critical piece of the puzzle.

-Rob

dan lavry

November 26th 03, 07:15 AM

Rob Adelman > wrote in message >...
> dan lavry wrote:
>
> Dan, thanks for all of the information, though I have to admit most of
> it is over my head. But somehow I am guessing that the above section is
> a pretty critical piece of the puzzle.
>
> -Rob

I am sorry. I have a massage and felt compelled to explain it with
some solid arguments. That I just got too technical. Let me rewind and
start again, though I may take some liberties to be less accurate and
more intuitive.

I someone told you that in order to draw a straight line, all you need
to know are 2 point, you will believe it. It talks to your common
sense. If some one tried to tell you that you need more points, you
will probably dismiss them.

Lets try drawing a circle through 3 points. Are 3 points enough to
draw that circle line? Slightly less intuitive is it not? But
manageable.

If I tell you that I have a curve that can go up or down or sideways
in a totally unpredictable way, you will realize that the more points
and the closer they are, the better the representation. So some folks
are saying: audio is like that complex curve, so give me more points –
increase the sample rate.

It may not be easy to grasp, but while audio is very complex, there
are some restrictions there, and it is not true that "anything goes".
The fact that we are dealing with some limited bandwidth (frequency
range) will, for example restrict that curved line representing the
sound from moving too fast (think of putting a restriction on the
maximum allowed slop). This is true for any wave, video, audio,
medical, instrumentation… The lower the bandwidth, the lower the
slope. I am not being completely accurate with slope, but higher
frequencies move faster.

The point is that the restrictions define the signal well enough so
that you do not need too many points to draw that line. Too few points
will not cut it, but you get to a certain level that allows you to
draw the line correctly. Just like 2 point for a straight line.

Some folks are trying to sell you on doubling the points, when you do
not need to. They call it 192KHz sampling. The extra points (samples)
take space, require you to double the processing power of your
machine, and in fact lower the quality of the outcome.

The argument is based on "more is better" which is often true, but not
always. Those that study EE and math know that it is not. Those that
do not have the background are just as likely to buy the BS, as they
are to buy the truth.

There are a lot of forces out there, from huge semiconductor houses to
huge workstation makers and their whole support network that have been
promoting that crock. With so much combined clout, few want to stand
up to it. And of course, such a myth gets propagated to the sales guys
that mostly lack the know how, and latch into that "more is merrier"
wrong explanation. We have a whole industry going in the wrong
direction.

But there are a lot of things in audio that are in the "gray area".
The above is not. We are dealing with fundamentals. The only argument
I can not deal with is: but I like it. Or It sounds great to me. Fine
if it does, but my point is: You do not need to go and double the data
and also double the processing to get that thing you like, If you have
a certain characteristic (distortion) you like, I can make it for you
with a 96KHz AD or lower.

I too want to improve quality, and there are things to do. But going
above 96KHz is screwing things up. Math engineering and science is on
my side of the argument. History will prove it, and hopefully very
soon. Meanwhile I am sorry to see folks pay good money to be taken to
a ride in the wrong direction.

Of course, those that got influenced to belive they are getting better
sound, are in a bind. It takes a "hack of a man" or a woman to go back
on it, to admitt you were wrong. Certainly such is the case in this
industry. And it is always "acceptble" to just say "but I like it". In
audio, you call it "an artistic decision" and no one will argue, well
almost no one...

I do not want to make anyone uncomfortable. I just want that 192 and
to have 96 accepted as morer than enough. I hoe it does and soon.
While observing some recording and mastering guys "go with the flow"
of faster is better, I am very pleased to see some top notch ears
that figured it out ""by ear". That is encouraging.

I hope this is clear and direct enough.

Dan Lavry

Tommi

November 26th 03, 09:38 AM

"dan lavry" > wrote in message
om...

> I do not want to make anyone uncomfortable. I just want that 192 and
> to have 96 accepted as morer than enough. I hoe it does and soon.
> While observing some recording and mastering guys "go with the flow"
> of faster is better, I am very pleased to see some top notch ears
> that figured it out ""by ear". That is encouraging.
>
> I hope this is clear and direct enough.
>
>
> Dan Lavry

Dan, you have explained your point about 192kHz problems very well!
It's sad to see some of these so-called 24-bit/192kHz converters which
really only have dynamics of about 100dB with bad distortion etc., marketed
with a 24/192 tag.
It appeals to some people because it's new, even though I must admit that I
myself didn't know about the 192kHz problems until you presented some
serious information.

Even though my head is still trying to figure out the various differences
between the variations of single- and multibit, shaping and non-shaping
converters,and their unique flaws/merits, your point is well made.

Mike Rivers

November 26th 03, 02:45 PM

In article > writes:

> If I tell you that I have a curve that can go up or down or sideways
> in a totally unpredictable way, you will realize that the more points
> and the closer they are, the better the representation. So some folks
> are saying: audio is like that complex curve, so give me more points –
> increase the sample rate.
>
> It may not be easy to grasp, but while audio is very complex, there
> are some restrictions there, and it is not true that "anything goes".
> The fact that we are dealing with some limited bandwidth (frequency
> range) will, for example restrict that curved line representing the
> sound from moving too fast (think of putting a restriction on the
> maximum allowed slop).

The way I like to explain this is to imagine that you're driving a
race car on a waveform-shaped track. If that track is a perfect sine
wave, you'll have to go at a certain speed in order to complete a
cycle in a given time. If you now put some more bends in the track,
you'll be going a greater distance from end to end. If you want
to match your time for the simple curved track, you'll have to drive
faster. Put more kinks in it and you'll have to drive still faster.
When the track gets sufficiently contorted, ignoring things like
centrifical force and coefficient of friction that cause you to slow
down for turns, eventually the path will be long enough so that you
simply can't get up enough speed to get to the end of the track within
the proscribed time.

If you can somehow increase your maximum speed (like by putting a
bigger engine in the car) you can then again meet your mark. The speed
of the car represents the sample rate, the turns in the track
represent the frequency because they increase the distance traveled in
a fixed amount of time.

--
I'm really Mike Rivers - )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

Hp Widmer

November 26th 03, 03:09 PM

some additional comments:

I did some measurements using an analog burst 10khz signal who is
complete not in sync with the sample clock.

To get a fine none-overshouting (pre/post) sampled signal a high
sample rate is required to get ride of the ringing (about 10 times
related to burst sine freq.).

I read some time ago in a paper that 1/100% ringing is audible....

Cheers

Hp

Len Moskowitz

November 26th 03, 05:03 PM

Tommi > wrote:

>It's sad to see some of these so-called 24-bit/192kHz converters which
>really only have dynamics of about 100dB with bad distortion etc., marketed
>with a 24/192 tag.

Well, there are other converters that are spec'd at 24/192 that offer
120+ dB of dynamic range, very fine distortion specs and very low
noise. And they sound fine.

What surprises me most these days are the folks who complain about the
sound of high res digital audio but extol LPs (vinyl) with their 30 dB
max of separation and surface noise.

--
Len Moskowitz PDAudio, Binaural Mics, Cables, DPA, M-Audio
Core Sound http://www.stealthmicrophones.com
Teaneck, New Jersey USA http://www.core-sound.com
Tel: 201-801-0812, FAX: 201-801-0912

Mike Rivers

November 26th 03, 05:11 PM

In article > writes:

> Dan, you have explained your point about 192kHz problems very well!
> It's sad to see some of these so-called 24-bit/192kHz converters which
> really only have dynamics of about 100dB with bad distortion etc., marketed
> with a 24/192 tag.

They get away with it because people see "24/192" and don't see the
actual dynamic range specifications.

In the early days of 96 kHz products, a lot converters (in both
directions) that sounded pretty good at 48 kHz didn't sound quite as
good at 96 kHz. I think those problems have been fixed, but the
numbers for S/N ratio and dynamic range are limited (at least with
good quality chips) by analog circuitry, short term clock stability,
board layout (which affects all of the above) and physics.

--
I'm really Mike Rivers - )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

dan lavry

November 26th 03, 09:18 PM

Hp Widmer > wrote in message >...
> some additional comments:
>
> I did some measurements using an analog burst 10khz signal who is
> complete not in sync with the sample clock.
>
> To get a fine none-overshouting (pre/post) sampled signal a high
> sample rate is required to get ride of the ringing (about 10 times
> related to burst sine freq.).
>
> I read some time ago in a paper that 1/100% ringing is audible....
>
> Cheers
>
> Hp

A burst is not a band limited signal. In fact, making a sharp corner
takes infinite bandwidth. So if you take a zero signal and all of a
sudden you "shoot up" into the first cycle of the burst, and also the
ending of it- sudenly to zero, it takes huge bandwidth.

That is why we window FFT's, and why we also the best FIR filters are
done with the window method. Loosly speaking, windowing anounts to a
very gradual taper at the start and end of the wave, sort of like fade
in and fade out that mastering and music editors do.

I am not sugesting you need to window andything, but I sugest that you
need to make sure that the burst is FILTERED with proper anti alaising
filter. If you did not filter the high frequency content of the burst,
you will have alaising.

If you filter it properly and the rest is done correctly, your
conclusion will change. If you get ANY more detail when sampling
faster there are 2 possibilities:
1. Something is wrong with the test or the setup
2. Nyquist was wrong,. Shannon was wrong, math does not work and
science are wrong.

I will not get into much back and forth regarding issues that are as
solid as the law of garvity.

I do not know why after my long post, I have to deal with such a
response. You are saying "I tested it and sampling theorm is wrong". I
would be inclined to figure out what is wrong with the test or the
setup.

Dan Lavry

Glenn Zelniker

November 27th 03, 01:20 AM

Arny Krueger wrote:

> You said:
>
> "For instance, attempting to design a 4th order Butterworth high
> pass filter with a cutoff of 20Hz at a sample rate of 48kHz is
> difficult in 32 bit floating point but it can be done in 64
> bit double precision floating point."
>
> So let's put your claim to the test. Cool Edit Pro has a variable frequency
> high pass Butterworth filter of any reasonable order including 4. It can be
> run at any sample rate up to 999,999 Hz including 48 Khz and 384 KHz. AFAIK
> it is implemented in 32 bit floating point arithmetic.

Aaahhh. A discussion that's near and dear to my day's work.

You *did* misinterpret what Erik said. His point was that moving to
sample rates *complicates* matters considerably. In other words, it's
something that a DSP guy dreads.

I can assure you from extensive personal experience that if you design,
say, a highpass filter of fixed order and a cutoff frequency that's very
low relative to Fs, implementing a filter with the same cutoff frequency
(in absolute Hz, *not* in rad/sample) at 2Fs is much more difficult. 8Fs
is even worse. And maintaining the same *transition* band makes it even
uglier, as the required order grows with Fs.

What do I mean by difficult? Specifically, I'm talking about the
roundoff noise power and its attendant spectral distribution. For an IIR
filter, this metric gives an indication of the filter's output's
dependence on and vulnerability to re-circulated round-off noise errors.
And it also gives some indication of the relative amplitudes of the
filter's internal states. Here's a simple experiment you can do yourself
if you have a good analysis tool like Matlab. Design a HP filter with a
cutoff of 28 Hz and a transition band of 10 Hz and a sampling rate of
44.1 kHz. Now use those same specs and do it for 88.2 kHz. Now do it for
176.4 kHz. On each of those filters, use Matlab's qfilt objects to
generate finite-wordlength equivalents and set the quantization format
to {float,32,8}, architect the thing as a direct-form filter and run a
noise-load analysis on it. You'll see that direct-form is horrible at
44.1 kHz, hopeless at 88.2 kHz, and lethal at 176.4 kHz. We're talking
about so much noise that the filter won't work.

One's first thought is to re-implement the thing as cascaded
second-order sections. The results are stil bad, terrible, useless,
respectively. Doing it as a lattice/ladder only helps a little. In fact,
the only way to make it "work" is to resort to cascaded exotic
section-optimal minimum-roundoff structures, a la Roberts and Mullis.

The point here is not that it can't be done. But one must take heroic
measures to make it work properly at higher sample rates. (By
*properly*, I mean in the worst case, stable; in the best case, as good
as the "easy" filters). The theory is very, very solid on this. Have a
look at Dick Roberts and Cliff Mullis's book -- they give the most
elegant geometric explanation of the phenomenon to date. But try the
expermient yourself. It's maddening!

(In case anybody is wondering, I have to live with these rates because
of DSD. It wasn't my choice!)

> What bad thing to look for when operating it? Bad frequency response?
> Nonlinear distortion? Bad phase response?

Overwhelming noise and graininess on the output, clipping, limit cycles,
grunge, etc. Just make sure you force the transition band to be tight.
Or, if you like, try a peaking filter centered at 28 Hz with a Q of 10
or greater and a boost of 8 dB. And just because you don't hear any
"nasties" doesn't mean the filter is really doing what you asked of it!
I've seen many a designer cheat and implement a fatter filter in order
to avoid the problem.

Or, it may indeed be the case that your program *does* work properly, in
which case the designer applied some real TLC to the DSP programming.
But it's still waaaaaaay harder to make it work at higher Fs.

enjoy!

Glenn @ Z-Systems

dan lavry

November 27th 03, 04:22 AM

(Mike Rivers) wrote in message news:<znr1069858374k@trad>...
> In article > writes:
>
> > Dan, you have explained your point about 192kHz problems very well!
> > It's sad to see some of these so-called 24-bit/192kHz converters which
> > really only have dynamics of about 100dB with bad distortion etc., marketed
> > with a 24/192 tag.
>
> They get away with it because people see "24/192" and don't see the
> actual dynamic range specifications.

You are correct. First, as I stated before, all the 192 gear and IC's
I saw, and I look all the time (!!!) is specfied with A weighting.
Most of the other gear 96KHz gear is is specified A weighted. Next,
many of the AD converter companies just copy the IC specifications as
if it "exsists in mid air". You look at the finished design, including
the front end circuitry, power supply, clocks... you will be surprised
at the discrapancy there.

I am not saying the "ordinary specs" are the only thing that matter.
There are a lot of things not on the spec sheet, and should be there,
that make it or break it.

Here is an example: I am often amazed by how folks insist on ,1dB
flatness response. Not that it is difficult to do. I glad to comply
with that. But did you look at a speaker resonse latley? It is up and
down by dB's. It is a mess. Somewhere smack in the middle, there is a
crossover network that breaks that tone made out of say 500Hz into
(Assume 1.2KHz cross over for the example):

A. 500Hz fundumental and 1KHz first harmonic go to to the large cone.
B. 1.5Khz, 2KHz,.... 3rd forth and so on harmonics to the smaller
speaker cone

And guess what, there is a huge phase shift in there around the cross
over range...

And guess what else, that single say 500Hz piano note is played so
that some of the sound comes from the lower cone and the rest from the
upper cone...
If you play 100Hz, now you have the fundumntal and 10 harmonics or so
come out of the bottom large speaker...

When folks demand flatness I give it, but I do think they are barking
up the wrong tree. When the talk 192 I know they are full of it. When
they like a well preserved vinal, I at least understand some of what
they like. .

But transfering vinal to a 192 cracks me up.

BR
Dan Lavry

Mike Rivers

November 27th 03, 03:03 PM

In article > writes:

> You are correct. First, as I stated before, all the 192 gear and IC's
> I saw, and I look all the time (!!!) is specfied with A weighting.

I've noticed that too. It's amazing how much power supply hum goes
away with that filter, and stray clock hash too. The first computer
audio interface that I reviewed, the Echo Layla, on quick measurement
had an outrageous amount of noise, though it didn't sound noisy. A
look at the output with a scope showed lots of stuff above 22 kHz.
Applying a 20 kHz filter ahead of the meter let me get closer to the
manufacturer's published noise spec.

> many of the AD converter companies just copy the IC specifications as
> if it "exsists in mid air".

I've noticed that, and also just using the theoretical number of
96 or 144 dB (depending on the word length) as well.

> I am not saying the "ordinary specs" are the only thing that matter.
> There are a lot of things not on the spec sheet, and should be there,
> that make it or break it.

The sad thing is that a lot of people who buy this stuff have nothing
to go on but specifications. When you're dealing with people with so
little experience and knowledge that they think the only difference
between a mic and a line input is the size of the connector, it's easy
to absorb a lot of irreleveant information and then be confused when
trying to sort it out.

> Here is an example: I am often amazed by how folks insist on ,1dB
> flatness response. Not that it is difficult to do. I glad to comply
> with that. But did you look at a speaker resonse latley?

There must be something to frequency response. I hear people talk all
the time about fixing a "problem" in mastering by boosting something a
couple of tenths of a dB. I can't say as I hear it myself, but then
I'm just average.

--
I'm really Mike Rivers - )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

Scott Dorsey

November 27th 03, 03:18 PM

Hp Widmer > wrote in message >...
> I did some measurements using an analog burst 10khz signal who is
> complete not in sync with the sample clock.
>
> To get a fine none-overshouting (pre/post) sampled signal a high
> sample rate is required to get ride of the ringing (about 10 times
> related to burst sine freq.).

This is what bandlimiting does. That ringing is the result of the
bandlimiting, and the stuff you are seeing is all well above 20 KHz.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Dave Collins

November 28th 03, 04:32 AM

In article <znr1069943194k@trad>, (Mike Rivers)
wrote:

> There must be something to frequency response. I hear people talk all
> the time about fixing a "problem" in mastering by boosting something a
> couple of tenths of a dB. I can't say as I hear it myself, but then
> I'm just average.

People say a lot of things. What the those engineers should realize, is
that a tenth dB "correction" is absoultely meaningless in the real
world! No speaker, ear, room, etc. is nearly that accurate...

It's just a desire on the part of the engineer to do _something_ rather
that cut it flat.

DC

Arny Krueger

November 28th 03, 11:19 AM

"Len Moskowitz" > wrote in message

> Tommi > wrote:
>
>> It's sad to see some of these so-called 24-bit/192kHz converters
>> which really only have dynamics of about 100dB with bad distortion
>> etc., marketed with a 24/192 tag.
>
> Well, there are other converters that are spec'd at 24/192 that offer
> 120+ dB of dynamic range, very fine distortion specs and very low
> noise. And they sound fine.
>
> What surprises me most these days are the folks who complain about the
> sound of high res digital audio but extol LPs (vinyl) with their 30 dB
> max of separation and surface noise.

It's got to be the sentimentality factor. It ain't the technical
performance, that's for sure!

Mike Rivers

November 28th 03, 12:32 PM

In article > writes:

> > There must be something to frequency response. I hear people talk all
> > the time about fixing a "problem" in mastering by boosting something a
> > couple of tenths of a dB.

> It's just a desire on the part of the engineer to do _something_ rather
> that cut it flat.

Wait a minute. Aren't you a mastering engineer? But then most of us
have "fixed" something by turning a control on something that we
didn't realize was bypassed.

--
I'm really Mike Rivers )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

Bob Cain

November 28th 03, 04:51 PM

Mike Rivers wrote:
>
> Wait a minute. Aren't you a mastering engineer? But then most of us
> have "fixed" something by turning a control on something that we
> didn't realize was bypassed.

Whew! I'm glad to hear you say that. I feared I was the
only one who had fallen into that trap. :-)

Bob
--

"Things should be described as simply as possible, but no
simpler."

A. Einstein

Dave Collins

November 28th 03, 11:08 PM

In article <znr1070019730k@trad>, (Mike Rivers)
wrote:

> Wait a minute. Aren't you a mastering engineer? But then most of us
> have "fixed" something by turning a control on something that we
> didn't realize was bypassed.

Yes, I am. One that is realistic enough to know that you can't "fix"
anything by boosting 0.1dB... I think it's more of a canine-equine show
for the client "Ooh, listen to the air when I add .2 @20k Q1" That's
bound to increase sales.

Back when everything sounded great, the eq's were in 2dB steps.....

DC

Roger W. Norman

November 29th 03, 01:21 PM

"Mike Rivers" > wrote in message
news:znr1070019730k@trad...
> Wait a minute. Aren't you a mastering engineer? But then most of us
> have "fixed" something by turning a control on something that we
> didn't realize was bypassed.

All too often.

--

Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.

Roger W. Norman

November 29th 03, 01:27 PM

Ah, but just the bigger engine (a step up in converter sampling) futzes with
the drag coefficient, weight and balance, and g force so that the bigger
engine doesn't turn in any better time. Like in drag racing, after 190 mph
it costs an addition 14 horse power to get even one mph more. But if you
don't spend the money to redesign the drive chain, etc., the extra horse
power won't get you going any faster.

--

Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.

"Mike Rivers" > wrote in message
news:znr1069853646k@trad...
>
> In article >
writes:
>
> > If I tell you that I have a curve that can go up or down or sideways
> > in a totally unpredictable way, you will realize that the more points
> > and the closer they are, the better the representation. So some folks
> > are saying: audio is like that complex curve, so give me more points –
> > increase the sample rate.
> >
> > It may not be easy to grasp, but while audio is very complex, there
> > are some restrictions there, and it is not true that "anything goes".
> > The fact that we are dealing with some limited bandwidth (frequency
> > range) will, for example restrict that curved line representing the
> > sound from moving too fast (think of putting a restriction on the
> > maximum allowed slop).
>
> The way I like to explain this is to imagine that you're driving a
> race car on a waveform-shaped track. If that track is a perfect sine
> wave, you'll have to go at a certain speed in order to complete a
> cycle in a given time. If you now put some more bends in the track,
> you'll be going a greater distance from end to end. If you want
> to match your time for the simple curved track, you'll have to drive
> faster. Put more kinks in it and you'll have to drive still faster.
> When the track gets sufficiently contorted, ignoring things like
> centrifical force and coefficient of friction that cause you to slow
> down for turns, eventually the path will be long enough so that you
> simply can't get up enough speed to get to the end of the track within
> the proscribed time.
>
> If you can somehow increase your maximum speed (like by putting a
> bigger engine in the car) you can then again meet your mark. The speed
> of the car represents the sample rate, the turns in the track
> represent the frequency because they increase the distance traveled in
> a fixed amount of time.
>
>
>
> --
> I'm really Mike Rivers - )
> However, until the spam goes away or Hell freezes over,
> lots of IP addresses are blocked from this system. If
> you e-mail me and it bounces, use your secret decoder ring
> and reach me here: double-m-eleven-double-zero at yahoo

Mike Rivers

November 29th 03, 04:14 PM

In article > writes:

> Yes, I am. [a mastering engineer] One that is realistic enough to know
> that you can't "fix" anything by boosting 0.1dB...

> Back when everything sounded great, the eq's were in 2dB steps.....

Back then, in mastering, what you "fixed" wasn't what's wrong that's
keeping the record from selling millions, you fixed what would keep
you from cutting it properly. When you got the pressings, you expected
them to be missing a little of what you heard in the recording studio.
You didn't expect them to sound a whole lot better.

--
I'm really Mike Rivers - )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

Roger W. Norman

November 29th 03, 11:38 PM

You're telling Dave?

--

Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.

"Mike Rivers" > wrote in message
news:znr1070112071k@trad...
>
> In article >
writes:
>
> > Yes, I am. [a mastering engineer] One that is realistic enough to know
> > that you can't "fix" anything by boosting 0.1dB...
>
> > Back when everything sounded great, the eq's were in 2dB steps.....
>
> Back then, in mastering, what you "fixed" wasn't what's wrong that's
> keeping the record from selling millions, you fixed what would keep
> you from cutting it properly. When you got the pressings, you expected
> them to be missing a little of what you heard in the recording studio.
> You didn't expect them to sound a whole lot better.
>
>
>
> --
> I'm really Mike Rivers - )
> However, until the spam goes away or Hell freezes over,
> lots of IP addresses are blocked from this system. If
> you e-mail me and it bounces, use your secret decoder ring
> and reach me here: double-m-eleven-double-zero at yahoo

dan lavry

November 30th 03, 01:05 AM

(Scott Dorsey) wrote in message >...
> Hp Widmer > wrote in message >...
> > I did some measurements using an analog burst 10khz signal who is
> > complete not in sync with the sample clock.
> >
> > To get a fine none-overshouting (pre/post) sampled signal a high
> > sample rate is required to get ride of the ringing (about 10 times
> > related to burst sine freq.).
>
> This is what bandlimiting does. That ringing is the result of the
> bandlimiting, and the stuff you are seeing is all well above 20 KHz.
> --scott

It is easier to relate to the ringing with low frequency squate waves.
That 10KHz burst is "mixing up" a lot of things.
Say you take a 1KHz square wave. Say we talk about 44.1KHz. Than there
are 11 harmonics (decaying in amplitude). For square wave only odds
exist (no even harmonics) and as you add them, you find the decaying
ringing, and yes it is at high frequency. If you go to 100Hz, you have
110 harmonics.
As scott pointed out, the added harmonics are never enough to express
a square wave. With limited bandwith, the missing high frequncy energy
shows up as deviation from perfect square, by means of high frequency
overshoot and ringing. the DIFFERENCE between the perfect and
bandlimited square wave IS HIGH FREQUENCY CONTENT.

BR

Dan Lavry

Hp Widmer

November 30th 03, 09:49 AM

Hi Dan,

>It is easier to relate to the ringing with low frequency squate waves.
>That 10KHz burst is "mixing up" a lot of things.

The real question will be : Bandlimit vs. transients given from some
natural instruments and the required sample rate. In other words how
do we really hear or we sime mask above 1xkHz.

When I listen to my 192khz/24bit Stereo DVD-A Eagles (Hotel
California) or Neil Young (Harvest) there is so much more MUSIC coming
out...

Hp

Roger W. Norman

November 30th 03, 01:57 PM

"Hp Widmer" > wrote in message
...
> When I listen to my 192khz/24bit Stereo DVD-A Eagles (Hotel
> California) or Neil Young (Harvest) there is so much more MUSIC coming
> out...
>
>
> Hp
>

Than what? How do you make a comparative when you give nothing by which the
comparison is made?

--

Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.

Scott Dorsey

November 30th 03, 04:13 PM

Hp Widmer > wrote:
>When I listen to my 192khz/24bit Stereo DVD-A Eagles (Hotel
>California) or Neil Young (Harvest) there is so much more MUSIC coming
>out...

How much of this is because the DVA-A is well done, and how much of it
is because the CD release was just butchered?

The CD release of Hotel California is just nasty-sounding. Harsh and
screechy, but with no real detail. It sounds nothing at all like the
original LP. That's no slam against the CD format, that is a slam against
the people who horribly bungled the CD release.

On the other hand, does it really matter? If the DVD-A sounds good, that
is a reason to go to DVD-A, whether it's a technical improvement or just
a social one. I mean, I'm the first one to acknowledge that there are serious
problems with LPs, but I'm not getting rid of my turntable because of the
number of recordings out there that sound better on LP because the CD release
was so poorly handled.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Rob Adelman

November 30th 03, 04:36 PM

Roger W. Norman wrote:

>>When I listen to my 192khz/24bit Stereo DVD-A Eagles (Hotel
>>California) or Neil Young (Harvest) there is so much more MUSIC coming
>>out...

> Than what? How do you make a comparative when you give nothing by which the
> comparison is made?

Then before he hits play? ;)

Rob Adelman

November 30th 03, 04:38 PM

Scott Dorsey wrote:

> The CD release of Hotel California is just nasty-sounding. Harsh and
> screechy, but with no real detail. It sounds nothing at all like the
> original LP. That's no slam against the CD format, that is a slam against
> the people who horribly bungled the CD release.

I have this LP. While the songs are classics, and some really good stuff
I was always disappointed with the overall sound. It seems a little 2d
to me. I never had the CD but I can imagine.

-Rob

Justin Ulysses Morse

November 30th 03, 05:20 PM

Scott Dorsey > wrote:

> How much of this is because the DVA-A is well done, and how much of it
> is because the CD release was just butchered?
>
> The CD release of Hotel California is just nasty-sounding.

Yeah, but I'm pretty sure the Eagles played on the DVD-A release also.

ulysses

Remixer

November 30th 03, 05:47 PM

After you've worked with analog and digital formats for a while, you know
that 44.1k sampled reproduction does not retain the detail, air, and open
sounding quality of an analog source. This is not the case with 2Fs or 4Fs
sampling, at least not in practice. You get to recognize the closed in,
blurry quality of CDs in general vs the more natural sound of DVD-A or SACD.
Before someone rises up to smack me down with the sampling theorem, and how
1Fs is all we'll ever need because of the bandwidth the human auditory blah
blah blah, the fact is that in *practical implementation* the promise of the
sampling theorem is not fulfilled, even with the best available 1Fs dacs.
I've recently been doing some tests with 44.1k 16-bit recordings upsampled
to DSD with a new "trellis" algorithm from Philips. The playback of the
upsampled-to-DSD files through a DSD dac has better detail and transient
response than what can be heard from the original 44k file played through a
variety of the best 1Fs dacs in the industry. This may be due to the fact
that there is actually more information encoded in a good 44.1k recording (a
la the sampling theorem) than we can playback with real world
top-of-the-line PCM dacs, but more of that information does come through the
upsampled DSD playback. Not trying to prove anything, only reporting what I
can hear.

"Roger W. Norman" > wrote in message
> Than what? How do you make a comparative when you give nothing by which
the
> comparison is made?
>

Rob Adelman

November 30th 03, 07:10 PM

Remixer wrote:

> Not trying to prove anything, only reporting what I
> can hear.

But, but, but..... ;)

dan lavry

November 30th 03, 09:36 PM

Hp Widmer > wrote in message >...
> Hi Dan,
>
> >It is easier to relate to the ringing with low frequency squate waves.
> >That 10KHz burst is "mixing up" a lot of things.
>
> The real question will be : Bandlimit vs. transients given from some
> natural instruments and the required sample rate. In other words how
> do we really hear or we sime mask above 1xkHz.
>
> When I listen to my 192khz/24bit Stereo DVD-A Eagles (Hotel
> California) or Neil Young (Harvest) there is so much more MUSIC coming
> out...
>
>
> Hp

You seem to be insisting that in suport of 192KHz despite the fact
that it is against all aspects of physics and math. There is a
research artical done with the greatest care, that after searching
around for tons of cases, talks about some special case showing a
muted trumpet yielding -55dB ay around 45KHz and almost zero at 90KHz.
(96KHz AD will be enough for that).
The mics used are special measuring types. Not the kind folks use for
recording. "Ordinary" mics drop off way below the 120KHz test mic. The
speakers do also. The human ear does too.

So here is the old BS about the mysterious transients. Transients are
subject to bandwidth limitations. It may be true that a Fourier series
is based on periodic waves so it is not the way to approach
transients. So what? So you use different math. Nyquist was still
right. You bandlimit the signal to a frequency called Nyquist, and
sample over twice that rate, and you got ALL the information in the
data. Not just sine waves that last for ever. The transients you hear
are made out of energy in the same bandwidth called audio.

I know I did not say it that way earlier. I though it would be enough
to say that if the mics don't pick it, the speakers do not sound it...
it places the same limitation on the bandwidth of all music - which
includes transients.
Anything outside constant preiodic waves that started at the beginng
of time and will last for ever, includes some transient energy! I
guess you are trying to talk about "fast transients". There are huge
misconceptions around that. For example, many folks thing that a
bandlimited (say to 22KHz) 1KHz square wave rises faster (more slew
rate - volts per second) than a sine wave at 20KHz. I can go on but
will not.

I have been designing for a long time and know about transient
problems such as capacitor dialectric absorption in sample hold. That
is not a 192KHz related. It is at the oposite side - low frequencies
way within the ausio band. If it fixing it took changing from 96 to
192K sampling, you would not hear it! That is the whole point. 192KHz
will not fix a thing for audio.

Folks keep throwing misleading garbage based on lack of understanding,
and I can not answer it for ever. I need to work for a living, but I
am right, so the billion dollar conglamerates will chock me and the
truth but constatly throwing garbage at such volume that it will be
unanswered.

And whan it all gets too much for them to handel, there will always be
that last resort. That last argument that no one can stand to: It
sounded better. This argument will yield best results if you get some
real big names. Guess what- I have some real big names agreeing with
me that 96KHz is better than 192. I did not ask permision to say who,
but I put my inegrity on the line.

In either case, I came accross a big name that refused to audition 96
on the grounds that 192 contains "something" that staed with the music
at 44.1 (after decimation. Now, I will not ever calim that I know 1%
about recording when compared to that gentelman. I just wish that I
was recipicated with the same respect regarding Math and Engineering.

I would like to belive that there is a differance between the
stereophile and the pro community. Not all stereophile stuff is bad,
but much of it is very ridiculess. You do not hear a differance
between a yellow and orange cable. Arrowa on speakers is insanly
stupid. The argument is always about "It sounds better to me"...
Followed by someone getting ripped off.

Folks this is about getting our industry back on track. Double the
data. double the processing and LOWERING THE QUALITY OF CONVERSION is
a high price to pay. Science, and egineering and math on one hand, led
by greats such as Shanon and Nyquist, and the billion dollars
industial companies using every trick in the book (including avoiding
the physics and engineering books) to sell thiere BS. I belive it is
about marketing and money.

I realize that if one fell for it, it is difficult to admit. But
192KHz for audio is a crock! The only argumet left is "but I like it".
No one can stand to that one. If I tell you you added 10% distortions,
and you say "I like it", you win. So lets compromise: I say 192KHz s a
crock. You agree it is but you get to like it.

Sorry for my tone. I put out some energy to do good. I lost a big sale
because I refuse to play that 192KHz game. I try to educte folks about
it. First it was "more dots are better". Than it was "better
antialiasing". Than "transients". All very wrong and based on far from
sufficient understanding of the basics. What's next?

I am still here, but getting tired of it.

Dan Lavry

Bobby Owsinski

December 1st 03, 12:41 AM

In article >,
(dan lavry) wrote:

> Hp Widmer > wrote in message
> >...
> > Hi Dan,
> >
> > >It is easier to relate to the ringing with low frequency squate waves.
> > >That 10KHz burst is "mixing up" a lot of things.
> >
> > The real question will be : Bandlimit vs. transients given from some
> > natural instruments and the required sample rate. In other words how
> > do we really hear or we sime mask above 1xkHz.
> >
> > When I listen to my 192khz/24bit Stereo DVD-A Eagles (Hotel
> > California) or Neil Young (Harvest) there is so much more MUSIC coming
> > out...
> >
> >
> > Hp
>
> You seem to be insisting that in suport of 192KHz despite the fact
> that it is against all aspects of physics and math. There is a
> research artical done with the greatest care, that after searching
> around for tons of cases, talks about some special case showing a
> muted trumpet yielding -55dB ay around 45KHz and almost zero at 90KHz.
> (96KHz AD will be enough for that).
> The mics used are special measuring types. Not the kind folks use for
> recording. "Ordinary" mics drop off way below the 120KHz test mic. The
> speakers do also. The human ear does too.
>
> So here is the old BS about the mysterious transients. Transients are
> subject to bandwidth limitations. It may be true that a Fourier series
> is based on periodic waves so it is not the way to approach
> transients. So what? So you use different math. Nyquist was still
> right. You bandlimit the signal to a frequency called Nyquist, and
> sample over twice that rate, and you got ALL the information in the
> data. Not just sine waves that last for ever. The transients you hear
> are made out of energy in the same bandwidth called audio.
>
> I know I did not say it that way earlier. I though it would be enough
> to say that if the mics don't pick it, the speakers do not sound it...
> it places the same limitation on the bandwidth of all music - which
> includes transients.
> Anything outside constant preiodic waves that started at the beginng
> of time and will last for ever, includes some transient energy! I
> guess you are trying to talk about "fast transients". There are huge
> misconceptions around that. For example, many folks thing that a
> bandlimited (say to 22KHz) 1KHz square wave rises faster (more slew
> rate - volts per second) than a sine wave at 20KHz. I can go on but
> will not.
>
> I have been designing for a long time and know about transient
> problems such as capacitor dialectric absorption in sample hold. That
> is not a 192KHz related. It is at the oposite side - low frequencies
> way within the ausio band. If it fixing it took changing from 96 to
> 192K sampling, you would not hear it! That is the whole point. 192KHz
> will not fix a thing for audio.
>
> Folks keep throwing misleading garbage based on lack of understanding,
> and I can not answer it for ever. I need to work for a living, but I
> am right, so the billion dollar conglamerates will chock me and the
> truth but constatly throwing garbage at such volume that it will be
> unanswered.
>
> And whan it all gets too much for them to handel, there will always be
> that last resort. That last argument that no one can stand to: It
> sounded better. This argument will yield best results if you get some
> real big names. Guess what- I have some real big names agreeing with
> me that 96KHz is better than 192. I did not ask permision to say who,
> but I put my inegrity on the line.
>
> In either case, I came accross a big name that refused to audition 96
> on the grounds that 192 contains "something" that staed with the music
> at 44.1 (after decimation. Now, I will not ever calim that I know 1%
> about recording when compared to that gentelman. I just wish that I
> was recipicated with the same respect regarding Math and Engineering.
>
> I would like to belive that there is a differance between the
> stereophile and the pro community. Not all stereophile stuff is bad,
> but much of it is very ridiculess. You do not hear a differance
> between a yellow and orange cable. Arrowa on speakers is insanly
> stupid. The argument is always about "It sounds better to me"...
> Followed by someone getting ripped off.
>
> Folks this is about getting our industry back on track. Double the
> data. double the processing and LOWERING THE QUALITY OF CONVERSION is
> a high price to pay. Science, and egineering and math on one hand, led
> by greats such as Shanon and Nyquist, and the billion dollars
> industial companies using every trick in the book (including avoiding
> the physics and engineering books) to sell thiere BS. I belive it is
> about marketing and money.
>
> I realize that if one fell for it, it is difficult to admit. But
> 192KHz for audio is a crock! The only argumet left is "but I like it".
> No one can stand to that one. If I tell you you added 10% distortions,
> and you say "I like it", you win. So lets compromise: I say 192KHz s a
> crock. You agree it is but you get to like it.
>
> Sorry for my tone. I put out some energy to do good. I lost a big sale
> because I refuse to play that 192KHz game. I try to educte folks about
> it. First it was "more dots are better". Than it was "better
> antialiasing". Than "transients". All very wrong and based on far from
> sufficient understanding of the basics. What's next?
>
> I am still here, but getting tired of it.
>
> Dan Lavry

While I don't even dream to have even a portion of the technical
knowledge and experience that you have, Dan, I have done emperical tests
of the same material at 48, 96 and 192k.

We used acoustic sources; piano, light percussion (bells, shakers,
etc.), acoustic guitar, cymbals, room tone and a ping-pong match. Every
person involved (5 of us) agreed that there was no comparison; all 192k
material from every source sounded far better than the others.

After listening to the 192k material, we were able to pick out
undesirable characteristics at the lower sampling rates that were not
that evident before. 96k had a slight hardness in the upper mids and
48k was really closed, small and hard (not scientific descriptions, I
admit) with even more of the upper mid edgyness. 192k had none of the
hardness that the lower sample rates exhibited. After listening to the
192k stuff, 48k was practically unlistenable and 96k only slightly
moreso. In blind tests, the musicians especially picked the 192k
material every time.

So why is this? I always thought it was because of the higher sampling
rate but if you can provide an alternative explanation I'm certainly
open to it.

Until then, I've got to say that 192k sounds A LOT better to me than
anything else.

--
Bobby Owsinski
Surround Associates
http://www.surroundassociates.com

dan lavry

December 1st 03, 03:45 AM

"Remixer" > wrote in message >...

> Before someone rises up to smack me down with the sampling theorem, and how
> 1Fs is all we'll ever need because of the bandwidth the human auditory blah
> blah blah...

And the laws of garvity? and basic math? and the whole of science just
because you hear something good enough to have every thing wrong. This
is pretty disrespectfull to engineering and math.

Why do not figure what is it about your setup blha blah blah blah...

> the fact is that in *practical implementation* the promise of the
> sampling theorem is not fulfilled, even with the best available 1Fs dacs.

We are talking 192KHz fs, not a 1fs DA. It has been years simce 1fs
DA. Stay on the subject. You are mixing it up blah blah blah blah

> I've recently been doing some tests with 44.1k 16-bit recordings upsampled
> to DSD with a new "trellis" algorithm from Philips. The playback of the
> upsampled-to-DSD files through a DSD dac has better detail and transient
> response than what can be heard from the original 44k file played through a
> variety of the best 1Fs dacs in the industry.

The best 1fs Dac in the in the industry is long gone. Wake up to
upsampling. The problem with 1 bit DA was the need for exreamly high
order anti imaging filter. So we went to upsdampling. Also, mixing 16
bits into the argument is out of place. Clearly 20 bits is better than
16. We are talking sample rates.

> This may be due to the fact
> that there is actually more information encoded in a good 44.1k recording (a
> la the sampling theorem) than we can playback with real world
> top-of-the-line PCM dacs, but more of that information does come through the
> upsampled DSD playback. Not trying to prove anything, only reporting what I
> can hear.

I really do not think you know how significent the sampling theorm is.
It is not about a couple of jerks drinking beer a spouting garbage. It
is about the fundumentals of modern signal theory, by great minds. Not
much of the electronics around you would work if the therory was
wrong. Which camp do I respect? Reality science math and engineering.
And a few top notch audio engineers that know their stuff.

Dan Lavry

>
>
> "Roger W. Norman" > wrote in message
> > Than what? How do you make a comparative when you give nothing by which
> the
> > comparison is made?
> >

Bob Cain

December 1st 03, 04:12 AM

dan lavry wrote:
>
> "Remixer" > wrote in message >...
>
> > Before someone rises up to smack me down with the sampling theorem, and how
> > 1Fs is all we'll ever need because of the bandwidth the human auditory blah
> > blah blah...
>
> And the laws of garvity?

Hey, dood, I dropped a feather and a BB that weighed the
same from my window yesterday and they definitely did _not_
fall at the same rate. Those Norton and Bernstein guys got
something wrong.

Seriously, it is a pleasure having someone with your
recognized authority debunking all the nonsense that plugs
the gullet here. I wish you luck in convincing people that
they should be looking for the factors that are really
causing the perceived differences.

Bob
--

"Things should be described as simply as possible, but no
simpler."

A. Einstein

Remixer

December 1st 03, 04:13 AM

"dan lavry" > wrote in message
om...
> And the laws of garvity? and basic math? and the whole of science just
> because you hear something good enough to have every thing wrong. This
> is pretty disrespectfull to engineering and math.
>
> Why do not figure what is it about your setup blha blah blah blah...
>

I realize my use of "blah blah blah" was sufficiently ambiguous to offend
just about anybody, but it was not intended for Mr. Lavry and was not meant
to disparage any of the well-reasoned arguments he has put forth in this
thread.

> We are talking 192KHz fs, not a 1fs DA. It has been years simce 1fs
> DA. Stay on the subject. You are mixing it up blah blah blah blah
>

This thread has already split off into tangents. My post was in response to
Mr. Norman's questioning of HP's ability to recognize the perceived
superiority of a DVD-A without the benefit of a direct comparison under
scientifically controlled circumstances. (That's why I back quoted Mr.
Norman's post.) Sorry if it was taken any other way.

> The best 1fs Dac in the in the industry is long gone. Wake up to
> upsampling. The problem with 1 bit DA was the need for exreamly high
> order anti imaging filter.

The pcm dacs in my test were all oversampling dacs. I guess I should have
been more explicit in comparing the DSD upsampling of a 44.1k source against
the same source played through an upsampling PCM dac. Yet another tangent
and another target for a flame war.

> I really do not think you know how significent the sampling theorm is.
> It is not about a couple of jerks drinking beer a spouting garbage. It
> is about the fundumentals of modern signal theory, by great minds. Not
> much of the electronics around you would work if the therory was
> wrong. Which camp do I respect? Reality science math and engineering.
> And a few top notch audio engineers that know their stuff.
>

Again, no attempt to question the validity of a proven mathematical theorem,
but it is all too often used as a blunt instrument to bludgeon those who
find fault with 1Fs recording and reproduction, upsampled or not.

Mr Lavry, please don't get caught by the remarkable power of newsgroups to
get one's dander up. With so many cross-currents going on at once, and
without the benefit of nuances, it is easy to take offense when none was
intended, or even directed.. Your recent posts have already caused me to
rethink a lot of what I have taken for granted about 4Fs. Take it all in
stride and keep contributing. Every little bit of good information helps
somebody somewhere in spite of the punting, trolling, and sniping that does
go on.

Rob Adelman

December 1st 03, 04:35 AM

dan lavry wrote:

> And the laws of garvity? and basic math? and the whole of science just
> because you hear something good enough to have every thing wrong. This
> is pretty disrespectfull to engineering and math.

The problem is that on paper, you could have 2 sets of criteria, one
seeming to be far more significant than the other. But in reality, the
lesser one could have far more significance in the way the brain
interprets it for reasons we just don't understand.

-Rob

Dave Collins

December 1st 03, 07:06 AM

> "Remixer" > wrote in message
> >...

> I've recently been doing some tests with 44.1k 16-bit recordings upsampled
> to DSD with a new "trellis" algorithm from Philips. The playback of the
> upsampled-to-DSD files through a DSD dac has better detail and transient
> response than what can be heard from the original 44k file played through a
> variety of the best 1Fs dacs in the industry.

First of all, you don't have a 1fs dac. 64fs is entry level, yours is
probaby more. Second, that "better detail and transient response" could
just be distortion that you like better. Nothng wrong with that. A
little 3rd harmonic added sounds like detail.... How much new music
would you say the upsampling creates?

I guess the question is: If you run it through the trellis 20 times,
does it just get better and better?

DC

Remixer

December 1st 03, 07:26 AM

> First of all, you don't have a 1fs dac. 64fs is entry level, yours is
> probaby more. Second, that "better detail and transient response" could
> just be distortion that you like better. Nothng wrong with that. A
> little 3rd harmonic added sounds like detail.... How much new music
> would you say the upsampling creates?
>
> I guess the question is: If you run it through the trellis 20 times,
> does it just get better and better?
>
> DC

I don't think it's 3rd harmonic distortion, it doesn't really sound like
that well known effect, (I've been doing this since before the Aphex Aural
Exciter and HEDD) and doubtful that Philips would let enough distortion
creep into their rather pricey trellis algorithm to make such an audible
difference.

Tommi

December 1st 03, 08:02 AM

"Bobby Owsinski" > wrote in message
...

> While I don't even dream to have even a portion of the technical
> knowledge and experience that you have, Dan, I have done emperical tests
> of the same material at 48, 96 and 192k.
>
> We used acoustic sources; piano, light percussion (bells, shakers,
> etc.), acoustic guitar, cymbals, room tone and a ping-pong match. Every
> person involved (5 of us) agreed that there was no comparison; all 192k
> material from every source sounded far better than the others.
>
> After listening to the 192k material, we were able to pick out
> undesirable characteristics at the lower sampling rates that were not
> that evident before. 96k had a slight hardness in the upper mids and
> 48k was really closed, small and hard (not scientific descriptions, I
> admit) with even more of the upper mid edgyness. 192k had none of the
> hardness that the lower sample rates exhibited. After listening to the
> 192k stuff, 48k was practically unlistenable and 96k only slightly
> moreso. In blind tests, the musicians especially picked the 192k
> material every time.
>
> So why is this? I always thought it was because of the higher sampling
> rate but if you can provide an alternative explanation I'm certainly
> open to it.
>
> Until then, I've got to say that 192k sounds A LOT better to me than
> anything else.

Where do I begin?
What converters did you use? What speakers? Mics? Preamps? How can you say
anything about the characteristics of different sample rates without any
valid data?
It's like saying your kitchen television has a better picture than the
monitors at Lucasfilm, no one can argue about your opinion, but there is no
science backing it up.

Remixer

December 1st 03, 08:22 AM

Presumably they used the same converters at each sample rate, eliminating
the converter variable, and compared with the feed from the studio,
eliminating the mic and pre-amp variables. BTW, when you go to the
supermarket and ask for change for your twenty do you demand proof of the
validity of integer arithmetic? At least some of the people who post here
are professional listeners who don't need to prove the earth is a planet to
take a walk. Dismissing the reasoned observations of working engineers
because they're not EEs under a double-blind test is such an old saw and not
particularly productive.

"Tommi" > wrote in message
...
>
> Where do I begin?
> What converters did you use? What speakers? Mics? Preamps? How can you say
> anything about the characteristics of different sample rates without any
> valid data?
> It's like saying your kitchen television has a better picture than the
> monitors at Lucasfilm, no one can argue about your opinion, but there is
no
> science backing it up.
>
>

KikeG

December 1st 03, 09:57 AM

Bobby Owsinski > wrote in message >...

> In blind tests, the musicians especially picked the 192k
> material every time.

So you are saying you were able to pick apart the 48/96/192 KHz
material, under blind conditions, with no problems at all, being the
differences evident, aren't you? This is an extraordinary claim, that
I'd like to know more about. Were the test samples properly level
matched (< 0.1 dB difference) and time aligned? Was the test
double-blind? How many trials and correct identifications did you get?
Could you give us more information about the equipment used at the
test? Could you provide us with same of the samples used at the test?

Dave Collins

December 1st 03, 11:14 AM

In article >,
"Remixer" > wrote:

> Dismissing the reasoned observations of working engineers
> because they're not EEs under a double-blind test is such an old saw and not
> particularly productive.

I'm a professional listener, not an EE, yet a working engineer. Somehow
it makes me more skeptical......

DC

Roger W. Norman

December 1st 03, 12:28 PM

"Remixer" > wrote in message
...
> This thread has already split off into tangents. My post was in response
to
> Mr. Norman's questioning of HP's ability to recognize the perceived
> superiority of a DVD-A without the benefit of a direct comparison under
> scientifically controlled circumstances. (That's why I back quoted Mr.
> Norman's post.) Sorry if it was taken any other way.

What, was my Dad here? <g>

The basic premise of language is to convey ideas and concepts, create common
ground and to be able to make verbal comparisons that represent intangible
items. An example of the latter would be something as "An apple tastes
better THAN ****." If one simply says "An apple tastes better" particularly
with the appended "..." then we don't have a comparison. Of course, the
other aspect is that when one makes a comparison they have the experience of
both the apple's taste and that of ****'s taste.

Now one could make the digital to digital assumption that Hp was saying his
comparison was against CDDA, but it might have been against Sony's PCM-F1
playback, or MP3, or any number of current or previous digital formats.

So, as you may see, I wasn't making a statement about his injection of
comment on the subject, nor questioning his ability to even be able to make
a comparison, I was simply saying that the language of the comment did not
make a comparison. He allowed you or I to infer one and make our own
decision about what the comparison meant.

--

Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.

Roger W. Norman

December 1st 03, 12:29 PM

And, of course, the old saw is "The Titanic looked good, on paper."
Unfortunately it had to float on the sea, of which it obviously did a pretty
**** poor job.

--

Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.

"Rob Adelman" > wrote in message
...
>
>
> dan lavry wrote:
>
> > And the laws of garvity? and basic math? and the whole of science just
> > because you hear something good enough to have every thing wrong. This
> > is pretty disrespectfull to engineering and math.
>
> The problem is that on paper, you could have 2 sets of criteria, one
> seeming to be far more significant than the other. But in reality, the
> lesser one could have far more significance in the way the brain
> interprets it for reasons we just don't understand.
>
> -Rob
>

Remixer

December 1st 03, 03:32 PM

So....how about sharing with the class then, DC? What has your own
experience been in comparing 1Fs, 2Fs, 4Fs, and DSD formats? What gear and
conditions have led to your opinions?

"Dave Collins" > wrote in message news:dcollins-
> I'm a professional listener, not an EE, yet a working engineer. Somehow
> it makes me more skeptical......
>
> DC

Remixer

December 1st 03, 03:47 PM

Context, son, context. I doubt he was comparing with an F-1 or a wax
cylinder either for that matter.

"Roger W. Norman" > wrote in message
...
> He allowed you or I to infer one and make our own
> decision about what the comparison meant.

Tommi

December 1st 03, 04:04 PM

"Remixer" > wrote in message
.. .
> Presumably they used the same converters at each sample rate, eliminating
> the converter variable, and compared with the feed from the studio,
> eliminating the mic and pre-amp variables. BTW, when you go to the
> supermarket and ask for change for your twenty do you demand proof of the
> validity of integer arithmetic? At least some of the people who post here
> are professional listeners who don't need to prove the earth is a planet
to
> take a walk. Dismissing the reasoned observations of working engineers
> because they're not EEs under a double-blind test is such an old saw and
not
> particularly productive.

Yes, but the point was that no-one knows _why_ the 192kHz sounded better.
If there is nothing to back it up except your ears, I don't think there's
any serious reason to think 192 kHz must be better. All sorts of tests are
done all over the world all the time, each with different results. Of
course, your ears are the most valuable tool at your disposal in the music
world, but ears can deceive you.
Take psychoacoustics for example; there actually were some "serious" tests
in the beginning of the mp3 era which claimed that downgrading a 16/44.1 wav
to an 128bps results in no audible change in most cases..

Likewise, if there's no science to back up 192, it all just comes down to
the "it sounds good" argument. What "sounds good" can be anything from
harmonic distortion to a gentle background hiss, it can be ultrasonic noise
downmixing to audible band in the air, etc. The point is that what "sounds
good" to the human ear doesn't mean that it is the truth about the sound
source. Heck, phaser can sound good when we use it in the mix to "widen" a
sound, but phase shift certainly isn't a thing we want in our amplifier
characteristics.

If a man who designs converters for a living says that 192kHz converters
_currently_ have much more jitter than 96kHz ones, that they're not good
enough for audio just yet, why couldn't you take his word for it? Science
usually beats our perception, even though we don't want always to believe
that.
The big leaps in our technology and knowledge always have always been due to
theoretical calculations, empirical measurements, and to a certain extent,
our own perception. There are always folks who base their opinion on one or
two of the above things, but the real pioneers always check that their
"thing" (whatever it is) is correct with all three aspects. Einstein was
one, but Aristotle for example wasn't since he relied mostly on perception
and logic..

Bobby Owsinski

December 1st 03, 04:29 PM

In article >,
"Tommi" > wrote:

> "Bobby Owsinski" > wrote in message
> ...
>
> > While I don't even dream to have even a portion of the technical
> > knowledge and experience that you have, Dan, I have done emperical tests
> > of the same material at 48, 96 and 192k.
> >
> > We used acoustic sources; piano, light percussion (bells, shakers,
> > etc.), acoustic guitar, cymbals, room tone and a ping-pong match. Every
> > person involved (5 of us) agreed that there was no comparison; all 192k
> > material from every source sounded far better than the others.
> >
> > After listening to the 192k material, we were able to pick out
> > undesirable characteristics at the lower sampling rates that were not
> > that evident before. 96k had a slight hardness in the upper mids and
> > 48k was really closed, small and hard (not scientific descriptions, I
> > admit) with even more of the upper mid edgyness. 192k had none of the
> > hardness that the lower sample rates exhibited. After listening to the
> > 192k stuff, 48k was practically unlistenable and 96k only slightly
> > moreso. In blind tests, the musicians especially picked the 192k
> > material every time.
> >
> > So why is this? I always thought it was because of the higher sampling
> > rate but if you can provide an alternative explanation I'm certainly
> > open to it.
> >
> > Until then, I've got to say that 192k sounds A LOT better to me than
> > anything else.
>
> Where do I begin?
> What converters did you use? What speakers? Mics? Preamps? How can you say
> anything about the characteristics of different sample rates without any
> valid data?
> It's like saying your kitchen television has a better picture than the
> monitors at Lucasfilm, no one can argue about your opinion, but there is no
> science backing it up.
>
>

There is no science, just listening under normal studio conditions on a
PT HD rig. All sources were recorded with the same signal chain and
reproduced with the same signal chain. Signal path was a very nice C12
through a Hardy mic amp directly into PTHD. Playback was out of the PT
HD into an SSl 9k out to Genelec 1031's, HD-1's and large soffit mounted
monitors (dual 15's with an SLS ribbon tweeter). No EQ or dynamics in
the signal path.

Before anyone else gets upset about this, please just go and listen for
yourself. I clearly heard a difference. Maybe your will or maybe you
won't. But for now, I can hear it.

--
Bobby Owsinski
Surround Associates
http://www.surroundassociates.com

Bobby Owsinski

December 1st 03, 04:30 PM

In article >,
(KikeG) wrote:

> Bobby Owsinski > wrote in message
> >...
>
> > In blind tests, the musicians especially picked the 192k
> > material every time.
>
> So you are saying you were able to pick apart the 48/96/192 KHz
> material, under blind conditions, with no problems at all, being the
> differences evident, aren't you? This is an extraordinary claim, that
> I'd like to know more about. Were the test samples properly level
> matched (< 0.1 dB difference) and time aligned? Was the test
> double-blind? How many trials and correct identifications did you get?
> Could you give us more information about the equipment used at the
> test? Could you provide us with same of the samples used at the test?

These were unscientific tests of a PT HD system, just to see if we could
hear any real differences between the sample rates. We could with no
problem, but there was no attempt to calibrate, level match, etc. Just
listening under somewhat normal studio conditions.

Agreed, you could drive a truck through the scientific holes in the
testing, but we all thought that the differences were not subtle.

--
Bobby Owsinski
Surround Associates
http://www.surroundassociates.com

Scott Dorsey

December 1st 03, 04:50 PM

Bobby Owsinski > wrote:
>
>Before anyone else gets upset about this, please just go and listen for
>yourself. I clearly heard a difference. Maybe your will or maybe you
>won't. But for now, I can hear it.

I did, and I heard a difference, but it was no greater than the difference
between different A/D units at 44.1. Admittedly this is a fairly big
difference, but if the converters all sound different then the sample
rate differences are the least of our worries.
--scott

--
"C'est un Nagra. C'est suisse, et tres, tres precis."

dan lavry

December 1st 03, 08:17 PM

Bobby Owsinski > wrote in message news:<polymedia-

> After listening to the 192k material, we were able to pick out
> undesirable characteristics at the lower sampling rates that were not
> that evident before....
>
> So why is this? I always thought it was because of the higher sampling
> rate but if you can provide an alternative explanation I'm certainly
> open to it.
>
> Until then, I've got to say that 192k sounds A LOT better to me than
> anything else.

Great! We are finaly getting somewhere! Alternative explanation! That
is what I have been saying all along.

It all started really bugging me when I heard folks say that when
recording at 192KHz, than decimating down to 44.1KHz for CD, they
still had that special 192KHz "quality" in the music. I heard it from
more than one person I rerspect.

I did not stop respecting them as great recording and mastering
engineers, but I did start on that path, claiming that there is no way
on earth that a 44.1KHz CD can ever contain anything with higher than
22.050KHz content.

So I thought it would be easy to state that it can not possibly be
about that "extra bandwidth". "Easy my foot"! I am now hearing
arguments similar to: I have the best supply and the best meter and it
measured 5.1V. It should be 5V. Can anyone there do something about
Ohms Law?

One of the last posting was calling someone to do something about that
sampling theorm. There are probably a quater million EE's in the IEEE
alone. Probably over a million EE's out there, in medical, telecom,
instrumentation, video, and yes audio and more. Are we talking about
50 million engineer years? And no one wants to step up to the plate
and chalange Nyquist?
It is like siding with the primitive tribal medicine man, against
what? Penicilin?

A couple of weeks ago, in Times science special publication, it stated
(based on a pole) that 1/2 of Americans belive in ghosts. So what am I
complaining about?

Back to the issue at hand - alternative explanations:

We need to agree on the bandwidth needed (for a given application,
video, audio, instrumentation...). There is no escape, we need to do
that! Going with too little, you loose importent data. Going with too
much you end up with too much data (large files) and difficult
processing. If your sensor can not go above 1MHz, sampling at much
higher than 2MHz is a waste. If your sound source and microphone are
limited to say 40KHz, than 96KHz is fine.

Once we agree on bandwidth, let us not get confuse how to get there. A
much higher front end sampling rate with less bits for further
decimation is a common method, and it helps make the anti aliasing
filter order into a non argument. The sampling rate will be the data
output rate.

Within the context of the given bandwidth, and that includes ALL music
including ALL the transient and horns and bells... there are 2
fundumental approaches:

1. Get the output waveform to track the input waveform as precisly as
possible.
2. Have some distortions.

To have 1. you need linear phase, low noise, quick recovery from
overdrive, lowest jitter possible, great components and material and
so on and on....

To have 2. opens a huge a vast field to talk about. Is there a type of
distortion that you like? You all know what I am saying - tube sound,
2nd and 3rd harmoics, wide main lobe of a SRC, jitter sidebands...

Some of such distortions are better understood than others in terms of
hearing and little is documented, thus a designer with more experience
can have some ideas for what combinations or tradeoffs to make. That
is where analog becomes importent...

I like to have that "waveform in" is the same as "waveform out", but I
do face tradoffs all the time. I do have some ideas of where the
tradoffs should be, and I am not going to spell it out and let the
competion catch up.

The 192KHz does create more distortions than 96KHz. Again, more speed
results in less accuracy, regardless to what voodo one throws at it.
There is a certain nature to those distortions, something to do with
"pushing" things a certain way. For example, If you are trying to
charge a cap via a reasistor, the longer you wait, the closer you will
be (EE call it "how many time constants"). This example is
logarithmic! If you have an OPamp with a certain Bode plot
(characteristics), how will it settle? The longer time (slower), the
further say 2nd order rolloff. It is a mess to figure out everything,
but I see some light, in terms of explaining it.

Clearly one way is to measure the outcome of 192KHz. 96, 44.1 and see
what happens. That is why I am dicusted with the A weighing specs of
192. It hides things.

So I will not answer your question as to what you hear in 192. I will
not answer what you hear in 44.1 or with a single transistor. But
since it is not about added conent that slower system can not
accomodate, than it is about DISTORTIONS.

I learned a long time ago that distortions is not always a bad word. I
can give you the same distortions you like at 192 with a 96K system,
without the penelty of twice file size, double processing requirnment.
The "Forces to be" are trying to say otherwise so they can sell you
that 192 gear. I think it is a bad thing to force a whole industry to
192KHz to get that distortion.

I do not tell anyone what to do. You want to use 100MHz and down
sample to 44.1KHz, fine! You like 1% distortion? Who am I to object?
If it sounds good, I'll buy the CD. I just don't like to have folks
twisted into beliving that the 192 gives better representaion of the
signal, when in fact it is the oposite.

I am of the opinion that folks that alrerady comitted to 192 will be
resisting my arguments. But with 192KHz it is distortions you hear,
and it does not take high sampling rate to make the sort of such
distortions. The 96KHz provides much better "waveform in" = "waveform
out", and if the job is about picking up air vibrations (sound) and
making it as identical as possible in playback, 192KHz is barking up
the wrong tree.

Again, I know some distortions are fine and fall under the category of
artistic decision. In fact, much of mastering is about just that -
musical taste. I apppreciate a well recorded and master CD. It is one
thing to have less distortions than make the adjustments you want.
Once stuck with distortionms, you can not remove them. I would hope
that there are other toys out there to play with, without the need to
fall for what I think is a marketing scheme attempting to move a whole
industry into a horrible mess - double the data and double the
processing (buy all new gear) and get more distortions for it.

So what is the next argument going to be? I already answered the
issues:
more bandwidth
more points is better
transients
Let's someone adjust that gravity stuff so we can fly

I do not want to say too much about the ego aspect (I can hear it so
no one tells me nothing).

BR

Dan Lavry

Glenn Booth

December 1st 03, 09:23 PM

Hi,

In message >, Scott Dorsey
> writes
>Bobby Owsinski > wrote:
>>
>>Before anyone else gets upset about this, please just go and listen for
>>yourself. I clearly heard a difference. Maybe your will or maybe you
>>won't. But for now, I can hear it.
>
>I did, and I heard a difference, but it was no greater than the difference
>between different A/D units at 44.1. Admittedly this is a fairly big
>difference, but if the converters all sound different then the sample
>rate differences are the least of our worries.

Taking it a step further, in my experience it's quite possible for the
same converter to sound different at different sample rates. Who's to
know how much testing a given 192kHz capable converter gets at 96kHz?

--
Regards,
Glenn Booth

Chris Hornbeck

December 1st 03, 10:41 PM

On 1 Dec 2003 12:17:52 -0800, (dan lavry)
wrote:

>We need to agree on the bandwidth needed (for a given application,
>video, audio, instrumentation...).

I'm having a very hard time following this discussion. The numbers
(44.1, 96, 192, etc.) are thrown around as if everyone but me
understands the useage.

With modern ADC's these are not the input sampling rate. No one
has discussed processing. If they're related to the output
resampling rate, that's been left unclear, and is also unlikely
with modern DAC's.

So, what ARE we talking about?

Thanks,

Chris Hornbeck
"That is my Theory, and what it is too."
Anne Elk

dan lavry

December 2nd 03, 12:50 AM

"Tommi" > wrote in message news:<KEJyb.2

> Yes, but the point was that no-one knows _why_ the 192kHz sounded better.
> If there is nothing to back it up except your ears, I don't think there's
> any serious reason to think 192 kHz must be better. All sorts of tests are
> done all over the world all the time, each with different results. Of
> course, your ears are the most valuable tool at your disposal in the music
> world, but ears can deceive you.

YOU DID NOT GET IT! ASSUMING PERFECT CONVERSION, THERE IS NO
DIFFERANCE BETWEEN A 48KHZ BANDLIMITED SIGNAL SAMPLED AT 96KHZ AND A
48KHZ SAMPLED AT 192KHZ. YOU CAN USE 2 POINT TO DESCRIBE A LINE, OR
DOUBLE TYHE FILE SIZE TO 4 POINTS. IT IS A STRIGHT LINE. IN THEORY
THERE IS NO DIFFERANCE.

IT IS NOT THAT WE CAN NOT EXPLAIN IT. ASUMMING PERFECT CONVERSION.
THEY WILL YIELD THE SAME SIGNAL OUT OF THE DA. NOT A FIMPTO VOLT OF
DIFFERANCE!!!

SO ANY DIFFERANCE IS ABOUT PRACTICE, NOT THEORY. AND WE CAN EXPALIN IT
JUST FINE:
1. 192 YIELDS MORE NOISE, THOUGH YOU CAN NOT HEAR MUCH OF IT
2. 192 IS LESS ACCURATE AND THAT IS WHAT YOU HEAR.

IF YOU LIKE IT IT IS FINE. BUT PLEASE GET OFF THAT "EAR IS BETTER THAN
EE". YES THE EARS ARE VERY VALUBLE FOR MUSIC. AND ALSO, DO NOT ASSUME
THAT A GOOD CIRCUIT DESIGN DOES NOT TAKE ENGINEERING AND TECHNICAL
KNOWHOW. THEY ARE BOTH IMPORTENT.

AGAIN: IN THEORY WE KNOW THAT THE OUTCOME WAVE OUT OF THE DA WILL BE
THE SAME IDENICAL OUTCOME. IT IS NOT THAT WE DO NOT KNOW HOW TO
EXPLAIN THINGS. THIS IS CAST IN STONE. IF I GIVE YOU A SCOPE PROBE AND
YOU SEE THE EXACT WAVE FORM IT IS ENDE OF CONVERSATION. THIS
ELECTRICAL WAVEFORM IS WHAT DRIVES AMPLIFIER AND THE SPEAKER. IF YOU
HAVE ZERO DIFFERANCE, THAN YOU CAN NOT HEAR IT EITHER.

"I CAN NOT EXPLAIN IT BUT THERE" IS NO STRANGER TO ALL OF US IN AUDIO.
BUT IS NOT ALWAYS THE CASE. WE DO KNOW A COUPLE OF THINGS! SUCH AS
NYQUIST, AND OHMS LAW AND 1+1=2. I AM NOT STATING THE THEORY OF
RELATIVITY HERE. JUST FUNDUMENTAL STUFF.

BR

DAN LAVRY

> Take psychoacoustics for example; there actually were some "serious" tests
> in the beginning of the mp3 era which claimed that downgrading a 16/44.1 wav
> to an 128bps results in no audible change in most cases..
>
> Likewise, if there's no science to back up 192, it all just comes down to
> the "it sounds good" argument. What "sounds good" can be anything from
> harmonic distortion to a gentle background hiss, it can be ultrasonic noise
> downmixing to audible band in the air, etc. The point is that what "sounds
> good" to the human ear doesn't mean that it is the truth about the sound
> source. Heck, phaser can sound good when we use it in the mix to "widen" a
> sound, but phase shift certainly isn't a thing we want in our amplifier
> characteristics.
>
> If a man who designs converters for a living says that 192kHz converters
> _currently_ have much more jitter than 96kHz ones, that they're not good
> enough for audio just yet, why couldn't you take his word for it? Science
> usually beats our perception, even though we don't want always to believe
> that.
> The big leaps in our technology and knowledge always have always been due to
> theoretical calculations, empirical measurements, and to a certain extent,
> our own perception. There are always folks who base their opinion on one or
> two of the above things, but the real pioneers always check that their
> "thing" (whatever it is) is correct with all three aspects. Einstein was
> one, but Aristotle for example wasn't since he relied mostly on perception
> and logic..

LeBaron & Alrich

December 2nd 03, 01:56 AM

Bob Cain wrote:

> Seriously, it is a pleasure having someone with your
> recognized authority debunking all the nonsense that plugs
> the gullet here. I wish you luck in convincing people that
> they should be looking for the factors that are really
> causing the perceived differences.

Amen. Now what's the best ADC for less than ten bucks? <g>

--
ha

LeBaron & Alrich

December 2nd 03, 03:57 AM

Bobby Owsinski wrote:

> Before anyone else gets upset about this, please just go and listen for
> yourself. I clearly heard a difference. Maybe your will or maybe you
> won't. But for now, I can hear it.

I'm not upset about this, nor do I have access to a Digi 192 rig. But it
does occur to me that a manufacturer offering a "more is more so it's
also better" product might well, in businesslike self-interest, might
try to make sure the "more" also sounded "better".

So one might wonder if PT 192 sounds "better" than Paris at 48, for
example.

--
ha

dan lavry

December 2nd 03, 07:09 AM

Chris Hornbeck > wrote in message >...
> On 1 Dec 2003 12:17:52 -0800, (dan lavry)
> wrote:

> I'm having a very hard time following this discussion. The numbers
> (44.1, 96, 192, etc.) are thrown around as if everyone but me
> understands the useage.
>
> With modern ADC's these are not the input sampling rate. No one
> has discussed processing. If they're related to the output
> resampling rate, that's been left unclear, and is also unlikely
> with modern DAC's.
>
> So, what ARE we talking about?
>
> Thanks,
>
> Chris Hornbeck
> "That is my Theory, and what it is too."
> Anne Elk

Sorry yiou are confused about it. I am pretty sure I made a clear
distinction between the input sampling and output sampling. I did it
at least twice, pointing out that the input sample is the one
determinimg the requirnment for anti aliasing filter. I also stated
that modern AD's mostly sample at 64fs to 256fs input rate thus the
antialiasing filter does not need be high order.

You sound like you understand it, so I will assume you just did not
read my comments. I do not balme you - they are long.

Regarding the procesing, input rate output rate and the rest, I have
been trying to explain here that most conversion starts with at most
few bits at high rate, and than we go through a process of "trading
off" output speed for
performance. The old 1 bit at 64fs can be made to say 105dB at 44.1KHz
(noise start rising at 22Khz). Of course one could have opted for say
a 96KHz output rate with 48KHz, but at lower accurcy. Or a 192KHz with
96KHz at yet lower accuracy. Folks tend to relate to the AD at the
output rate, so a 192KHz is the AD with 192KHz output (what you called
resampled, I called decimated, and we both call output rate).

Your comments give me one more chance to say the same thing but from a
different point of view: The whole concept of the modern AD, be it 1
bit or a few bits at high oversampling is based on "moving unwanted
energy" of low bits perfrmance from a low frequncy band (audio) to
high frequency (above audio). The amount of tradeoff (thus the final
outcome) is greatly a matter of oversampling ratio. It also depends on
the noise shaping order and other factors. But all things being equal,
when the ratio of sampling rate to what we decide to call audio
bandwidth is high, we get better results. If you started at 64fs 3
bits and wanted to keep 64fs your audio bandwidth, you get 3 bits! Go
to 32fs, and you get a few more bits and so on. 192KHz is twice the
rate of 96KHz so it will not yield as good performance to say 40KHz...

So the same converter at 96 is better than 192KHz. And it should sound
so unless someone did something to skew the test. Who would do such a
thing? Even those that stand to gain a lot of money selling that
192KHz crock would not go that low. I did not test for that, so I will
assume they did not.

BR

Dan Lavry

Tommi

December 2nd 03, 07:29 AM

"dan lavry" > wrote in message
om...

> YOU DID NOT GET IT! ASSUMING PERFECT CONVERSION, THERE IS > NO
> DIFFERANCE BETWEEN A 48KHZ BANDLIMITED SIGNAL SAMPLED > AT 96KHZ AND A
> 48KHZ SAMPLED AT 192KHZ. YOU CAN USE 2 POINT TO DESCRIBE A LINE, OR
> DOUBLE TYHE FILE SIZE TO 4 POINTS. IT IS A STRIGHT LINE. IN THEORY
> THERE IS NO DIFFERANCE.
> IT IS NOT THAT WE CAN NOT EXPLAIN IT. ASUMMING PERFECT CONVERSION.
> THEY WILL YIELD THE SAME SIGNAL OUT OF THE DA. NOT A FIMPTO VOLT OF
> DIFFERANCE!!!

Erm..You're either replying to the wrong post, or missed the tone of my
text.
I did never say that 192kHz sounds better to me, OR that faster sampling
results in better accuracy! I was just replying to the post where someone
said that.

> SO ANY DIFFERANCE IS ABOUT PRACTICE, NOT THEORY. AND WE CAN EXPALIN IT
> JUST FINE:
> 1. 192 YIELDS MORE NOISE, THOUGH YOU CAN NOT HEAR MUCH OF IT
> 2. 192 IS LESS ACCURATE AND THAT IS WHAT YOU HEAR.

> IF YOU LIKE IT IT IS FINE. BUT PLEASE GET OFF THAT "EAR IS BETTER THAN
> EE". YES THE EARS ARE VERY VALUBLE FOR MUSIC. AND ALSO, DO NOT ASSUME
> THAT A GOOD CIRCUIT DESIGN DOES NOT TAKE ENGINEERING AND TECHNICAL
> KNOWHOW. THEY ARE BOTH IMPORTENT.

Did I assume that? Did I EVER say that "the ears are better than EE"??
My point was EXACTLY the opposite, my point was that I understand people who
say that their ears hear a difference at 192, but that if there's no science
to back it up, then the "better" results they hear are due to
psychoacoustics! Harmonic distortion can have a pleasing effect, but
converters should be as close to the truth as possible without adding any
elements.

> AGAIN: IN THEORY WE KNOW THAT THE OUTCOME WAVE OUT OF THE DA WILL BE
> THE SAME IDENICAL OUTCOME. IT IS NOT THAT WE DO NOT KNOW HOW TO
> EXPLAIN THINGS. THIS IS CAST IN STONE. IF I GIVE YOU A SCOPE PROBE AND
> YOU SEE THE EXACT WAVE FORM IT IS ENDE OF CONVERSATION. THIS
> ELECTRICAL WAVEFORM IS WHAT DRIVES AMPLIFIER AND THE SPEAKER. IF YOU
> HAVE ZERO DIFFERANCE, THAN YOU CAN NOT HEAR IT EITHER.
> "I CAN NOT EXPLAIN IT BUT THERE" IS NO STRANGER TO ALL OF US IN AUDIO.
> BUT IS NOT ALWAYS THE CASE. WE DO KNOW A COUPLE OF THINGS! SUCH AS
> NYQUIST, AND OHMS LAW AND 1+1=2.

I really don't understand this reply. Either I didn't express my opinion
correctly, or it was severely misunderstood. Either way, I don't disagree
with any of the above since my point was more or less the same.

Luke Kaven

December 2nd 03, 10:41 AM

"Remixer" > wrote
> Presumably they used the same converters at each sample rate, eliminating
> the converter variable, and compared with the feed from the studio,
> eliminating the mic and pre-amp variables. BTW, when you go to the
> supermarket and ask for change for your twenty do you demand proof of the
> validity of integer arithmetic? At least some of the people who post here
> are professional listeners who don't need to prove the earth is a planet to
> take a walk. Dismissing the reasoned observations of working engineers
> because they're not EEs under a double-blind test is such an old saw and not
> particularly productive.

There are a lot of fallacies in this:

1) Using the same converters at different sample rates does not
eliminate the "converter variable" if there were such a thing. There
may be a quantitative shift in performance of the same device that can
not be explained by the mere doubling/halving of sample rate. Those
quantitive shifts are what the experimenter needs to somehow control.
The secondary problem is how one knows that one has controlled all
relevent variables, and that may not even be possible to know.

2) Asking for proof of integer arithmetic when getting change at the
supermarket is not the same as verifying the validity of one's
auditory judgments. First of all, integer arithmetic is not an
empirical phenomenon. It is true in all possible worlds. What is at
issue here are *empirical* questions, which are contingent, contingent
upon all kinds of things.

3) Being a professional listener says nothing about the validity of
one's auditory judgments. Auditory judgments are among the most
unreliable there are, even among experts, because they are subject to
many kinds of short term to long-term context effects.

4) Nobody here claimed that only electrical engineers were qualified
to make such judgments. Anyone schooled in experimental method can
suggest a set of controls. But because the phenomena in question is
empirical, we can only control for things *so far as we know*, unless
there are *analytical* facts (true in all possible worlds) supporting
it.

I've got to conclude that your "argument from common sense" does not
work.

Luke

Luke Kaven

December 2nd 03, 10:47 AM

Bobby Owsinski > wrote
[...]
> There is no science, just listening under normal studio conditions on a
> PT HD rig. All sources were recorded with the same signal chain and
> reproduced with the same signal chain. Signal path was a very nice C12
> through a Hardy mic amp directly into PTHD. Playback was out of the PT
> HD into an SSl 9k out to Genelec 1031's, HD-1's and large soffit mounted
> monitors (dual 15's with an SLS ribbon tweeter). No EQ or dynamics in
> the signal path.
>
> Before anyone else gets upset about this, please just go and listen for
> yourself. I clearly heard a difference. Maybe your will or maybe you
> won't. But for now, I can hear it.

One needn't deny that you *heard* a qualitative difference. The claim
is that you haven't successfully *explained* what you heard, and you
have given no grounds to support any claim of what the *causes* are.
This is not to deny your *practical knowledge*, but as we know,
practical knowledge is partly the act of reasoning under uncertainty,
with all the error that entails.

Luke

Luke Kaven

December 2nd 03, 10:51 AM

Bobby Owsinski > wrote [...]

> Agreed, you could drive a truck through the scientific holes in the
> testing, but we all thought that the differences were not subtle.

You're committing the same fallacy again. You have no grounds to
attribute those difference to any particular causes, whether sampling
frequency, or elsewise. Your implied reference to sampling rate in
citing "the differences" is vacuous. Undoubtedly, you are an
impressively skilled audio engineer. But that is not the same thing
as science, and science is what you are trying to argue.

Luke

Luke Kaven

December 2nd 03, 12:07 PM

(dan lavry) wrote:

>"Tommi" > wrote in message news:<KEJyb.2
>
>> Yes, but the point was that no-one knows _why_ the 192kHz sounded better.
>> If there is nothing to back it up except your ears, I don't think there's
>> any serious reason to think 192 kHz must be better. All sorts of tests are
>> done all over the world all the time, each with different results. Of
>> course, your ears are the most valuable tool at your disposal in the music
>> world, but ears can deceive you.
>
>YOU DID NOT GET IT!

Dan, wait...I think Tommi was agreeing with you.

Roger W. Norman

December 2nd 03, 01:00 PM

"KikeG" > wrote in message
om...
>
> So you are saying you were able to pick apart the 48/96/192 KHz
> material, under blind conditions, with no problems at all, being the
> differences evident, aren't you? This is an extraordinary claim, that
> I'd like to know more about. Were the test samples properly level
> matched (< 0.1 dB difference) and time aligned? Was the test
> double-blind? How many trials and correct identifications did you get?
> Could you give us more information about the equipment used at the
> test? Could you provide us with same of the samples used at the test?

I can assure anyone here that Bobby Owsinski knows equipment and results and
how to set up double blind tests. Not only does he know equipment, he has
some of the best available and is a very talented person.

On the other hand, Dan Lavry is also a person of high esteem, knowledgable
in electronics design and also it's usage.

And Dave Collins' name shouldn't have to go with any additional comments.

So rather than argue these things out in terms of "listening" situations,
I'd suggest that one sit back and listen to the arguments these very able
people are putting forth. To put it into perspective, any of us here would
be very lucky indeed to have Dan's converters on the front end of a
recording, with Bobby doing the tracking and mixing and Dave doing the final
touchup in mastering, preferrably output through another set of Dan's
converters. And I mean these people represent the highest quality we all
hope to achieve in audio.

Sometimes I think we really need to add a Who's Who section to
www.recaudiopro.net .

--

Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.

Roger W. Norman

December 2nd 03, 01:05 PM

"dan lavry" > wrote in message
m...
> A couple of weeks ago, in Times science special publication, it stated
> (based on a pole) that 1/2 of Americans belive in ghosts. So what am I
> complaining about?

Now wait a minute here, Dan. I just spent some time praising your abilities
and equipment and here you go, picking on me? <g>

--

Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.

> Bobby Owsinski > wrote in message news:<polymedia-
>
> > After listening to the 192k material, we were able to pick out
> > undesirable characteristics at the lower sampling rates that were not
> > that evident before....
> >
> > So why is this? I always thought it was because of the higher sampling
> > rate but if you can provide an alternative explanation I'm certainly
> > open to it.
> >
> > Until then, I've got to say that 192k sounds A LOT better to me than
> > anything else.
>
> Great! We are finaly getting somewhere! Alternative explanation! That
> is what I have been saying all along.
>
> It all started really bugging me when I heard folks say that when
> recording at 192KHz, than decimating down to 44.1KHz for CD, they
> still had that special 192KHz "quality" in the music. I heard it from
> more than one person I rerspect.
>
> I did not stop respecting them as great recording and mastering
> engineers, but I did start on that path, claiming that there is no way
> on earth that a 44.1KHz CD can ever contain anything with higher than
> 22.050KHz content.
>
> So I thought it would be easy to state that it can not possibly be
> about that "extra bandwidth". "Easy my foot"! I am now hearing
> arguments similar to: I have the best supply and the best meter and it
> measured 5.1V. It should be 5V. Can anyone there do something about
> Ohms Law?
>
> One of the last posting was calling someone to do something about that
> sampling theorm. There are probably a quater million EE's in the IEEE
> alone. Probably over a million EE's out there, in medical, telecom,
> instrumentation, video, and yes audio and more. Are we talking about
> 50 million engineer years? And no one wants to step up to the plate
> and chalange Nyquist?
> It is like siding with the primitive tribal medicine man, against
> what? Penicilin?
>
>
> Back to the issue at hand - alternative explanations:
>
> We need to agree on the bandwidth needed (for a given application,
> video, audio, instrumentation...). There is no escape, we need to do
> that! Going with too little, you loose importent data. Going with too
> much you end up with too much data (large files) and difficult
> processing. If your sensor can not go above 1MHz, sampling at much
> higher than 2MHz is a waste. If your sound source and microphone are
> limited to say 40KHz, than 96KHz is fine.
>
> Once we agree on bandwidth, let us not get confuse how to get there. A
> much higher front end sampling rate with less bits for further
> decimation is a common method, and it helps make the anti aliasing
> filter order into a non argument. The sampling rate will be the data
> output rate.
>
> Within the context of the given bandwidth, and that includes ALL music
> including ALL the transient and horns and bells... there are 2
> fundumental approaches:
>
> 1. Get the output waveform to track the input waveform as precisly as
> possible.
> 2. Have some distortions.
>
> To have 1. you need linear phase, low noise, quick recovery from
> overdrive, lowest jitter possible, great components and material and
> so on and on....
>
> To have 2. opens a huge a vast field to talk about. Is there a type of
> distortion that you like? You all know what I am saying - tube sound,
> 2nd and 3rd harmoics, wide main lobe of a SRC, jitter sidebands...
>
> Some of such distortions are better understood than others in terms of
> hearing and little is documented, thus a designer with more experience
> can have some ideas for what combinations or tradeoffs to make. That
> is where analog becomes importent...
>
> I like to have that "waveform in" is the same as "waveform out", but I
> do face tradoffs all the time. I do have some ideas of where the
> tradoffs should be, and I am not going to spell it out and let the
> competion catch up.
>
> The 192KHz does create more distortions than 96KHz. Again, more speed
> results in less accuracy, regardless to what voodo one throws at it.
> There is a certain nature to those distortions, something to do with
> "pushing" things a certain way. For example, If you are trying to
> charge a cap via a reasistor, the longer you wait, the closer you will
> be (EE call it "how many time constants"). This example is
> logarithmic! If you have an OPamp with a certain Bode plot
> (characteristics), how will it settle? The longer time (slower), the
> further say 2nd order rolloff. It is a mess to figure out everything,
> but I see some light, in terms of explaining it.
>
> Clearly one way is to measure the outcome of 192KHz. 96, 44.1 and see
> what happens. That is why I am dicusted with the A weighing specs of
> 192. It hides things.
>
> So I will not answer your question as to what you hear in 192. I will
> not answer what you hear in 44.1 or with a single transistor. But
> since it is not about added conent that slower system can not
> accomodate, than it is about DISTORTIONS.
>
> I learned a long time ago that distortions is not always a bad word. I
> can give you the same distortions you like at 192 with a 96K system,
> without the penelty of twice file size, double processing requirnment.
> The "Forces to be" are trying to say otherwise so they can sell you
> that 192 gear. I think it is a bad thing to force a whole industry to
> 192KHz to get that distortion.
>
> I do not tell anyone what to do. You want to use 100MHz and down
> sample to 44.1KHz, fine! You like 1% distortion? Who am I to object?
> If it sounds good, I'll buy the CD. I just don't like to have folks
> twisted into beliving that the 192 gives better representaion of the
> signal, when in fact it is the oposite.
>
> I am of the opinion that folks that alrerady comitted to 192 will be
> resisting my arguments. But with 192KHz it is distortions you hear,
> and it does not take high sampling rate to make the sort of such
> distortions. The 96KHz provides much better "waveform in" = "waveform
> out", and if the job is about picking up air vibrations (sound) and
> making it as identical as possible in playback, 192KHz is barking up
> the wrong tree.
>
> Again, I know some distortions are fine and fall under the category of
> artistic decision. In fact, much of mastering is about just that -
> musical taste. I apppreciate a well recorded and master CD. It is one
> thing to have less distortions than make the adjustments you want.
> Once stuck with distortionms, you can not remove them. I would hope
> that there are other toys out there to play with, without the need to
> fall for what I think is a marketing scheme attempting to move a whole
> industry into a horrible mess - double the data and double the
> processing (buy all new gear) and get more distortions for it.
>
> So what is the next argument going to be? I already answered the
> issues:
> more bandwidth
> more points is better
> transients
> Let's someone adjust that gravity stuff so we can fly
>
> I do not want to say too much about the ego aspect (I can hear it so
> no one tells me nothing).
>
> BR
>
> Dan Lavry

Roger W. Norman

December 2nd 03, 01:18 PM

"dan lavry" > wrote in message
om...
> Regarding the procesing, input rate output rate and the rest, I have
> been trying to explain here that most conversion starts with at most
> few bits at high rate, and than we go through a process of "trading
> off" output speed for
> performance.

And you pretty much opened your line of comments with the "desired" sampling
rate working out to be somewhere around 60 kHz, which I then presume, taken
with your less bits equals better performance, means that Sony's 14 bit 50
kHz PCM F1 was pretty much an idea converter set parameters, sans mechanical
tape storage method?

--

Roger W. Norman
SirMusic Studio
Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
See how far $20 really goes.

>

LeBaron & Alrich

December 2nd 03, 03:30 PM

Roger W. Norman wrote:

> So rather than argue these things out in terms of "listening" situations,
> I'd suggest that one sit back and listen to the arguments these very able
> people are putting forth. To put it into perspective, any of us here would
> be very lucky indeed to have Dan's converters on the front end of a
> recording, with Bobby doing the tracking and mixing and Dave doing the final
> touchup in mastering, preferrably output through another set of Dan's
> converters. And I mean these people represent the highest quality we all
> hope to achieve in audio.

Where can I sign up? <g>

--
ha

John La Grou

December 2nd 03, 03:31 PM

On 1 Dec 2003 16:50:33 -0800, (dan lavry)
wrote:

>
>IF YOU LIKE IT IT IS FINE. BUT PLEASE GET OFF THAT "EAR IS BETTER THAN
>EE". .... IF I GIVE YOU A SCOPE PROBE AND
>YOU SEE THE EXACT WAVE FORM IT IS ENDE OF CONVERSATION. THIS
>ELECTRICAL WAVEFORM IS WHAT DRIVES AMPLIFIER AND THE SPEAKER. IF YOU
>HAVE ZERO DIFFERANCE, THAN YOU CAN NOT HEAR IT EITHER.

Hi Dan,

I appreciate your perspective, and of course agree that 2F sample rate
can achieve "mathematically perfect" representation of F. I must
disagree, however, that we can measure "zero differences" of two
(analog) signals and assume that said signals sound identical.

Too many times, I've heard two circuits with essentially "zero test
differences" (THD, IMD, SN, SR, FR, CMR, PH, etc...) sound quite
different. Some wise psychoacoustic-physicist types are attempting to
find new algorithms to better correlate these subjective differences
with repeatable / measurable metrics, such as:

http://www.gedlee.com/distortion_perception.htm

Any linear function is subject to these illusive variables, including
ADC front-ends (modulators, etc..). I'm not sure if this speaks to
your argument, but certainly there is more to consider here than
sampling theory and bench tests.

JL

Tommi

December 2nd 03, 03:32 PM

"Luke Kaven" > wrote in message
...
> (dan lavry) wrote:
>
> >"Tommi" > wrote in message news:<KEJyb.2
> >
> >> Yes, but the point was that no-one knows _why_ the 192kHz sounded
better.
> >> If there is nothing to back it up except your ears, I don't think
there's
> >> any serious reason to think 192 kHz must be better. All sorts of tests
are
> >> done all over the world all the time, each with different results. Of
> >> course, your ears are the most valuable tool at your disposal in the
music
> >> world, but ears can deceive you.
> >
> >YOU DID NOT GET IT!
>
> Dan, wait...I think Tommi was agreeing with you.

Yes, I was.

Remixer

December 2nd 03, 03:35 PM

> >IF YOU LIKE IT IT IS FINE. BUT PLEASE GET OFF THAT "EAR IS BETTER THAN
> >EE". .... IF I GIVE YOU A SCOPE PROBE AND
> >YOU SEE THE EXACT WAVE FORM IT IS ENDE OF CONVERSATION. THIS
> >ELECTRICAL WAVEFORM IS WHAT DRIVES AMPLIFIER AND THE SPEAKER. IF YOU
> >HAVE ZERO DIFFERANCE, THAN YOU CAN NOT HEAR IT EITHER.
>

Dan,

Here is a link to a published waveform set, albeit from a manufacturer with
vested interest, that endeavors to show that impulse response improves as
sample rate increases. Perhaps there were some EEs involved in producing
the scope traces. Would you care to comment?

http://www.merging.com/ (at the site, click on the link to DSD/SACD in the
left column.)

Scott Dorsey

December 2nd 03, 04:05 PM

In article >,
Remixer > wrote:
>> >IF YOU LIKE IT IT IS FINE. BUT PLEASE GET OFF THAT "EAR IS BETTER THAN
>> >EE". .... IF I GIVE YOU A SCOPE PROBE AND
>> >YOU SEE THE EXACT WAVE FORM IT IS ENDE OF CONVERSATION. THIS
>> >ELECTRICAL WAVEFORM IS WHAT DRIVES AMPLIFIER AND THE SPEAKER. IF YOU
>> >HAVE ZERO DIFFERANCE, THAN YOU CAN NOT HEAR IT EITHER.
>>
>
>Here is a link to a published waveform set, albeit from a manufacturer with
>vested interest, that endeavors to show that impulse response improves as
>sample rate increases. Perhaps there were some EEs involved in producing
>the scope traces. Would you care to comment?

You haven't been listening.

The impulse response improves as the sample rate increases, BECAUSE the
bandwidth increases. All of the effects of the higher sample rate are
the result of the wider bandwidth.

And the improved impulse response only results in more ultrasonic components.
Is this significant? I don't know. But I do know that it's not an issue
below 20 KHz.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Per Karlsson

December 2nd 03, 06:23 PM

"Arny Krueger" > wrote:
> What bad thing to look for when operating it?[CEP] [...] Bad phase response?

Yes, since the Cooledit filters are infinite impulse response (IIR)
filters the phase response is necessarily strange. Strange in the same
way analog filters are though, so it doesn't mean they must sound
bad...

dan lavry

December 2nd 03, 06:46 PM

"Tommi" > wrote in message >...
> "dan lavry" > wrote in message
> om...
>
> > YOU DID NOT GET IT! ASSUMING PERFECT CONVERSION...
>
> Erm..You're either replying to the wrong post, or missed the tone of my
> text.
> I did never say that 192kHz sounds better to me, OR that faster sampling
> results in better accuracy! I was just replying to the post where someone
> said that.

I must have misread you. My appologies.

BR

Dan Lavry

> I really don't understand this reply. Either I didn't express my opinion
> correctly, or it was severely misunderstood. Either way, I don't disagree
> with any of the above since my point was more or less the same.

Again, sorry if I misunderstood your comments. I certainly do not want
to ofend anyone. I must be getting weary...

Regards

Dan Lavry

Mike Rivers

December 2nd 03, 07:01 PM

In article > writes:

> I can assure anyone here that Bobby Owsinski knows equipment and results and
> how to set up double blind tests. Not only does he know equipment, he has
> some of the best available and is a very talented person.

Yes. but he didn't do a double blind test here. He had a jury of
people who can be reasonably expected to have a valid opinion of which
sounds better than what. It's so difficult to control conditions for
an audio listening test that you rarely find one that's statistically
valid. You can choose to trust or distrust the opinion of this group,
or another group, or your own ears. I suspect that by the time it's
important enough for me to really care, there really won't be a
choice. (I eventually HAD to get a second speaker for stereo)

> On the other hand, Dan Lavry is also a person of high esteem, knowledgable
> in electronics design and also it's usage.

As I understand Dan's argument, audio that's band-limited to 48 kHz
doesn't sound any better when recorded at 192 kHz than at 96 kHz, and
this is certainly what's predicted by sampling theory. Obviously with
96 kHz sample rate, it's imperitive that the bandwidth be limited to
48 kHz, however this is not so when using a higher sample rate. All
arguments of the unobtrusiveness and no group delay of oversampling
filters aside, if there's no reason to limit the input bandwidth to
48 kHz, why do so? Is the argument that there's no usable audio above
48 kHz so we don't need to record it any more valid than the argument
of 25 years ago that there's no usable audio above 22 kHz?

If one believes that the only extension beyond a frequency they choose
as the highest useful frequency consists of noise, given the disk
space and other overhead, as long as it's recorded accurately, it can
be filterd on playback without being destructive. Other than the
pocketbook issue, there's no reason not to sample at 192 kHz (or
whatever), limit the bandwidth going in so that the sampling theorem
works right, and limit the playback to a narrower bandwidth if you
feel it's necessary. If listeners think that it sounds better with the
full bandwidth, give it to them. If listeners think it sounds better
with the bandwidth restricted, they can have that too.

> So rather than argue these things out in terms of "listening" situations,
> I'd suggest that one sit back and listen to the arguments these very able
> people are putting forth. To put it into perspective, any of us here would
> be very lucky indeed to have Dan's converters on the front end of a
> recording, with Bobby doing the tracking and mixing and Dave doing the final
> touchup in mastering, preferrably output through another set of Dan's
> converters. And I mean these people represent the highest quality we all
> hope to achieve in audio.

Well, you need something to listen to. And since, at least with
present technology, there's no way to change nothing but the sample
rate unless the "high" sample rate is in excess required by the
sampling theorem (Dan's argument) we can only compare the sound of
different systems. I think that's a valid comparison, but it shouldn't
be called "192 vs. 96," but rather "A vs. B."

--
I'm really Mike Rivers )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

Remixer

December 2nd 03, 07:22 PM

"Scott Dorsey" > wrote in message
...

> You haven't been listening.
>
> The impulse response improves as the sample rate increases, BECAUSE the
> bandwidth increases. ...
> And the improved impulse response only results in more ultrasonic
components.
> Is this significant? I don't know. But I do know that it's not an issue
> below 20 KHz.

The traces on the Merging site are of a 3usec pulse shown to be better
reproduced at 192k than at 96k or 48k. The human ear is said to be able to
resolve arrival times of a few microseconds so would it not stand to reason
the the blurring of the pulse at lower rates can correspond to what
listeners report as reduced detail and reduced sense of space?

I have, in fact, been listening, but when I report what I hear than someone
from the bully pulpit rises up to insist that I do not know how to match
levels, or to hide the identity of sources in a blind test, or I am the dupe
of an evil manufactruring empire's nefarious plot to cripple lower-res
formats to be able to sell more new gear. If I report something I do not
like about 1Fs encoding it is because of "distortion" if I report something
I do like about 4Fs encoding it is because of "distortion." Dan's thesis
that properly implemented 96k sounds better than 192k has been intriguing
and I intend to re-test myself, but it might be more objective to use
equipment from another source, such as dcs or Mytek, to avoid the
possibility of any skewing that might have found its way into Mr. Lavry's
own designs. I appreciate a designer of Dan's stature participating in the
mosh pit that is rec.audio.pro, I'm sure that anybody following this thread
has learned a lot. On the other hand, the zealous shouting and misreading
of posts shows that even Mr. Lavry is human. When other designers of Dan's
caliber, such as MJ or JL have posted differing viewpoints, their posts have
been ignored. I would like to see what transpires if Dan would try his
findings out on pro-aud. I don't really think my reports deserved to be
ridiculed and dismissed out of hand. I took pains not to question the
mathematical truth of the sampling theorem, only with imperfect devies and
implementations. I have used Dan's designs on critical projects and respect
what he has accomplished, but I do not regard anyone as knowing all there is
about human auditory response.

Roger W. Norman

December 2nd 03, 07:27 PM

There is no context in a comparison when there's no comparison. It's simply
language. You cannot say the sky is brighter than night. Is it that the
daylight sky is brighter or the nighttime sky is darker? What's the
comparison?

And, at 52, don't call me son. I'd take that from someone older, but my
guess is that you aren't. However, if you've got something to teach, then
I'm willing to learn.

--

Roger W. Norman
SirMusic Studio
RAP FAQ and Purchase your copy of the Fifth of RAP CD set at
www.recaudiopro.net.
See how far $20 really goes.

"Remixer" > wrote in message
...
> Context, son, context. I doubt he was comparing with an F-1 or a wax
> cylinder either for that matter.
>
>
>
> "Roger W. Norman" > wrote in message
> ...
> > He allowed you or I to infer one and make our own
> > decision about what the comparison meant.
>
>

Arny Krueger

December 2nd 03, 07:47 PM

"Hp Widmer" > wrote in message

> Hi Dan,
>
>> It is easier to relate to the ringing with low frequency squate
>> waves. That 10KHz burst is "mixing up" a lot of things.
>
> The real question will be : Bandlimit vs. transients given from some
> natural instruments and the required sample rate. In other words how
> do we really hear or we sime mask above 1xkHz.
>
> When I listen to my 192khz/24bit Stereo DVD-A Eagles (Hotel
> California) or Neil Young (Harvest) there is so much more MUSIC coming
> out...

It's pretty simple.

Record the analog outputs of your DVD-A player with a really good 24/192
sound card like a LynxTWO. Obviously, this preserves the full frequency
range of the DVD-A track. It also does very little to its actual dynamic
range.

Then downsample the 24/192 to 44/16 with CoolEdit or something like it.

Then upsample the 44/16 back up to 24/192 to facilitate the DBT comparison.

Then do a DBT comparison of the two 24/192 files with one of the DBT
comparator programs you can download from www.pcabx.com .

Report back the results...

BTW, if you think that your 24/192 sound card is trashing the music, loop
your music through it about 5 times and compare the before to the after.

If you want to preview your probable results, just download the relevant
files from:

http://www.pcabx.com/technical/sample_rates/index.htm

and

http://www.pcabx.com/product/cardd_deluxe/index.htm

Arny Krueger

December 2nd 03, 07:49 PM

"Bobby Owsinski" > wrote in message

> In article >,
> (KikeG) wrote:
>
>> Bobby Owsinski > wrote in message
>> >...
>>
>>> In blind tests, the musicians especially picked the 192k
>>> material every time.
>>
>> So you are saying you were able to pick apart the 48/96/192 KHz
>> material, under blind conditions, with no problems at all, being the
>> differences evident, aren't you? This is an extraordinary claim, that
>> I'd like to know more about. Were the test samples properly level
>> matched (< 0.1 dB difference) and time aligned? Was the test
>> double-blind? How many trials and correct identifications did you
>> get? Could you give us more information about the equipment used at
>> the test? Could you provide us with same of the samples used at the
>> test?
>
>
> These were unscientific tests of a PT HD system, just to see if we
> could hear any real differences between the sample rates. We could
> with no problem, but there was no attempt to calibrate, level match,
> etc. Just listening under somewhat normal studio conditions.
>
> Agreed, you could drive a truck through the scientific holes in the
> testing, but we all thought that the differences were not subtle.

So why not do things a little more scientifically?

Are you aware of all the free tools for doing this that you can find via
links posted at www.pcabx.com?

Scott Dorsey

December 2nd 03, 08:26 PM

Remixer > wrote:
>"Scott Dorsey" > wrote in message
...
>
>> You haven't been listening.
>>
>> The impulse response improves as the sample rate increases, BECAUSE the
>> bandwidth increases. ...
>> And the improved impulse response only results in more ultrasonic
>components.
>> Is this significant? I don't know. But I do know that it's not an issue
>> below 20 KHz.
>
>The traces on the Merging site are of a 3usec pulse shown to be better
>reproduced at 192k than at 96k or 48k. The human ear is said to be able to
>resolve arrival times of a few microseconds so would it not stand to reason
>the the blurring of the pulse at lower rates can correspond to what
>listeners report as reduced detail and reduced sense of space?

It's possible, although somewhat doubtful. BUT, that does not change the
fact that the blurring at lower rates is due entirely to the bandlimiting.

>I have, in fact, been listening, but when I report what I hear than someone
>from the bully pulpit rises up to insist that I do not know how to match
>levels, or to hide the identity of sources in a blind test, or I am the dupe
>of an evil manufactruring empire's nefarious plot to cripple lower-res
>formats to be able to sell more new gear.

Well, this is because you're making some statements that aren't really
relevant, and some that aren't accurate.

>If I report something I do not
>like about 1Fs encoding it is because of "distortion" if I report something
>I do like about 4Fs encoding it is because of "distortion."

Well, it might be, and that is the basic problem. There are so many sources
of distortion that it is very difficult to straighten them all out.

As several people have pointed out, the distortion produced by ANY of the
converters is high enough that different converters sound different at
the same sample rate. This being the case, there is NO accurate way to
compare conversion at two different sample rates.

Nobody has done a clean and well-conducted study on the subject, and I
don't think anyone can until converters get a lot better.

>Dan's thesis
>that properly implemented 96k sounds better than 192k has been intriguing
>and I intend to re-test myself, but it might be more objective to use
>equipment from another source, such as dcs or Mytek, to avoid the
>possibility of any skewing that might have found its way into Mr. Lavry's
>own designs.

But it's possible that equipment from another source might sound even
more different.

I mean, after using the Panasonic gear which sounded totally different at
44.1 than ar 48 ksamp/sec, because of poor filter design, how can anyone
be anything but skeptical about any of these tests either way?

All you can say is that converter A running at rate B sounds better than
converter C running at rate D. You can't make any more generalizations than
that. One converter might sound better at one rate while another sounds
better at another rate. They all have so many different distortion sources
that you can't generalize anything.

> I appreciate a designer of Dan's stature participating in the
>mosh pit that is rec.audio.pro, I'm sure that anybody following this thread
>has learned a lot. On the other hand, the zealous shouting and misreading
>of posts shows that even Mr. Lavry is human. When other designers of Dan's
>caliber, such as MJ or JL have posted differing viewpoints, their posts have
>been ignored.

No, because they have basically explained what they hear, and why they think
they hear it. You are saying what you hear and making generalizations about
it, and then providing very doubtful information to support it. If you just
said "this thing set this way sounds good to me" nobody would be yelling at
you.

>I would like to see what transpires if Dan would try his
>findings out on pro-aud. I don't really think my reports deserved to be
>ridiculed and dismissed out of hand. I took pains not to question the
>mathematical truth of the sampling theorem, only with imperfect devies and
>implementations. I have used Dan's designs on critical projects and respect
>what he has accomplished, but I do not regard anyone as knowing all there is
>about human auditory response.

They aren't. Your attempts to generalize your reports are being ridiculed
and dismissed. I don't in any way doubt what you are hearing, what I doubt
is that you can generalize this to say something that is true about every
system.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

dan lavry

December 2nd 03, 11:52 PM

John La Grou > wrote in message >...
> On 1 Dec 2003 16:50:33 -0800, (dan lavry)
> wrote:
>
> >
> >IF YOU LIKE IT IT IS FINE. BUT PLEASE GET OFF THAT "EAR IS BETTER THAN
> >EE". .... IF I GIVE YOU A SCOPE PROBE AND
> >YOU SEE THE EXACT WAVE FORM IT IS ENDE OF CONVERSATION. THIS
> >ELECTRICAL WAVEFORM IS WHAT DRIVES AMPLIFIER AND THE SPEAKER. IF YOU
> >HAVE ZERO DIFFERANCE, THAN YOU CAN NOT HEAR IT EITHER.
>
>
> Hi Dan,
>
> I appreciate your perspective, and of course agree that 2F sample rate
> can achieve "mathematically perfect" representation of F. I must
> disagree, however, that we can measure "zero differences" of two
> (analog) signals and assume that said signals sound identical.
>
> Any linear function is subject to these illusive variables, including
> ADC front-ends (modulators, etc..). I'm not sure if this speaks to
> your argument, but certainly there is more to consider here than
> sampling theory and bench tests.
>
> JL

John,

I wonder if you are joining the conversation late in the game. I agree
that there are diffrences between circuits, converters and so on.
Also, I can not be a "one man show" here. BUt I know you are a good
engineer, so I would expect you to agree that IN THEORY, the extra
sample points carry zero information. I would also expect you to agree
that if audio is say 48KHz in musical content, mic pickup, speaker
reproduction and the ear itself, Than the differances you hear are not
about bandwidth.

I know about slew rates and fet vs bipolar and the rest. My point is
that it is not about conversion bandwidth. I can not go over it again
and again. My posts are still there.

BR

Dan Lavry

dan lavry

December 2nd 03, 11:59 PM

"Roger W. Norman" > wrote in message >...
> "dan lavry" > wrote in message
> om...
> > Regarding the procesing, input rate output rate and the rest, I have
> > been trying to explain here that most conversion starts with at most
> > few bits at high rate, and than we go through a process of "trading
> > off" output speed for
> > performance.
>
> And you pretty much opened your line of comments with the "desired" sampling
> rate working out to be somewhere around 60 kHz, which I then presume, taken
> with your less bits equals better performance, means that Sony's 14 bit 50
> kHz PCM F1 was pretty much an idea converter set parameters, sans mechanical
> tape storage method?
>
> --
>
>
> Roger W. Norman
> SirMusic Studio
> Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
> See how far $20 really goes.
>

The considerations at the time were different. That was before up and
down sampling. That 44.1KHz was "high enough" to be over twice 20KHz,
and low enough to cram a symphony on a disk - storage vs pain - the
pain of imposible anti alaising filters. Frankly, even with a 3db at
20KHz, how far does a pole you get to at 22KHz? A 20 pole would not be
enough, and that is 10 OPamps in series... will kill your noise and
distortions...

So 50KHz was "huge breating room" relative to 50KHz. It still sucked.
Nowdays, with input rates in the MHz, life is easier.

BR

Dan Lavry

Chris Hornbeck

December 3rd 03, 12:04 AM

On 2 Dec 2003 15:26:37 -0500, (Scott Dorsey) wrote:

>As several people have pointed out, the distortion produced by ANY of the
>converters is high enough that different converters sound different at
>the same sample rate. This being the case, there is NO accurate way to
>compare conversion at two different sample rates.
>
>Nobody has done a clean and well-conducted study on the subject, and I
>don't think anyone can until converters get a lot better.

Hi Scott,
I can't seem to get a handle on what everybody's talking about. Could
you help by correcting where I'm falling off the tracks?

1. Modern ADC's are very high speed, few-bit oversamplers, maybe
352.8 or 384 kHz. This implies that the same bandlimiting, input
sampling, dither and quantization are applied to all "sample rates".

2. The above first "sample rate" is decimated to another, lower
"sample rate" in software. This requires integration, hence
further bandlimiting.

3. This second "sample rate" may or may not be converted to a
third "sample rate" for DAC.

Questions:
1. Which of the three above "sample rates" are we talking about?

2. If noticable differences occur between different nominal
sample rates, are the differences in step 2 above, IOW, in
software?

Thanks very much for whatever light you can shed on my darkness.

Chris Hornbeck
"That is my Theory, and what it is too."
Anne Elk

Arny Krueger

December 3rd 03, 12:18 AM

"Remixer" > wrote in message

> "Scott Dorsey" > wrote in message
> ...
>
>> You haven't been listening.
>>
>> The impulse response improves as the sample rate increases, BECAUSE
>> the bandwidth increases. ...
>> And the improved impulse response only results in more ultrasonic
>> components. Is this significant? I don't know. But I do know that
>> it's not an issue below 20 KHz.
>
> The traces on the Merging site are of a 3usec pulse shown to be better
> reproduced at 192k than at 96k or 48k.

Look at it this way. A 3 usec pulse followed by 3 usec of silence is a
166,666 Hz square wave with a DC offset. Rule-of-thumb is that it takes
through the 11th harmonic to reproduce a nice square wave. This signal will
keep looking better and better until you sample it at 3.66 MHz or more.

>The human ear is said to be
> able to resolve arrival times of a few microseconds so would it not
> stand to reason the blurring of the pulse at lower rates can
> correspond to what listeners report as reduced detail and reduced
> sense of space?

The temporal resolution of 16/44 is 350 picoseconds (3.5 x 10E-10 seconds).
That makes 3 uSec look HUGE.

> I have, in fact, been listening, but when I report what I hear than
> someone from the bully pulpit rises up to insist that I do not know
> how to match levels, or to hide the identity of sources in a blind
> test, or I am the dupe of an evil manufacturing empire's nefarious
> plot to cripple lower-res formats to be able to sell more new gear.

I think you now know how to do all of the above. However, I'm looking to you
to say that you've actually done it.

Luke Kaven

December 3rd 03, 12:18 AM

"Remixer" > wrote:
[...]
>I have, in fact, been listening, but when I report what I hear than someone
>from the bully pulpit rises up to insist that I do not know how to match
>levels, or to hide the identity of sources in a blind test, or I am the dupe
>of an evil manufactruring empire's nefarious plot to cripple lower-res
>formats to be able to sell more new gear. If I report something I do not
>like about 1Fs encoding it is because of "distortion" if I report something
>I do like about 4Fs encoding it is because of "distortion."

On scientific grounds, you made some methodological mistakes and
errors in analysis. If your conclusions are based on those, then we
should call you on it. The criticism isn't motivated by religious
beliefs.

Scott Dorsey

December 3rd 03, 12:31 AM

Chris Hornbeck > wrote:
>
>1. Modern ADC's are very high speed, few-bit oversamplers, maybe
>352.8 or 384 kHz. This implies that the same bandlimiting, input
>sampling, dither and quantization are applied to all "sample rates".

Probably. I'm not going to say this is the case for everything around
because there are still folks designing around lower rate systems with
more bits.

>2. The above first "sample rate" is decimated to another, lower
>"sample rate" in software. This requires integration, hence
>further bandlimiting.

Right. So, the ultimate storage rate determines the overall system
bandwidth.

>3. This second "sample rate" may or may not be converted to a
>third "sample rate" for DAC.

Right, but if it is upsampled, the total system bandwidth is limited
by the lowest sample rate in the system.

>Questions:
>1. Which of the three above "sample rates" are we talking about?

The lowest rate in the system, which would presumably be the storage
rate if the system were designed optimally.

>2. If noticable differences occur between different nominal
>sample rates, are the differences in step 2 above, IOW, in
>software?

The decimation could be done in software or it could be done with
dedicated hardware, and there could also be goofy stuff going on
in the process with dither and with things to eliminate idle tones.

I think in this discussion we're considering the A/D and D/A systems
as black boxes that produce output at a given sample rate which is
stored, and then reconstruced after storage. So, whatever oversampling
is done is something that goes on inside the black box that we don't
have to worry about.

In fact, changing the rate of the clocks inside the converters may very
well affect the sound, but we're only worried about storage rates here.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Luke Kaven

December 3rd 03, 12:31 AM

"Roger W. Norman" > wrote:
[...]
>I can assure anyone here that Bobby Owsinski knows equipment and results and
>how to set up double blind tests. Not only does he know equipment, he has
>some of the best available and is a very talented person.
[...]

I don't doubt Bobby's skills. But he made errors in scientific method
and analysis, and if he wants to assert his conclusions based on the
reasons he offered, we have to call him on it. And it is well worth
doing, especially since one would expect him to do a lot with his
knowledge and skills.

Luke

dan lavry

December 3rd 03, 12:44 AM

"Roger W. Norman" > wrote in message >...
> "dan lavry" > wrote in message
> m...
> > A couple of weeks ago, in Times science special publication, it stated
> > (based on a pole) that 1/2 of Americans belive in ghosts. So what am I
> > complaining about?
>
> Now wait a minute here, Dan. I just spent some time praising your abilities
> and equipment and here you go, picking on me? <g>

> Roger W. Norman
> SirMusic Studio
> Purchase your copy of the Fifth of RAP CD set at www.recaudiopro.net.
> See how far $20 really goes.

I am trying to have fun with it, but I am an old man now, so I guess I
get to be a carmugeon (spelling?). I belive some of the problems here
about is having a difficult time explaining how less points will yield
as good a detail as more points. I know about good sounding vs bad
sounding circuits, and do not want to argue that a high slew rate amp
can sound better, or that some caps have memory. I do not want to say
there is no differance in sound between anything to anything. I know
hwo sensitive the ear is.

I am not arguing about the bandwidth or slew rate or noise or recovery
of an analog front end. I am saying that if there is no energy over
say 48KHz, do not go to 192KHz. Yes, what is in the data depends on
mics, mic pre, analog filter and drive. But the same circuitry can be
used for 96 and 192Khz.

I read with intrest some analogies about cars going around corners and
so on. A better way to understand it is is as follows. Some theory
first:

1. Take a low pass filter and hit it with a "zero width" impulse and
what do you get? Say the filter is a brick wall low pass at Nyquist:
You get a wave that is like a "rounded pulse" with decaying ringing on
both sides. It is the SINC function (sin(x)/x). This is a fundumental
relationship between frequency and time domain - a brick wall low pass
filter in the time domain is manifeted as a Sinc in the time domain.

When the sync is centered on a sample time, the decaying ringing on
the sides cross the zero values at all the adjacent sample times.

Here is the intersting point. As strange looking as that Sinc is, it
is a "magic wave form". Can you use it to say make DC? Not easy to
picture, but if you took a bunch of equal amplitude of those Sincs,
say a million in a row, and added them togeteher. The middle range
would be DC. The peaks are seperated by sample time and it is easy to
see that things are the same at each sample time (at the middle of the
bunch). What happens between samples? Some go up, others down... a
"complicated mess" that adds to the same value.

Say we only took 100 Sincs. We end up with some rise, where the first
Sinc started, the next one start "correction" the third... by the time
we are closer to center it is a bit closer to "DC" than as we get
towards the end we start having less DC like behaviour. The last Sinc
is where we have a fall...
This is the represenstation of a bandlimited square wave, with
harmonics above nyquist removed. Overshoot than rise, than decaying
ringing, more ringing overshoot and fall ....

Now, Take any band limited signal (under a certain Nyquist) abd all
you have to do is to line up the time domain Sincs with sample times,
and scale each Sinc amplitud to the sample value. If you add the whole
"mess", you end up with smooth perfect representation of the original
signal.

This is just the fundumental concept. True you need a lot of samples
to get there, infinite in theory, much less in practice.

No one can take a simple point and say where the signal will go. Take
2 points and it is still far from enough. Take say 1000 points in each
direction, and you get a pretty good value at the "center of the
mess". So in practice the question can be: how many points, or what
kind of filter is needed to get 60dB? 100dB? 120dB?

If audio signal (not the analog circuit, JL) is limited to say 48KHz,
you do not need to double the Sinc functions to get a smooth curve -
approach perfect (say 120dB?) representation of the analog wave. In
fact, doubnling of the processing power will add no improvment at all.

I hope some of you understand anfd find it intersting. I will make a
document and post it on my web - a picture is worth 1000 words. But it
will take a while to do.

BR

Dan Lavry

Chris Hornbeck

December 3rd 03, 12:45 AM

Thanks Scott,
Couldn't be clearer.

ps: Congratulations and Our Royal Felicitations on that
other matter.

Chris Hornbeck
"That is my Theory, and what it is too."
Anne Elk

Vladan

December 3rd 03, 12:57 AM

On 2 Dec 2003 15:26:37 -0500, (Scott Dorsey) wrote:

>Nobody has done a clean and well-conducted study on the subject, and I
>don't think anyone can until converters get a lot better.

Problem is, if you want to have results in "hard print" and analyze
technical details, you can compare only different (kinds of)
recordings, picture of the picture, which is meaningles in determining
fidelity, because you can never know how close to the original, live
source, it realy is. All you can say is this one has nicer looking
curve, or less peaks, or..., or I like this sound better than the
other. It may be, inspite all the logic, lower technical specs give
results truer to the real thing (as reported by vynil lovers, for
instance).
Of course logic tels us higher specs equipment is able of reproducing
signals recorded by lover specs equipment, exactly as they are, but
may need aditional effort to remove unnescessary data.

Scott Dorsey

December 3rd 03, 02:15 AM

Vladan > wrote:
>On 2 Dec 2003 15:26:37 -0500, (Scott Dorsey) wrote:
>
>>Nobody has done a clean and well-conducted study on the subject, and I
>>don't think anyone can until converters get a lot better.
>
>Problem is, if you want to have results in "hard print" and analyze
>technical details, you can compare only different (kinds of)
>recordings, picture of the picture, which is meaningles in determining
>fidelity, because you can never know how close to the original, live
>source, it realy is. All you can say is this one has nicer looking
>curve, or less peaks, or..., or I like this sound better than the
>other. It may be, inspite all the logic, lower technical specs give
>results truer to the real thing (as reported by vynil lovers, for
>instance).

If that is the case, we should be able to measure it. Until we can
measure it, we can't talk about it in a meaningful way, yes. That
is the point of this discussion.

>Of course logic tels us higher specs equipment is able of reproducing
>signals recorded by lover specs equipment, exactly as they are, but
>may need aditional effort to remove unnescessary data.

Only if the specifications are irrelevant. Which is sadly often the
case, because people don't know what to measure.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Bobby Owsinski

December 3rd 03, 04:44 AM

In article >,
Luke Kaven > wrote:

> "Roger W. Norman" > wrote:
> [...]
> >I can assure anyone here that Bobby Owsinski knows equipment and results and
> >how to set up double blind tests. Not only does he know equipment, he has
> >some of the best available and is a very talented person.
> [...]
>
> I don't doubt Bobby's skills. But he made errors in scientific method
> and analysis, and if he wants to assert his conclusions based on the
> reasons he offered, we have to call him on it. And it is well worth
> doing, especially since one would expect him to do a lot with his
> knowledge and skills.
>
> Luke

Luke,

There was no scientific method and analysis. There was no attempt to do
so whatsoever. We were just curious if we could here the difference,
and all the people involved could with no problem. Was there another
reason other than the higher sampling rate? Maybe. Was it because our
methodology was flawed? We had no methodology so maybe that was the
reason, or maybe not.

All I know is that with a PT HD system, 192k sounded better to my ears
(and all others involved) when we did the experiment. I'm actually
shocked that more people don't try this, considering the number of HD
systems out there.

I have no interest of doing a "scientific" experiment. It's too boring
for me (and doesn't make me any money). I'll just rely on the empirical
for now and leave the scientific to people who enjoy it and are better
at it than I ever could be. Then I'll read the AES paper and hopefully
learn something new.

The only reason why I posted in the first place was because there seemed
to be a lot of purely theoretical discussion on the subject yet no one
talking about what they actually heard.

--
Bobby Owsinski
Surround Associates
http://www.surroundassociates.com

Tommi

December 3rd 03, 08:18 AM

"dan lavry" > wrote in message
om...
> "Tommi" > wrote in message
>...
> > "dan lavry" > wrote in message
> > om...
> >
> > > YOU DID NOT GET IT! ASSUMING PERFECT CONVERSION...
> >
> > Erm..You're either replying to the wrong post, or missed the tone of my
> > text.
> > I did never say that 192kHz sounds better to me, OR that faster sampling
> > results in better accuracy! I was just replying to the post where
someone
> > said that.
>
> I must have misread you. My appologies.

No worries, Dan.

Arny Krueger

December 3rd 03, 08:25 AM

"Bobby Owsinski" > wrote in message

> There was no scientific method and analysis. There was no attempt to
> do so whatsoever. We were just curious if we could here the
> difference, and all the people involved could with no problem.

The problem is that due to the generally-agreed-upon absence of scientific
method and analysis, exactly what you perceived is extremely easy to
question.

> Was there another reason other than the higher sampling rate? Maybe.

There's no doubt that sighted cues, inadequate level matching, and
inadequate time synchronization of musical samples will lead to positive
identification, even when both sonic alternatives are exactly identical.

> Was it because our methodology was flawed? We had no methodology so
> maybe that was the reason, or maybe not.

The obvious thing to do is to repeat the experiment with proper controls and
see what happens.

> All I know is that with a PT HD system, 192k sounded better to my ears
> (and all others involved) when we did the experiment. I'm actually
> shocked that more people don't try this, considering the number of HD
> systems out there.

I'm shocked how many people who claim the skills and resources to do this
with proper controls, don't. I'm not the sharpest knife in the drawer, but
I've done this sort of thing 100's of times. Thousands of people have taken
advantage of the files from my www.pcabx.com web site and done the same
thing, as well.

> I have no interest of doing a "scientific" experiment. It's too
> boring for me (and doesn't make me any money).

Well, that's a common response. It's exciting to brag about your equipment,
and its not exciting to put it to a proper test. After all, you might find
out that the bigger numbers don't actually give better sound, and that could
be a real drag.

> I'll just rely on the
> empirical for now and leave the scientific to people who enjoy it and
> are better at it than I ever could be. Then I'll read the AES paper
> and hopefully learn something new.

If you did listening tests that you knew were seriously flawed, why didn't
you report that they were flawed initially?

> The only reason why I posted in the first place was because there
> seemed to be a lot of purely theoretical discussion on the subject
> yet no one talking about what they actually heard.

That's simply not true.

KikeG

December 3rd 03, 08:33 AM

John La Grou > wrote in message >...

> >IF YOU LIKE IT IT IS FINE. BUT PLEASE GET OFF THAT "EAR IS BETTER THAN
> >EE". .... IF I GIVE YOU A SCOPE PROBE AND
> >YOU SEE THE EXACT WAVE FORM IT IS ENDE OF CONVERSATION. THIS
> >ELECTRICAL WAVEFORM IS WHAT DRIVES AMPLIFIER AND THE SPEAKER. IF YOU
> >HAVE ZERO DIFFERANCE, THAN YOU CAN NOT HEAR IT EITHER.

> Too many times, I've heard two circuits with essentially "zero test
> differences" (THD, IMD, SN, SR, FR, CMR, PH, etc...) sound quite
> different.

Now, did you listen to them under blind conditions? If the measured
differences (there are always some measurable differences to talk of,
if you measure adequately) were much below known thresholds of
audibility of human ear, it's most likely it was just placebo effect,
or in other words, it was your brain knowing what you were listening
to, playing tricks in your perception. This is a very serious issue,
that is surprisingly taken in count very little times in the world of
audio.

There are fields of audio development where this is a fundamental
issue that is taken care of (by means of blind tests), such as
psychoacoustic lossy compression, just because it's fundamental for
efficient and reliable development.

Now, there's people that are convinced that there are audible
differences that can't be measured. I'm really skeptic about that, for
several reasons. First, because limits of human hearing are pretty
well known from several years ago. Second, because all kind of
distortions that can happen over an electric signals are known from
many years ago. Third, because measurement devices surpassed known ear
capabilities many years ago, by a very wide margin. And fourth,
because people claiming such things, usually have not performed a
controlled, rigorous, reliable blind test, that definitely proves
their claims. You know, as Carl Sagan said, "extraordinary claims
demand extraordinary proof", or something like that. Anecdotal
evidence, with sighted listening, or unexisting or poor time and/or
level matching, or very little repeatability is just not good proof
enough to take it seriously, and I've been witness of such things
happening relatively often.

> Some wise psychoacoustic-physicist types are attempting to
> find new algorithms to better correlate these subjective differences
> with repeatable / measurable metrics, such as:

But they always talk about both readily audible and measurable
differences. They just want to correlate better the measurements with
the perceivable degradation.

> Any linear function is subject to these illusive variables, including
> ADC front-ends (modulators, etc..). I'm not sure if this speaks to
> your argument, but certainly there is more to consider here than
> sampling theory and bench tests.

Dan is just saying that sampling rate, by itself, can't be the cause
of audible differences. If there is really an audible difference, then
it must be somewhere else. I must add, one should make sure the
differences are not *outside* the equipment, thinking about things as
simple as slightly different output levels, or just the listener brain
influencing his perceptions.

I have personally degraded a piece of music extracted from a CD by
playing at recording it several times through an inexpensive semi-pro
soundcard, and challenged people to be able to tell it from the
original, under blind conditions, in a repeatable manner. Nobody, from
dozens of person, some with very good equipment, listening training
and ears, has been able so far, even when the degradation caused is
easily measurable. If you are interested in trying such a test, just
tell me, and I'll make it available for anyone that is interested.

Enrique G.

KikeG

December 3rd 03, 08:47 AM

"Remixer" > wrote in message >...

> Here is a link to a published waveform set, albeit from a manufacturer with
> vested interest, that endeavors to show that impulse response improves as
> sample rate increases. Perhaps there were some EEs involved in producing
> the scope traces. Would you care to comment?

It'll comment a little bit. For one, the main difference in those
impulses is just related to the different bandwith of the systems, and
as Dan explained, there's no advantage in extending the bandwidth of
the systems beyond the speakers or mics actual bandwidth, or even over
the ear's bandwidth. Another thing that can be observed is the
"smearing" or pre- and post-ringing of the impulse when using lower
sample rates. Even when impulse response helps characterizing a
device, there's no such thing as a perfect impulse in real-world
music, and the observed "smearing" over the impulse response just
won't happen with real-world music. Second, this pre- and
post-ringing, that is so easy to see at the graph, is hardly audible,
because of its nature. For a better explanation, just read my first
post in this thread about pre-ringing due to FIR brickwall filters.

Enrique G.

Jay - atldigi

December 3rd 03, 09:17 AM

In article >,
Bobby Owsinski > wrote:

> All I know is that with a PT HD system, 192k sounded better to my ears
> (and all others involved) when we did the experiment. I'm actually
> shocked that more people don't try this, considering the number of HD
> systems out there.

Interestingly, some people have tried and have come to the opposite
conclusion. Certainly any experience, one way or the other, deserves
consideration for what it's worth; but these experiences are far from
conclusive and certainly would not trump a larger body of evidence
should it develop on one possible explanation or another.

While the experience may indicate some difference, it certainly does not
explain the difference. You can't observe one example and conclude that
it must globally be a function of the sample rate when so many other
variables are unaccounted for. It's just one piece of the puzzle.

--
Jay Frigoletto
Mastersuite
Los Angeles
promastering.com

Jay - atldigi

December 3rd 03, 09:34 AM

In article >, "Arny Krueger"
> wrote:

> If you did listening tests that you knew were seriously flawed, why
> didn't you report that they were flawed initially?
>

Bobby's experience is worth hearing, but simply needs to be put into
context of the circumstances, additional differing reports from other
working professionals, and the rest of the body of evidence. I think it
was clear what the circumstances of his listening experiences were and
you can judge accordingly, but there's no reason to ridicule the
offering of his observation, no matter how trivial it seems. It's just a
clue. One can draw no conclusions from it, but if it were me, it would
intrigue me and make me want to look at the issue more closely, with a
more controlled test, and perhaps more wide ranging examples. Where I
would find the time to do this is another question, but I'd be (and I
am) on the lookout for opportunities to do a little more investigation.

--
Jay Frigoletto
Mastersuite
Los Angeles
promastering.com

Arny Krueger

December 3rd 03, 03:39 PM

"KikeG" > wrote in message
om

> John La Grou > wrote in message
> >...
>
>>> IF YOU LIKE IT IT IS FINE. BUT PLEASE GET OFF THAT "EAR IS BETTER
>>> THAN EE". .... IF I GIVE YOU A SCOPE PROBE AND
>>> YOU SEE THE EXACT WAVE FORM IT IS ENDE OF CONVERSATION. THIS
>>> ELECTRICAL WAVEFORM IS WHAT DRIVES AMPLIFIER AND THE SPEAKER. IF YOU
>>> HAVE ZERO DIFFERANCE, THAN YOU CAN NOT HEAR IT EITHER.

>> Too many times, I've heard two circuits with essentially "zero test
>> differences" (THD, IMD, SN, SR, FR, CMR, PH, etc...) sound quite
>> different.

> Now, did you listen to them under blind conditions? If the measured
> differences (there are always some measurable differences to talk of,
> if you measure adequately) were much below known thresholds of
> audibility of human ear, it's most likely it was just placebo effect,
> or in other words, it was your brain knowing what you were listening
> to, playing tricks in your perception. This is a very serious issue,
> that is surprisingly taken in count very little times in the world of
> audio.

There's another issue. IME "measures the same on the test bench" and
"measures the same in actual use" can be two different things. I got my nose
rubbed in this many times while setting up level-matched DBTs. The equipment
was supposed to have flat response but...

> There are fields of audio development where this is a fundamental
> issue that is taken care of (by means of blind tests), such as
> psychoacoustic lossy compression, just because it's fundamental for
> efficient and reliable development.

Note that many interested parties were able to do DBTs when they were trying
to skewer audio watermarking, but the same parties can't seem to figure out
how do a DBT to show the benefits of the extra bandwidth and DR in SACD and
DVD-A.

> Now, there's people that are convinced that there are audible
> differences that can't be measured. I'm really skeptic about that, for
> several reasons. First, because limits of human hearing are pretty
> well known from several years ago. Second, because all kind of
> distortions that can happen over an electric signals are known from
> many years ago. Third, because measurement devices surpassed known ear
> capabilities many years ago, by a very wide margin. And fourth,
> because people claiming such things, usually have not performed a
> controlled, rigorous, reliable blind test, that definitely proves
> their claims.

Agreed. "Ear versus Gear" always makes a lot more sense when the "Ear" is
bias-controlled.

> You know, as Carl Sagan said, "extraordinary claims
> demand extraordinary proof", or something like that. Anecdotal
> evidence, with sighted listening, or unexisting or poor time and/or
> level matching, or very little repeatability is just not good proof
> enough to take it seriously, and I've been witness of such things
> happening relatively often.

Case in point, this thread.

>> Some wise psychoacoustic-physicist types are attempting to
>> find new algorithms to better correlate these subjective differences
>> with repeatable / measurable metrics, such as:

> But they always talk about both readily audible and measurable
> differences. They just want to correlate better the measurements with
> the perceivable degradation.

Agreed.

>> Any linear function is subject to these illusive variables, including
>> ADC front-ends (modulators, etc..). I'm not sure if this speaks to
>> your argument, but certainly there is more to consider here than
>> sampling theory and bench tests.

> Dan is just saying that sampling rate, by itself, can't be the cause
> of audible differences. If there is really an audible difference, then
> it must be somewhere else. I must add, one should make sure the
> differences are not *outside* the equipment, thinking about things as
> simple as slightly different output levels, or just the listener brain
> influencing his perceptions.

(1) Take two identical sources and slightly mismatch the levels. It is
nearly trivial to obtain perfect scores in a DBT.

(2) Take two identical sources and slightly mismatch their timing. It is
nearly trivial to obtain perfect scores in a DBT.

(3) Take two identical sources match them up as perfectly as you can,
technically speaking . It is nearly trivial to obtain perfect scores in a
sighted evaluation.

Why are so many people who claim to be sophisticated with respect to audio
production so reality-challenged when it comes to covering these three
relatively simple points?

> I have personally degraded a piece of music extracted from a CD by
> playing at recording it several times through an inexpensive semi-pro
> soundcard, and challenged people to be able to tell it from the
> original, under blind conditions, in a repeatable manner. Nobody, from
> dozens of person, some with very good equipment, listening training
> and ears, has been able so far, even when the degradation caused is
> easily measurable. If you are interested in trying such a test, just
> tell me, and I'll make it available for anyone that is interested.

Good idea. There's also a lot of relevant evaluations in many other places.
One easy one is posted at http://www.pcavtech.com/test_data/ .

Scott Dorsey

December 3rd 03, 04:16 PM

Jay - atldigi > wrote:
>In article >,
>Bobby Owsinski > wrote:
>
>> All I know is that with a PT HD system, 192k sounded better to my ears
>> (and all others involved) when we did the experiment. I'm actually
>> shocked that more people don't try this, considering the number of HD
>> systems out there.
>
>Interestingly, some people have tried and have come to the opposite
>conclusion. Certainly any experience, one way or the other, deserves
>consideration for what it's worth; but these experiences are far from
>conclusive and certainly would not trump a larger body of evidence
>should it develop on one possible explanation or another.

Hell, I can't even decide which dither algorithm sounds better. One day
one seems to sound better, another day another one seems to sound better.

Let's not even talk about the Ampex 440 sitting next to the ATR-100, because
sometimes maybe it sounds better so I don't have the heart to get rid of it.

>While the experience may indicate some difference, it certainly does not
>explain the difference. You can't observe one example and conclude that
>it must globally be a function of the sample rate when so many other
>variables are unaccounted for. It's just one piece of the puzzle.

I am always hearing differences that don't make sense and I can't explain,
so I'm not surprised at all when other people hear differences they can't
explain.

What worries me is when people try to explain them and get the explanations
all wrong. Better just to say "in this case, this sounded better" and be
done with it.

If I could generalize these things, my 440 would be up for sale in a heartbeat.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

John La Grou

December 3rd 03, 05:01 PM

On 2 Dec 2003 15:52:31 -0800, (dan lavry)
wrote:

>John,
>
>I wonder if you are joining the conversation late in the game. I agree
>that there are diffrences between circuits, converters and so on.
>Also, I can not be a "one man show" here. BUt I know you are a good
>engineer, so I would expect you to agree that IN THEORY, the extra
>sample points carry zero information.

Dan,

If I could slightly rephrase your statement -- yes. In theory,
anything over 2F sample points beyond usable audio is unnecessary.
However, in practice, we're far from achieving a theoretically perfect
delta-sigma modulator.

If you're saying that, in ALL design cases, higher sample rates will
negatively impact a DSM's inherent accuracy, I'll defer to your
experience. I don't know this to be true.

>I would also expect you to agree
>that if audio is say 48KHz in musical content, mic pickup, speaker
>reproduction and the ear itself, Than the differances you hear are not
>about bandwidth.

What I'm saying is that we've not yet correlated the "differences we
hear" with the "differences we measure." We need to be careful about
making black and white statements in these matters.

>I know about slew rates and fet vs bipolar and the rest. My point is
>that it is not about conversion bandwidth.

Ultimately, I think these kinds of arguments should come back to
perceived sonic performance, rather than specifications. If a producer
thinks that a certain ADC sounds more appropriate at 192kHZ than 48kHz
(etc..) for a particular program, no amount of technical hand waving
or protestations will change that reality.

In my opinion, giving recording engineers the choice of 44.1 / 48 /
88.2 / 96 / 176.4 / 192 / DSD / etc.. is better than restricting
their choices. But if you have a strong personal position against
higher sample rates, I can certainly respect that.

JL

John La Grou

December 3rd 03, 05:38 PM

On 3 Dec 2003 00:33:03 -0800, (KikeG) wrote:

>John La Grou > wrote in message >...
>
>> >IF YOU LIKE IT IT IS FINE. BUT PLEASE GET OFF THAT "EAR IS BETTER THAN
>> >EE". .... IF I GIVE YOU A SCOPE PROBE AND
>> >YOU SEE THE EXACT WAVE FORM IT IS ENDE OF CONVERSATION. THIS
>> >ELECTRICAL WAVEFORM IS WHAT DRIVES AMPLIFIER AND THE SPEAKER. IF YOU
>> >HAVE ZERO DIFFERANCE, THAN YOU CAN NOT HEAR IT EITHER.
>
>> Too many times, I've heard two circuits with essentially "zero test
>> differences" (THD, IMD, SN, SR, FR, CMR, PH, etc...) sound quite
>> different.
>
>Now, did you listen to them under blind conditions?

Single blind, not double blind, using a randomizing AB switching box
in a 40kHz analog path.

>If the measured
>differences (there are always some measurable differences to talk of,
>if you measure adequately) were much below known thresholds of
>audibility of human ear, it's most likely it was just placebo effect,
>or in other words, it was your brain knowing what you were listening
>to, playing tricks in your perception. This is a very serious issue,
>that is surprisingly taken in count very little times in the world of
>audio.

You are correct. At these levels of differences, it can become very
difficult and illusive. I've written more about this on my website, so
I'll not get into it again here. I've done enough of these kinds of AB
tests to convince myself that measurements do not always correlate
with predictable sonic performance. To the contrary, sometimes
circuits with better all-around specifications can sound pretty bad,
and visa versa.

>There are fields of audio development where this is a fundamental
>issue that is taken care of (by means of blind tests), such as
>psychoacoustic lossy compression, just because it's fundamental for
>efficient and reliable development.
>
>Now, there's people that are convinced that there are audible
>differences that can't be measured. I'm really skeptic about that, for
>several reasons.

I agree with you. Someday, we will probably have a more comprehensive
suite of tools to measure and characterize audible differences. Today,
however, our tools do not fully correlate human experience with
metrics. Geddes and Lee have effectively proven this with their recent
series of controlled experiments.

> First, because limits of human hearing are pretty
>well known from several years ago. Second, because all kind of
>distortions that can happen over an electric signals are known from
>many years ago.

Perhaps, but we're certainly not testing for all known sources of
distortion. G&L have shown that THD and IMD offer vague and limited
correlation with our real-world experience of music. They are looking
at non-linear and dynamic (vs. static) correlations not found in
today's test equipment. While some of these non-linear phenomenon may
have been known for some time, to my knowledge G&L are the first to
develop accurate connections between theory and perception.

I've also spoken with a number of today's leading audio test engineers
and, to a person, there is general agreement that current audio test
parameters do not sufficiently correlate to audio perception.

>> Any linear function is subject to these illusive variables, including
>> ADC front-ends (modulators, etc..). I'm not sure if this speaks to
>> your argument, but certainly there is more to consider here than
>> sampling theory and bench tests.
>
>Dan is just saying that sampling rate, by itself, can't be the cause
>of audible differences.

I would agree with that statement, with two caveats:

1.) we're talking about sampling theory, not modulator performance

2.) we're sampling all necessary information (insert ultrasonic
arguments here).

Excellent thoughts, Enrique. Nice to see you here.

JL

LeBaron & Alrich

December 3rd 03, 06:45 PM

Arny Krueger wrote:

> Why are so many people who claim to be sophisticated with respect to audio
> production so reality-challenged when it comes to covering these three
> relatively simple points?

Because when not in a lab, but out in the real musical world handling
production, we go by what we hear, if we're paying attention. Based on
what we hear we make gadzillions of decisions to try to achieve whatever
we're after. This is a good thing, in context, and while such
observations do not hold in a scientific sense, it stands to reason that
people doing that kind of work rely on what they think they hear. The
bad things happen when we attempt to attribute causality to what we
hear, outside of the production environment.

--
ha

Steven Sullivan

December 3rd 03, 08:07 PM

Bobby Owsinski > wrote:
> In article >,
> Luke Kaven > wrote:

> > "Roger W. Norman" > wrote:
> > [...]
> > >I can assure anyone here that Bobby Owsinski knows equipment and results and
> > >how to set up double blind tests. Not only does he know equipment, he has
> > >some of the best available and is a very talented person.
> > [...]
> >
> > I don't doubt Bobby's skills. But he made errors in scientific method
> > and analysis, and if he wants to assert his conclusions based on the
> > reasons he offered, we have to call him on it. And it is well worth
> > doing, especially since one would expect him to do a lot with his
> > knowledge and skills.
> >
> > Luke

> Luke,

> There was no scientific method and analysis. There was no attempt to do
> so whatsoever. We were just curious if we could here the difference,
> and all the people involved could with no problem.

The thing is, people will tend to hear a difference if they are
comparing things. Not always, but often. And it can get reinforced
when *groups* are doing the comparing. That's all just human nature,
as described by decades of psychophysics research.
But it doesn't necessarily mean there was a real difference.
So if you are curious to know if there's a real difference, you
have to rule out the possibility that what you perceived
was a 'false positive' due to psychological effects.

Sometimes it's safe to assume a perceived difference was a
real difference -- when compariing things that can be expected
from first principles to sound different, e.g., speakers.
But in the case you describe, where audible difference is
controversial, an uncontrolled comparison, even by
respected audio professionals such as yourself,
unlikley to convince anyone who's aware of the
caveats raised by perceptual psychology.

> Was there another
> reason other than the higher sampling rate? Maybe. Was it because our
> methodology was flawed? We had no methodology so maybe that was the
> reason, or maybe not.

> All I know is that with a PT HD system, 192k sounded better to my ears
> (and all others involved) when we did the experiment. I'm actually
> shocked that more people don't try this, considering the number of HD
> systems out there.

> I have no interest of doing a "scientific" experiment. It's too boring
> for me (and doesn't make me any money). I'll just rely on the empirical
> for now and leave the scientific to people who enjoy it and are better
> at it than I ever could be. Then I'll read the AES paper and hopefully
> learn something new.

Science *is* empirical at its core. It takes observations of the real
world and builds theoretical models from them. But great pains are
taken not to over-intrepret the observations. You have made an
observation, and interpreted it as an audible difference. However,
you haven't ruled out *other* likely interpetations; the
conditions under which you made the observations did not rule
these interpretations out. So you've over-interpreted your
observation.

> The only reason why I posted in the first place was because there seemed
> to be a lot of purely theoretical discussion on the subject yet no one
> talking about what they actually heard.

You have described what you *perceived*. What you actually heard, has
not been established.

--

-S.

"They've got God on their side. All we've got is science and reason."
-- Dawn Hulsey, Talent Director

dan lavry

December 3rd 03, 09:19 PM

"Remixer" > wrote in message news:<3j2zb.179095
> Dan,
>
> Here is a link to a published waveform set, albeit from a manufacturer with
> vested interest, that endeavors to show that impulse response improves as
> sample rate increases. Perhaps there were some EEs involved in producing
> the scope traces. Would you care to comment?
>
> http://www.merging.com/ (at the site, click on the link to DSD/SACD in the
> left column.)

First, I will not talk about other folks gear. Positive remarks, yes,
sometimes. Negative remarks, no.

Second, I do know some folks in audio, and it has come to my attention
that a couple of posters here have been asked by some industry
salesmen to try and shoot my arguments down, or at least wear me out
so I go away. I guess they want to sell thier BS and I am in the way.
Shame on those that do it.

I admit I was having more fun with it a few days ago, so they may be
wearing me out a bit. Regarding the chalanges to what I say, it is a
non issue. As long as I stick with what is totaly well founded, it is
easy to confront it all. Remember the story "the king has no clothes"?
I am the kid that is saying "the king has no cloth".

Well, I did take a look at the web you pointed to. There is not much
there to talk about. Those are "sort of theoretical waves" and it does
not say much.

First, it shows a near spike waveform, and they call it analog. Well
it can be and audio impulse, depending on the scales. If your analog
front end is 100MHz, you can send a 1 usec spike through it. If your
analog front end is say 100KHz, the 1 usec spike will "almost
disapear" - tiny amplitude and slow rise and fall. What does that
analog wave on the web represent? How wide or narrow should it be?

The DSD wave is clearly not a scope picture either. It does not show
the high frequency noise. In fact the other "PCM" impulses are not a
"scope probe picture". To me, all but the "analog implulse" are a time
domain representaion of a low pass filter - for various bandwidth. You
quickly notice that as you increase the bandwidth, the impulse
response gets narrower.

My post from yesterday touched on the SINC finction and the impulses
shown are of that wave shape. If you read my other post (about the
SINC), you find that I mentioned that the zero crossings of the SINC
(the ringings on the impulse) happen at sample times. If you sample
slower, the time between samples is longer, thus you "stretch" the
SINC (impulse) wider. And of course vica versa.

Are the impulses real or not? They are fundumental math tools to
understand sampling, and also used for DSP work (such as design of
FIR's). From and analog standpoint, they are less real in a sense. If
you hit an analog low pass filter with a narrow pulse, can you have
the output to have riniging before the pulse occured? (the ringing on
the left). It is physicaly not realizable. Brick wall filters at
Nyquist are not possible either. So those impulses are not exactly
what you get.

But it does show that there is a relasionship between impulse width
and bandwidth. I can tell you the faster rise times and narrow pulses
require more bandwidth. Indeed after I saw a reasearch showing under
some extream cases and with extream hardware you may find some -50dB
energy at 48KHz. Say we accepted 48KHz as good goal, than your impulse
will be pinned down to so and so width...

We could talk bandwidth or impulse width and be saying the same thing.
If you try and run a 1 nano second pulse into your speaker, will your
speaker (or ear or mic) handel it? Not with 10 Mega Hz system...

When folks wanted to sell you that 192KHz crock, and they realized
that they can not convince anyone that you can hear anything at 96KHz
(or 40KHz or less). They tried , some started talking about that time
resolution thing. The argument was that if you can pin down where the
sound comes from very accuratly, you have great image. Now I am a guy
that researched imaging for 10 years. I know some things regarding
what "kills it" and what does not. With 3usec impulse foing to your
speaker, you need not worry about imaging... Your speaker will not
play it...

I am well aware of the importance of keeping a tight watch on sample
time, and I know that there are situations where a few usec can botch
up a job. But this is about mixing waves out of phase. A 20KHz tone is
only 50usec long. Say you have 2 mics both picking up that 20KHz and
you try to add the signals. If one is delayed by 25usec, yo get
cancelation instead of addition. So yes keep the AD timing (clocks and
delays) together, be it at 44.1KHz or 1Mhz....
But that is about properly processing of electrical signals.

When it comes to sound, what is 25us mean? Sound speed in air is about
1000 feet/second. This is .012 inch per usec. So 25usec acoustic delay
is equivalent to moving a mic, or speaker or your ear by .3 inch.

So if someone is telling me that I get better time resolution, better
image, when the impulse is say 3usec and not 10usec. I know of someone
very credible that reasearch it, and told me that they found a few
folks (out of many) capable of discerning 15usec delay step on say the
R speaker (holding the L fixed). I asked him how they were able to
hold the distaces fixed to .18 inch and the answer was: The listners
agreed to be tested with thier heads in a mechanical tight grip.

So I am going to stop now. I am gald to see those pictures of the
analod, dsd and pcm impulses there without too much BS attached to it.
Those that are looking for explanations for whay 192K or 384K 0r
384000K is the best thing can sort of "fall or it". There is some
idirect sugestion there that the 192 impulse looks more "analog like"
at 192Khz. In fact the first thing to notice is that the musical
instrument gives out a HUGELY WIDE analog impulse compare to not just
the shown "analog impulse", but also compare to all the other SINC
shaped impulses...

So we have dealt with:
More dots are better
Nyquist is a bum (PHD from Yale, 3 major theorm, one of Bell labs
greats...)
We can not explain nothing
It is antialias filter
Very narrow pulses
Nobody tells me nothing, I know

I probably forgot some. What is next

BR
Dan Lavry

Peter L. Pollack

December 3rd 03, 09:42 PM

Mr. Lavry,

I just wanted to say -- before you tire of our little group -- that I'm
enjoying reading your posts and have learned quite a lot. I suspect I'm
not alone.

Thank you for the excellent information. Please do not let a few
argumentative types ruin rec.audio.pro for you.

-Pete Pollack

In article >,
says...
> "Tommi" > wrote in message >...
> > "dan lavry" > wrote in message
> > om...
> >
> > > YOU DID NOT GET IT! ASSUMING PERFECT CONVERSION...
> >
> > Erm..You're either replying to the wrong post, or missed the tone of my
> > text.
> > I did never say that 192kHz sounds better to me, OR that faster sampling
> > results in better accuracy! I was just replying to the post where someone
> > said that.
>
> I must have misread you. My appologies.
>
> BR
>
> Dan Lavry
>
> > I really don't understand this reply. Either I didn't express my opinion
> > correctly, or it was severely misunderstood. Either way, I don't disagree
> > with any of the above since my point was more or less the same.
>
> Again, sorry if I misunderstood your comments. I certainly do not want
> to ofend anyone. I must be getting weary...
>
> Regards
>
> Dan Lavry
>

Justin Case

December 3rd 03, 10:01 PM

"Roger W. Norman" > wrote in message >...
> And, of course, the old saw is "The Titanic looked good, on paper."
> Unfortunately it had to float on the sea, of which it obviously did a pretty
> **** poor job.

Actually, the Titanic did a pretty good job of 'floating on the sea'.
At least until that damn iceberg got in the way...

Arny Krueger

December 3rd 03, 10:17 PM

"LeBaron & Alrich" > wrote in message

> Arny Krueger wrote:
>
>> Why are so many people who claim to be sophisticated with respect to
>> audio production so reality-challenged when it comes to covering
>> these three relatively simple points?

> Because when not in a lab, but out in the real musical world handling
> production, we go by what we hear, if we're paying attention.

Which is fine for what we do in that context.

>Based on what we hear we make gadzillions of decisions to try to achieve
> whatever we're after. This is a good thing, in context, and while such
> observations do not hold in a scientific sense, it stands to reason
> that people doing that kind of work rely on what they think they
> hear.

Which is fine for what they do in that context.

> The bad things happen when we attempt to attribute causality to
> what we hear, outside of the production environment.

Agreed, but that's not exactly the answer to the question I'm asking. Maybe
I'm not asking it clearly enough.

Given that people are now informed about the need to play by different rules
during certain equipment evaluations, and given that the new rules have been
laid out in detail and are even explained by means of practical examples,
why aren't people applying their production skills in this other context?

Arny Krueger

December 3rd 03, 10:20 PM

"Justin Case" > wrote in message
om
> "Roger W. Norman" > wrote in message
> >...
>> And, of course, the old saw is "The Titanic looked good, on paper."
>> Unfortunately it had to float on the sea, of which it obviously did
>> a pretty **** poor job.
>
> Actually, the Titanic did a pretty good job of 'floating on the sea'.
> At least until that damn iceberg got in the way...

The Titanic looked bad on paper. The low bulkheads and the missing lifeboats
were in the paper plans.

Chris Hornbeck

December 3rd 03, 11:04 PM

On Wed, 3 Dec 2003 17:17:45 -0500, "Arny Krueger" >
wrote:

>Given that people are now informed about the need to play by different rules
>during certain equipment evaluations, and given that the new rules have been
>laid out in detail and are even explained by means of practical examples,
>why aren't people applying their production skills in this other context?

Not certain how useful this might be as an answer, but there is a
school of thought that different listening tasks use different
brain pathways (whatever that means) and even different parts of
the brain.

Music listening for enjoyment, for example, is a very different
task than speech listening for comprehension. Analytical listening
to music may be like either or neither of these; beats me.

I suspect that *some* of the gulf of viewpoints in these
religious arguments between smart, capable and informed listeners
lies in different modes of perception, some important ones
even being non-verbal.

Makes for a tricky post-game wrap-up.
Anyway, just a thought.

Chris Hornbeck
"That is my Theory, and what it is too."
Anne Elk

Arny Krueger

December 4th 03, 12:54 AM

"John La Grou" > wrote in message

> Perhaps, but we're certainly not testing for all known sources of
> distortion. G&L have shown that THD and IMD offer vague and limited
> correlation with our real-world experience of music. They are looking
> at non-linear and dynamic (vs. static) correlations not found in
> today's test equipment. While some of these non-linear phenomenon may
> have been known for some time, to my knowledge G&L are the first to
> develop accurate connections between theory and perception.

Geddes and Lee recent AES papers (preprints 5890 and 5891) describe a scheme
for weighting measurements of various orders of nonlinear distortion in such
a way that the resulting number increases and decreases in a way that is not
inconsistent with human perception of audible degradation.

The means described by Geddes and Lee for actually analyzing the results of
a fairly conventional signal analysis while novel, is not surprising. It
deals with errors that have been known about for a long time and that are
already routinely measured by other means. It is a weighting scheme for
orders of nonlinear distortion that is proportional to the square of the
order of the distortion, and inversely proportional to the amplitude of the
signal at which the distortion is apparent.

Heavier weighting of higher orders of distortion has been suggested by
various investigators for decades. It has long been known that
discontinuities at low levels, such as crossover distortion, are highly
annoying.

The Geddes-Lee metric does not predict the threshold of audibility of
nonlinear distortion.

The Geddes-Lee metric was designed for the analysis of the performance audio
components such as loudspeakers that have relatively large amounts of
nonlinear distortion. This is consistent with Geddes previous years study of
loudspeakers, and many of his earlier papers.

Therefore, it is highly questionable that the Geddes-Lee metric is
applicable to equipment that are already known to have extremely low levels
of nonlinear distortion, such as high quality analog-digital converters.

About a month ago I reviewed this analysis of the Geddes-Lee papers with one
of the authors:

Basically, (the recent Geddes-Lee AES papers) propose weighting nonlinear
coefficients (e.g., harmonics)
based on the cosine of the amplitude times order squared.

The claim is made that the audibility of nonlinear distortion is inversely
related to amplitude (nonlinearities at low levels are more audible) and
also related to order squared (nonlinearities that generate spurious
responses that are more broadly dispersed from the original signal are far
more audible). They back their new metric up with listening test results
showing that this criteria does a better job of fitting a variety of
synthetic distortion sources to perceptions of degraded sound quality.

In short, the Geddes/Lee papers show that the way most people formally
characterize audio gear nonlinear distortion today, which focuses on high
outputs and unweighted harmonics, is about as close to irrelevant as one
could imagine. The papers report experimental studies of current metrics
that support the idea that they are irrelevant or at least uncorrelated with
human perceptions of sound quality.

My own simplistic research into the subject and the scientific literature
was highly supportive of Geddes/Lee general thinking before they got the far
more polished and complete results that are published in the articles. My
point here is that it seems reasonable to view the Geddes/Lee results as
being orthodox and reasonable, as far as they go. To their credit they seem
to have found a far better metric, but maybe not the best metric. I'll take
better as long as it is the best we have!

Mike Rivers

December 4th 03, 01:16 AM

In article > writes:

> I am always hearing differences that don't make sense and I can't explain,
> so I'm not surprised at all when other people hear differences they can't
> explain.
>
> What worries me is when people try to explain them and get the explanations
> all wrong. Better just to say "in this case, this sounded better" and be
> done with it.

What worries me is that things can sound so good these days and still
people are worried that there's something better so they're not
satisfied with their own good tools. It's a wonder that some people
get anything done (and no wonder that others get plenty done).

--
I'm really Mike Rivers - )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

John La Grou

December 4th 03, 01:41 AM

On Wed, 3 Dec 2003 19:54:45 -0500, "Arny Krueger" >
wrote:

>The Geddes-Lee metric was designed for the analysis of the performance audio
>components such as loudspeakers that have relatively large amounts of
>nonlinear distortion. This is consistent with Geddes previous years study of
>loudspeakers, and many of his earlier papers.
>
>Therefore, it is highly questionable that the Geddes-Lee metric is
>applicable to equipment that are already known to have extremely low levels
>of nonlinear distortion, such as high quality analog-digital converters.

Arny,

Good overview. Thanks.

I spent some time talking with Earl (Geddes) before the last AES
Convention. He seems convinced that their work can, and will, evolve
into a generalized metric applicable to all audio distortion testing.

I hope he's right.

JL

Arny Krueger

December 4th 03, 02:29 PM

"John La Grou" > wrote in message

> On Wed, 3 Dec 2003 19:54:45 -0500, "Arny Krueger" >
> wrote:
>
>> The Geddes-Lee metric was designed for the analysis of the
>> performance audio components such as loudspeakers that have
>> relatively large amounts of nonlinear distortion. This is consistent
>> with Geddes previous years study of loudspeakers, and many of his
>> earlier papers.

>> Therefore, it is highly questionable that the Geddes-Lee metric is
>> applicable to equipment that are already known to have extremely low
>> levels of nonlinear distortion, such as high quality analog-digital
>> converters.

> Arny,

> Good overview. Thanks.

You're welcome.

> I spent some time talking with Earl (Geddes) before the last AES
> Convention. He seems convinced that their work can, and will, evolve
> into a generalized metric applicable to all audio distortion testing.

> I hope he's right.

Earl and I (& the rest of the *Detroit Audio Mafia*) have been talking about
this topic for years... A good background for these papers can be found in a
couple of chapters of his book about speakers.
http://www.gedlee.com/Audio_trans.htm /

Earl thinks a lot more about this sort of thing these days than I do... I
think his two AES papers represented a lot of progress.

For a theory to be generalized it has to handle both the middle ground and
the boundaries. I see a lot in his current direction that relates to the
middle ground, but I don't see anything that relates to the most important
boundary condition: The Threshold Of Audibility.

LeBaron & Alrich

December 4th 03, 03:35 PM

dan lavry wrote:

> First, I will not talk about other folks gear. Positive remarks, yes,
> sometimes. Negative remarks, no.

> Second, I do know some folks in audio, and it has come to my attention
> that a couple of posters here have been asked by some industry
> salesmen to try and shoot my arguments down, or at least wear me out
> so I go away. I guess they want to sell thier BS and I am in the way.
> Shame on those that do it.

> I admit I was having more fun with it a few days ago, so they may be
> wearing me out a bit. Regarding the chalanges to what I say, it is a
> non issue. As long as I stick with what is totaly well founded, it is
> easy to confront it all. Remember the story "the king has no clothes"?
> I am the kid that is saying "the king has no cloth".

Ignore those naked dweebs and hold in mind that you have presented
material for cogitation and that those who engage in that sort of
activity are both grateful and intrigued. Only those with open minds can
have a chance of seeing the light. The rest will remain in darkness and
there's little one can do about them. More is more except when it isn't.
But not many people in the US today pay attention to details.

Thank you for your contributions. Understand that threads here sometimes
have a settling time of months and eventually some of them are no longer
worth sampling.

I'll wager you will sell some of those nifty little USB preamps. Thanks
for making those and putting the emphasis where you have.

--
hank alrich * secret mountain
audio recording * music production * sound reinforcement
"If laughter is the best medicine let's take a double dose"

Scott Dorsey

December 4th 03, 04:31 PM

In article <znr1070493402k@trad>, Mike Rivers > wrote:
>In article > writes:
>
>> I am always hearing differences that don't make sense and I can't explain,
>> so I'm not surprised at all when other people hear differences they can't
>> explain.
>>
>> What worries me is when people try to explain them and get the explanations
>> all wrong. Better just to say "in this case, this sounded better" and be
>> done with it.
>
>What worries me is that things can sound so good these days and still
>people are worried that there's something better so they're not
>satisfied with their own good tools. It's a wonder that some people
>get anything done (and no wonder that others get plenty done).

I have been working on a compilation album of tracks made at a local folk
festival over the past 40 years or so.

One of the tracks on the album was originally recorded with a Wollensak
recorder and an Altec PA mike (probably a rebadged EV) being used as an
area mike. This would have been around 1970 or so, I think. 3 3/4 ips.

I played the recording back for the original group, which was playing at
an event this past weekend in Maryland, and they were amazed at how it
sounded. "This sounds great! What kind of equipment was used?" one of
the guys asked.

To be fair, Don Grossinger spent about two days getting it to sound that
good. And it would have been better if the mike placement had been a bit
closer. But you know, it's good enough to go on the album, and I predict
people will buy it.

Now I am waiting on the Plangent Technologies people to see if they can
clean up a slightly later track recorded on a Uher 4000 with some problems.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Earl Geddes

December 4th 03, 04:35 PM

"Arny Krueger" > wrote in message >...
>

Arny's "been their, done that" attitude aside he is not correct about
all he says about my papers on subjective assement of nonlinear
distortion.

> Geddes and Lee recent AES papers (preprints 5890 and 5891) describe a scheme
> for weighting measurements of various orders of nonlinear distortion in such
> a way that the resulting number increases and decreases in a way that is not
> inconsistent with human perception of audible degradation.
>
> The means described by Geddes and Lee for actually analyzing the results of
> a fairly conventional signal analysis while novel, is not surprising. It
> deals with errors that have been known about for a long time and that are
> already routinely measured by other means. It is a weighting scheme for
> orders of nonlinear distortion that is proportional to the square of the
> order of the distortion, and inversely proportional to the amplitude of the
> signal at which the distortion is apparent.

If there errors were know and dealt with "by other means" please lead
me to those results.

>
> Heavier weighting of higher orders of distortion has been suggested by
> various investigators for decades. It has long been known that
> discontinuities at low levels, such as crossover distortion, are highly
> annoying.

Correct, but it has never been quantified in a metric. weighting the
orders alone achieves only a small improvement in correlation compared
to the entire scheme that I use.

>
> The Geddes-Lee metric does not predict the threshold of audibility of
> nonlinear distortion.

Incorrect. Read the papers.

>
> The Geddes-Lee metric was designed for the analysis of the performance audio
> components such as loudspeakers that have relatively large amounts of
> nonlinear distortion. This is consistent with Geddes previous years study of
> loudspeakers, and many of his earlier papers.

Incorrect. Read the papers - they only breifly mention loudspeakers.
The work is applicable to all audio components.

>
> Therefore, it is highly questionable that the Geddes-Lee metric is
> applicable to equipment that are already known to have extremely low levels
> of nonlinear distortion, such as high quality analog-digital converters.

Incorrect.

>
> About a month ago I reviewed this analysis of the Geddes-Lee papers with one
> of the authors:

And I warned you that you did not have the correct understanding and
this may come back and bite you.

>
> Basically, (the recent Geddes-Lee AES papers) propose weighting nonlinear
> coefficients (e.g., harmonics)
> based on the cosine of the amplitude times order squared.

Close but no cigar. It does not deal with harmonics at all. Read the
papers.

>
> The claim is made that the audibility of nonlinear distortion is inversely
> related to amplitude (nonlinearities at low levels are more audible) and
> also related to order squared (nonlinearities that generate spurious
> responses that are more broadly dispersed from the original signal are far
> more audible). They back their new metric up with listening test results
> showing that this criteria does a better job of fitting a variety of
> synthetic distortion sources to perceptions of degraded sound quality.

This you got right.

>
> In short, the Geddes/Lee papers show that the way most people formally
> characterize audio gear nonlinear distortion today, which focuses on high
> outputs and unweighted harmonics, is about as close to irrelevant as one
> could imagine. The papers report experimental studies of current metrics
> that support the idea that they are irrelevant or at least uncorrelated with
> human perceptions of sound quality.

Correct.

>
> My own simplistic research into the subject and the scientific literature
> was highly supportive of Geddes/Lee general thinking before they got the far
> more polished and complete results that are published in the articles.

And I too have been studying this problem for nearly twenty years. I
only published my work when it was conclusive.

My
> point here is that it seems reasonable to view the Geddes/Lee results as
> being orthodox and reasonable, as far as they go. To their credit they seem
> to have found a far better metric, but maybe not the best metric. I'll take
> better as long as it is the best we have!

When you have a better metric (proven) please let me know, but thanks
for the "qualified" support.

Earl Geddes
www.gedlee.com

Arny Krueger

December 4th 03, 06:11 PM

"Earl Geddes" > wrote in message
om
> "Arny Krueger" > wrote in message
> >...
>
>> The Geddes-Lee metric does not predict the threshold of audibility
>> of nonlinear distortion.

> Incorrect. Read the papers.

Earl, you talk about my attitude, this is a good example of yours. We've
both got the papers sitting in front of us (and you know it!) and I don't
see what you're claiming. Therefore, it's appropriate for you to quote
yourself to support your own claim, as opposed to attacking my reading
comprehension.

One of these days Earl you're going to learn that communications failures
reflect on both the reader and the writer.

Not everything written in English is perfectly understood by all competent
readers of English.

>> The Geddes-Lee metric was designed for the analysis of the
>> performance audio components such as loudspeakers that have
>> relatively large amounts of nonlinear distortion. This is consistent
>> with Geddes previous years study of loudspeakers, and many of his
>> earlier papers.
>
> Incorrect. Read the papers - they only briefly mention loudspeakers.
> The work is applicable to all audio components.

I'll reserve my comments about that until you support your claims about
thresholds. There's a rather obvious logical point here that you seem to
keep missing Earl. Distortion audibility only makes sense when the
distortion is audible. Not all distortion that is measurable, is audible.

>> Therefore, it is highly questionable that the Geddes-Lee metric is
>> applicable to equipment that are already known to have extremely low
>> levels of nonlinear distortion, such as high quality analog-digital
>> converters.

> Incorrect.

Ditto.

>> About a month ago I reviewed this analysis of the Geddes-Lee papers
>> with one of the authors:

> And I warned you that you did not have the correct understanding and
> this may come back and bite you.

See my former comments about communication failures.

>> Basically, (the recent Geddes-Lee AES papers) propose weighting
>> nonlinear coefficients (e.g., harmonics)
>> based on the cosine of the amplitude times order squared.

> Close but no cigar. It does not deal with harmonics at all. Read the
papers.

You're splitting hairs, Earl. We both know that nonlinear coefficients and
harmonics are not the identically same thing, but they are related. I know
from experience that a lot more people can relate to harmonics than
nonlinear coefficients, so I was trying to make this as understandable as I
can to people who don't get into the details. Notice that I restated the
same idea in more detail in the next paragraph and got an A-OK from you.

>> The claim is made that the audibility of nonlinear distortion is
>> inversely related to amplitude (nonlinearities at low levels are
>> more audible) and also related to order squared (nonlinearities that
>> generate spurious responses that are more broadly dispersed from the
>> original signal are far more audible). They back their new metric up
>> with listening test results showing that this criteria does a better
>> job of fitting a variety of synthetic distortion sources to
>> perceptions of degraded sound quality.

> This you got right.

>> In short, the Geddes/Lee papers show that the way most people
>> formally characterize audio gear nonlinear distortion today, which
>> focuses on high outputs and unweighted harmonics, is about as close
>> to irrelevant as one could imagine. The papers report experimental
>> studies of current metrics that support the idea that they are
>> irrelevant or at least uncorrelated with human perceptions of sound
>> quality.

> Correct.

>> My own simplistic research into the subject and the scientific
>> literature was highly supportive of Geddes/Lee general thinking
>> before they got the far more polished and complete results that are
>> published in the articles.

> And I too have been studying this problem for nearly twenty years. I
> only published my work when it was conclusive.

Of course, that's a good thing.

> My
>> point here is that it seems reasonable to view the Geddes/Lee
>> results as being orthodox and reasonable, as far as they go. To
>> their credit they seem to have found a far better metric, but maybe
>> not the best metric. I'll take better as long as it is the best we
>> have!

> When you have a better metric (proven) please let me know, but thanks
> for the "qualified" support.

Earl, the fact that you have a metric that you've shown is better than
something that is very bad does not mean that you've got the best possible
metric. For example, does squared fit the subjective data better than
magnitude of first order or third order? I see no experimental data either
way. Is cosine better than Sinc going up to the first zero-crossing...

Bob Cain

December 4th 03, 11:10 PM

LeBaron & Alrich wrote:
>
> I'll wager you will sell some of those nifty little USB preamps. Thanks
> for making those and putting the emphasis where you have.

Sorry to be ignorant, but which are those?

Bob
--

"Things should be described as simply as possible, but no
simpler."

A. Einstein

Arny Krueger

December 5th 03, 12:23 AM

"Bob Cain" > wrote in message

> LeBaron & Alrich wrote:
>>
>> I'll wager you will sell some of those nifty little USB preamps.
>> Thanks for making those and putting the emphasis where you have.
>
> Sorry to be ignorant, but which are those?

I had the same question and the answer is...

http://www.lavryengineering.com/miniprs2.html

Ain't google wonderful?

;-)

Jay - atldigi

December 5th 03, 04:52 AM

In article >, "Arny Krueger"
> wrote:

> >> I'll wager you will sell some of those nifty little USB preamps.
> >> Thanks for making those and putting the emphasis where you have.
> >
> > Sorry to be ignorant, but which are those?
>
> I had the same question and the answer is...
>
> http://www.lavryengineering.com/miniprs2.html
>
> Ain't google wonderful?
>
> ;-)

Sure is. God save the queen - and google, and Dan Lavry too!

--
Jay Frigoletto
Mastersuite
Los Angeles
promastering.com

Mike Rivers

December 5th 03, 12:00 PM

In article > writes:

> You're committing the same fallacy again. You have no grounds to
> attribute those difference to any particular causes, whether sampling
> frequency, or elsewise.

Why does he have to attribute what he heard to anything? He liked X
better than Y. Isn't that OK? Why do you like Glenfiddich better than
Tullimore?

The fact that in the case of Bobby's listening test, X happened to be
chocolate . . . sampled at 192 kHz may or may not mean that all
chocolate is better, just what he listened to that day. As he said,
not very scientific, just an observation that a particular system
running at 192 kHz sounded better than a similar system running at
96 kHz.

Until everything but sample rate can be held constant (which it can't,
presently), or until you're ready to accept a comparison of two
different systems, you really can't do a valid test. It's like
comparing two mic preamps and concluding that tubes sound better than
solid state because the one you liked best had tubes.

--
I'm really Mike Rivers - )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

Arny Krueger

December 5th 03, 05:49 PM

"Mike Rivers" > wrote in message
news:znr1070591864k@trad

> Until everything but sample rate can be held constant (which it can't,
> presently), or until you're ready to accept a comparison of two
> different systems, you really can't do a valid test. It's like
> comparing two mic preamps and concluding that tubes sound better than
> solid state because the one you liked best had tubes.

I don't know about that. I think that it's pretty easy to hold practically
everything but sample rate constant, particularly considering the outcome of
the experiment, every time it's actually done.

The basic approach is to take a high sample rate original recording,

(I.) downsample it,
(II.) upsample it, and
(III.) compare the original and the downsampled/upsampled copy. They are
both high sample rate files. Only the data in them differs.

This can easily be done using PCABX or something like it, which maintains
the entire reproduction chain in an electrically identical condition for
either alternative.

Now, if there were a positive outcome for audible differences, it wouldn't
be known whether:

(a) The audible difference is due to the change in sample rate, or whether
(b) the audible difference is due to the upsampling & downsampling, or
whether
(c) the audible difference is due to both the change in sample rate and the
resampling.

However, with good upsampling and downsampling there have NEVER been any
positive outcomes for differences in a blind, time-synched, level-matched
test.

(Note to Fletcher, these tests were always done without customers waiting.
;-) )

Therefore we must conclude that:

(1) The lower sample rates that were investigated and produced no audible
differences don't cause audible degradation,

-or that-

(2) The downsampling and upsampling some how mysteriously exactly
compensates for the degradation caused by the lower sample rates.

==================

(2) seems very unlikely!

dan lavry

December 5th 03, 06:15 PM

John La Grou > wrote in message >...
> On 2 Dec 2003 15:52:31 -0800, (dan lavry)
> wrote:
>
> >John,
> >
> >I wonder if you are joining the conversation late in the game. >
>
> Dan,
>
> >
> If you're saying that, in ALL design cases, higher sample rates will
> negatively impact a DSM's inherent accuracy, I'll defer to your
> experience. I don't know this to be true.

I guess you are comming late into the conversation. I will repeat some
of the points:

Yes, there is speed accuracy tradeoff. Go to any manufacturer at any
given year and check the state of the art conversion (AD and DA's).
You find out that you do not get a lot of bits, much dynamic range or
good distortions at very high speeds. Today, at 1GHz it is 1 bit. 10
bits around 50MHz or so, and approaching 24 bits at a few Hz.

I do not know all that you do, but I know you do good analog work. So
you should be well equipped to know that some of the video amps just
don't cut it for audio... I assume you check amplifier settling time,
and so you know that if I give it a step, the longer I wait the closer
it (the output) convarge toward the correct value. That is true for
exponetial charge, for overshoot and ringing... For fast 8 bits AD,
you may want some circuit that settles to say .1% very fast. That will
not be the choice for audio... Clearly, you can go anywhere and see
that speed and accuracy is a tradeoff. And it is not just about
settling. If your sample hold (or switched cap filter) is meant to
work faster, than you need to reduce the cap values ending up with
more feed through. If your circuit depends on feedback, you may have
more open loop gain at lower frequencies. I am not even taking time to
think about it, just typing as fast as I can...

Now, the AD's out there can be based on say (for example), 3 bits at
high speed. The feedback DA circuit in them is not a 3 bit +/- 1/2 LSB
circuit. If you want to end up with good say 20 bits, the 3 bits DA
accuracy needs to be 3 bit +/- .00000001LSB accuracy. So settling is
importent.

Even with DSD, when everyone is talking about that easy single
comperator circuit, speed is importent. Comperators do not switch "at
zero". You need to "overcome something", pumping some charge to get
over zero. If it is a bipolar, you need to both overcome the
equivalent input capacity, and the fT (gain bandwidth limitations of
the transistor)... and so on. The switching time does depend on the
slope (rise time) of the driving signal. There is memory effect
against you in terms of thermal effects. When the input transistor is
off, it is cooler. You "go over zero" and turn it on and you are in
2.2mV per degree C world, and the time constants are fast! the
dimensions are tiny... It is worse with FETs..... and so on.

It is possible that your area is strickly analog. That is fine. You do
good work and I respect it. But when going into AD conversion, the
WHOLE CONCEPT is to trade bits for bandwidth. That 1 bit at 64fs needs
to go through 6 octave sample rate reduction to get to be in that
17-18 bit dynamic range area. Do you think one can take a 1 bit at
2.88Mhz and make it into 16 bits at 1.44MHz? The same is true for the
multibit noise shaping stuff. If your goal is say 48KHz bandwidth, and
not 96KHz, you have a lot of advatage. A whole octave of decimation
advatage.

The tradeoff for sigma delta and noise shaping (on a "per octave"
basis) gets bigger when you increase the order of the feedback. This
is very complicated stuff.

Lets keep the analog specific requirnments for amps and device types
out of it. Lets agree thogh, that the same designer making an amp with
say .001% large signal distortions at say 100KHz, will not be able to
get such performance at 100MHz. No one says that if you want to say
amplify a signal with 20KHz bandwidth, you need and amp with 21KHz
bandwidth :-)

BR

Dan Lavry

Kurt Albershardt

December 5th 03, 07:40 PM

dan lavry wrote:
>
> when going into AD conversion, the
> WHOLE CONCEPT is to trade bits for bandwidth. That 1 bit at 64fs needs
> to go through 6 octave sample rate reduction to get to be in that
> 17-18 bit dynamic range area. Do you think one can take a 1 bit at
> 2.88Mhz and make it into 16 bits at 1.44MHz? The same is true for the
> multibit noise shaping stuff. If your goal is say 48KHz bandwidth, and
> not 96KHz, you have a lot of advatage. A whole octave of decimation
> advatage.

So what if we say our goal is 24 kHz bandwidth, but using an 88.2k or
96k samplerate?

Mike Rivers

December 5th 03, 11:18 PM

In article > writes:

> I think that it's pretty easy to hold practically
> everything but sample rate constant, particularly considering the outcome of
> the experiment, every time it's actually done.
>
> The basic approach is to take a high sample rate original recording,
>
> (I.) downsample it,
> (II.) upsample it, and
> (III.) compare the original and the downsampled/upsampled copy. They are
> both high sample rate files. Only the data in them differs.

But this isn't what we do in real life. You might be able to prove
something, but is it something useful?

What I was looking for was to run a signal from a live sound source
through a microphone and preamp (obviously those can be held constant)
into two sets of back-to-back A/D and D/A converters, one running at
each sample rate under test and compare each of the outputs with the
source to see if it's possible to say that one sounds more like the
original than the other.

Now if you take a 96 kHz converter and simply double the sample rate
of the A/D and D/A conversion without changing anything else, then you
may be able to conduct this experiment. But you haven't really made a
typical 192 khz converter because you're still filtering the input
below 48 kHz. If there was actually a difference in sound, then this
would disprove Dan's argument. And if they sounded the same, you're
not proving that you can (or can't) make a more accurate conversion by
working with a wider bandwidth.

--
I'm really Mike Rivers - )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

Luke Kaven

December 6th 03, 02:03 AM

(Mike Rivers) wrote:
> writes:
>
>> You're committing the same fallacy again. You have no grounds to
>> attribute those difference to any particular causes, whether sampling
>> frequency, or elsewise.
>
>Why does he have to attribute what he heard to anything? He liked X
>better than Y. Isn't that OK? Why do you like Glenfiddich better than
>Tullimore?
>
>The fact that in the case of Bobby's listening test, X happened to be
>chocolate . . . sampled at 192 kHz may or may not mean that all
>chocolate is better, just what he listened to that day. As he said,
>not very scientific, just an observation that a particular system
>running at 192 kHz sounded better than a similar system running at
>96 kHz.

It was implicit in Bobby's report of the test (however casual he might
have said it was) that the perceived differences in results were
attributable to the change in sampling rate. Otherwise, his remarks
would have been vacuuous.

For example, If he had said "I like the blue converters over the red
converters", which involves what we presume to be two properties that
are wholly inert for all we know, then his remarks would have been
vacuuous, and we might have asked why he bothered straying off topic.
It was no mere coincidence that he identified the two tests solely by
a change in sample rate.

Arny Krueger

December 6th 03, 02:40 AM

"Mike Rivers" > wrote in message
news:znr1070661584k@trad
> In article >
> writes:

>> I think that it's pretty easy to hold practically
>> everything but sample rate constant, particularly considering the
>> outcome of the experiment, every time it's actually done.

>> The basic approach is to take a high sample rate original recording,

>> (I.) downsample it,
>> (II.) upsample it, and
>> (III.) compare the original and the downsampled/upsampled copy. They
>> are both high sample rate files. Only the data in them differs.

> But this isn't what we do in real life.

In real life we don't compare music at different sample rates, either.

> You might be able to prove something, but is it something useful?

I think so. Regrettably you snipped the explanation of the reasons why.

> What I was looking for was to run a signal from a live sound source
> through a microphone and preamp (obviously those can be held constant)
> into two sets of back-to-back A/D and D/A converters, one running at
> each sample rate under test and compare each of the outputs with the
> source to see if it's possible to say that one sounds more like the
> original than the other.

Why bother with two sets of converters? Why not just compare the analog feed
to something that has been round-tripped through back-to-back converters?
Been there done that.

I think it can be done with a Lynx L22... You just need to double-check the
levels and have an hidden assistant to swap cables behind a curtain and
another assistant to play the music.

All the PCABX stuff does is eliminate the need for any assistants.

Everybody I've known or heard of who has tried something like this, blind,
has been very impressed with the results. It takes what most people find to
be shockingly few bits to be the subjective equal of a straight wire. IME
44/16 is more than enough.

Arny Krueger

December 6th 03, 02:40 AM

"Luke Kaven" > wrote in message

> (Mike Rivers) wrote:
>> writes:
>>
>>> You're committing the same fallacy again. You have no grounds to
>>> attribute those difference to any particular causes, whether
>>> sampling frequency, or elsewise.
>>
>> Why does he have to attribute what he heard to anything? He liked X
>> better than Y. Isn't that OK? Why do you like Glenfiddich better than
>> Tullimore?
>>
>> The fact that in the case of Bobby's listening test, X happened to be
>> chocolate . . . sampled at 192 kHz may or may not mean that all
>> chocolate is better, just what he listened to that day. As he said,
>> not very scientific, just an observation that a particular system
>> running at 192 kHz sounded better than a similar system running at
>> 96 kHz.
>
> It was implicit in Bobby's report of the test (however casual he might
> have said it was) that the perceived differences in results were
> attributable to the change in sampling rate. Otherwise, his remarks
> would have been vacuuous.
>
> For example, If he had said "I like the blue converters over the red
> converters", which involves what we presume to be two properties that
> are wholly inert for all we know, then his remarks would have been
> vacuuous, and we might have asked why he bothered straying off topic.
> It was no mere coincidence that he identified the two tests solely by
> a change in sample rate.

Aye, there's the rub.

Arny Krueger

December 6th 03, 03:21 AM

"Arny Krueger" > wrote in message

> "Earl Geddes" > wrote in message
> om
>> "Arny Krueger" > wrote in message
>> >...
>>
>>> The Geddes-Lee metric does not predict the threshold of audibility
>>> of nonlinear distortion.
>
>> Incorrect. Read the papers.

Well, Earl had you said something like take another look at Figure 5 in
preprint 5891 you would have saved a lot of grief.

The text a bit below Figure 5 says:

"The results indicate that systems where Gm < 1.0 can be expected to yield
subjective ratings of "imperceptible" and that levels of Gm < 3.0 can be
expected to yield subjective ratings of "barely perceptible but not
annoying".

Gm is of course the Geddes-Lee metric. This effectively corrects my
misunderstanding, and shows that the Geddes-Lee metric DOES have the
potential to predict the threshold of audibility of various kinds of
nonlinear distortion. Therefore it can be used to determine which equipment
has audible distortion and which does not.

Whether the Geddes-Lee metric is applicable to judging differences among low
distortion equipment such as good analog-digital converters and power
amplifiers that are operating in their nominally linear range, then depends
on what their Geddes-Lee metric actually is.

If two amplifiers have Geddes-Lee metric values < 1.0, then the metric says
they both have imperceptible distortion. The metric then sheds no further
light on their sound quality. However, if amplifier distortion is
imperceptible, then no further light would seem to need to be shed!*

A real-world clue to the levels of nonlinear distortion that are
"imperceptible" comes from the fact that the monitoring system that was used
for the subjective evaluations was composed of a Turtle Beach Santa Cruz
sound card and Etymotic ER-4 earphones. Fine equipment but by no means
exotic. This test setup must itself have a metric value of < 1.0 since
figure 5 and Table 1 show five instances where Gm < 1.0. That the Turtle
Beach Santa Cruz sound card has inaudible distortion is confirmed by over
two years of listener experience with
http://www.pcabx.com/product/santa_cruz/index.htm .

* note - when equipment is used in a cascade, audible distortion for the
cascaded equipment can increase. There is then some utility for a metric
that somehow commutes and/or associates, so that the audible performance of
the cascaded equipment can be predicted. For example, the Turtle beach Santa
Cruz has audible effects when cascaded 5 times. The DAL Card Deluxe does not
have audible effects when cascaded 5 times.

Mike Rivers

December 6th 03, 02:06 PM

In article > writes:

> It was implicit in Bobby's report of the test (however casual he might
> have said it was) that the perceived differences in results were
> attributable to the change in sampling rate.

That's what he said, but I was able (and willing) to interpret it to
mean that he compared systems that (among other things) differed in
sample rate.

> It was no mere coincidence that he identified the two tests solely by
> a change in sample rate.

Of course not, because that's why he bothered to make the comparison.
And if he was going to make a purchase (or production tool) choice
based on his listening test, I'm sure he would listen to exactly what
he intended to use and make his choice based on that.

--
I'm really Mike Rivers - )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

Mike Rivers

December 6th 03, 02:06 PM

In article > writes:

> In real life we don't compare music at different sample rates, either.

That's true, so why are we even arguing the point. We may compare
Brand X with Brand Y. If we like Brand X better, it happens to use
twice as much disk space, and we can afford it, we use it. If not, we
use what we got and go to work.

> > What I was looking for was to run a signal from a live sound source
> > through a microphone and preamp (obviously those can be held constant)
> > into two sets of back-to-back A/D and D/A converters, one running at
> > each sample rate under test and compare each of the outputs with the
> > source to see if it's possible to say that one sounds more like the
> > original than the other.
>
> Why bother with two sets of converters? Why not just compare the analog feed
> to something that has been round-tripped through back-to-back converters?
> Been there done that.

Because one converter is working at one sample rate and the other
converter is working at another sample rate. I don't know what happens
inside when I flip that sample rate switch. If, at the higher sample
rate, it opens up the bandwidth, that in itself might make a
difference. Or if it changes something in the PLL, that might make a
difference. I don't have enough information to know that I can change
absolutely nothing but sample rate, so the best I can do is compare
performance of two units.

--
I'm really Mike Rivers - )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

Arny Krueger

December 6th 03, 11:55 PM

"Mike Rivers" > wrote in message
news:znr1070713429k@trad
> In article >
> writes:
>
>> In real life we don't compare music at different sample rates,
>> either.
>
> That's true, so why are we even arguing the point. We may compare
> Brand X with Brand Y. If we like Brand X better, it happens to use
> twice as much disk space, and we can afford it, we use it. If not, we
> use what we got and go to work.
>
>>> What I was looking for was to run a signal from a live sound source
>>> through a microphone and preamp (obviously those can be held
>>> constant) into two sets of back-to-back A/D and D/A converters, one
>>> running at each sample rate under test and compare each of the
>>> outputs with the source to see if it's possible to say that one
>>> sounds more like the original than the other.
>>
>> Why bother with two sets of converters? Why not just compare the
>> analog feed to something that has been round-tripped through
>> back-to-back converters? Been there done that.
>
> Because one converter is working at one sample rate and the other
> converter is working at another sample rate. I don't know what happens
> inside when I flip that sample rate switch.

I do. The converter goes away for a while, and comes back running at the new
sample rate. If the converter is good and both sample rates are sufficient,
that's all that you hear happening.

>If, at the higher sample
> rate, it opens up the bandwidth, that in itself might make a
> difference.

My point is that the significance of comparing one converter to another
pales in comparison to comparing a converter to *nothing*. If there was a
big difference when one compares a good converter to *nothing* then
comparing two good converters could be interesting.

>Or if it changes something in the PLL, that might make a
> difference. I don't have enough information to know that I can change
> absolutely nothing but sample rate, so the best I can do is compare
> performance of two units.

I've found that most people find the experience of comparing a good
converter to *nothing* blind, synched, and level-matched to be a bit of a
mind-expanding experience, the first time. Just the first time.

dan lavry

December 7th 03, 06:28 AM

Kurt Albershardt > wrote in message >...
> dan lavry wrote:
> >
> > when going into AD conversion, the
> > WHOLE CONCEPT is to trade bits for bandwidth. That 1 bit at 64fs needs
> > to go through 6 octave sample rate reduction to get to be in that
> > 17-18 bit dynamic range area. Do you think one can take a 1 bit at
> > 2.88Mhz and make it into 16 bits at 1.44MHz? The same is true for the
> > multibit noise shaping stuff. If your goal is say 48KHz bandwidth, and
> > not 96KHz, you have a lot of advatage. A whole octave of decimation
> > advatage.
>
> So what if we say our goal is 24 kHz bandwidth, but using an 88.2k or
> 96k samplerate?

I would say GREAT. I would even go as far as pushing it to 60KHz or so
overcome any objections reagding preshoot Which at 60KHz sampling is
down in the mud, and also to be cooperating with folks that claimed
that some folks react to the presence of up to 26KH.

I doubt it that anyone wants to do 60KHz. But we already had 88.2 and
96KHz and that was a bit high. But if one wanted to make a 44.1KHz CD,
th 88.2KHz was good in one sense: you do a syncromous sample rate
conversion down by 2. That mrans you do a digital process of fitering,
than simple removal of every other sample.

The synchronus conversion (factors of 2,4,8...) are the best and make
less distortions. Asynchronus sample rate conversions (say non integer
ratio such as 192 to 44.1) are much more complex and thus yield lesser
performance. So as long as CD is the main thing out there, 88.2KHz may
be a good choice. If we really needed to have 48, I would be happy
with it - there we meet the 24KHz goal. 96 could be be the high rate
version, and that is pretty wastefull already.

Let me say something outside the science and engineering area, for a
change: I read a lot of the posts and keep thinking in the back of my
mind: I have some FANTASTIC 44.1KHz CD's. It CAN BE DONE GREAT WITH
44.1KHz. My wife has a great ear. Many times she says somerthing like:
I enjoyed the concert (or the acoustic was terrible).. or whatever...
but often she says: "I much rather listen to that music at home on a
CD". This is always compliment to me - we listen to my DA924, but also
a compliment to the recording and mastering engineers. You do not
always get the best seat in the house, but the recording guy "does get
that seat in a sense". The mastering guy can fix things. I listen to
most of my music at home on CD's. Some of it is so great, I can not
imagine it being much better if any...

BR

Dan Lavry

BR
Dan Lavry

Mike Rivers

December 7th 03, 12:20 PM

In article > writes:

> My point is that the significance of comparing one converter to another
> pales in comparison to comparing a converter to *nothing*. If there was a
> big difference when one compares a good converter to *nothing* then
> comparing two good converters could be interesting.

I'm pretty sure I mentioned this not terribly scientific demonstration
before, but I have compared converters operating at 48 and 96 kHz to
the mic preamp output and each one was identifiable every time after a
few stabs to get used to what each path sounded like.

When I suggested that there might be more happening than just the
sample rate changing, I was alluding to differences in how the clock
locks up (in addition to the band-limiting filters, which are pretty
obvious). This is not necessarily the same when you change the basic
clock frequency. A given converter may have more jitter when operating
at one sample rate than the other. This may not fit your definition of
"good converter" but if that's the way it is in the real world, at
least for now we have to accept and deal with it.

--
I'm really Mike Rivers )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

Scott Dorsey

December 7th 03, 03:17 PM

In article <znr1070762390k@trad>, Mike Rivers > wrote:
>
>I'm pretty sure I mentioned this not terribly scientific demonstration
>before, but I have compared converters operating at 48 and 96 kHz to
>the mic preamp output and each one was identifiable every time after a
>few stabs to get used to what each path sounded like.

It's a lot easier to tell that two things are different than to tell which
one is better, though.

>When I suggested that there might be more happening than just the
>sample rate changing, I was alluding to differences in how the clock
>locks up (in addition to the band-limiting filters, which are pretty
>obvious). This is not necessarily the same when you change the basic
>clock frequency. A given converter may have more jitter when operating
>at one sample rate than the other. This may not fit your definition of
>"good converter" but if that's the way it is in the real world, at
>least for now we have to accept and deal with it.

Sadly, yes.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Mike Rivers

December 7th 03, 08:23 PM

In article > writes:

> It's a lot easier to tell that two things are different than to tell which
> one is better, though.

Isn't that what it's all about? If you can't decide which one is
better, it's time to stop worrying about it and use the one you have
until you can decide that there's something not only discernibly
better, but enough better to justify a change.

--
I'm really Mike Rivers )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

Rob Adelman

December 7th 03, 11:01 PM

Mike Rivers wrote:

>
> Isn't that what it's all about? If you can't decide which one is
> better, it's time to stop worrying about it and use the one you have
> until you can decide that there's something not only discernibly
> better, but enough better to justify a change.

Kind of like the NFL challenge. If there isn't conclusive evidence to
overturn the field judge, the play has to stand...

Arny Krueger

December 8th 03, 03:02 AM

"Mike Rivers" > wrote in message
news:znr1070762390k@trad
> In article >
> writes:
>
>> My point is that the significance of comparing one converter to
>> another pales in comparison to comparing a converter to *nothing*.
>> If there was a big difference when one compares a good converter to
>> *nothing* then comparing two good converters could be interesting.
>
> I'm pretty sure I mentioned this not terribly scientific demonstration
> before, but I have compared converters operating at 48 and 96 kHz to
> the mic preamp output and each one was identifiable every time after a
> few stabs to get used to what each path sounded like.

I'm sure you know my thoughts and experiences related to *unscientific*
tests, so I'll not belabor the point.

> When I suggested that there might be more happening than just the
> sample rate changing, I was alluding to differences in how the clock
> locks up (in addition to the band-limiting filters, which are pretty
> obvious).

There's no reason why the band-limiting filters can't be identically the
same for every clock frequency, and they often are. Actually, making them
different for every clock is a more recent *refinement*.

> This is not necessarily the same when you change the basic
> clock frequency. A given converter may have more jitter when operating
> at one sample rate than the other.

We can hypothesize that anything conceivable *may* happen, but whether
there's audible jitter under any given set of conditions is a testable
hypothesis.

In the end it all comes down to what happens when one does blind, synched,
level-matched listening tests comparing a fair selection of modern
converters to the proverbial straight wire. There seems to be a lot more
interest in talking about what *might* happen than actually finding out what
DOES happen. It's not that hard to take a scientific listen at what actually
happens when there's no customers around and the paying work is done for the
moment. I've always be a "cut to the chase" kinda guy.

> This may not fit your definition of
> "good converter" but if that's the way it is in the real world, at
> least for now we have to accept and deal with it.

I think that In the real world what we really want to know what happens when
we compare our *whatsit* to *nothing*.