Re: Toob sound?

gwhite wrote:
> Kevin Aylward wrote:

>>>
>>> I am not aware of any scientific study that conclusively shows that
>>> humans can reliably detect different characteristic distortions
>>> under
>>> 0.5 to 1.0%, for as good or poor as the gross THD spec is (a "paper
>>> spec," as you rightly call it).
>>
>> Well, I have no doubt whatsoever that 0.5% 2nd harmonic sounds
>> different from 0.5% 3rd. I don't see that as being debatable. Those
>> sorts of test have been done. Its imd that thats at issue as well. A
>> 1k and 1.4k generating a new 400hz at 0.5%,
>>
>> You are actully way, way, out there on this.
>
> Assuming there is a threshold of detection, I don't really care if it
> is 1% or 0.1% or 0.01% in absolute terms, since hi-fi amps that meet
> any of these specs are quite easy to come by, and cheap too.

This is besides the point of your lack of knowledge on proof that a
gross thd spec is insufficient, if the thd/imd is too large, especially
for value as large as 0.5%. I do agree, that if the thd/imd is low
enough, differences between them cannot be detected.

>
> If you can hear the difference between two different amps respectively
> rated at "0.5% 2nd" and "0.5% third" in actual music program material,
> then good for you. If you can only detect it with pure tone input
> test signals, it doesn't count. Like I said, I am not aware of I am
> not aware of any scientific study that conclusively shows that humans
> can reliably detect different characteristic distortions under 0.5 to
> 1.0%. Maybe it exists, but where is it?

Well, you should have read the whole post first, because the evidence
for this is everywhere, and you seem to be completely oblivious of what
the files referenced proved.

>
>>> I could never make the claim that
>>> "different types of distortion sound different"
>>
>> I can. Its trivially obvious. 2nd harmonic is twice the fundamental,
>> so a g harmonic sounds likes root g. 3rd harmonic is a d on top of
>> the g. You bet your booty you can hear that extra d if its large
>> enough.
>
> I like your "IF." "[I]f its large enough." Of course, if it is large
> enough. No one is arguing that. You should have know from *context*
> I wasn't talking about some guitar amp deliberately driven to high
> distortion levels. After all, aren't you the guy who wrote "Producing
> an amplifier with say, 0.001% thd/imd at 20Khz, is straightforward?"

Yep.

>
> When I say I can't claim it, I'm talking about hi-fi amps, with the
> characteristically low distortions available. Obviously I can hear the
> difference at some higher level, but I don't know what my own
> threshold is (don't really care). Whether it is "absolutely" 1.0% or
> 0.1% or
> 0.01% is not what I am referring to.

But that's what I am taking issue with your "not aware...at 0.5%
level..etc"

This is blatantly incorrect. You phrased your assertion in such a way as
to imply that there was no reasonable doubt that such distortion levels
could not be distinguished, when in fact, only those pretty much
completely ignorant of the field would imply such a fact.

>
> You can argue some one-size-fits-all figure-of-merit and that 0.5% is
> too high, but I think that misses the main point: Humans likely have a
> threshold of audibility and hi-fi amps are available below the
> theshold of humans.

I don't see that there is a threshold, but there is a smeared limit
range, below which distortion is inaudible. It depends on frequency for
starters.

>
>>> because I am aware of
>>> no evidence that such a statement is true *in general*.
>>
>> I am. There is plenty of evidence for that. Load and loads.
>
> Have out with it. Please constrain your "loads" to hi-fi
> amplification.

I am confining my self to your assertions and implied assertions of
facts. Whether its hi-if or not is irrelevant.

Your implied facts are incorrect. It is those facts I am dealing with.
Stop trying to weasel out by changing the subject.

>
>> Download the files and listen mate. Hint, a 0.1% thd way worse then
>> 12.5% thd
>>
>> http://www.gedlee.com/distortion_perception.htm
>
> Nice. The audio industry is apparently 20 years behind the
> communications and control industries.

But apparently, way ahead of you. It is aware of much of what you
aren't.

>That's impressive. I am not
> universally fond of simplistic THD or intermod (THD, as used, is
> simply a special case of intermod) representations either. That's
> why I wrote: "for as good or poor as the gross THD spec is."

I use the claim that a *real* amplifier with distortion < 0.01% thd/imd
at 20 Kkz, has distortion that is audiable undetectable.

Such an amplifier, invariable has low frequency distortion in the 0.002%
range, just because of the mechanics of amplifier design.

I don't claim that this is a maximum lower bound. It *might* be 0.05%.

>
> _The Volterra and Wiener Theories of Nonlinear Systems_, by Martin
> Schetzen
> ISBN: 0471044555
> http://www.amazon.com/exec/obidos/tg/detail/-/0471044555/
>
> I was looking for papers and references on non-linear system theory in
> the context of predistortion for RF amplifiers some years back.
> Schetzen's text was almost universally referenced in all the papers I
> found. Schetzen studied under Wiener at MIT (it was Wiener who first
> recognized the applications of Volterra's equations for non-linear
> systems). Wiener is also the so-called father of cybernetics. It
> took me about 4 years to find and grab a used Schetzen text off the
> market (at a price commensurate with supply). There are some
> relevent IEEE papers by Schetzen too.

Note that Spice small signal distortion analysis uses the Volterra
model.

>
> In any case, for loudspeakers, Gedlee talks about using the memoryless
> polynomial model of Volterra rather than the complete Volterra model
> which includes memory. That's nice, because not including memory
> effects makes it easy. But my concerns here are mostly about amps. I
> think speakers might be the most difficult problem in hi-fi.
>
> You could use your 0.001% amp and then sum with non-linear blocks
> providing 2nd *xor* 3rd order distortion at 0.5% at rated power
> (before hard clipping), listen to actual musical program material in a
> double-blind test, and prove your claim of being able to hear 0.5%
> characteristic difference with music program material. I almost
> assume you've done this, since it is your apparent claim.

What am I missing here? What are you missing here?

I have just pointed you to files with 0.1% thd, and the distortion in
them was trully dreadfull. The case is closed.

>
> Incidentally, can you think of a problem when applying the
> Volterra-Wiener polynomial model to something like a typical audio
> power amp (basically a powerful op-amp). I can. Maybe that gross
> THD/IMD spec isn't all that bad. Why?
>
>> You need to become a bit more aware, mate.
>
> Oh, for sure. But what is your excuse? You pointed me to a
> reference, which itself referenced something (incorrectly) I had
> already mentioned. What is up with that?

I haven't read hardly any of this thread.


Kevin Aylward
informationEXTRACT@xxxxxxxxxxxxx
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.


.

Quantcast