16-bit dithered, 24-bit

[Tommi]

It is said that dithering from 24-bits to a final 16-bit cd master uses more
bits in the final product.

Then again, if 16-bit is dithered properly, in it's available 96dB range
it's resolution is in theory infinite.

Suppose you record the whole album with BOTH at 16 bits, properly dithered,
and at 24 bits, dithered, and none of the tracks used in the whole product
have over 96dB's worth of dynamics. (internal processing would be done with
floating point 32 or 64)

In such case, *how* could the 24-bit version of the final mixes dithered
further down to final 16 bits, sound any better than than its 16-bit
counterpart?

[WillStG]

>"Tommi"
>It is said that dithering from 24-bits to a final 16-bit cd master uses
>more
>bits in the final product.

No, it doesn't. Dithering adds hard-to-hear noise at the bottom where they
cut the lowest/quietest bits off (the "least signifigant bits"/LSB ) so that
the errors caused by the lowest bits toggling on and off get smoothed out (at
the fade ins and fade outs, on the reverb tails.)

To hear this, record a sample of a 24 bit reverb tail from a snare hit and
let it trail off into silence. Now bit reduce a copy of that sample to 16 bits
without dither (bounce to disk with faders at 0 will work). Normalize the end
of the bit reduced sample and play back, you'll hear all kind of ugly
distortion at the tail as the bits toggle between 0's and 1's.

Now bit reduce the original reverb sample to 16 bits with dither, normalize
and play back. You'll hear where the
sound tails off it's much more pleasant and smooth.

Will Miho
NY Music & TV Audio Guy
Off the Morning Show! & sleepin' In... / Fox News
"The large print giveth and the small print taketh away..." Tom Waits

[Mike Rivers]

In article > writes:

> It is said that dithering from 24-bits to a final 16-bit cd master uses more
> bits in the final product.
>
> Then again, if 16-bit is dithered properly, in it's available 96dB range
> it's resolution is in theory infinite.

Um . . who says all this?

> In such case, *how* could the 24-bit version of the final mixes dithered
> further down to final 16 bits, sound any better than than its 16-bit
> counterpart?

If you plugged a microphone into a true 16-bit converter and did
nothing else but write the resulting file on to a CD (no level
adjustments, no edits with crossfades, no EQ or dynamics processing),
then a "converted" 24-bit recording would sound a trifle noisier,
though in practice I doubt you could tell the difference.

If you manipulate a 16-bit recording in ways that cause its word
length to increase, and they you dither (or worse, truncate) the final
product, it will probalby sound worse than a similar recording that
went through the same processes but started out at 24-bit resolution.


--
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

[Tommi]

"WillStG" > wrote in message
...
> >"Tommi"
> >It is said that dithering from 24-bits to a final 16-bit cd master uses
> >more
> >bits in the final product.
>
> No, it doesn't. Dithering adds hard-to-hear noise at the bottom where
they
> cut the lowest/quietest bits off (the "least signifigant bits"/LSB ) so
that
> the errors caused by the lowest bits toggling on and off get smoothed out
(at
> the fade ins and fade outs, on the reverb tails.)

With "more bits" I mean exactly that. I should've been more clear and state
something like: "dithering from 24-bits to a final 16-bit essentially
corrects the quantization errors of the least significant bits used in the
16-bit, thus resulting ideally in a linear level response in the system's
whole dynamic range".
That means, with dither, a digital system has practically no "stairsteps" in
its available dynamic range, making it as analogous to the original signal
as a tape system.

[Tommi]

"Mike Rivers" > wrote in message
news:znr1072028938k@trad...
>
> In article >
writes:
>
> > It is said that dithering from 24-bits to a final 16-bit cd master uses
more
> > bits in the final product.
> >
> > Then again, if 16-bit is dithered properly, in it's available 96dB range
> > it's resolution is in theory infinite.
>
> Um . . who says all this?

With more bits (when dithered 16bit)I mean that the quantization errors of
least significant bits are removed(see my reply to Will's post), this means
that a properly dithered 16-bit file has a better resolution than 16 bits
would allow in theory. It wasn't being very clear, but I meant that when I
said that dithering allows "more bits".

As for the second argument, I will quote Bob Cain's post(pardon me if I've
in any way misunderstood it) in a recent thread which discussed 16bits vs
24bits:

"Many think that the increased
resolution results in less perception of some kind of
stairstep effect. That is not the case. The preceived
situation with an N bit converter done properly and going
through the A/D and then the D/A process is _exactly_ the
same as an infinite resolution conversion at both stages
with a digital adder in between just adding in a noise
signal comprised of a random variable with values of 0 or
2^-N at each sample time. What is heard is additive noise
and only that iff the conversion is done without correlation
between the value of that bit and the value of the sample.
This is practically achievable."

There were also many other who informed that any system which is properly
dithered has, in theory, an infinite range in the system's own available
dynamic range. (thus more bits only allow a greater dynamic range, there is
no stairstep effect when the system is dithered).
I know that LSB, MSB(every bit doubles the choice of values) thing, I'm just
curious.

[WillStG]

>"Tommi"

>With "more bits" I mean exactly that. I should've been more clear and state
>something like: "dithering from 24-bits to a final 16-bit essentially
>corrects the quantization errors of the least significant bits used in the
>16-bit, thus resulting ideally in a linear level response in the system's
>whole dynamic range".
>That means, with dither, a digital system has practically no "stairsteps"
>in
>its available dynamic range, making it as analogous to the original signal
>as a tape system.

I think you are confusing two concepts with the "stair steps" idea,
resolution is determined by sample rate not bit-length. The difference between
24 and 16 bit files has to do with how quiet the signal can get. Bascially
bit-reduction, with or without dither, is truncation plain and simple. You're
cutting off the quietest bits, for most pop music you really need a dead quiet
room and playback system to hear the difference between a 24 bit file and a
truncated 16 bit file. With acoustic classical music it's easier to hear, if
you try the practical examples I gave you, listening to normalized reverb tails
that trail into silence, you will be able to put what the actual sonic
difference is between a dithered and non-dithered file under the microscope.
Many people have so much noise in their audio chains that they can forego
dither when recording a file digitally, because the noise of the mics or
micpres are enough to dither, that's "self dithering".

Will Miho
NY Music & TV Audio Guy
Off the Morning Show! & sleepin' In... / Fox News
"The large print giveth and the small print taketh away..." Tom Waits

[Bob Cain]

Tommi wrote:
>
> With more bits (when dithered 16bit)I mean that the quantization errors of
> least significant bits are removed(see my reply to Will's post), this means
> that a properly dithered 16-bit file has a better resolution than 16 bits
> would allow in theory. It wasn't being very clear, but I meant that when I
> said that dithering allows "more bits".

In a sense it is infinite but that isn't really relevant
because the uncorrelated noise that appears down at the
bottom after proper dither puts a floor on the resolution
you can make out above that noise. It's just that the noise
after dither has no relationship to the music and it is
relationships between the noise and the music, such as that
caused by simple truncation or rounding, that cause audible
unpleasantries.


Bob
--

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

A. Einstein

[Scott Dorsey]

Tommi > wrote:
>It is said that dithering from 24-bits to a final 16-bit cd master uses more
>bits in the final product.

What does "uses more bits" mean?

>Then again, if 16-bit is dithered properly, in it's available 96dB range
>it's resolution is in theory infinite.

What does "resolution" mean?

>Suppose you record the whole album with BOTH at 16 bits, properly dithered,
>and at 24 bits, dithered, and none of the tracks used in the whole product
>have over 96dB's worth of dynamics. (internal processing would be done with
>floating point 32 or 64)
>
>In such case, *how* could the 24-bit version of the final mixes dithered
>further down to final 16 bits, sound any better than than its 16-bit
>counterpart?

If the converters are perfect, they won't. BUT, in the real world, some
converters are better than others. AND, it can be nice to have some slop.
If you record a 2-track session with a 16 bit converter for release on 16
bits, you must get the levels just perfect to get 16 bits of significant data
on the final release. If you record on a 20 bit medium and then dither down
to 16 bits, you have 12 dB additional slop.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Andrew M.]

Scott Dorsey wrote:
> Tommi > wrote:
>
>>It is said that dithering from 24-bits to a final 16-bit cd master uses more
>>bits in the final product.
>
>
> What does "uses more bits" mean?
>
>
>>Then again, if 16-bit is dithered properly, in it's available 96dB range
>>it's resolution is in theory infinite.
>
>
> What does "resolution" mean?
>
>
>>Suppose you record the whole album with BOTH at 16 bits, properly dithered,
>>and at 24 bits, dithered, and none of the tracks used in the whole product
>>have over 96dB's worth of dynamics. (internal processing would be done with
>>floating point 32 or 64)
>>
>>In such case, *how* could the 24-bit version of the final mixes dithered
>>further down to final 16 bits, sound any better than than its 16-bit
>>counterpart?
>
>
> If the converters are perfect, they won't. BUT, in the real world, some
> converters are better than others. AND, it can be nice to have some slop.
> If you record a 2-track session with a 16 bit converter for release on 16
> bits, you must get the levels just perfect to get 16 bits of significant data
> on the final release. If you record on a 20 bit medium and then dither down
> to 16 bits, you have 12 dB additional slop.
> --scott

That's an excellent explanation Scott.

[Mike Rivers]

In article > writes:

> With more bits (when dithered 16bit)I mean that the quantization errors of
> least significant bits are removed(see my reply to Will's post), this means
> that a properly dithered 16-bit file has a better resolution than 16 bits
> would allow in theory. It wasn't being very clear, but I meant that when I
> said that dithering allows "more bits".

That's still kind of fuzzy. If you're talking about recording a file
through a 16-bit converter, to be fair, you have to assume that it's a
perfect 16-bit converter, and that the lowest order bit will be
accurate. Unfortunately they don't make 'em like that. If you record
with a 24-bit converter, the chances are that the 17th or better bit
will be accurate, so when you shorten the word length, you'll
essentially be making a better 16-bit converter.

I'll leave it to the statisticians to decide whether the low order bit
of a dithered-and-truncated file is any more accurate than the low
order bit that's as accurate as an ideal converter can make it. I
don't really know, and I doubt that you'd be able to convince me with
whatever mathematical proof you can write. But I'm not saying that
you're wrong either, just that when it comes to small differences like
this, I'm too lazy to worry about the math.

The good news is that 24-bit hardware allows us to have 16-bit
accuracy, which is sufficient for most practical purposes as a final
delivery medium.

> There were also many other who informed that any system which is properly
> dithered has, in theory, an infinite range in the system's own available
> dynamic range.

I hate to use the term "infinite" since the system noise will limit
the dynamic range if nothing else does first.



--
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

[Tommi]

"Scott Dorsey" > wrote in message
...

> What does "uses more bits" mean?

Suppose you have a 3-bit A/D converter. That's only 8 values to choose from.
If the 3-bit converter is properly dithered, by adding noise you will
essentially also eliminate the quantization effects. An input signal of
0.324 volts will have an average ADC output of exactly 0.324, not something
like 0.35. Again, it wasn't very clear from me to say it uses "more bits",
but when dither is done properly, it is my understanding that a dithered
3-bit A/D converter has _not_ 8 values in its 18dB FS range, but an infinite
number of values to choose from in it's own dynamic range(that is, after
dither, slightly less than 18 decibels).


> >Then again, if 16-bit is dithered properly, in it's available 96dB range
> >it's resolution is in theory infinite.
>
> What does "resolution" mean?

That there are is no audible quantization, no audible "jump" between the
voltage values of a converter. It behaves like an analog system.

> >In such case, *how* could the 24-bit version of the final mixes dithered
> >further down to final 16 bits, sound any better than than its 16-bit
> >counterpart?
>
> If the converters are perfect, they won't. BUT, in the real world, some
> converters are better than others. AND, it can be nice to have some slop.
> If you record a 2-track session with a 16 bit converter for release on 16
> bits, you must get the levels just perfect to get 16 bits of significant
data
> on the final release. If you record on a 20 bit medium and then dither
down
> to 16 bits, you have 12 dB additional slop.

I was thinking something like that, thank you for such a clear answer.

[Tommi]

"Bob Cain" > wrote in message
...
>
>
> Tommi wrote:
> >
> > With more bits (when dithered 16bit)I mean that the quantization errors
of
> > least significant bits are removed(see my reply to Will's post), this
means
> > that a properly dithered 16-bit file has a better resolution than 16
bits
> > would allow in theory. It wasn't being very clear, but I meant that when
I
> > said that dithering allows "more bits".
>
> In a sense it is infinite but that isn't really relevant
> because the uncorrelated noise that appears down at the
> bottom after proper dither puts a floor on the resolution
> you can make out above that noise. It's just that the noise
> after dither has no relationship to the music and it is
> relationships between the noise and the music, such as that
> caused by simple truncation or rounding, that cause audible
> unpleasantries.


When properly dithered, the noise floor goes up, but above that noise floor,
the resolution(the choice of values for a given voltage)is in effect
infinite, just like an analog tape is. This means that an input signal of
2.666 volts has an average A/D output of 2.666, not something like 2.7
volts.
He who knows this dither stuff better, by all means correct if I'm wrong.