Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!accuvax.nwu.edu!ferret.Berkeley.EDU!sandell
From: sandell@ferret.Berkeley.EDU (Greg Sandell)
Newsgroups: comp.music
Subject: Re: Compact Disks with Musical Instruments on them
Message-ID: <6304@accuvax.nwu.edu>
Date: 12 Apr 90 17:54:23 GMT
References: <680@bbxsda.UUCP>
Sender: news@accuvax.nwu.edu
Lines: 58

> Jim Hurley writes:
> >                                                            I suppose
> >that every fourth note was just a concession to space or time constraints
> >on the CD or the performers.
> 
> This is a valid compromise since most of the time
> you can create an *almost* seamless transition from on sample to the
> next.  If adjacent samples are that distinct from each other then
> the sample library itself was poorly created. 

The question is, how much is `too distinct'?  As I pointed out, there
are many changes in character and spectral quality in musical instruments
from one note to the next that are a natural outcome of acoustical
properties (that arise from mechanical and manufacturing problems).
Starting with D4 on the Bb clarinet...*all* Bb clarinets...the notes get
increasingly thin, and the notes F4-G#4 are so thin that this group
of notes has a name ("throat register").  (I am referring to sounding
rather than written pitches, by the way.) Then at A4 there is a 
dramatic increase in fullness, and this is referred to as the break
between the chalumeau (lower register) and the clarino (higher register).
This may be one of the more dramatic discontinuities in the family
of orchestral instruments, but all instruments have them to one degree
or another.  The point is, that these bugs have been turned into
features by composers for a long time, and it's part of the character
of the instrument.

> Samples should be
> as vanilla as possible (with exceptions) so you can add your own
> vibrato and articulation.  The end result might not be *exactly*
> like the original instrument but can be musically pleasing.

I guess sampler users want to `cue' certain sounds..."Oh, that's a
clarinet".  There's a world of difference between that and the 
activities of people at IRCAM, CCRMA, CNMAT, and so on, with their
instrument modeling projects, where the idea is to make as
authentic-sounding a simulation as possible.  But I think that sampler
users and manufacturers ought to recognize how short-lived the
effectiveness of the `cueing' approach is in synthesis techniques.
When FM first came out, and for a few years after the DX-7 came out,
the brass sounds were real enough that a listener would say, "oh, 
that's a trumpet or trombone or french horn or whatever."  But now
it's just a dated, phony-sounding imitation brass sound, and people
expect something better.  Whether we know it or not, we are all pupils
in a massive ear-training project.  Each time a new synthesis technique
comes out, it has a lifetime of its own, governed by how long users
are willing to accept it as being lifelike and realistic...and the
outcome is that our ears are getting more sophisticated and demanding.

Please don't flame me for concentrating on the re-creation of natural
sounds...that's how this discussion began.  Many of the issues of what
makes a imitation of an acoustic instrument lifelike and pleasing is
directly applicable to what makes a newly-created sound interesting.

Greg Sandell
****************************************************************
* Greg Sandell (sandell@ils.nwu.edu)                           *
* Institute for the Learning Sciences, Northwestern University *
****************************************************************