The $10000 Hi-Fi

On Wed, 6 May 2015 17:28:46 +0000 (UTC), glen herrmannsfeldt
<gah@ugcs.caltech.edu> wrote:

>Done on unsigned (that is, truncate toward + or - infinity), the only
>difference should be a 0.5lsb offset. But if you truncate toward zero,
>that, I believe, could make a difference.

There are definite harmonic distortion products resulting from
truncation. See, for example, Figures 3, 4, 9, 10 here:
http://www.users.qwest.net/~volt42/cadenzarecording/DitherExplained.pdf

Note that the above uses RPDF dither, according to the author.

Greg Berchin <gjberchin@chatter.net.invalid> writes:

> On Wed, 06 May 2015 11:19:00 -0400, Randy Yates
> <yates@digitalsignallabs.com> wrote:
>
>>I presume that "highpass" means it's further spectrally-shaped and not
>>"white." I seem to recall he wasn't a fan of spectrally-shaped,
>>non-subtractive dithers.
>
> (current random value - previous random value)
>
> It is a 1st-order digital differentiator, with a zero at DC and a gain
> of 2 at half the sampling frequency. Scale to suit desired dither level.
>
> If my ears disagree with Dr. Wannamaker's ears, so be it. Your ears may
> disagree with mine. That's OK, too.

Your (our) ears are the final judge. I absolutely agree.
-- 
Randy Yates
Digital Signal Labs
http://www.digitalsignallabs.com

One thing we can all agree on is that any posting debating the audibility of various impairments generates 5x more responses than any other subject. 

If I were running for president I would run on a pro-192khz Fs platform and would sweep the oft-overlooked audiophile voting block. Our numbers may be small but we have a way of making ourselves heard above the clatter:) 

Bob

On 5/6/15 1:50 PM, Greg Berchin wrote:
> On Wed, 6 May 2015 17:28:46 +0000 (UTC), glen herrmannsfeldt
> <gah@ugcs.caltech.edu>  wrote:
>
>> Done on unsigned (that is, truncate toward + or - infinity), the only
>> difference should be a 0.5 LSB offset. But if you truncate toward zero,
>> that, I believe, could make a difference.

yes.  dead zone at zero crossings.

but, in my opinion, there is little sonic difference between 
rounding-to-nearest and rounding-toward-minus-infinity except for a DC 
constant of 1/2 LSB (which is not sonic).

> There are definite harmonic distortion products resulting from
> truncation. See, for example, Figures 3, 4, 9, 10 here:
> http://www.users.qwest.net/~volt42/cadenzarecording/DitherExplained.pdf
>
> Note that the above uses RPDF dither, according to the author.

i would have a few little bones to pick with this paper.  first of all 
the author repeats the common mistake of confusing power spectrum 
("white" vs. "colored") and p.d.f. (RPDF vs. TPDF vs. Gaussian).  them's 
about different properties of a random process.  you most certainly 
**can** have TPDF *and* some kinds of colored noise.  my favorite

-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."

On 5/6/15 1:25 PM, glen herrmannsfeldt wrote:
> rickman<gnuarm@gmail.com>  wrote:
>
> (snip)
>
>> Maybe I am misunderstanding what you are saying.   You don't add
>> dithering to make anything audible.  You add the dithering so the
>> truncation does not inter-modulate with the signal.  Instead of getting
>> tones from the quantization or truncation, adding the minimum amount of
>> dither (1 lsb max) spreads the noise over the full range of frequency
>> making it much lower than the tones you would get otherwise.

those tones can cover up very low-level sounds even if they might have 
less total energy than the TPDF dither.

>
> If you make the noise less audible, the signal is more audible.

    http://drewdaniels.com/dither.pdf

(an early and seminal AES paper in the 80s.  dunno who drew daniels is.)

>
> I once made two CDs, one truncate, one dither.  The tracks were all the
> same source, but shifted successively one bit more.

were they originall 16-bit sound?  if so, not a particularly useful 
experiment, me thinks.

> I don't remember now
> how far down it went, but listening carefully and turning the volume
> control up, it was obvious that dither had less audible noise.
>
> -- glen
>
>


-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."

On 5/6/15 1:45 AM, Steve Pope wrote:
> rickman<gnuarm@gmail.com>  wrote:
>
>> The issue is not how many bits are used for the signal, but how many are
>> used for the noise.  With 16 bits, yeah, you may hear the difference.
>> But at 24 I just can't see how you would  ever hear it.  I'd bet that if
>> you fed a 1 bit tone into a 24 bit ADC with the volume on the amp at a
>> normal listening level you can't hear the tone at all.
>
> Well, that isn't the only scenario in which a 24-bit signal might
> get used.

whether it's with an old Freescale 56K or some other processor, a lot of 
internal arithmetic in audio algorithms is done at 24 bits.  32-bit 
floats have a 25-bit mantissa.

> The real point is, when you quantize from analog and/or reduce precision,
> dithering is wise.  If there is not already enough noise in the
> source signal to auto-dither the thing, then add explicit dithering.
> Don't waste brain cycles arguing "it might not matter", just do it,
> unless the cost is prohibitive, which it almost never is, especially
> in audio.
>
> If you've got a K-band radar signal you are debating whether to
> dither, then that is a plausible debate.  At audio it's a no brainer.

wellllll....

i do audio. in audio algorithms i deal with, there are several, 
sometimes *many* points of quantization.  like, often, every gain block 
and more.

dither isn't that cheap.  a good RNG costs instruction cycles.  cheap 
RNGs (like linear congruance) usually sound like crap.  building TPDF 
dither doubles the cost (unless you're building that high-pass TPDF 
which does not double the cost).

i have found that undithered truncating *with* noise-shaping (like that 
fixed-point DC blocking trick at dspguru.com) which Randy calls 
"fraction saving" (a *very* good name for it) works *very* well for most 
of these internal nodes (at 24 bit).  it steers the quantization noise 
into the top octave (and away from the bottom 8 or 9 octaves) where i am 
deaf anyway.  and it solves the DC limit-cycle problem which is why it's 
useful in a DC-blocking filter or any other audio filter with high-Q 
poles (an annoyance is when absolute silence goes into your filter after 
some sound and the level comes down to -70 dB and gets stuck there).

and, with floating-point, i most often don't do anything.  even with 
single-precision float.  i'm trying to understand how i can cheaply get 
those lost 8 bits in the SHArC when it goes from 40 internal bits to 32. 
  i would think that rounding error can be used with noise shaping 
feedback in the same manner as with fixed-point error shaping.

however, like when going from 24-bit to 16-bit (like in mastering), 
really good dither and noise-shaping (or using "colored dither" like 
UV22) is mandatory.  dunno what should be done if mastering to MP3.  i 
thought the MP3 coding alg figgered it out anyway.

so, in my opinion, some brainy judgment is useful.

-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."

robert bristow-johnson <rbj@audioimagination.com> wrote:

(snip, I wrote)
>> I once made two CDs, one truncate, one dither.  The tracks were all the
>> same source, but shifted successively one bit more.
 
> were they originall 16-bit sound?  if so, not a particularly useful 
> experiment, me thinks.

They were originally 24, but it doesn' matter, as more and more bits
were shifted off.

What I do is read in a 24 bit value, fix the sign appropriately,
add appropriate dither, shift right a user selectable amount, 
fix the sign again, and write out the right 16 bits.

If I get the record level right, the shift value is 8, but I can
select anything from 0 to 31. Any value over 8 results in a smaller
amplitude signal, up to 23 when just one bit is left. 

For any shift value greater than zero, the dither has an effect.

As well as I remember, I did it up to the point that only six
bits were left. 

-- glen

On 5/5/15 4:47 PM, glen herrmannsfeldt wrote:
>
...
>
> I use a pretty simple LFSR generator, based on CRC32, to generate
> bits, shift and add.

oooh, glen, that can't be good.  are you using the LFSR to just get 
random *bits* (like +1 and -1) that are white.  then it works good.  but 
if you're using the whole shift register for a random number, that's not 
so good.  that's because 50% of the time the following register value is 
related to the current by a factor of 2.  and the spectrum of this is 
*not* white but is low-pass.


> Cedron<103185@dsprelated>  wrote:
>
>> Do you apply software audio compression? (not data compression).  Many
>> years ago I derived the formulas for a compression curve that extended the
>> straight line portion with a hyperbola

something like  x - (1-1/r)*log(k + e^x) ?

("r" is the compression ratio after the bend. "k" is the knee softness.)


>> that matched the first derivative
>> so there was no knee effect.

a bend without a knee?  what does that mean?

do you actually mean a "soft knee"?  a knee with programmable softness?



-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."

On 5/6/15 9:23 AM, Steve Pope wrote:
> Bob Masta<NoSpam@daqarta.com>  wrote:
>
>
>> As I recall, this became
>> a problem only when the S/N improved so much that
>> self-dither wasn't taking place.  (Sony, I think, was first
>> with this.)  All this was in the 16-bit era.
>>
>> So yeah, definitely needed on real-world signals.
>
> Not sure Sony was the first, but they had an early 16-bit
> quantizer (I believe marketed as a "PCM unit", although that's
> a misnomer) that piggybacked onto a VCR used as a data recorder.
> This was around 1978.


i think it's the Sony F1.  and they used betamax, not VHS.  and this is 
the reason why 44.1 kHz (or, more precisely, 44.056 kHz with the F1) 
became the CD sample rate standard.  very icky.  too bad they didn't go 
with 48 kHz.

> A friend was an engineer at Crystal Sound. The liked the Sony
> unit, until it glitched once, then they abandoned it.  No room
> for error in this biz.

immature technology often glitches.  that's why i like using an old Mac 
rather than whatever is the newest thingie.


-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."

On 5/6/15 4:51 PM, radams2000@gmail.com wrote:
> One thing we can all agree on is that any posting debating the audibility of various impairments generates 5x more responses than any other subject.
>
> If I were running for president I would run on a pro-192khz Fs platform and would sweep the oft-overlooked audiophile voting block. Our numbers may be small but we have a way of making ourselves heard above the clatter:)
>

if i were prez, i would advocate summary execution of monster cable 
promoters and shitty code writers.  don't need no 192 kHz.  96 should 
be good enuf.


-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."