Conceptual problem: Dither vs. Backlash

Archived from groups: rec.audio.pro (More info?)

Chris Hornbeck wrote:
> Several audio related devices exhibit zero-crossing
> "flat-spots", regions in their transfer curves near
> zero crossing where the Y axis output is transitionally
> flat. Poorly designed push-pull amplifiers exhibit
> "crossover distortion", magnetic storage has poorly
> characterized related effects, etc.
>
> Question: is this electronic "backlash", to use an old
> mechanist's term, conceptually related (at all?) to the
> well understood issue of quantizing distortion/ error?
> And, if so, could classical dither be considered to be
> a possible remedy?
>
> Or, to put the question another way, could dither (of any
> amount) correct a ferinstance problem of a D/A conversion
> with zero output below some absolute level, but with correct output
> above that level? Not an error in estimating a quantized
> level, but rather zero output below some (unsigned) level.
>
> I'll raise a secondary question about very large "bias"
> signals' effects after feedback and research.
>
>
> Thanks to all for any possible comments and corrections. Sorry
> it's so wishy-washy. As was once said (better) in this very
newsgroup,
> if I could properly ask the question, I wouldn't have to ask it.
> I think that might have been Logan; can't be sure now.
>
> Thanks,
>
> Chris Hornbeck

That is a great question, I've been thinking about the same thing
myself.

The answer must be YES. High frequency bias used to linearize
magnetic tape recording is really the same principle as dither!

I would think that any device with a dead-zone dead-band or backlash to
use any of the common phrases would benefit from dither or other form
of bias.

In the case of crossover distortion in a class AB amplifier, adding
audible noise would linearize but the noise is obviously undesirable.
A high frequency bias might work if it didn't cause intermodulation.
Noise outside the audio band might be interesting.

In A/D converters, the non-linearities are very small, i.e 1 LSB so it
takes a very small and acceptable amount of dither noise to linearize
them. In magnetic tape, the non-linearity is very large and it takes a
very large dither to linearize it so they use high frequency bias.

I once worked with ultrasonic transducers, and it was interesting that
when they were energized, they appeared very slippery because the
static friction was removed.

Great question... I look forward to seeing other comments...

Mark

Archived from groups: rec.audio.pro (More info?)

"Chris Hornbeck" <chrishornbeckremovethis@att.net> wrote in message
news:nbo141t4b02pf6aslqrbgsi6jn2lhvnere@4ax.com
> Several audio related devices exhibit zero-crossing
> "flat-spots", regions in their transfer curves near
> zero crossing where the Y axis output is transitionally
> flat. Poorly designed push-pull amplifiers exhibit
> "crossover distortion", magnetic storage has poorly
> characterized related effects, etc.
>
> Question: is this electronic "backlash", to use an old
> mechanist's term, conceptually related (at all?) to the
> well understood issue of quantizing distortion/ error?
> And, if so, could classical dither be considered to be
> a possible remedy?

Yes. I've actually tried this with a SS amp whose bias circuits I
modified to be a poor class B. The addition of a dithering noise of
suitable amplitude eliminated the gross, nasty-sounding distortion due
to crossover distortion.

> Or, to put the question another way, could dither (of any
> amount) correct a ferinstance problem of a D/A conversion
> with zero output below some absolute level, but with correct output
> above that level? Not an error in estimating a quantized
> level, but rather zero output below some (unsigned) level.

Yes. The only problem is that the dynamic range of the resulting
equipment ain't all that wonderful. Guess what - its noisy! ;-)

Archived from groups: rec.audio.pro (More info?)

In article <nbo141t4b02pf6aslqrbgsi6jn2lhvnere@4ax.com>,
Chris Hornbeck <chrishornbeckremovethis@att.net> wrote:

> Several audio related devices exhibit zero-crossing
> "flat-spots", regions in their transfer curves near
> zero crossing where the Y axis output is transitionally
> flat. Poorly designed push-pull amplifiers exhibit
> "crossover distortion", magnetic storage has poorly
> characterized related effects, etc.
>
> Question: is this electronic "backlash", to use an old
> mechanist's term, conceptually related (at all?) to the
> well understood issue of quantizing distortion/ error?
> And, if so, could classical dither be considered to be
> a possible remedy?
>

In terms of analog magnetic recording, AC bias does in fact accomplish something
similar to dither. I don't think it is conceptually the same, though, since the
amplitude of the bias signal exceeds the signal amplitude. In A/D dither, the
"bias" signal is a fraction of the LSB in amplitude.

-Jay
--
x------- Jay Kadis ------- x---- Jay's Attic Studio ------x
x Lecturer, Audio Engineer x Dexter Records x
x CCRMA, Stanford University x http://www.offbeats.com/ x
x---------- http://ccrma.stanford.edu/~jay/ ------------x

Archived from groups: rec.audio.pro (More info?)

On Wed, 23 Mar 2005 08:23:13 -0800, Jay Kadis <jay@ccrma.stanford.edu>
wrote:

>In article <nbo141t4b02pf6aslqrbgsi6jn2lhvnere@4ax.com>,
> Chris Hornbeck <chrishornbeckremovethis@att.net> wrote:
>
>> Several audio related devices exhibit zero-crossing
>> "flat-spots", regions in their transfer curves near
>> zero crossing where the Y axis output is transitionally
>> flat. Poorly designed push-pull amplifiers exhibit
>> "crossover distortion", magnetic storage has poorly
>> characterized related effects, etc.
>>
>> Question: is this electronic "backlash", to use an old
>> mechanist's term, conceptually related (at all?) to the
>> well understood issue of quantizing distortion/ error?
>> And, if so, could classical dither be considered to be
>> a possible remedy?

The analogy between tape bias and dither has been mentioned here
before. What these all appear to have in common is the system has one
or more "dead bands" and adding the appropriate "non-signal" to the
signal reduces or eliminates this dead band(s) in the output.
Magnetics and push-pull amplifiers have one fairly large dead band,
and digital systems have lots of tiny dead bands, one between every
two consecutive digital values.

>In terms of analog magnetic recording, AC bias does in fact accomplish something
>similar to dither. I don't think it is conceptually the same, though, since the
>amplitude of the bias signal exceeds the signal amplitude.

It depends on how you define "conceptually" (sorry, I always feel a
like President Clinton saying that...) - at the basic level of "an
added signal fixes a dead band" they are the same.

>In A/D dither, the
>"bias" signal is a fraction of the LSB in amplitude.

It's actually a LARGE fraction, roughly equal to one LSB, though
it's still quite small relative to the peak signal, whereas (as you
say) tape bias is quite large relative to the signal.
Digital systems have very small "dead spots" whereas ferromagnetics
have one very large "dead spot." The magnetic dead spot is unique in
that it MOVES in the direction you want to magnetize something (the
hysteresis), so the bias (or dither) signal needs an even higher
amplitude to compensate for it. Basically, the bias/dither signal is
made large enough to get rid of the dead band.

I've not heard of a mechanical system that uses vibration
(coparable to dither or AC tape bias) to get rid of a deadband, but
I've seen gears for variable capacitors (used for tuning radios, where
backlash is very undesirable) where the larger gear is actually two
gears side-by-side with a spring pushing the teeth on to both sides of
each pinion tooth (hope I described that adequately). This is
analogous to a "correctly biased" class AB push-pull amplifier
circuit.

>-Jay

-----
http://mindspring.com/~benbradley

Archived from groups: rec.audio.pro (More info?)

In article <66b341lm0bf4sldl2ollgkc2e6l0ch0rn0@4ax.com>,
Ben Bradley <ben_nospam_bradley@frontiernet.net> wrote:

> >In terms of analog magnetic recording, AC bias does in fact accomplish
> >something
> >similar to dither. I don't think it is conceptually the same, though, since
> >the
> >amplitude of the bias signal exceeds the signal amplitude.
>
> It depends on how you define "conceptually" (sorry, I always feel a
> like President Clinton saying that...) - at the basic level of "an
> added signal fixes a dead band" they are the same.


The difference is that the dither becomes part of the signal while the bias does
not. While it's a significant difference, I'm not sure if it rises to the level
of "conceptual", either.

-Jay
--
x------- Jay Kadis ------- x---- Jay's Attic Studio ------x
x Lecturer, Audio Engineer x Dexter Records x
x CCRMA, Stanford University x http://www.offbeats.com/ x
x---------- http://ccrma.stanford.edu/~jay/ ------------x

Archived from groups: rec.audio.pro (More info?)

AC bias used in tape recording is not at all like DC bias used in
amplifers.
Tape recording is already bi-polar, the AC bias is needed to linearize
it.

Dither is not based on audibility, dither is used in many other
non-audio related devices to linearizse the effects of quantizing.
Dither is NOT just masking the distortion with noise. Dither
fundamentally does improve the linearity, not just make the distortion
inaudably by masking it with noise.


Mark

Archived from groups: rec.audio.pro (More info?)

On Sat, 26 Mar 2005 04:18:32 GMT, Les Cargill <lNOcargill@cfl.Arr.com>
wrote:

>Really, crossover distortion is best managed from first principles
>- match the components, and it goes away. Provide enough current,
>and good slew rate.
>
>*Is* there crossover distortion in modern, post-statistical QC
>electronics? I think it's extinct.

Thanks for your comments. Crossover distortion in electronics
isn't fundamental, but economic pressures are.

But that's not my interest; iron-cored transformers exhibit a similar
issue, and I'd like to learn about their B-H zero-crossing flat-spot
issues.

But, much more importantly, the conceptual problem puzzles me.
The math approach is apparently over my head, and I've approached
the one person I know who could attack it that way, and he
didn't have an off-the-cuff answer.

FWIW, he suggested I ask Randy Yates. Are you out there, sir?

Thanks again to all for their input. All is valued.

Chris Hornbeck
"Excuse me, since when is getting paid for it not USING the property?"
-Bob Olhsson

Archived from groups: rec.audio.pro (More info?)

On 25 Mar 2005 21:03:28 -0800, "Mark" <makolber@yahoo.com> wrote:

>AC bias used in tape recording is not at all like DC bias used in
>amplifers.
>Tape recording is already bi-polar, the AC bias is needed to linearize
>it.

Some early tape recorders actually used DC bias. Not what
you mean, of course.

DC bias is also used to attack an iron-cored transformer's
B-H zero-crossing flat-spot in the oft-maligned "single
ended triode" amplifier philosophy.

But the very large signal "bias" of magnetic tape recording
has *got* to be fundamentally different from classical dither,
doesn't it? Yet another question beyond my reach.


>Dither is not based on audibility, dither is used in many other
>non-audio related devices to linearize the effects of quantizing.
>Dither is NOT just masking the distortion with noise. Dither
>fundamentally does improve the linearity, not just make the distortion
>inaudably by masking it with noise.

At first blush, I'd guess that you and Les are saying the same thing
in two different ways, but I'll bow out for comments. Actually, that
topic should be a whole thread; it's great stuff.

Much thanks for everyone's comments,

Chris Hornbeck
"Excuse me, since when is getting paid for it not USING the property?"
-Bob Olhsson

Archived from groups: rec.audio.pro (More info?)

Chris Hornbeck wrote:

> < ...snip.. >
>
> But that's not my interest; iron-cored transformers exhibit a similar
> issue, and I'd like to learn about their B-H zero-crossing flat-spot
> issues.
>
> But, much more importantly, the conceptual problem puzzles me.
> The math approach is apparently over my head, and I've approached
> the one person I know who could attack it that way, and he
> didn't have an off-the-cuff answer.

Chris,

I've done a little thinking on mechanical analogies and may have come up
with
some useful visualizations. First off, you seem to be dealing with 3
different cases:

[1] a quantized system
[2] a dead band system
[3] a threshold system

Starting backward,
Case [3] the threshold system (ie: B-H crossing; magnetic hysteresis)
may be thought of as a wood block (or maybe a clutch) with static
friction. At zero
velocity it takes extra force to get it to move. When the velocity drops
below threshold
the block sticks and requires extra force to get it to move again.

Case [2] the dead band system: imagine a clothesline loop with a fixed
line below it. Picture
a slider on the fixed line and two clothespins on the loop with the
pointer between.
The distance between the clothespins is your dead band.

Case [1] a quantized system: Again, two clotheslines as above but this
time the fixed line
has equally spaced bumps that force the pointer to snap between the bumps.
In this case
the clothespins are close together.

Now, apply dither to the above cases and average appropriately. (Note that
in case [3]
you'd only apply dither below the threshold)

Later...

Ron Capik
--

Archived from groups: rec.audio.pro (More info?)

On Sat, 26 Mar 2005 05:01:06 GMT, Chris Hornbeck
<chrishornbeckremovethis@att.net> wrote:


>FWIW, he suggested I ask Randy Yates. Are you out there, sir?

Randy is a regular on comp.dsp, look/ask for him there. I've only
seen him here on RAP in crossposts.

>Thanks again to all for their input. All is valued.
>
>Chris Hornbeck
>"Excuse me, since when is getting paid for it not USING the property?"
> -Bob Olhsson

-----
http://mindspring.com/~benbradley