Re: [ARSCLIST] Listening Test - was: Sampling Theory (was Fred Layn's post on the Studer list re: Quantegy)

["Steven C. Barr" <stevenc@xxxxxxxxxxxxxx>]

Wouldn't there be problems with "false signals" created by heterodyning as
wave frequencies approached simple rates, or their harmonics and
subharmonics?
It seems to me if you sampled a 11025 kHz sine wave at 44.1, you would get
the same four levels at every sample, which would depend on phase
differences
between the two but would repeat as long as the wave was sampled (and would
apply to any number of possible analog waveforms!). If you try an 11030 kHz
sine wave, the sampling points (and thus the sample levels) will slowly
drift out of alignment, generating a 5 Hz component that isn't actually
there...right?
Steven C. Barr

----- Original Message -----
From: "Dave Bradley" <db65@xxxxxxxxxxx>
To: <ARSCLIST@xxxxxxxxxxxx>
Sent: Wednesday, January 19, 2005 8:29 PM
Subject: Re: [ARSCLIST] Listening Test - was: Sampling Theory (was Fred
Layn's post on the Studer list re: Quantegy)


> Hi Eric,
>
> I will reply to your private e-mail shortly, but for now the public one.
>
> >I know an analog function generator does not produce music,
> >but it does produce a very repeatable signal for basic
> >listening tests, and as it is analog, it contains all the
> >harmonics that can affect timbre.
>
> Just because it is "analog" does not mean it contains harmonics. If you
are
> playing it through a loudspeaker and testing that output with a
microphone,
> then there may be harmonics, but if you are simply generating a tone with
> the generator and it's direct wired into your testing equipment, then
there
> won't be harmonics unless the device is designed to create harmonics
> intentionally.  A sine wave isn't a sine wave if there are harmonics mixed
> with it.
>
> >At 10 kHz, the sine and triangle were not clearly differentiated
> >by their sound, but rather by their intensity or volume level.
> >The triangle being slightly louder than the sine.  The square
> >still had a distinct timbre to it that was different from the
> >sine and triangle.
>
> Which disproves the issue of a 10 kHz square wave with 44.1 kHz sampling
at
> 16-bits that you had originally written about.  For the record, after
> receiving some clarification from Eric about the choice of the 10 kHz
> frequency after writing to him about a test I did, I tried the test with a
> 20 kHz sine wave and a 20 kHz square wave (I didn't try this particular
> test with a triangular wave).  The 20 kHz square wave couldn't be
> reproduced by my pro ADC / DAC (RME-Audio PAD 96) at those settings, but
> could be at 96 kHz / 24-bit.  The sine wave was fine at the lower digital
> resolution.
>
> >At 16 kHz, I could no longer differentiate the sine and triangle,
> >but I could still consistently pick out the square wave by its
> >intensity - but not by its timbre.
>
> Yes, but could you hear a difference between the analog and the digital
> versions of those waves that you could hear?
>
> >To make things more interesting, we repeated the entire listening
> >exercise by monitoring the ADC-DAC output sampled at 88.2 kHz, and
> >the results were the same as the analog listening test!  There
> >may have been subtle differences in loudness and timbre between
> >analog and digital, but that wasn't the goal of this test.
>
> Actually, I would have expected that it was part of the goal of the test
> since the original hypothesis posted on the list was that a 10 kHz square
> wave couldn't be properly sampled at 44.1 kHz 16-bit resolution.
>
> >However, at 44.1 kHz I had much more difficulty differentiating the
> >square wave at 10 kHz - but I still consistently could.
>
> Did you try with frequencies slightly off from 10 kHz?  Like 9,542 Hz as a
> random example?  Just curious if that had any impact?  You could also go
> above that 10 kHz if you think it's a matter of hearing limitations. You
> could do 10,458 Hz instead of 9,542 and see if you could differentiate it
> better or worse.  The reason I propose this is that if there is an actual
> frequency or type of wave form that 44.1 kHz 16-bit can't reproduce
> properly, it's possible that it's mathematical and a slight variation
could
> clarify that.  For example, a sine wave at 22,050 Hz should be able to be
> sampled and reproduced by a 44.1 kHz 16-bit ADC DAC combination, but if it
> couldn't, then try 22,049 and see if that is better.  Nyquist theory and
> all that jazz.
>
> >I could not tell the 10 kHz sine and triangle apart.  At 16 kHz I could
no
> >longer differentiate the square wave from the triangle or the sine,
> >whereas I could at 88.2 kHz sampling.
>
> That may be similar to not being able to distinguish something at 15 ips
> half track analog, but being able to distinguish it at 30 ips half track
> analog.
>
> >The most interesting conclusion for me from this experiment is that
> >the waveform clearly impacts the EQ.
>
> That's one way to describe the effect you witnessed. I'd think it's not
> actually the "EQ" but a matter of the properties of the waveform. The
> reason it's a different waveform is the same reason it sounds different.
> Square waves do definitely have a different sound than a triangle wave or
a
> sine wave.
>
> >If a non-sinusoidal wave or impulse has its waveform significantly
altered
> >in the ADC-DAC chain, then the perceived loudness of those frequencies
> >will change as well, and hence the EQ!
>
> Again, maybe not the EQ, but you are correct that their perceived loudness
> will change, which will change the overall combined sound.
>
>
>
> -----------------
> Diamond Productions
> Specializing in analog tape & film preservation / restoration in the
> digital domain.
> Dave Bradley   President
>