QUOTE(SiriusCG @ Dec 13 2005, 08:31 PM)
The secret is to get multiple clean samples throughout the range of the instrument, ie; getting C3, C4, C5... then "layering" them, at least in the case of soundfonts, together.
For an ideal soundfont (or other sample-based library), you should sample chromatically -- or at most a third between samples, to minimize formant artefacts. If you're doing a stringed instrument, you'll probably also want to make sure you have one sample for the open string, and other samples for the fretted/fingered notes immediately above and below them, for greater realism.
For an even better soundfont, you should also create samples for each note at several different velocities, normalize the samples, then layer each note based on a velocity map: e.g.
if you sampled four different velocities for each note, then velocities 0-31 trigger the "softest" sample (I put it in quotes because once you normalize the samples, they may be roughly the same volume), 32-63 trigger the next loudest sample, 64-95 trigger the next loudest, and 96-127 trigger the loudest sample.
This way, if you're entering the MIDI data with a velocity-sensitive keyboard, then how hard you strike the key will change the character of the note. So if you're doing a string bass, you can get a light pluck by lightly pressing the key, or a nice slap if you really hit it hard. The more velocity layers you have, the more "alive" it will sound when you play it.
Of course, using multiple samples for each chromatic note will make for a huge soundfont, but with memory as cheap and plentiful as it is today (and with the advent of VST soundfont players such as sfz
, which can stream the data "direct from disk" to conserve memory), that's not really as much of a problem as it once was.