The $10000 Hi-Fi

Bob Masta <NoSpam@daqarta.com> wrote:

>From what I recall, Back In The Day the need for dither was
>first noted on long piano decays at the end of a track,
>where instead of fading smoothly into the noise floor you'd
>hear a buzz  just before fade-out.  

I didn't see that exact effect, but it is completely consisten.

>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.

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.


Steve

On Wed, 06 May 2015 12:39:51 GMT, N0Spam@daqarta.com (Bob Masta) wrote:

>Count me among the skeptics on this.  First, I seriously
>doubt that Stereophile magazine would know *everything* that
>the manufacturer did in the change, and I'd be more inlined
>to think it was something else that turned the tide... I'd
>guess even the semi-mythical capacitor replacement would
>have a higher audibility.

Again, I am working from old memory here, but as I recall the very point
of the article being that the designers themselves were astonished by
the effect of just changing from rounding to truncation at the 24th bit,
and brought it to the attention of the Stereophile editors.

I doubt that I have the issue any more, else I'd reference the article
directly. It was long enough ago that it may even have occurred before
dither was well understood. If so, then it's still amazing that any
effect would be audible at the 24th bit.

And that's my point -- we don't understand everything about how and what
we hear, and sometimes mathematical perfection doesn't lead to the best
sounding results.

On Tue, 05 May 2015 15:20:03 -0500, "Cedron" <103185@DSPRelated> wrote:

>Is the dither significant soundwise?  In other words, can you hear the
>difference with or without?  

Not only can you hear the difference with or without, you can hear the
difference between different dither types.

Years ago I received some Matlab code from RB-J that implemented several
different kinds of dither -- RPDF, TPDF, spectrally-shaped, noise-shaped
via feedback, etc. I converted the code from monophonic to multichannel
for an Audio Engineering Society project, and in the process I listened
to all of the different dither types with typical program material
(music on CD that I had in my collection). Since this program material
was presumably already dithered at the 16th bit, I redithered at bit 14
or lower to make an actual difference.

To my ears, the best sounding dither was TPDF highpass, which is nice
because it's also one of the easiest to implement. Meanwhile, the
super-sophisticated, noise-shaped to correspond to Fletcher and Munson
and who-knows-who-else dither schemes sounded pretty bad by comparison.

YMMV.

>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 that matched the first derivative
>so there was no knee effect.  I am considering building it into my
>recording program, but not sure it would be worthwhile or better to leave
>it as an external process.

I have designed dynamic range compressors, both soft-knee and hard-knee.
Matching the first derivative with a smooth curve (not necessarily
hyperbolic, but a hyperbola is fine) definitely changes the sound. Some
people prefer the soft knee, others the hard knee. Mustang/Camaro.

On 5/6/2015 9:35 AM, Greg Berchin wrote:
> On Tue, 05 May 2015 15:20:03 -0500, "Cedron" <103185@DSPRelated> wrote:
>
>> Is the dither significant soundwise?  In other words, can you hear the
>> difference with or without?
>
> Not only can you hear the difference with or without, you can hear the
> difference between different dither types.
>
> Years ago I received some Matlab code from RB-J that implemented several
> different kinds of dither -- RPDF, TPDF, spectrally-shaped, noise-shaped
> via feedback, etc. I converted the code from monophonic to multichannel
> for an Audio Engineering Society project, and in the process I listened
> to all of the different dither types with typical program material
> (music on CD that I had in my collection). Since this program material
> was presumably already dithered at the 16th bit, I redithered at bit 14
> or lower to make an actual difference.

When you say you "dithered" it at 14 bits, did you truncate/round the 
data to 14 bits?  If not, you were just listening to added noise which 
is very different.

-- 

Rick

Greg Berchin <gjberchin@chatter.net.invalid> writes:
> [...]
> To my ears, the best sounding dither was TPDF highpass, 

TPDF is what Robert Wannamaker, in his PhD thesis, concludes is a good
distribution because it decorrelates the first and second moments from
the signal.

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.
-- 
Randy Yates
Digital Signal Labs
http://www.digitalsignallabs.com

On Wed, 06 May 2015 10:36:45 -0400, rickman <gnuarm@gmail.com> wrote:

>When you say you "dithered" it at 14 bits, did you truncate/round the 
>data to 14 bits?  If not, you were just listening to added noise which 
>is very different.

The bits below 14 (or whatever "N" was selected) were all 0.

The code works in double precision floating point, scales as necessary
to make 2^(N-1) represent full-scale, adds the dither (a fractional
value), rounds [add 0.5 then floor()], and then re-scales to 16 bits.

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.

On 06 May 2015 11:20:11 GMT, Allan Herriman
<allanherriman@hotmail.com> wrote:

>On Tue, 05 May 2015 17:46:00 +0000, Eric Jacobsen wrote:
>
>> On Tue, 05 May 2015 12:02:41 GMT, N0Spam@daqarta.com (Bob Masta) wrote:
>> 
>>>On Mon, 04 May 2015 22:43:01 GMT, eric.jacobsen@ieee.org (Eric Jacobsen)
>>>wrote:
>>>
>>><snip>
>>>
>>>>I haven't seen anybody claim that to be true for even a new Ferrari or
>>>>Porsche yet, but I suspect they're getting close.
>>>
>>>My engineering career started back in the early '70s at GM's Cadillac
>>>division.  One epiphany from that was the huge engineering advantage of
>>>high production volumes. I was surprised to discover that Roll Royce,
>>>Lamborghini, etc bought GM transmissions. But when you are only turning
>>>out a handful of cars per year, you can't afford a huge engineering
>>>staff for just one part of the car.  Even more important, you can't get
>>>decent field data on failure modes so you can improve your product.
>> 
>> It's interesting that for a lot of the performance technology, including
>> things like ABS, stability control, active handling, and
>> performance-optimized traction control, the technology often comes from
>> ultra-low volume race applications.   Companies that have big race
>> efforts as part of their core, like Ferrari, McLaren, Porsche, Honda,
>> Audi, etc., etc., and in this case specifically GM with the Corvette
>> (and also Cadillac lately), the technology gets developed or really
>> optimized as part of the racing effort.
>
>
>One racing innovation that I found fascinating (and also related to 
>filter design, keeping it on-topic) is the "inerter".
>
>Suspension designers use filter synthesis techniques borrowed from 
>electronics.  They often use the "Mobility analogy" :
>
> mass == capacitance to ground
> spring (or tyres) == inductor
> shock absorber == resistor
>
>One can synthesise low pass filters using the usual method(s) for analog 
>filter design.  But to get sharper rolloffs, e.g. as in an elliptic 
>filter, the designer needs a pair of zeros on or near the jw axis.  In a 
>filter, that is done with a series element consisting of an inductor in 
>parallel with a capacitor.
>Can't be done mechanically as there isn't a mechanical element analogous 
>to a "flying" (i.e. ungrounded) capacitor ... that is, until someone 
>invented the "inerter".  Amazingly, this didn't happen until early this 
>century.
>
>
>"Flying" capacitor in parallel with inductor giving transmission zeros:
>http://en.wikipedia.org/wiki/File:RLC_parallel_band-stop.svg
>
>"Inerter":
>http://en.wikipedia.org/wiki/Inerter_%28mechanical_networks%29
>
>Regards,
>Allan

Very cool.   Thanks for posting the links.


Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com

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.

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

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

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

(snip)

> Again, I am working from old memory here, but as I recall the very point
> of the article being that the designers themselves were astonished by
> the effect of just changing from rounding to truncation at the 24th bit,
> and brought it to the attention of the Stereophile editors.

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.
 
> I doubt that I have the issue any more, else I'd reference the article
> directly. It was long enough ago that it may even have occurred before
> dither was well understood. If so, then it's still amazing that any
> effect would be audible at the 24th bit.
 
> And that's my point -- we don't understand everything about how and what
> we hear, and sometimes mathematical perfection doesn't lead to the best
> sounding results.

-- glen