Hello People,

With SACD and DVD-A out, I've thought a bit about the senseless overkill
these standards promote. In the process, I have acknowledged that there
may be a few people in the world who truly do hear over 20 kHz (1 in 1000?
Anyone got any stats on this?). So the 44.1 kHz of CD may not quite get it
if you account for things like the antialiasing filter rolloff.

So if we were to up the sample rate, where should we put it? And what about
resolution? Should it change as well?

Here's what I've come up with (at least on the back of a napkin): 64 kHz
sample rate, 32-bit data. This should allow for more than ample overhead
in spectrum width and resolution without going completely crazy (as in,
e.g., 192 kHz). It just seems to have a really nice symmetry about it and
utility to it. The SHARC already has 32-bit fixed-point processing capability,
and I can see future DSPs geared toward audio having that same width. If
you can hear about 30 kHz, you're not human. So this oughta cover it, right???

Just pipe dreaming, but was wondering what the group thinks. Comments?
-- 
%  Randy Yates                  % "...the answer lies within your soul
%% Fuquay-Varina, NC            %       'cause no one knows which side
%%% 919-577-9882                %                   the coin will fall."
%%%% <y...@ieee.org>           %  'Big Wheels', *Out of the Blue*, ELO
http://home.earthlink.net/~yatescr

Hi Randy,

Isn't the reason that the sampling frequency has increased, that auditorial
perception goes beyond 20kHz? For pure tone signals 20 Khz is the limit, but
for transients and/or harmonic components the human is sensitive for higher
frequencies?

Cheers
Thomas S


"Randy Yates" <y...@ieee.org> wrote in message
news:36DFb.9119$w...@newsread1.news.atl.earthlink.net...
> Hello People,
>
> With SACD and DVD-A out, I've thought a bit about the senseless overkill
> these standards promote. In the process, I have acknowledged that there
> may be a few people in the world who truly do hear over 20 kHz (1 in 1000?
> Anyone got any stats on this?). So the 44.1 kHz of CD may not quite get it
> if you account for things like the antialiasing filter rolloff.
>
> So if we were to up the sample rate, where should we put it? And what
about
> resolution? Should it change as well?
>
> Here's what I've come up with (at least on the back of a napkin): 64 kHz
> sample rate, 32-bit data. This should allow for more than ample overhead
> in spectrum width and resolution without going completely crazy (as in,
> e.g., 192 kHz). It just seems to have a really nice symmetry about it and
> utility to it. The SHARC already has 32-bit fixed-point processing
capability,
> and I can see future DSPs geared toward audio having that same width. If
> you can hear about 30 kHz, you're not human. So this oughta cover it,
right???
>
> Just pipe dreaming, but was wondering what the group thinks. Comments?
> -- 
> %  Randy Yates                  % "...the answer lies within your soul
> %% Fuquay-Varina, NC            %       'cause no one knows which side
> %%% 919-577-9882                %                   the coin will fall."
> %%%% <y...@ieee.org>           %  'Big Wheels', *Out of the Blue*, ELO
> http://home.earthlink.net/~yatescr

Heureka wrote:

> Hi Randy,
> 
> Isn't the reason that the sampling frequency has increased, that auditorial
> perception goes beyond 20kHz? For pure tone signals 20 Khz is the limit, but
> for transients and/or harmonic components the human is sensitive for higher
> frequencies?

Hi Heureka,

I don't remember ever hearing that. In fact, I've heard the opposite, but
in general. That is, generally, a tone is easier to detect than anything
else. I don't see why it would make a difference whether it's above or
below 20 kHz.

Do you have a reference?

--Randy

> 
> Cheers
> Thomas S
> 
> 
> "Randy Yates" <y...@ieee.org> wrote in message
> news:36DFb.9119$w...@newsread1.news.atl.earthlink.net...
> 
>>Hello People,
>>
>>With SACD and DVD-A out, I've thought a bit about the senseless overkill
>>these standards promote. In the process, I have acknowledged that there
>>may be a few people in the world who truly do hear over 20 kHz (1 in 1000?
>>Anyone got any stats on this?). So the 44.1 kHz of CD may not quite get it
>>if you account for things like the antialiasing filter rolloff.
>>
>>So if we were to up the sample rate, where should we put it? And what
> 
> about
> 
>>resolution? Should it change as well?
>>
>>Here's what I've come up with (at least on the back of a napkin): 64 kHz
>>sample rate, 32-bit data. This should allow for more than ample overhead
>>in spectrum width and resolution without going completely crazy (as in,
>>e.g., 192 kHz). It just seems to have a really nice symmetry about it and
>>utility to it. The SHARC already has 32-bit fixed-point processing
> 
> capability,
> 
>>and I can see future DSPs geared toward audio having that same width. If
>>you can hear about 30 kHz, you're not human. So this oughta cover it,
> 
> right???
> 
>>Just pipe dreaming, but was wondering what the group thinks. Comments?
>>-- 
>>%  Randy Yates                  % "...the answer lies within your soul
>>%% Fuquay-Varina, NC            %       'cause no one knows which side
>>%%% 919-577-9882                %                   the coin will fall."
>>%%%% <y...@ieee.org>           %  'Big Wheels', *Out of the Blue*, ELO
>>http://home.earthlink.net/~yatescr
> 
> 
> 


-- 
%  Randy Yates                  % "...the answer lies within your soul
%% Fuquay-Varina, NC            %       'cause no one knows which side
%%% 919-577-9882                %                   the coin will fall."
%%%% <y...@ieee.org>           %  'Big Wheels', *Out of the Blue*, ELO
http://home.earthlink.net/~yatescr

Randy Yates wrote:
> With SACD and DVD-A out, I've thought a bit about the senseless
> overkill these standards promote. In the process, I have acknowledged
> that there may be a few people in the world who truly do hear over 20
> kHz (1 in 1000?  Anyone got any stats on this?). So the 44.1 kHz of CD
> may not quite get it if you account for things like the antialiasing
> filter rolloff.

As well as the obvious reasons ("people will spend the money", "my dog
likes to listen too", and anything beginning with "it's more complicated
than that")...  Isn't part of the reason to compensate for incompetent
engineers doing a bad job within the limitations imposed on them by the
format?  Children playing too close to the road, if you will.

Moving things out an extra few kHz isn't going to save them; you have to
give them so much space they can't possibly hurt themselves (or the
consumer).

Heureka wrote:
> Isn't the reason that the sampling frequency has increased, that auditorial
> perception goes beyond 20kHz? For pure tone signals 20 Khz is the limit, but
> for transients and/or harmonic components the human is sensitive for higher
> frequencies?

Randy Yates wrote:
> I don't remember ever hearing that. In fact, I've heard the opposite, but
> in general. That is, generally, a tone is easier to detect than anything
> else. I don't see why it would make a difference whether it's above or
> below 20 kHz.

If you'll excuse me for talking all airy-fairy, pseudo-scientific,
metaphysical, and without references for just a moment...

It is a possibility that even if your brain doesn't say "there is a tone
at 21kHz, 30 degrees to your right", it may still consider that
information in relation to other frequencies as part of its complex
analysis and modelling of the world around you.  That tone on its own is
more likely an internal hardware fault than anything created in the
natural world, something which your brain will automatically discard or
rationalise.

So, you may not be able to consciously say that there is or is not a
tone at some frequency above 20kHz, but that tone could still affect
your perception of other sound below 20kHz.  Proving or disproving the
effect would be very difficult, because you're not simply looking for
yes/no answers but changes in perception... perhaps only slight changes
in perception in practiced listeners.

In article FXNFb.5578$I...@newsread3.news.atl.earthlink.net, Randy Yates
at y...@ieee.org wrote on 12/22/2003 21:41:

> Heureka wrote:
> 
>> Isn't the reason that the sampling frequency has increased, that auditorial
>> perception goes beyond 20kHz? For pure tone signals 20 Khz is the limit, but
>> for transients and/or harmonic components the human is sensitive for higher
>> frequencies?
> 
> I don't remember ever hearing that. In fact, I've heard the opposite, but
> in general. That is, generally, a tone is easier to detect than anything
> else. I don't see why it would make a difference whether it's above or
> below 20 kHz.
> 
> Do you have a reference?

i don't really have a reference
(maybe http://www.mytekdigital.com/DSDvsPCMdebate.txt ), but years ago we
were debating on some mailing list or another the merits of DSD (what became
SACD) vs. PCM .  one experiment that some of us were gonna try (and i think
that Bob Katz did) was whether or not we could really hear above 20 kHz.  i
thought that the meaning of the question is synonymous with whether or not
we could really hear the difference between some audio that is filtered with
a virtual brick wall filter (at 20 kHz) and the same audio not filtered.  we
thought that the experiment needed some audio that is sampled at a very high
sampling rate (96 or 192 kHz), has significant content above 20 kHz (so
there is something to filter out), some well designed (flat, flat, flat
below 20 kHz) linear phase FIR LPFs, and a good testing protocol (blind or
double blind, AB or ABX).

while discussing it, there was a lot of talk about that pre-echo issue (from
pass-band ripples) and other transient stuff (that people thought might
leave artifacts, but i dunno how you can talk about removing this stuff
above 20K and not leave transients, that's what it's about!) so we decided
on phase-linear FIR filters that were designed with MATLAB's firls()
(instead of remez()), had damn flat passband (0.002 dB) up to somewhere
around 20 kHz and the transition band between 20 and 22 kHz (so above 22.05
kHz it would be virtually totally wiped out spectrum and eligible for
resampling at 44.1).  i still got the lpf.m file (appended below).  i think
they got a good recording of castanets or something.  and, because of the
dithered quantization that happened in the FIR LPF, they added the same
dither to and requantized the control (unfiltered) sound so we could compare
apples to apples as much as possible.

anyway, Bob Katz (digido.com, who was a bit predisposed that he *would* hear
a difference) was gonna administer the test in his studio to other subjects
and, to his surprise, *he* could not hear it.  so he didn't even do the
test!  i don't know where he is now about DSD, SACD, etc. but i thought that
was interesting.

of course, what would be more interesting is if someone (say a school or
similar institution) with a very modern Pro Tools setup would do that
double-blind AB test with both golden eared and clay eared people (and
everyone in between) as test subjects.  they really should settle, at what
frequency that everything above can be lost (resulting in whatever
transients) and statistically no one can hear the difference.  that should
settle the PCM sample-rate issue once and for all (except for the Monster
Cable disciples).

r b-j



% lpf.m
% Basic Parks-McClellan (or Least Squares) Low-Pass Filter Design

if exist('filename') ~= 1,
    filename = 'lpfcoefs.txt';        % filename for FIR coefs output
end;

% the input parameters are not changed if they already exist

if exist('SR') ~= 1,
    SR = 192;    % sampling rate
end;

if exist('PB') ~= 1,
    PB = 20;    
% top of passband region (where filter should be flat and 0 dB)
end;

if exist('PBR') ~= 1,
    PBR = 0.002;   
% max passband ripple in dB which is twice the max error in passband
end;

if exist('SB') ~= 1,
%    SB = SR/2 - PB;
    SB = 22.5;   
 % bottom of stopband region (where filter should be flat and -inf dB)
end;

if exist('SBG') ~= 1,
    SBG = -150;    % max stop band gain in dB
end;

R = (PB + SB)/SR;

% Number of required FIR taps, from a formula in Oppenhiem & Schafer p. 480
%   the more accurate MATLAB function remlpord() could be used instead.
% for nicety, we always round up to the nearest odd number of FIR taps.
%   when N is odd, the delay (N-1)/2 is always an integer number of samples
    
N = 2*ceil(0.0342416*SR*(-0.5*SBG - 10*log10(10^(0.5*PBR/20)-1) - ...
     13)/(SB-PB) - 0.5) + 1

if 0               
% set this to 1 if you wanna get coefficients for
% windowed sinc LPF (for testing)

    h = zeros(1, N+1);            % zero the whole thang

    t = linspace(-R*(N-1)/2, R*(N-1)/2, N);
    h = [0 (R*sinc(t)) .* kaiser(N, 0.1102*(-SBG-8.7))'] ;
% kaiser windowed sinc()
    clear t;

else
    
    W = ( (10^(0.5*PBR/20) - 1) )/( 10^(SBG/20) )
                  % linear stopband/passband weighting ratio

    f = [0   PB/(0.5*SR)  SB/(0.5*SR)  1];
                  % Define passband and stopband regions
    m = [1   1            0            0];
                  % Define passband and stopband magnitudes
    w = [1                W             ];
                  % Define passband and stopband error weights
    
    
    h = zeros(1, N+1);
    
    h = [0 firls(N-1, f, m, w)];
                  % Compute impulse response from Least Squares
    
%   h = [0 remez(N-1, f, m, w)];
                  % Compute impulse response from Parks-McClellan

end

                % h(1) = 0 and h((N+1)/2) = max
                % symmetry:  h((N+1)/2 + k) = h((N+1)/2 - k)
                %   this should result in zero phase
                %   in the frequency response.
                
H = fft([h([(N+1)/2+1:N+1]) zeros(1, 3*(N+1)) h([1:(N+1)/2])]);
         % fft after zero-padding to extend length by factor of 4
freq = linspace(0, 1-2/(N+1), 2*N+1);
         % set of frequencies from DC to just under Nyquist

plot(h/R, '.');                      % plot scaled impulse response

pause;                               % dB by linear freq plot
plot((SR/2)*freq, 20*log10(abs(H(1:2*N+1))+1.0e-10));

pause;                               % look closely at passband ripple
axis([0 0.5*(SR/2) -0.01 0.01]);

pause;                               % dB by log freq plot, skip DC
semilogx((SR/2)*freq(2:2*N+1), 20*log10(abs(H(2:2*N+1))+1.0e-10), 'g');

pause;                               % look closely at passband ripple
axis([0.05*(SR/2) 0.5*(SR/2) -0.01 0.01]);

outfile = fopen(filename,'wt+');
fprintf(outfile, '%1.8e\n', h(2:N+1));
fclose(outfile);

In article b...@toast.void.nil, Simon Hosie at
g...@learcay.etnay.nzsomething.invalid wrote on 12/22/2003 22:49:

> Heureka wrote:
>> Isn't the reason that the sampling frequency has increased, that auditorial
>> perception goes beyond 20kHz? For pure tone signals 20 Khz is the limit, but
>> for transients and/or harmonic components the human is sensitive for higher
>> frequencies?
> 
> Randy Yates wrote:
>> I don't remember ever hearing that. In fact, I've heard the opposite, but
>> in general. That is, generally, a tone is easier to detect than anything
>> else. I don't see why it would make a difference whether it's above or
>> below 20 kHz.
> 
> If you'll excuse me for talking all airy-fairy, pseudo-scientific,
> metaphysical, and without references for just a moment...
> 
> It is a possibility that even if your brain doesn't say "there is a tone
> at 21kHz, 30 degrees to your right", it may still consider that
> information in relation to other frequencies as part of its complex
> analysis and modelling of the world around you.  That tone on its own is
> more likely an internal hardware fault than anything created in the
> natural world, something which your brain will automatically discard or
> rationalise.
> 
> So, you may not be able to consciously say that there is or is not a
> tone at some frequency above 20kHz, but that tone could still affect
> your perception of other sound below 20kHz.  Proving or disproving the
> effect would be very difficult, because you're not simply looking for
> yes/no answers but changes in perception... perhaps only slight changes
> in perception in practiced listeners.

Simon, that is not what we mean when we say that it is unlikely that we can
really hear above 20 or 22 or 2x kHz.  what we mean is that if *all* of the
content above that threshold was removed (or some spurious content added, at
reasonable levels), that we could not hear the difference.  that is, will a
brick-wall LPF make a sonic difference?

r b-j

Simon Hosie wrote:

> Heureka wrote:
> 
>>Isn't the reason that the sampling frequency has increased, that auditorial
>>perception goes beyond 20kHz? For pure tone signals 20 Khz is the limit, but
>>for transients and/or harmonic components the human is sensitive for higher
>>frequencies?
> 
> 
> Randy Yates wrote:
> 
>>I don't remember ever hearing that. In fact, I've heard the opposite, but
>>in general. That is, generally, a tone is easier to detect than anything
>>else. I don't see why it would make a difference whether it's above or
>>below 20 kHz.
> 
> 
> If you'll excuse me for talking all airy-fairy, pseudo-scientific,
> metaphysical, and without references for just a moment...
> 
> It is a possibility that even if your brain doesn't say "there is a tone
> at 21kHz, 30 degrees to your right", it may still consider that
> information in relation to other frequencies as part of its complex
> analysis and modelling of the world around you.  That tone on its own is
> more likely an internal hardware fault than anything created in the
> natural world, something which your brain will automatically discard or
> rationalise.
> 
> So, you may not be able to consciously say that there is or is not a
> tone at some frequency above 20kHz, but that tone could still affect
> your perception of other sound below 20kHz.  Proving or disproving the
> effect would be very difficult, because you're not simply looking for
> yes/no answers but changes in perception... perhaps only slight changes
> in perception in practiced listeners.

A test is easy. In the continuous presence of an audible tone, turn a
near-ultrasonic tone on and off. See if you or others can tell when it's
on. Traditional theory implies that even marginally audible high tones
will be masked by the lower ones, but there have been reports to the
contrary. Wishful thinking may be the root, but any claim is weak
without a test. Do we want to allow the existence of a phenomenon which
we agree can never be manifest?

Jerry
-- 
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

"Randy Yates" <y...@ieee.org> wrote in message
news:36DFb.9119$w...@newsread1.news.atl.earthlink.net...
> [...]
> Here's what I've come up with (at least on the back of a napkin): 64 kHz
> sample rate, 32-bit data. This should allow for more than ample overhead
> in spectrum width and resolution without going completely crazy (as in,
> e.g., 192 kHz). It just seems to have a really nice symmetry about it and
> utility to it. [...]

That would be with 8 channels, I presume? How about the regular 5.1 plus up
and down?  You could create whole new markets for speaker ceiling mounts and
sunken floor enclosures. OK, maybe that's a bit much ;-)

64 kHz seems like a good rate, though.  As Simon says, 44.1 is cutting it
too close for comfort, and 64 isn't excessively more if good reproduction is
important to you.  Wouldn't you be afraid to drive over a bridge that was
designed to be only a few % stronger than necessary?  Me too, because I
wouldn't trust the engineers to know everything significant there is to know
about the requirements, just like I don't really believe that we know
everything significant there is to know about hearing.

Matt Timmermans wrote:
> "Randy Yates" <y...@ieee.org> wrote in message
> news:36DFb.9119$w...@newsread1.news.atl.earthlink.net...
> 
>>[...]
>>Here's what I've come up with (at least on the back of a napkin): 64 kHz
>>sample rate, 32-bit data. This should allow for more than ample overhead
>>in spectrum width and resolution without going completely crazy (as in,
>>e.g., 192 kHz). It just seems to have a really nice symmetry about it and
>>utility to it. [...]
> 
> 
> That would be with 8 channels, I presume? How about the regular 5.1 plus up
> and down?  You could create whole new markets for speaker ceiling mounts and
> sunken floor enclosures. OK, maybe that's a bit much ;-)

Yeah. Let's just stick with above-floor systems. ;)

The channel count question is a whole 'nuther issue. I was just focusing on this
sample rate insanity, and to a lesser extent, bit width.

> 64 kHz seems like a good rate, though.  As Simon says, 44.1 is cutting it
> too close for comfort, and 64 isn't excessively more if good reproduction is
> important to you.  Wouldn't you be afraid to drive over a bridge that was
> designed to be only a few % stronger than necessary?  Me too, because I
> wouldn't trust the engineers to know everything significant there is to know
> about the requirements, just like I don't really believe that we know
> everything significant there is to know about hearing.

But even in bridges and such, how much is enough? 10x? 100x?
-- 
%  Randy Yates                  % "...the answer lies within your soul
%% Fuquay-Varina, NC            %       'cause no one knows which side
%%% 919-577-9882                %                   the coin will fall."
%%%% <y...@ieee.org>           %  'Big Wheels', *Out of the Blue*, ELO
http://home.earthlink.net/~yatescr