NetNews Usenet Archive 1992 #27/NN_1992_27.iso/spool/rec/audio/highend/2275

Internet Message Format | 1992-11-16 | 5.4 KB

Path: sparky!uunet!zaphod.mps.ohio-state.edu!darwin.sura.net!sgiblab!sgigate!rutgers!cs.utexas.edu!uwm.edu!csd4.csd.uwm.edu!info-high-audio-request
From: [e-mail] (A. J. Dean)
Newsgroups: rec.audio.high-end
Subject: Re: Why digitize: some bunk
Message-ID: <[e-mail]>
Date: 15 Nov 92 13:13:32 GMT
References: <[e-mail]>
Organization: Electrical Engineering, University of Canterbury, New Zealand
Lines: 75
Approved: [e-mail]
 <[e-mail]>; Mon, 16 Nov 1992 02:13:43 +1300
 16 Nov 92 02:13:36 NZD
NNTP-Posting-Host: 129.89.7.4
Content-Transfer-Encoding: 7BIT
Originator: [e-mail]

Bill Alford ([e-mail]) wrote:
: I've been pondering why it is that I've yet to hear a digital system image
: anywhere near as well as a modest LP system and the conclusion that I've
: come is that the basic assumptions in the current CD standard must be wrong.
: I was reading in a magazine about Pioneer's Legata CD player and it was
: stated, from memory, that the processes involved in human sound location
: (like when we use time differences between the two ears) were equivalent
: to 90KHz! Can anyone supply more information on this, as it ties in very
: nicely with Prof Johnson's HDCD process (as mentioned in Novembers Sterophile
: and the New York Times article for October 25, 1992).

I'm starting to wonder myself, even more than before. I'm not disputing the
mathematical correctness of the fourier transform, just the models we (as in
engineers) use. After all, the ear is essentially a parallel array of
sensors, each (well, some) capable of very accurate resolution of events in
time (if that makes sense). For instance, to represent a tone, nerves can
fire in 'volley' and even the simple process of ORing them together is
enough to recover the oringinal waveform that one nerve alone could not
carry. This happens in some creatures without cochlears (ie the mechanical
frequency differentiating bit) though apparently it happens in humans too
though we rely on it less. Anyway, this parallel arrangement coupled with a
processor can sense higher-bandwidth signals than our frequency sensing
cochlears may be attributed with (it's getting late, excuse wording :-)
Nyquist tells us that a parallel array of say 30000 independent channels
each with a firing rate in the region of 30 to 400 Hz would be able to
detect and discriminate between frequencies up to and beyond 10 MHz!
Obviously this would require each nerve to fire with an accuracy of
1/30000th sec and neurons and stuff to process it accurately, so is a 'bit'
optomistic + my rough theory is, well, rough.

Supporting this extended bandwidth view is the fact that us humans have one
of the poorest upper frequency responses (what we say we can hear) of the
mammals (ie highly developed ears), despite similar ear structure. All the
bones and fluids and bits have a good frequency response well over 20kHz but
we 'can't' hear those sounds. We have one of the best LF responses for some
reason (not entirely size tho).

[Wading into deep water here...] Also supporting this extended bandwith view
is the 'fact' that our cochlear is there to give us better pitch
discrimination and better sensitivity, but it does not rule out some sort of
_additional_ time domain or transient sensitivity - after all, the waves
still travel along the cochlear. Perhaps a different and additional form of
sensing occurs for complex time-varying waveforms at higher volumes. This
means that (a) the sensitivity imparted by the cochlear is no longer that
important in dertermining what gets sensed where, and (b) a complex waveform
is more likely to be picked up as a 'moving pattern' than a single tone. The
volley firing of say a 40kHz tone travelling along the cochlear _may_ be
picked up as a random firing of the sensors, while 40 kHz frequency
components of a complex pattern (ie including lower frequencies) may
eventually be noticed after it has travelled far enough. Meaning - hearing
is no longer limited to 20kHz... A bit far fetched maybe (my references were
a bit old) but perhaps something similar happens in a diluted form which
allows us to hear even the tiniest amount of this 'out of band' info under
certain circumstances.

While I'm at it I might as well postulate that we hear a lot more than we
counciously realise. After all, we are primarily visual creatures. Our
brains are wired that way, compared to creatures like bats and dolphins, or
cats whose brains are wired 'a bit of both'. Apparently positioning clues
from our ears go to areas in our brain which are closely coupled with the
movement of our eyes, while in dolphins etc it goes to different places
(perhaps linked with visual images themselves). Medial superiory olives or
something. Experiments have shown that the position of a sound can be
pin-pointed very arracurately by eye position, while the listener may be
pretty unsure. To say we are fully concious of everything we sense is a bit
of an arrogant view of how we work I think.

Whoops. This is a bit long, sorry for going on (and on...). But perhaps the
related physics and numbers are best in this group. Along with software
engineering, high-end audio must be one of the worst researched and
understood things I have come across (not intending to insult anyone having
a good bash at it...). Whatever happens, numbers and theory are never going
to change how we hear (just maybe what we hear!).

Antony