source file: m1491.txt Date: Fri, 31 Jul 1998 05:56:26 -0500 Subject: Re: Tuning Spectra program. From: Gary Morrison Somebody asked me a question stemming from a discussion of the instrument spectra I posted a few days ago. I suspect that some of the rest of you might want to ponder the following comments as well. > Differences in timbre are accounted for by the attack transient (which is > what makes realtime pitch detection such a bear) and the relative > strengths of harmonics in the sound. Correct? For a given tone from that instrument, yes, those are definitely the main factors. But I think that there are two factors that are equally important: The patterns of timbres (as defined by those attributes) across the entire instrument, and how the instrument is performed. Along the first lines - or perhaps both lines - I have recently become aware of a curious consideration with regard to synthesizing purely "fantasy" timbres - timbres not intended to resemble anything we're particularly familiar with: It's a whole lot more difficult to do than most of us realize. Synthesizer programmers tend to think of the various instrumental timbres as pillars of recognizability in a vast sea of timbral possibilities. I've found that the opposite is much closer to the truth: Viewed as individual tones, instrumental timbres are *anything but* narrow pillars in timbral space. There is instead a huge amount of variation in what we would call, for example, a clarinet tone, and that varied gamut overlaps profusely with flute, saxophone, trumpet, and even oboe tones. So our ears use several other cues to identify an instrument, not the least of which is its timbral register content. As an example, the fact that saxophones have an especially prominent dulling of its timbre upon crossing from notated C# to D (second space to second line) is a very big clue to our ears that we're hearing a saxophone. An instrument losing high-harmonic content in their upper registers is a pretty strong clue that you're more likely listening to a wind instrument than a string instrument. A closely-related, but probably even more important characteristic in identifying an instrument, is its range. Probably the deadest of giveaways to suggest that you're listening to an electronic simulation of an instrument rather than the proverbial real thing is playing a note out of its range, especially below its range, since the lower-end of most instruments' range is more clearly defined than the upper end. This is one of the very few places where a pianist actually has a disadvantage in composition: They're used to a continuous gamut spanning almost our entire range of hearing. Instrumental range limitations seem difficult to accept, whereas to performers of orchestral instruments, that's just a normal fact of life. Another related rare case where pianists seem to be at a disadvantage in composition, is in the concept of transposing instruments. I remember it once taking about five minutes to explain the concept to one of the best musicians - a lifetime pianist - in one of my ear-training classes. In her defense, I should point out that she understood what I was describing almost immediately, but promptly dismissed that idea as almost surely a misunderstanding. When she finally realized that she did indeed understand my suggestion correctly the first time, she spent a few minutes convinced that I kidding. Thankfully only occasionally, I sometimes hear another rather amusing dead giveaway that you're actually using a sampled simulation of an instrument: Failure to stick to a fixed complement of instrument parts. I have heard, in what was intended as an orchestral or small-ensemble simulation, duet of ... flutes, say ... suddenly blossom into four or five just for couple notes or so. You can semiarbitrarily add notes to fill in harmony at will on a keyboard instrument, but there's just no way that a symphony orchestra, or most other ensembles, will hire three or four extra flutists just to play a note or two here and there! You have to think in terms of a predefined combination of monophonic parts from the start. But anyway, another consideration that can make a lot of difference in deciding what instrument we're listening to is how we play the instrument. An extreme case of this is perhaps the Rhodes-style electric piano. At its theoretical basis, there's no reason at all why it should sound even mildly reminiscent of a real piano: As I understand it, they use - essentially - tuning forks, whereas real pianos of course use courses of strings. But even though piano strings produce very complex overtone structures and tuning forks put out almost exact sinewaves, the fact that it's being played stylistically similarly with a piano reminds us so much of a piano that we're fooled into thinking that it sounds kind of like a real piano. Granted, another big part of that electric piano illusion is due to the fact that the two instruments' attack and envelope characteristics are similar. Even still though, a surprising amount of what we identify as characteristic of a particular instrument are embodied in playing style. Often these are things that go easily unnoticed: As another example, classical brass playing style doesn't allow for much vibrato. The same is true for classical clarinet style, but vibrato is fine for other woodwinds. The result is that a saxophone can sound surprisingly brass-like by simply playing a bright tone with no vibrato. Since saxophone technique normally uses plenty of vibrato, when you forego it, the effect starts to remind us more of a brass instrument. The really unfortunate part of all this is that there's not very much room for interpretation in our audience's ears. You might be tempted to say, "hey, I don't have to stick with exact simulations of acoustic instruments; I'd like to instead use slight variations on them, like a violin without vibrato so as not to pollute the beatlessness of just intonation". I reeeeeeally don't recommend that approach! People are extremely picky about such things. If a tone you create sounds sufficiently like a violin, they'll immediately say "AAAAACCCCK, where's the vibrato?!", and the result will be something that they don't like. In summary then, I've found that, in coming up with electronic "fantasy sounds" it's a lot more difficult than it might seem to *NOT* give people the impression of it being a well-known instrument. And if you invoke that impression, you pretty much need to "go all the way" and reproduce that instrument sound perfectly, or some people will get upset. ------------------------------ End of TUNING Digest 1491 *************************