Re: [ARSCLIST] Sampling and bit rates, was 78 Listening tests

[Mike Richter <mrichter@xxxxxxx>]

Scott Phillips wrote:
> 	Perhaps if the same blind tests were run on 8 year old kids they
> would hear the difference. Age and exposure to loud noises take their
> toll on the rest of us. I still think there are subtle interactions with
> 18khz-35khz information that a CD loses. The fact that my ears
> reasonable top end response measures now at 14khz means I'm not going to
> notice much of that at best case. I spent years in studios and was very,
> very careful about loudness... to the point of carrying an SPL meter
> most of the time, along with ear plugs. Until I was in my early 40's, TV
> monitors used to drive me nuts with the squeal.
>
> I'd have to agree personally, the difference between 16 and 24 bit is
> plain to me most of the time... my imagination ?? Hope not.
>
> Scott

The difference does depend on the material and the listener as suggested 
here - and that applies to sampling rate as well as to depth. With older 
material having poor SNR, even 16 bits seems excessive. Perhaps it was 
recognition of the need for greater depth that caused the MPEG to define 
MP3 at 32 bits depth - in that sense, lossy compression can sound better 
than CD-DA if the source is good enough.

I believe I've stood on my soapbox on the question of sample rate 
before, so feel free to stop here.

The 'ear' is a non-linear device. I used to conduct a simple experiment 
on high-frequency response which demonstrated (based on minimal 
understanding of audio engineering) that high-end response does matter.

Say the subject knows his hearing falls off the cliff at 15 KHz. Provide 
him with two oscillators, one generating a square wave, the other a sine 
wave. Set both for 8 KHz and give him control of amplitude and frequency 
of the sine source. His objective: match the two sounds.

Now, we know that there is no difference between the two below 24 KHz. 
And we know his hearing does not reach 24 KHz. So the task should be 
simple. Instead, it's impossible. The two will always sound different. 
Repeat the test with sine and triangle waves and again they will be 
different, but not so clearly. (And, yes, I checked once that triangle 
and square cannot be matched either.)

The ear is sensitive to wave shape at least to an octave below the 
high-frequency cutoff. Given the right material and the right sort of 
reproduction, the effect of 44.1 ksps sampling is evident. 
(Incidentally, take a look at what that sampling produces for 8 KHz 
square waves. The third harmonic is, of course, somewhat over the 
Nyquist limit and the waveform becomes very odd and highly variable as 
the fundamental varies even slightly.)

--
mrichter@xxxxxxx
http://www.mrichter.com/