Raimund Specht wrote:
> Yes, ATRAC (and MP3) works very well for human listening. However,
> these systems benefit from the masking effects, that occur in our
> ears. A soft signal will be masked by a louder signal, if it is
> close in frequency (simultaneous masking). Also, a soft signal that
> occurs before and after a louder signal will not be heard (temporal
> masking). The ATRAC algorithm can therefore reduce the bit-depth for
> such masked signals without any perceptible loss of fidelity.
> However, even a cheap microphone would reproduce those signals, that
> are ignored by ATRAC. So, if want to get the entire information
> supplied by the microphone (not only the part, that is perceptible
> to the human ear), compression as ATRAC or MP3 should not be used.
> This is especially important if you want to analyze complex and
> rapidly modulated sounds on a spectrogram display (or if the
> properties of the human hearing system have to be examined).
That is why from the very first time I used ATRAC encoding (many years
ago now) I've looked at the recordings with sonograms. And, you know
what, I hear all this explanation, but don't find the effects assumed to
be there in the sonograms. I do not depend on my hearing to give me the
details of sound, particularly in the upper frequencies. I'm too old to
hear those very well. Try finding what you think ATRAC does using things
like Sonograms. Do that for a few years and you may find it's not what
you thought. I started out many years ago with a lot of worries about
ATRAC, but the longer I've used it and analyzed the sounds recorded with
it the more impressed I am with it.
ATRAC works well for human listening and for sonograms. And probably
lots of other analysis. It's a red herring to try and keep acting like
it's only been looked at by listening to it. Those of us that use it in
science, like myself, have done a lot more to check it out.
Note I do not claim that ATRAC encoded sound can be used for all
scientific purposes, just most of them. My feeling is that while all
this bit about ATRAC is going on that the rest of the variations in
recording are often ignored. Which makes me wonder if accuracy is really
the goal, or is something else at work.
> I agree. As far as I know, there is no evidence, that humans can
> hear sounds above 20 kHz at all. The only argument for sample rates
> higher than 48 kHz seems to be, that the cutoff-frequency of the
> anti-aliasing filter would be not so close to the highest signal
> frequencies (and the low-pass filter can be less steep). This could
> prevent some potential distortions at these high frequencies. I have
> the impression, that many manufacturers of consumer products use the
> new 24/96 or 24/192 capabilities for marketing reasons only (large
> numbers sell).
Indeed, this is the main reason for the use of these higher sampling
rates, marketing, egos. A great many of the folks I see going to this
have not even come close to utilizing the sound quality available at
16bit 44kHz. Going to higher sampling rate will not improve their
recordings because other things are what's limiting.
There is a good reason to carry greater depth when working sound
processing in a computer. Computers are much more inaccurate than most
think. Very few significant figures get through computer calculations,
and many of the processes are chains of calculations, each making even
more error. So, using greater bit depth and even sample rate while doing
software processing does have justification. But it's not necessary to
record at the higher rate to do this.
Higher sampling rate does have a place when you are dealing with
analyzing higher frequencies. For really detailed analysis it may be
better to up the sample frequency for the upper range of human hearing.
You would have more accurate waveforms. I'm not sure it matters much for
listening. Even 20kHz is something that only pretty young humans hear at
all well. By the time we are adults there is considerable falloff of
upper frequency hearing. In terms of absolute sound levels upper range
sounds have to be much, much louder to be heard at all. Our ears are
primarily designed to work well in the frequency range of human speech.
Since the ugly sounds under discussion seem to be heard fairly
universally, I'm thinking it won't be all that high a frequency that
causes the effects.
Walt
________________________________________________________________________
________________________________________________________________________
|