THD and Noise: What else is there? [Archive]

Don Pearce[_3_]

October 19th 09, 10:23 AM

On Mon, 19 Oct 2009 11:13:50 +0200, Markus Mietling
> wrote:

>(1) I understand that with most cheap gear today, you'd measure THD less
>than .01% and SNR better than 100 dB.
>
>(2) Now you can't hear a difference between .01% and .001% THD, and
>neither a difference between 100 dB and 110 dB SNR.
>
>(3) If this is so, but nevertheless people can hear a difference between
>cheap gear and expensive (= "good") gear, then obviously THD and SNR
>don't tell the whole story.
>
>Therefore I wonder: What else is there that separates the junk from the
>boutique?
>
>-m-

Generally, people can only hear the difference when they know which
they are listening to. There are differences, of course, but not in
the sound. They will be in aspects like the amount of heat sinking,
which determines what happens during extended loud passages. Also
functionality, "feel" of controls etc.

As for basic audible differences - a thing of the past.

Caveat - it is probably still possible to find gear that does sound
inferior, but that will certainly show up in the basic measurements of
noise, distortion and frequency response.

d

Don Pearce[_3_]

October 19th 09, 11:32 AM

On Mon, 19 Oct 2009 06:22:57 -0400, "Soundhaspriority"
> wrote:

>THD is probably not worth measuring.
>More significant arguments can be made about the audibility of IMD -
>intermodulation distortion.

These two are actually manifestations of the same thing -
nonlinearity. In other words you can't have a device that exhibits
one, but not the other of them. So a good THD spec implies a good IMD
spec too.

d

Markus Mietling

October 19th 09, 12:20 PM

(Don Pearce) wrote:

>On Mon, 19 Oct 2009 11:13:50 +0200, Markus Mietling
> wrote:
>
>>(3) If [...] people can hear a difference between
>>cheap gear and expensive (= "good") gear, then obviously THD and SNR
>>don't tell the whole story.
>
>There are differences, of course, but not in
>the sound.

Oops. I would have expected that there are additional physical
dimensions that are relevant to the sound, and not reflected in THD nor
SNR.

But if you are correct, then a high-end A/D converter from Prism or
Weiss doesn't generate any better results than prosumer gear (e.g. RME)
that can be bought for a fraction of the price.

Are you sure about that?

-m-

Don Pearce[_3_]

October 19th 09, 12:23 PM

On Mon, 19 Oct 2009 13:20:40 +0200, Markus Mietling
> wrote:

(Don Pearce) wrote:
>
>>On Mon, 19 Oct 2009 11:13:50 +0200, Markus Mietling
> wrote:
>>
>>>(3) If [...] people can hear a difference between
>>>cheap gear and expensive (= "good") gear, then obviously THD and SNR
>>>don't tell the whole story.
>>
>>There are differences, of course, but not in
>>the sound.
>
>Oops. I would have expected that there are additional physical
>dimensions that are relevant to the sound, and not reflected in THD nor
>SNR.
>
>But if you are correct, then a high-end A/D converter from Prism or
>Weiss doesn't generate any better results than prosumer gear (e.g. RME)
>that can be bought for a fraction of the price.
>
>Are you sure about that?
>
Converters have other stuff going on besides THD and noise, but by and
large yes. A decent mid-range item like something from M-Audio, Echo
etc, has just as good sound quality as anything considerably more
expensive.

d

Arny Krueger

October 19th 09, 12:54 PM

"Markus Mietling" > wrote in
message

> (1) I understand that with most cheap gear today, you'd
> measure THD less than .01% and SNR better than 100 dB.

True of many inexpensive electronic products.

> (2) Now you can't hear a difference between .01% and
> .001% THD, and neither a difference between 100 dB and
> 110 dB SNR.

Generally true.

> (3) If this is so, but nevertheless people can hear a
> difference between cheap gear and expensive (= "good")
> gear, then obviously THD and SNR don't tell the whole
> story.

There are many cases where people have not been able to hear the difference
between good inexpensive equipment and very expensive equipment. One
*secret* is to keep people from knowing the details of the signal path
through which they are listening.

In some cases, such as transformer-coupled preamps, expensive equipment may
have non-flat frequency response and more nonlinear distortion than more
inexpensive, all-electronic alternatives.

> Therefore I wonder: What else is there that separates the
> junk from the boutique?

Who says that all inexpensive equipment is junk?

Who says that only sound quality under ideal conditions is the full measure
of audio equipment?

Do not things like usability and durability matter?

Arny Krueger

October 19th 09, 01:03 PM

"Soundhaspriority" > wrote in message

> I can say that, at
> the recent AES show, I got into a conversation with a
> chip designer, who was manning the booth displaying his
> new preamp chips. And without the slightest hesitation,
> he told me that preamps using power devices would always
> have the potential for better measurements compared to
> monolithic, because thermal modulation is less of an
> effect with large geometry devices.

This is not a reasonable claim as related to commercial products. It may be
true in a lab setting, but that matters little to us practitioners.

The gain of most audio gear is still established by external discrete
resistors that are off-chip and therefore not susceptible to thermal
modulation to anything like the degree that a similar on-chip part might be.
As long as reasonably good op amps, whether discrete or monolithic are used
with typical amounts of inverse feedback, the off-chip parts set the overall
linearity of the equipment.

Since monolithic audio chips are capable of power levels in excess of 100
watts into speaker impedance loads, current levels of many amps, and
voltages of well over 60- volts, there is no reason why dynamic range would
be an issue. The use of discrete power transistors in anything but high
powered amplifiers and power supplies is more fashion and image than
advanced technology.

Besides with SMT technology, even sub-$100 stereo receivers can be built
with discrete output stages.

William Sommerwerck

October 19th 09, 01:15 PM

Herr Mietling, you need to listen before you buy anything. Carefully, with a
variety of material, over a period of days.

Ignore the specs, ignore the manufacturer's name, ignore the salesman
(especially the salesman). Don't be rushed, take your time, and listen. Make
sure the store will let you return the amplifier if you don't like it.

At the time I bought them, the brand P amplifiers retailed for $2000.
Despite what Arny might say, that's not a lot of money for amplifier that
puts out 250W continuous average power * per channel. (This amp is Class A
up to about 8 watts, which covers most of your listening, especially when
bi-amped.) One reviewer said "This isn't a good $2000 amplifier. It's a good
amplifier."

There are other good, not-horribly-expensive amps, such as brand A.

* This is the correct term. RMS power is not correct.

William Sommerwerck

October 19th 09, 01:15 PM

> Generally, people can only hear the difference when they know
> which they are listening to. There are differences, of course, but
> not in the sound. They will be in aspects like the amount of heat
> sinking, which determines what happens during extended loud
> passages. Also functionality, "feel" of controls etc.
> As for basic audible differences -- a thing of the past.

I wish this were true -- but it isn't. Not by a long shot.

I used to own brand K amplifiers, which were sold for reasons I don't want
to discuss. Their sound had a lot depth and space, and the kind of
"slammin'" bass brand K is known for.

I auditioned brand C switching amplifiers, hoping they'd be terrific. (Some
switching amps have gotten rave reviews, with comments such as "They don't
sound like tube or transistor. They sound like nothing at all.") They were
horrible-sounding, so God-awful they were unlistenable. Coarse, grainy,
flat. A man who used to be a company C honcho, and whom I worked for for two
years, told me that if he'd known I was going to buy these, he would have
stopped me. He said that if he'd still worked at company C, they never would
have gotten into production.

I finally wound up with brand P amplifiers, designed by a friend. These
sound the polar opposite of Brand K. They are more "forward"-sounding, and
have less of a sense of depth and space, and are a bit drier and grainier
sounding. The bass is good, but not "slammin'". Overall, I prefer these to
Brand K, as they have a better sense of liveness and presence, and even
subjectively more detail -- but I still miss brand K's sense of depth.

It seems that most "good" amplifiers fall in one of these two classes --
deep and spacious, or flattish and "present". I'm not the only person to
have observed this.

These three amplifiers were designed by three of the best-known and
best-respected designers in the business. They sound different -- very
different. They are so different that, given recordings I am familiar with,
I can tell which is which, in a few seconds, by listening to a single disk.
Blind or double-blind tests, as you prefer.

> Caveat -- it is probably still possible to find gear that does sound
> inferior, but that will certainly show up in the basic measurements
> of noise, distortion and frequency response.

Simply untrue. All these amplifiers have excellent specs.

According to my designer friend, advances are being made in reducing what I
call "self-modulation", and he calls "phase modulation". I've been arguing
for over 20 years about the need to look for distortion caused by the
variation in

Don Pearce[_3_]

October 19th 09, 01:31 PM

On Mon, 19 Oct 2009 05:15:27 -0700, "William Sommerwerck"
> wrote:

>> Generally, people can only hear the difference when they know
>> which they are listening to. There are differences, of course, but
>> not in the sound. They will be in aspects like the amount of heat
>> sinking, which determines what happens during extended loud
>> passages. Also functionality, "feel" of controls etc.
>> As for basic audible differences -- a thing of the past.
>
>I wish this were true -- but it isn't. Not by a long shot.
>
>I used to own brand K amplifiers, which were sold for reasons I don't want
>to discuss. Their sound had a lot depth and space, and the kind of
>"slammin'" bass brand K is known for.
>
>I auditioned brand C switching amplifiers, hoping they'd be terrific. (Some
>switching amps have gotten rave reviews, with comments such as "They don't
>sound like tube or transistor. They sound like nothing at all.") They were
>horrible-sounding, so God-awful they were unlistenable. Coarse, grainy,
>flat. A man who used to be a company C honcho, and whom I worked for for two
>years, told me that if he'd known I was going to buy these, he would have
>stopped me. He said that if he'd still worked at company C, they never would
>have gotten into production.
>
>I finally wound up with brand P amplifiers, designed by a friend. These
>sound the polar opposite of Brand K. They are more "forward"-sounding, and
>have less of a sense of depth and space, and are a bit drier and grainier
>sounding. The bass is good, but not "slammin'". Overall, I prefer these to
>Brand K, as they have a better sense of liveness and presence, and even
>subjectively more detail -- but I still miss brand K's sense of depth.
>
>It seems that most "good" amplifiers fall in one of these two classes --
>deep and spacious, or flattish and "present". I'm not the only person to
>have observed this.
>
>These three amplifiers were designed by three of the best-known and
>best-respected designers in the business. They sound different -- very
>different. They are so different that, given recordings I am familiar with,
>I can tell which is which, in a few seconds, by listening to a single disk.
>Blind or double-blind tests, as you prefer.
>
>
>> Caveat -- it is probably still possible to find gear that does sound
>> inferior, but that will certainly show up in the basic measurements
>> of noise, distortion and frequency response.
>
>Simply untrue. All these amplifiers have excellent specs.
>
>According to my designer friend, advances are being made in reducing what I
>call "self-modulation", and he calls "phase modulation". I've been arguing
>for over 20 years about the need to look for distortion caused by the
>variation in
>

It is very easy to tell you once reviewed for Stereophile, I'm afraid.

d

William Sommerwerck

October 19th 09, 01:37 PM

> It is very easy to tell you once reviewed for Stereophile, I'm afraid.

I'm not embarrassed about it. Nor am I embarrassed to say that I badly miss
the late J Gordon Holt. He's one of the few really careful thinkers I've
ever known.

When was the last time you sat down with a bunch of amps and auditioned
them?

I might add that the brand K amplifiers cost more than three times what the
brand P amps cost -- yet I preferred the latter.

Arny Krueger

October 19th 09, 02:23 PM

"William Sommerwerck" > wrote in
message
>> It is very easy to tell you once reviewed for
>> Stereophile, I'm afraid.
>
> I'm not embarrassed about it. Nor am I embarrassed to say
> that I badly miss the late J Gordon Holt. He's one of the
> few really careful thinkers I've ever known.

The problem here is not that Holt thought badly - heck he even agreed with
the concept of ABX in his way. He was clearly intellectually and ethically
head and shoulders above most of the current rabble that have run his
magazine into the ground, credibility-wise.

The problem is one of criterial biasing - William you obviously equate
*careful* with *understandable by you*. Holt was good, but not *that* good.

William Sommerwerck

October 19th 09, 02:56 PM

>>> It is very easy to tell you once reviewed for
>>> Stereophile, I'm afraid.

>> I'm not embarrassed about it. Nor am I embarrassed to
>> say that I badly miss the late J Gordon Holt. He's one
>> of the few really careful thinkers I've ever known.

> The problem here is not that Holt thought badly - heck he even
> agreed with the concept of ABX in his way. He was clearly
> intellectually and ethically head and shoulders above most
> of the current rabble that have run his magazine into the ground,
> credibility-wise.

Since when was "Stereophile" ever credible in your sight, Arny?

> The problem is one of criterial biasing -- William, you obviously equate
> *careful* with *understandable by you*. Holt was good, but not *that*
good.

I DO NOT make such an equation, Arny.

About a decade ago I worked in the COM Apps department at Microsoft,
cleaning up their documentation. One of the people there was a German
computer scientist of great intelligence.

I spent some time with him trying to understand his theories of software
design, and related issues. I simply could not understand them. That is,
they never fell together in a coherent way in my mind, which is what I mean
by "understanding". Okay?

Could it have been that he was wrong? Maybe. Was it possible that he wasn't
explaining them well? Probably. Was it that I was too stupid to understand
them? Not likely.

Regardless, I did not interpret my failure to understand him as proving --
or even suggesting -- that he was wrong. I simply accepted the fact that
there was a failure to communicate, and let it go at that.

[The following is for Arny alone. Others may ignore it. You and Don Pearce
are confusing time quantization with amplitude quantization. This is a
common point of confusion, and at the risk of sounding unduly rude, this
point should have cleared up in your mind at least 20 years ago. I remember
a Sharp (?) ad explaining digital, which made this same error. When I
contacted Sharp, the guy thought about it a minute or so, and then went "Oh,
no". The ad was later rewritten when it was used as a POP flyer.

Here's another point. Time quantization (sampling) does not cause
information loss. Amplitude quantization does. How does filtering restore
it? If Dr. Shannon were playing Hemo to your Dr. Science, how would you
answer him?]

Anahata

October 19th 09, 03:05 PM

On Mon, 19 Oct 2009 05:15:15 -0700, William Sommerwerck wrote:

> continuous average power * per channel.
> [...]
> * This is the correct term. RMS power is not correct.

[with apologies for thread drift]
You are right about "RMS power" being incorrect, but while your
replacement term at least means something, it can still be abused.

I think the best description of what's almost invariably *meant* by
either of those terms is average sine wave power. Without specifying the
waveform it's ambiguous.

For example, an amplifier capable of 100W undistorted sine wave output
can usually do nearly 200W "continuous average power" with a square wave,
assuming supply rail voltage is the limiting factor.

[now someone will nitpick about specifying load impedance too...]

--
Anahata
==//== 01638 720444
http://www.treewind.co.uk ==//== http://www.myspace.com/maryanahata

Arny Krueger

October 19th 09, 03:13 PM

"William Sommerwerck" > wrote in
message
>>>> It is very easy to tell you once reviewed for
>>>> Stereophile, I'm afraid.
>
>>> I'm not embarrassed about it. Nor am I embarrassed to
>>> say that I badly miss the late J Gordon Holt. He's one
>>> of the few really careful thinkers I've ever known.
>
>> The problem here is not that Holt thought badly - heck
>> he even
>> agreed with the concept of ABX in his way. He was clearly
>> intellectually and ethically head and shoulders above
>> most
>> of the current rabble that have run his magazine into
>> the ground, credibility-wise.
>
> Since when was "Stereophile" ever credible in your sight,
> Arny?

For example, when I was a charter subscriber. That "1 year" subscription was
one of the longest-lasting subscriptions I ever had! ;-)

As far as piercing the veil of high end audio, that didn't happen until
after I had been swilling TAS's cool aid for seveal years. I had to invent
ABX in order to disabuse myself of that sort of thing. So, we're talking
from like 1962 to maybe the mid-late 1970s.

>> The problem is one of criterial biasing -- William, you
>> obviously equate *careful* with *understandable by you*.
>> Holt was good, but not *that* good.

> I DO NOT make such an equation, Arny.

Actually, all of us who are insightful know we do it all the time. ;-)

> About a decade ago I worked in the COM Apps department at
> Microsoft, cleaning up their documentation. One of the
> people there was a German computer scientist of great
> intelligence.

> I spent some time with him trying to understand his
> theories of software design, and related issues. I simply
> could not understand them. That is, they never fell
> together in a coherent way in my mind, which is what I
> mean by "understanding". Okay?

Whatever.

> Could it have been that he was wrong? Maybe. Was it
> possible that he wasn't explaining them well? Probably.
> Was it that I was too stupid to understand them? Not
> likely.

I never said that JGH was stupid or intellectually lazy.

His biggest problem was likely that like many of us, he was a victim of his
own success.

> Here's another point. Time quantization (sampling) does
> not cause information loss. Amplitude quantization does.

Wrong. They both inherently cause information loss. Sampling has no
generalized ability to accurately gather data above Nyquist.

However this is arguing over nits. Information loss is only a serious
problem when the receiver can reliably respond to the lost information and
the source is actually capable of reliably sending the lost information.

If the receiver is the human ear, then it is often the weakest link both the
amplitude and frequency domains. If the transmitter is the physical world,
then there are usually serious problems in the amplitude domain due to the
fact that non-information-bearing noise is endemic.

Wecan do it

October 19th 09, 03:19 PM

William:

I believe you mentioned the desire for some work/income in a
previous post. Since you have special skills, specifically
hearing the difference between different brand/models of
competent amplifiers you may be interested in picking up some
easy money.

There is a guy looking to hire someone with your skills. As a
matter of fact he will pay you ten large ($10,000.00) if you
pass his interview. This will not be an easy interview. He has
been looking for nine years and no one has been able to hear
these differences you can hear, so he will be really excited
to meet you.

Here is a link to his web site.
http://www.talkaudio.co.uk/vbb/archive/index.php/t-18815.html

Peace
dawg

"Don Pearce" > wrote in message
...
> On Mon, 19 Oct 2009 05:15:27 -0700, "William Sommerwerck"
> > wrote:
>
>>> Generally, people can only hear the difference when they
>>> know
>>> which they are listening to. There are differences, of
>>> course, but
>>> not in the sound. They will be in aspects like the amount
>>> of heat
>>> sinking, which determines what happens during extended
>>> loud
>>> passages. Also functionality, "feel" of controls etc.
>>> As for basic audible differences -- a thing of the past.
>>
>>I wish this were true -- but it isn't. Not by a long shot.
>>
>>I used to own brand K amplifiers, which were sold for
>>reasons I don't want
>>to discuss. Their sound had a lot depth and space, and the
>>kind of
>>"slammin'" bass brand K is known for.
>>
>>I auditioned brand C switching amplifiers, hoping they'd be
>>terrific. (Some
>>switching amps have gotten rave reviews, with comments such
>>as "They don't
>>sound like tube or transistor. They sound like nothing at
>>all.") They were
>>horrible-sounding, so God-awful they were unlistenable.
>>Coarse, grainy,
>>flat. A man who used to be a company C honcho, and whom I
>>worked for for two
>>years, told me that if he'd known I was going to buy these,
>>he would have
>>stopped me. He said that if he'd still worked at company C,
>>they never would
>>have gotten into production.
>>
>>I finally wound up with brand P amplifiers, designed by a
>>friend. These
>>sound the polar opposite of Brand K. They are more
>>"forward"-sounding, and
>>have less of a sense of depth and space, and are a bit drier
>>and grainier
>>sounding. The bass is good, but not "slammin'". Overall, I
>>prefer these to
>>Brand K, as they have a better sense of liveness and
>>presence, and even
>>subjectively more detail -- but I still miss brand K's sense
>>of depth.
>>
>>It seems that most "good" amplifiers fall in one of these
>>two classes --
>>deep and spacious, or flattish and "present". I'm not the
>>only person to
>>have observed this.
>>
>>These three amplifiers were designed by three of the
>>best-known and
>>best-respected designers in the business. They sound
>>different -- very
>>different. They are so different that, given recordings I am
>>familiar with,
>>I can tell which is which, in a few seconds, by listening to
>>a single disk.
>>Blind or double-blind tests, as you prefer.
>>
>>
>>> Caveat -- it is probably still possible to find gear that
>>> does sound
>>> inferior, but that will certainly show up in the basic
>>> measurements
>>> of noise, distortion and frequency response.
>>
>>Simply untrue. All these amplifiers have excellent specs.
>>
>>According to my designer friend, advances are being made in
>>reducing what I
>>call "self-modulation", and he calls "phase modulation".
>>I've been arguing
>>for over 20 years about the need to look for distortion
>>caused by the
>>variation in
>>
>
> It is very easy to tell you once reviewed for Stereophile,
> I'm afraid.
>
> d

William Sommerwerck

October 19th 09, 03:21 PM

>> Here's another point. Time quantization (sampling) does
>> not cause information loss. Amplitude quantization does.

> Wrong. They both inherently cause information loss.

Hello? Hello? Sampling at or above the Nyquist rate is lossless. The fact
that one might need to filter out (presumably unnecessary) information --
such as ultrasound -- does not make the sampling itself lossy.

William Sommerwerck

October 19th 09, 03:22 PM

>> continuous average power * per channel.
>> * This is the correct term. RMS power is not correct.

> [with apologies for thread drift]
> You are right about "RMS power" being incorrect, but while your
> replacement term at least means something, it can still be abused.
> I think the best description of what's almost invariably *meant* by
> either of those terms is average sine wave power. Without specifying
> the waveform it's ambiguous.

Agreed. It's assumed we're talking about sine wave, but one could just as
well use a square wave.

> For example, an amplifier capable of 100W undistorted sine wave output
> can usually do nearly 200W "continuous average power" with a square
wave...

For how long? <grin>

Arny Krueger

October 19th 09, 03:32 PM

"William Sommerwerck" > wrote in
message
>>> Here's another point. Time quantization (sampling) does
>>> not cause information loss. Amplitude quantization does.

>> Wrong. They both inherently cause information loss.

> Hello? Hello? Sampling at or above the Nyquist rate is
> lossless. The fact that one might need to filter out
> (presumably unnecessary) information -- such as
> ultrasound -- does not make the sampling itself lossy.

You're going to have to explain this business of "sampling at or above the
Nyquist rate".

Since the Nyquist frequency is *always* half of the sample rate, it can
never be even equal the sample rate, let alone be above the sample rate.

You've also changed the issue being discussed.

I'm sticking to my guns - quantization in both the time and amplitude domain
inherently cause data loss if there is data to be lost. An example of a
situation where there is no data to be lost would be an analog signal that
has previously been brick-wall filtered at a lower frequency than the
current Nyquist, which is not impossible or even all that uncommon these
days.

William Sommerwerck

October 19th 09, 03:39 PM

> I believe you mentioned the desire for some work/income in
> a previous post. Since you have special skills, specifically
> hearing the difference between different brand/models of
> competent amplifiers you may be interested in picking up
> some easy money.

> There is a guy looking to hire someone with your skills. As
> a matter of fact, he will pay you ten large ($10,000) if you
> pass his interview. This will not be an easy interview. He has
> been looking for nine years and no one has been able to hear
> these differences you can hear, so he will be really excited
> to meet you.

> Here is a link to his web site.
> http://www.talkaudio.co.uk/vbb/archive/index.php/t-18815.html

I know you're not in the least sincere about this, but I appreciate your
referrring me to this page.

It looked very interesting -- until I got to the point about testing car
amplifiers! The following requires were also "turn-offs":

12. Although anyone is welcome to take the test, only subjects employed in
the car audio industry or Car Sound subscribers are eligible for the $10,000
prize. [That leaves me out.]

13. Cost to take the test is $100. $300 for people representing companies.
Payable in advance, scheduled appointments only. Done correctly the test
takes several hours and I don't have the time if you aren't serious.

I'd be happy to do such testing, but it would have to be in my home, with my
choice of amplifiers. I would also insist that those running the test also
do controlled and uncontrolled listening.

There's also a fundamental error in the structure of the test -- it's
limited to controlled double-blind testing. The testing should also included
sighted and unsighted uncontrolled listening. Why? To see what happens.

I've decided to make the following quote my signature. If you don't
understand it, you don't understand the process of scientific discovery.

William Sommerwerck

October 19th 09, 03:46 PM

>> Hello? Hello? Sampling at or above the Nyquist rate is
>> lossless. The fact that one might need to filter out
>> (presumably unnecessary) information -- such as
>> ultrasound -- does not make the sampling itself lossy.

> You're going to have to explain this business of "sampling
> at or above the Nyquist rate".

Explain what?

> Since the Nyquist frequency is *always* half of the sample rate, it can
> never be even equal the sample rate, let alone be above the sample rate.

Read the following from Wikipedia, which "gets the point":
"The Nyquist–Shannon sampling theorem is a fundamental result in the field
of information theory, in particular telecommunications and signal
processing. Sampling is the process of converting a signal (for example, a
function of continuous time or space) into a numeric sequence (a function of
discrete time or space). The theorem states:[1]

If a function x(t) contains no frequencies higher than B hertz, it is
completely determined by giving its ordinates at a series of points spaced
1/(2B) seconds apart.

In essence, the theorem shows that a bandlimited analog signal that has been
sampled can be perfectly reconstructed from an infinite sequence of samples
if the sampling rate exceeds 2B samples per second, where B is the highest
frequency in the original signal. If a signal contains a component at
exactly B hertz, then samples spaced at exactly 1/(2B) seconds do not
completely determine the signal, Shannon's statement notwithstanding."

If you sampled at exactly twice the highest frequency in the signal, then
that frequency would be aliased as a DC component in the samples. The proof
of this is absolutely trivial.

> I'm sticking to my guns - quantization in both the time and amplitude
domain
> inherently cause data loss if there is data to be lost. An example of a
> situation where there is no data to be lost would be an analog signal that
> has previously been brick-wall filtered at a lower frequency than the
> current Nyquist, which is not impossible or even all that uncommon these
> days.

I think you're confusing band-limiting -- which is, of course, a necessary
prerequisite of sampling -- with the sampling itself. The sampling does not,
per se, cause the loss of information.

William Sommerwerck

October 19th 09, 03:48 PM

Whoops...

William Sommerwerck

October 19th 09, 03:50 PM

Double whoops. It appears that Outlook doesn't consider a UseNet post a
message.

"We already know the answers -- we just haven't asked the right
questions." -- Dr. Edwin H. Land

Arny Krueger

October 19th 09, 04:13 PM

"William Sommerwerck" > wrote in
message
>>> Hello? Hello? Sampling at or above the Nyquist rate is
>>> lossless. The fact that one might need to filter out
>>> (presumably unnecessary) information -- such as
>>> ultrasound -- does not make the sampling itself lossy.
>
>> You're going to have to explain this business of
>> "sampling
>> at or above the Nyquist rate".
>
> Explain what?
>
>
>> Since the Nyquist frequency is *always* half of the
>> sample rate, it can never be even equal the sample rate,
>> let alone be above the sample rate.
>
> Read the following from Wikipedia, which "gets the point":
> "The Nyquist–Shannon sampling theorem is a fundamental
> result in the field of information theory, in particular
> telecommunications and signal processing. Sampling is the
> process of converting a signal (for example, a function
> of continuous time or space) into a numeric sequence (a
> function of discrete time or space). The theorem
> states:[1]
>
> If a function x(t) contains no frequencies higher than B
> hertz, it is completely determined by giving its
> ordinates at a series of points spaced 1/(2B) seconds
> apart.
>
> In essence, the theorem shows that a bandlimited analog
> signal that has been sampled can be perfectly
> reconstructed from an infinite sequence of samples if the
> sampling rate exceeds 2B samples per second, where B is
> the highest frequency in the original signal. If a signal
> contains a component at exactly B hertz, then samples
> spaced at exactly 1/(2B) seconds do not completely
> determine the signal, Shannon's statement
> notwithstanding."

> If you sampled at exactly twice the highest frequency in
> the signal, then that frequency would be aliased as a DC
> component in the samples. The proof of this is absolutely
> trivial.

And this applies to the discussion at hand, how?

>> I'm sticking to my guns - quantization in both the time
>> and amplitude domain inherently cause data loss if there
>> is data to be lost. An example of a situation where
>> there is no data to be lost would be an analog signal
>> that has previously been brick-wall filtered at a lower
>> frequency than the current Nyquist, which is not
>> impossible or even all that uncommon these days.

> I think you're confusing band-limiting -- which is, of
> course, a necessary prerequisite of sampling -- with the
> sampling itself. The sampling does not, per se, cause the
> loss of information.

Of course sampling causes data loss. Input data at frequencies at or above
the Nyquist frequency cannot be reliably coded. It's lost data, pure and
simple.

Arny Krueger

October 19th 09, 04:23 PM

"William Sommerwerck" > wrote in
message
>> I believe you mentioned the desire for some work/income
>> in
>> a previous post. Since you have special skills,
>> specifically hearing the difference between different
>> brand/models of competent amplifiers you may be
>> interested in picking up
>> some easy money.
>
>> There is a guy looking to hire someone with your skills.
>> As
>> a matter of fact, he will pay you ten large ($10,000) if
>> you pass his interview. This will not be an easy
>> interview. He has been looking for nine years and no one
>> has been able to hear these differences you can hear, so
>> he will be really excited to meet you.
>
>> Here is a link to his web site.
>> http://www.talkaudio.co.uk/vbb/archive/index.php/t-18815.html
>
> I know you're not in the least sincere about this, but I
> appreciate your referring me to this page.
>
> It looked very interesting -- until I got to the point
> about testing car amplifiers! The following requires were
> also "turn-offs":
>
> 12. Although anyone is welcome to take the test, only
> subjects employed in the car audio industry or Car Sound
> subscribers are eligible for the $10,000 prize. [That
> leaves me out.]

Car Sound magazine subscriptions are readily and economically available:

http://www.carsound.com/news/archives/11_11_99_1.shtml

> 13. Cost to take the test is $100. $300 for people
> representing companies. Payable in advance, scheduled
> appointments only. Done correctly the test takes several
> hours and I don't have the time if you aren't serious.

Seems reasonable enough. I think he'll be doing ABX testing or something
like it. Been there, done that and that's what it takes.

> I'd be happy to do such testing, but it would have to be
> in my home, with my choice of amplifiers. I would also
> insist that those running the test also do controlled and
> uncontrolled listening.

I don't know if it is clear from the page referenced, but this would be a
DBT.

> There's also a fundamental error in the structure of the
> test -- it's limited to controlled double-blind testing.

How else do you propose to avoid the even more severe and fundamental
problems related to sighted evaluations?

> I've decided to make the following quote my signature. If
> you don't understand it, you don't understand the process
> of scientific discovery.

I read your comments as you declining to take the Richard Clark challenge
as presented. Join the other millions of other no-shows...

Scott Dorsey

October 19th 09, 04:40 PM

Markus Mietling > wrote:
>(1) I understand that with most cheap gear today, you'd measure THD less
>than .01% and SNR better than 100 dB.

Right.

>(2) Now you can't hear a difference between .01% and .001% THD, and
>neither a difference between 100 dB and 110 dB SNR.

Sure you can. In fact, the .001% THD might sound worse than the .01% THD,
because the spectrum might be different.

I bet you can't hear less than maybe 1% third harmonic distortion. I
bet you can hear .001% sixth harmonic distortion.

>(3) If this is so, but nevertheless people can hear a difference between
>cheap gear and expensive (= "good") gear, then obviously THD and SNR
>don't tell the whole story.

No, THD is a useless number. It's still useful when comparing fairly
high distortion equipment with similar design topologies. Unfortunately
it doesn't take into account the distortion spectrum, and some harmonics
are far more audible than others.

Also, that SNR number is usually doubtful too. You'll see consoles rated
with very high SNR numbers given a perfect gain structure, but when you
actually use them and are forced to adjust the gains to deal with real world
signals, they sometimes aren't so quiet.

>Therefore I wonder: What else is there that separates the junk from the
>boutique?

There are lots of useful measurements out there for distortion, stuff like
distortion spectra and TIMD measurements. Vendors won't put that stuff on
the datasheet, though.

There are also some much more useful ways of measuring SNR too, but vendors
won't put that stuff on the datasheet either. An actual noise spectrum
can be very enlightening.

In the case of transducers like microphones and loudspeakers, the measured
response plots don't ever look even vaguely like the smoothed plots that
the vendors put on the data sheet. And the vendors won't tell you that they
usually get a lot worse once you get off axis (and when area miking, everything
is off-axis).

So there are plenty of ways to measure and characterize the way things sound
objectively, and designers often use them, but the marketing people won't
tell you anything about them.
--scott

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

Scott Dorsey

October 19th 09, 04:43 PM

William Sommerwerck > wrote:
>
>I finally wound up with brand P amplifiers, designed by a friend. These
>sound the polar opposite of Brand K. They are more "forward"-sounding, and
>have less of a sense of depth and space, and are a bit drier and grainier
>sounding. The bass is good, but not "slammin'". Overall, I prefer these to
>Brand K, as they have a better sense of liveness and presence, and even
>subjectively more detail -- but I still miss brand K's sense of depth.
>
>It seems that most "good" amplifiers fall in one of these two classes --
>deep and spacious, or flattish and "present". I'm not the only person to
>have observed this.
>
>These three amplifiers were designed by three of the best-known and
>best-respected designers in the business. They sound different -- very
>different. They are so different that, given recordings I am familiar with,
>I can tell which is which, in a few seconds, by listening to a single disk.
>Blind or double-blind tests, as you prefer.

I bet a nickel that if you sent me data files of a 1KHz square wave going
through each with the same (real speaker) load, that I could tell you which
was which, too.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Ian Bell[_2_]

October 19th 09, 04:46 PM

Markus Mietling wrote:
> (1) I understand that with most cheap gear today, you'd measure THD less
> than .01% and SNR better than 100 dB.
>
> (2) Now you can't hear a difference between .01% and .001% THD, and
> neither a difference between 100 dB and 110 dB SNR.
>
> (3) If this is so, but nevertheless people can hear a difference between
> cheap gear and expensive (= "good") gear, then obviously THD and SNR
> don't tell the whole story.
>
> Therefore I wonder: What else is there that separates the junk from the
> boutique?
>
> -m-

THD+N measurements are made using a single sine wave input signal.

Music is not a single sine wave, it is a complex waveform. If you were
to test with a similarly complex waveform you would be able to measure
the differences.

Cheers

Ian

Adrian Tuddenham[_2_]

October 19th 09, 05:36 PM

William Sommerwerck > wrote:

> >> continuous average power * per channel.
> >> * This is the correct term. RMS power is not correct.
>
> > [with apologies for thread drift]
> > You are right about "RMS power" being incorrect, but while your
> > replacement term at least means something, it can still be abused.
> > I think the best description of what's almost invariably *meant* by
> > either of those terms is average sine wave power. Without specifying
> > the waveform it's ambiguous.
>
> Agreed. It's assumed we're talking about sine wave, but one could just as
> well use a square wave.
>
>
> > For example, an amplifier capable of 100W undistorted sine wave output
> > can usually do nearly 200W "continuous average power" with a square
> wave...
>
> For how long? <grin>

Probably for at least as long, if not longer. The dissipation of the
output devices is lower on a rail-to-rail squarewave than it is on a
rail-to-rail sinewave. If there is a weak point elsewhere (e.g.
under-rated emitter resistors) that ought to have been shown up by the
manufacturers' own tests and is very cheap to address at the design
stage.

This is where reviewers could help by running the amplifier into a dummy
load at full power sinewave and, if it blows up, saying so in their
reviews. That would soon sort out the false advertising claims.

--
~ Adrian Tuddenham ~
(Remove the ".invalid"s and add ".co.uk" to reply)
www.poppyrecords.co.uk

William Sommerwerck

October 19th 09, 05:36 PM

>> There's also a fundamental error in the structure of the
>> test -- it's limited to controlled double-blind testing.

> How else do you propose to avoid the even more severe and
> fundamental problems related to sighted evaluations?

Simply to see what happens.

> I read your comments as you declining to take the Richard Clark
> challenge as presented. Join the other millions of other no-shows...

Absolutely. I'm not interested in testing unfamiliar car amplifiers. I would
LIKE to be challenged to demonstrate that I CAN hear differences among
amplifiers that I believe had strong differences in sound quality. What
would be the point of the testing otherwise?

William Sommerwerck

October 19th 09, 05:38 PM

> I bet a nickel that if you sent me data files of a 1KHz square
> wave going through each with the same (real speaker) load,
> that I could tell you which was which, too.

Unfortunately, I have only the brand P amplifiers at this time.

William Sommerwerck

October 19th 09, 05:46 PM

>>> For example, an amplifier capable of 100W undistorted sine wave
>>> output can usually do nearly 200W "continuous average power"
>>> with a square wave...

>> For how long? <grin>

> Probably for at least as long, if not longer. The dissipation of the
> output devices is lower on a rail-to-rail squarewave than it is on
> a rail-to-rail sinewave. If there is a weak point elsewhere (e.g.
> under-rated emitter resistors) that ought to have been shown up
> by the manufacturers' own tests and is very cheap to address
> at the design stage.
> This is where reviewers could help by running the amplifier into a
> dummy load at full power sinewave and, if it blows up, saying so
> in their reviews. That would soon sort out the false advertising claims.

The testing procedures forced by the FTC on the audio industry about 30
years ago included a pre-conditioning warmup at 1/3 full power.
Unfortunately, this is close to the point where a class B (or near-class B)
amplifier dissipates maximum power. It's only about 5dB from full output, so
unless you're listening to highly compressed material at high volume levels,
it doesn't reflect "real-life" usage.

Wecan do it

October 19th 09, 05:54 PM

"William Sommerwerck" > wrote in
message ...
>>> There's also a fundamental error in the structure of the
>>> test -- it's limited to controlled double-blind testing.
>
>> How else do you propose to avoid the even more severe and
>> fundamental problems related to sighted evaluations?
>
> Simply to see what happens.
>
>
>> I read your comments as you declining to take the Richard
>> Clark
>> challenge as presented. Join the other millions of other
>> no-shows...
>
> Absolutely. I'm not interested in testing unfamiliar car
> amplifiers. I would
> LIKE to be challenged to demonstrate that I CAN hear
> differences among
> amplifiers that I believe had strong differences in sound
> quality. What
> would be the point of the testing otherwise?

I recall some of the early takers were testing amps like a
dyanco st70 against a Pass aleph(x) and could not tell the
difference.

If you were really serious I bet you a nickle that Richard
would have no problem betting your hundred bux to his ten
thousand that you could not tell the difference between any
old car amp he has laying around and and your Krell or Pass
amps.

If you want to do his test and agree to his conditions I can
ask him for you. Surely this is childs play for someone with a
critical ear like yours. No?

Objecting shows me and everyone you are attempting to persuade
here that you are unwilling to bet a hundred to one that what
you say is not BULL**** from your deluded mind. Deep down you
know you are sick dont you?

Peace
dawg

Adrian Tuddenham[_2_]

October 19th 09, 06:03 PM

Markus Mietling > wrote:

> (1) I understand that with most cheap gear today, you'd measure THD less
> than .01% and SNR better than 100 dB.
>
> (2) Now you can't hear a difference between .01% and .001% THD, and
> neither a difference between 100 dB and 110 dB SNR.
>
> (3) If this is so, but nevertheless people can hear a difference between
> cheap gear and expensive (= "good") gear, then obviously THD and SNR
> don't tell the whole story.
>
> Therefore I wonder: What else is there that separates the junk from the
> boutique?

Apply pulses of sinewave and look with an oscilloscope to see if there
is a false LF component appearing at the beginning and end of the pulse.
This is a technique which gets resurrected from time to time, but
usually only after suspicions cannot be allayed by other methods.

It wouldn't be picked up by any of the above tests, but it can produce a
thud in a loudspeaker system with a really good bass response. It can
also cause momentary distortion if the signal is large and near to
clipping or if its cause also produces intermodulation distortion.

There are two main causes of this effect:

1) LF coupling into the signal circuit from unstabilised power supply
rails, which 'duck' on load. That doesn't usually cause intermodulation
unless it is very bad.

2) Capacitor coupling into base-emitter impedances (or similar
diode-like devices), which exhibit signal-dependent loading and
partially rectify the signal. This can cause momentary
intermodulation and harmonic distortion, but it will not show up with
conventional steady-state measurement techniques.

Both of the above could be reduced by an outer feedback loop, provided
it is effective down to DC; not all of them are.

Another important point is how an amplifier behaves when it is pushed
beyond its normal limits. Some degrade gracefully (some valve designs
were excellent in this respect), others behave disgracefully and take
ages to recover. It may not happen very often, but given the random
nature of real sounds, it is almost certain to happen from time to time;
when it does, that is when you will hear big differences between one
design and another.

--
~ Adrian Tuddenham ~
(Remove the ".invalid"s and add ".co.uk" to reply)
www.poppyrecords.co.uk

William Sommerwerck

October 19th 09, 06:15 PM

> I recall some of the early takers were testing amps like
> a Dyanco ST-70 against a Pass aleph(x) and could not
> tell the difference.

Then "something is wrong". Do you really believe that both amps are
"perfect", or that they have essentially the same sets of errors &
colorations? It's not likely.

> If you were really serious I bet you a nickel that Richard
> would have no problem betting your hundred bux to his ten
> thousand that you could not tell the difference between any
> old car amp he has laying around and and your Krell or Pass
> amps.

Do you believe that any randomly selected car amplifier is indistinguishable
from a Krell or Pass amplifier? (By the way, I don't own Pass Labs
amplifiers.) I don't. Nor do I believe that Krell and Pass amplifiers are
necessarily indistinguishable from each other.

> If you want to do his test and agree to his conditions I can
> ask him for you. Surely this is child's play for someone with
> a critical ear like yours. No?

There is no point in doing such a test unless you're using equipment the
listener already thinks he hears differences among.

I would gladly take part in a controlled, double-blind test using the
amplifiers I mentioned, in my own home, with my own program source and
speakers. But I would insist that the people running the test take it as
well, including uncontrolled listening.

Why? TO SEE WHAT HAPPENS. But you don't get this, do you? In your mind,
"science" is limited to double-blind testing -- and nothing else.

> Objecting shows me and everyone you are attempting to
> persuade here that you are unwilling to bet a hundred to
> one that what you say is not BULL**** from your deluded
> mind. Deep down you know you are sick, don't you?

That's a pretty nasty thing to say.

You support philosophically invalid testing (such as ABX), then condemn
people who want to test equipment as it is actually used, with controls that
don't interfere with the test.

Neither of us is sick. But one of us is deluded -- he has a very parochial
view of how one determines what is true and what is not.

Your problem is that you (and others) believe that certain types of
"scientific" testing define truth. What you don't seem to understand is that
the best of kind of science is done when the scientist says "Let's try this
and see what happens".

Arny Krueger

October 19th 09, 06:24 PM

"Adrian Tuddenham" >
wrote in message
valid.invalid

> This is where reviewers could help by running the
> amplifier into a dummy load at full power sinewave and,
> if it blows up, saying so in their reviews. That would
> soon sort out the false advertising claims.

IME very few amplifiers would find that test to be daunting.

Wecan do it

October 19th 09, 07:28 PM

"William Sommerwerck" > wrote in
message ...
>> I recall some of the early takers were testing amps like
>> a Dyanco ST-70 against a Pass aleph(x) and could not
>> tell the difference.
>
> Then "something is wrong". Do you really believe that both
> amps are
> "perfect", or that they have essentially the same sets of
> errors &
> colorations? It's not likely.

I am sure that Richard Clark is interested in your talent to
be able to tell the two appart within the limitations of a
double blind test. He will even pay you Ten thousand Dollars!
If it is so easy why dont you do it Willaim?

Waiting for answer William?

Your neglecting to take this easy test shows me that you
BELIEVE 100:1 that you could not tell the difference between
your beloved amp and any run-of-the-mill amp with comparable
specs. PERIOD. You BELIEVE you will fail or you would take the
challenge and pick up the TEN GRAND.

peace
dawg

Scott Dorsey

October 19th 09, 09:15 PM

Arny Krueger > wrote:
>"Adrian Tuddenham" >
>wrote in message
valid.invalid
>
>> This is where reviewers could help by running the
>> amplifier into a dummy load at full power sinewave and,
>> if it blows up, saying so in their reviews. That would
>> soon sort out the false advertising claims.
>
>IME very few amplifiers would find that test to be daunting.

Arny, I don't think you realize just how bad some amplifiers out there
really are.

The car stereo world is even worse, too.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

Adrian Tuddenham[_2_]

October 19th 09, 09:30 PM

Arny Krueger > wrote:

> "Adrian Tuddenham" >
> wrote in message
> valid.invalid
>
> > This is where reviewers could help by running the
> > amplifier into a dummy load at full power sinewave and,
> > if it blows up, saying so in their reviews. That would
> > soon sort out the false advertising claims.
>
> IME very few amplifiers would find that test to be daunting.

There are some bad one around and that would soon weed them out.

--
~ Adrian Tuddenham ~
(Remove the ".invalid"s and add ".co.uk" to reply)
www.poppyrecords.co.uk

Arny Krueger

October 19th 09, 09:42 PM

"Scott Dorsey" > wrote in message

> Arny Krueger > wrote:

>> "Adrian Tuddenham" >
>> wrote in message
>> valid.invalid

>>> This is where reviewers could help by running the
>>> amplifier into a dummy load at full power sinewave and,
>>> if it blows up, saying so in their reviews. That would
>>> soon sort out the false advertising claims.

>> IME very few amplifiers would find that test to be
>> daunting.

> Arny, I don't think you realize just how bad some
> amplifiers out there really are.

My favorite benchmark for mediocre amplifiers has been 100 wpc receivers
normally street priced about $75. Used to be Pioneer, now are Sherwood. True
bottom feeder bait, but not all that bad on the test bench.

I've blown some of them up, but the conditions were well beyond reasonable
expectations, such as clipping @ 20 KHz for well over an hour. Just a half
hour wouldn't do it. This was, BTW an accident - got called away and forgot
to power down the test.

> The car stereo world is even worse, too.

I've tested a number of low-priced bottom feeders from that world as well.

The most common problem I've experienced with bottom-feeder equipment is
optimistic power ratings. If you define "full power" as clipping or a few
tenths below, they are often remarkably good. However the actual power is a
dB plus less than what the ads say.

PStamler

October 20th 09, 03:23 AM

Meanwhile, as the usual argument rages on, let's get back to the
original question.

Scott's right about the uselessness of THD as a measure of quality.
I've found high-frequency IMD tests more useful at separating sheep
from goats, at least for line-level devices. For example, I ran tests
on two graphic EQs from a company with a Japanese name (they also make
pianos and motorcycles) and one with a German name. The latter showed
horrendous levels of distortion when hit by a signal of 19kHz +
19.5kHz, equal levels, at about +10dBu if I remember correctly. The
former showed about the same level of distortion as the residual of
the test setup. The one with the German name, incidentally, sounds
hideously bad in use. The one with the Japanese name doesn't seem to
change the sound much at all, set flat.

Don P. (I think) suggests that high-frequency THD maps well to high-
frequency IMD tests. I don't know, but intend to do some tests to try
and find out.

Meanwhile, about the suggestion that amplifiers using discrete devices
suffer less from thermal distortion than integrated circuit
chips...Samuel Groner has done a lot of tests of IC amplifiers (and
some discrete amplifiers). Some ICs, including the common 5532, show
significant amounts of thermal distortion at LF, suggesting that the
idea that an opamp's behavior is determined by its external components
isn't correct. On the other hand, his tests also show a lot of ICs
that *don't* have significant thermal distortion, proving that you
don't necessarily need discrete devices to avoid the problem.

By the way, Mr. Groner includes in his measurement series midfrequency
HD spectra for all of the opamps he tests. There's a huge amount of
variety.

Peace,
Paul

Meindert Sprang

October 20th 09, 08:22 AM

"William Sommerwerck" > wrote in message
...
> > For example, an amplifier capable of 100W undistorted sine wave output
> > can usually do nearly 200W "continuous average power" with a square
> wave...
>
> For how long? <grin>

Longer than with a sine wave. With a square wave, the dissipation in the
output stage is lower.

Meindert

Anahata

October 20th 09, 08:43 AM

On Mon, 19 Oct 2009 09:46:19 -0700, William Sommerwerck wrote:

> The testing procedures forced by the FTC on the audio industry about 30
> years ago included a pre-conditioning warmup at 1/3 full power.
> Unfortunately, this is close to the point where a class B (or near-class
> B) amplifier dissipates maximum power. It's only about 5dB from full
> output, so unless you're listening to highly compressed material at high
> volume levels, it doesn't reflect "real-life" usage.

Surely better to test under (almost) worst case conditions, for
confidence that it won't fail when used normally, or the first time it's
rn hard at high ambien temperature.

(1/4 power square wave would be the worst for the output devices, or full
power square wave if you want to prove the power supply is unbreakable.)

--
Anahata
==//== 01638 720444
http://www.treewind.co.uk ==//== http://www.myspace.com/maryanahata

William Sommerwerck

October 20th 09, 12:06 PM

>>> For example, an amplifier capable of 100W undistorted
>>> sine wave output can usually do nearly 200W "continuous
>>> average power" with a square wave...

>> For how long? <grin>

> Longer than with a sine wave. With a square wave, the dissipation
> in the output stage is lower.

I'm not sure. That would definitely be true if the amp were driven into
saturation/cutoff. But it isn't, in this case.

You'd have to do a calculation to confirm this, I think. How's your integral
calculus?

William Sommerwerck

October 20th 09, 12:07 PM

>> The testing procedures forced by the FTC on the audio industry about
>> 30 years ago included a pre-conditioning warmup at 1/3 full power.
>> Unfortunately, this is close to the point where a class B (or near-class
>> B) amplifier dissipates maximum power. It's only about 5dB from full
>> output, so unless you're listening to highly compressed material at
>> high volume levels, it doesn't reflect "real-life" usage.

> Surely better to test under (almost) worst case conditions, for
> confidence that it won't fail when used normally, or the first time it's
> rn hard at high ambien temperature.

Yes, but... 1/3 power doesn't correspond to common usage. Listeners tend to
listen either at moderate levels, or push the amplifier to near its limits.

> (1/4 power square wave would be the worst for the output devices, or full
> power square wave if you want to prove the power supply is unbreakable.)

Meindert Sprang

October 20th 09, 12:49 PM

"William Sommerwerck" > wrote in message
...
> >>> For example, an amplifier capable of 100W undistorted
> >>> sine wave output can usually do nearly 200W "continuous
> >>> average power" with a square wave...
>
> >> For how long? <grin>
>
> > Longer than with a sine wave. With a square wave, the dissipation
> > in the output stage is lower.
>
> I'm not sure. That would definitely be true if the amp were driven into
> saturation/cutoff. But it isn't, in this case.

Ah, I assumed (from a comment further up in the postm I believe) saturation.

Meindert

Arny Krueger

October 20th 09, 02:13 PM

"PStamler" > wrote in message

> Meanwhile, as the usual argument rages on, let's get back
> to the original question.

> Scott's right about the uselessness of THD as a measure
> of quality.

Kinda. One of the last well-known authorities to study this matter is Dr.
Earl Geddes, who I see personally fairly frequently including last night at
a local AES chapter meeting (speaker: James Johnson).

Earl, who is well known for his AES papers in this matter will tell you that
THD is useless in those cases where the amount of measured THD is at audible
levels. IOW, if you have two devices (e.g. speakers) that have audible
amounts of distortion, then aTHD measurements will give you no useful clue
as to which might actually sound better. It is possible that one will sound
far better than the other depending on the spectral content of the spurious
products due to nonlinear distoriton.

Earl is quick to point out that if a THD measurement is so low that there
will be no audible distortion (e.g. THD 100 dB+ down) then the THD
measurement is very valuable. It tells you not to worry!

> I've found high-frequency IMD tests more
> useful at separating sheep from goats, at least for
> line-level devices. For example, I ran tests on two
> graphic EQs from a company with a Japanese name (they
> also make pianos and motorcycles) and one with a German
> name. The latter showed horrendous levels of distortion
> when hit by a signal of 19kHz +
> 19.5kHz, equal levels, at about +10dBu if I remember
> correctly. The former showed about the same level of
> distortion as the residual of the test setup. The one
> with the German name, incidentally, sounds hideously bad
> in use. The one with the Japanese name doesn't seem to
> change the sound much at all, set flat.

I'm guessing Yamaha and Behringer. Big difference in price.

However, I agree that HF nonlinear distortion can be one of those tests that
separates sheep from goats. Given the pervasive use of op amps, it is easy
for an unskilled designer to pick the wrong chip and run out of open loop
gain at high frequencies.

> Don P. (I think) suggests that high-frequency THD maps
> well to high- frequency IMD tests. I don't know, but
> intend to do some tests to try and find out.

The first problem you run into HF THD measurements is that some or all of
the harmonics can easily fall outside the operational or measurement
equipment bandpass. Multitones (including some twin tones) will produce
in-band distortion products, which are relevant and useful.

> Meanwhile, about the suggestion that amplifiers using
> discrete devices suffer less from thermal distortion than
> integrated circuit chips...Samuel Groner has done a lot
> of tests of IC amplifiers (and some discrete amplifiers).
> Some ICs, including the common 5532, show significant
> amounts of thermal distortion at LF, suggesting that the
> idea that an opamp's behavior is determined by its
> external components isn't correct.

Please see my previous comments about problems related to running out of
open loop gain, particularly at high frequencies. The external components
only set performance when there is a goodly surplus of open loop gain. It's
up to the designer to ensure that adequate reserves of open loop gain are
present. One rather obvious way to get tripped up this way is with mic
preamps. However, the HF bands of a graphic equalizer can also make demands
that an inexperienced desgner didn't anticipate.

> On the other hand, his
> tests also show a lot of ICs that *don't* have
> significant thermal distortion, proving that you don't
> necessarily need discrete devices to avoid the problem.

That was my point.

> By the way, Mr. Groner includes in his measurement series
> midfrequency HD spectra for all of the opamps he tests.
> There's a huge amount of variety.

http://www.edn.com/blog/1700000170/post/610040861.html

http://www.sg-acoustics.ch/analogue_audio/ic_opamps/pdf/opamp_distortion.pdf

Arny Krueger

October 20th 09, 02:15 PM

"William Sommerwerck" > wrote in
message
>>>> For example, an amplifier capable of 100W undistorted
>>>> sine wave output can usually do nearly 200W "continuous
>>>> average power" with a square wave...
>
>>> For how long? <grin>
>
>> Longer than with a sine wave. With a square wave, the
>> dissipation in the output stage is lower.

> I'm not sure. That would definitely be true if the amp
> were driven into saturation/cutoff. But it isn't, in this
> case.

Really? Don't you think that the square wave is due to the output stage
being driven from saturation to cutoff?

> You'd have to do a calculation to confirm this, I think.

No, just practical experience with power amps.

> How's your integral calculus?

Good enough to know it isn't needed.

William Sommerwerck

October 20th 09, 02:24 PM

>>>>> For example, an amplifier capable of 100W undistorted
>>>>> sine wave output can usually do nearly 200W "continuous
>>>>> average power" with a square wave...

>>>> For how long? <grin>

>>> Longer than with a sine wave. With a square wave, the
>>> dissipation in the output stage is lower.

>> I'm not sure. That would definitely be true if the amp
>> were driven into saturation/cutoff. But it isn't, in this
>> case.

> Really? Don't you think that the square wave is due to the output
> stage being driven from saturation to cutoff?

I interpreted the remark as referring to a square wave input, not to
clipping. And a sine wave that clipped the amp would have to severely
overdrive it before it could be donsidered a "true" square wave.

Anahata

October 20th 09, 02:50 PM

On Tue, 20 Oct 2009 06:24:42 -0700, William Sommerwerck wrote:

> I interpreted the remark as referring to a square wave input, not to
> clipping.

Correct.
And I assumed the the output stage was just at the onset of clipping so
the square ware was beign reproduced correctly.
Under those conditions the output transistors don't actually get very
hot, because they alternately have no current passing through them or
very little voltage across them.

Worst case for output stage dissipation is a square wave at half clipping
voltage, or strictly speaking half the supply voltage.

That's with a resistive load - a pure reactive load is even worse for
output stage dissipation.

--
Anahata
==//== 01638 720444
http://www.treewind.co.uk ==//== http://www.myspace.com/maryanahata