Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/12/84; site nbs-amrf.UUCP
Path: utzoo!linus!decvax!ucbvax!ucdavis!lll-crg!gymble!umcp-cs!nbs-amrf!manheimer
From: manheimer@nbs-amrf.UUCP (Ken Manheimer)
Newsgroups: net.audio
Subject: Re: Re: Re: Tighter bass and edgeless piano
Message-ID: <19@nbs-amrf.UUCP>
Date: Sun, 22-Sep-85 03:46:54 EDT
Article-I.D.: nbs-amrf.19
Posted: Sun Sep 22 03:46:54 1985
Date-Received: Wed, 25-Sep-85 09:17:44 EDT
References: <1636@druxu.UUCP> <16538@watmath.UUCP>
Organization: National Bureau of Standards
Lines: 139

> In article <1636@druxu.UUCP> tlz@druxu.UUCP (ZrustTL) writes:
> >Heh, heh!  Ear training? Flat frequency response with mid-fi or
> >lo-fi results?  What has happened here?
> > [...]
> >Is there another domain of sound reproduction that has escaped 
> >detection by electronic instruments but is be obvious to the ear?  
> >Don't give me this "you can hear it but I don't know why" stuff.
> >I hear that from stereo store sales people and its garbage.
> >Everything happens for a reason and anything that can be heard
> >can be measured.
> 
> probably an infinite number.  all these specifications printed are
> known to correlate with listening quality, but there is nothing that
> says that we've measured all there is to measure.
> 
> >Everything happens for a reason and anything that can be heard
> >can be measured.
> 
> but there is more than just measurement going on here; there is
> interpretation.  a parallel can be drawn with the current research in
> computer vision.  a camera can record and measure far more than the
> human eye can, but all existing computer vision systems fall far short
> of the human visual system because there is an immense amount of both
> information processing and knowledge being utilized to interpret an
> image.  
> 
> the human hearing system is also doing the same type of processing and
> is also using a huge knowledge base.  we cannot help but interpret
> everything we see and everything we hear.  our senses are not objective
> in the sense that instruments are.  we bring too much of our own
> experiences into the listening.  it maps sounds back to something that
> we have experienced before.  that is the reason why the human hearing
> system is poor at differentiation between signals that are extremely
> similar.  it tries to adapt to something recognizable and previously
> experienced.  
> 
> suppose that you have never heard a musical instrument before and have
> never had anyone describe it to you either, for instance, a flute.  how
> would you determine if the reproduction of it is correct purely by
> listening to the recording.  yet someone who has heard it being played
> live may be able to tell you that the person is blowing it from the
> wrong end.  that is the crux, that the person has heard it before.

I feel that the above is a very fine response to the issue raised in the
original article.  However, it seems to me to verge on clearly expressing
a crucial point about the difficulties in parameterizing real world events,
but then the issue seems to become a bit clouded:

> 
> ever notice how different music is when you're at a concert when compared
> to a recording of the same concert being made under the same conditions
> (i.e. same microphone location as yourself and audience).  the differences
> are not in the sound so much as the experience, and that cannot yet be
> reproduced.
> [...]
> 
> Herb Chong...
> 
> I'm user-friendly -- I don't byte, I nybble....
> 
> (will disappear Spetember 30)
> UUCP:  {decvax|utzoo|ihnp4|allegra|clyde}!watmath!hachong
> CSNET: hachong%watmath@waterloo.csnet
> ARPA:  hachong%watmath%waterloo.csnet@csnet-relay.arpa

I think the clouding may be in my interpretation of the notion of the
distinction between experience and facsimile.  This could implies a few
things, among them one that i'd like to emphasize (as briefly as i can)
as my candidate for the guts of the measurability issue.

First, i consider the misleading interpretation of 'actual experience'
vs.  facsimile to be in the incidental, non-sound components of in person
musical experiences - e.g.  the nostalgic odor of perfume/pot wafting
gently in the breeze, the depth of emotion in the performers/audiences/
hot dog vendors expression, etc.  I hope this wasn't the intended idea of
the article - it's too peripheral an argument, there's enough to consider
in the fact that any facsimile is bound to have lost to *some* degree the
sound content of a musical envent without being diverted by the
incidentals (pleasant though they may be).

The issue that holds so much interest for me is the difficulties in
parameterizing, not experience, but the underlaying events.  I make
this distinction because i think that the way people apprehend events is
limited at the onset by parameterization that is a necessary component of
human experiencing.  The catch is that we have the capacity to extend the
finess of our ability to apprehend in this way, meaning that we're
constantly being let down by growing out of previously useful ways of
parameterizations.  This doesn't only happen for individuals; the
refinement and revolution of science represents the same process of
resolution.

What does this have to do with analyzing the accuracy of sound
reproduction, where sound can be entirely captured in terms of varying
sound pressure waves?  Even getting a close aproximation to a 'simple'
event using mathematical function introduces complicated, never entirely
resolved issues of accuracy.  Given that any facsimile is bound to differ
from the original event it represents to some degree, we have to decide
on ways to compare the merits of different variations from the original,
and that's where the difficulties come in.  The frequency content of the
signal does *not* constitute the whole story.

The example herb gave (above) for the misplayed flute comes near to
expressing my point - rather than considering the difference between
players who blows on different (right and wrong ends) of the flute,
consider the difference between an talented intermediate player's and a
talented master's rendition of some specific piece of music - both of
their performances will vary from the piece as intended by the composer,
yet there will usually be a clear consensus in the audience as to
preference for the renditions.  Furthermore, the master's presentation
may, on occasion, diverge more dramatically from what is embodied in the
strict interpretation of the written tablature, yet could still be
apprehended to be closer to the intentions of the composer.  The
imperfection lies in the narrowness of the analytic expression of the
music (the tablature), not in the interpretation of the master.  This
argument may be more metaphoric than analogous to the characterization of
sound by mathematical analysis, but i think it is important.

Easuring the pure tones that a flute player is able to produce is not
sufficient to decide how good the player is.  Furthermore, measuring the
frequency content of the performances isn't going to tell you much about
the relative merits of the performances, because we *don't know how to
characterize the quality of a piece of music in terms of it's harmonic
content*, or even harmonic/ melodic/etc.  progress.  SIMILARLY, we're
limited in our analytic grasp on the cogent features crucial to sounds.
I feel it's in the course of experience that we engage our intuitive
sense to guide us in tuning our analytic sense of what's important, not
the other way around; that increased resolution can be gained by
profiting on the knowledge encoded in the analytic representation, but
the analytic representation generally (always?) has room for extension.

(Pre P.S.  You may notice that this article indicates something about the
meaning of the silly little epithet i always have after my signiture
line.)  

(and cheers)

Ken Manheimer	...!seismo!nbs-amrf!manheimer or manheimer@nbs-amrf.uucp

(Everything leaks.)


Brought to you by Super Global Mega Corp .com