If imitative synthesis took a major step forward when it incorporated sampled attack transients, why shouldn't a similar approach be used with samples themselves?

With a little bit of time spent editing your existing sample library, it's easy to obtain more expressive samples - when you know how. Here, for those of you who don't, are a few pointers to put more wellie in your winds, more snap in your strings, and more bite in your brass...

It's generally accepted that the first few fractions of a second of a natural musical sound - that part which contains what are called the attack transients - tell the human brain a considerable amount about the sound. These transients tell us, for example, whether an instrument is being blown, bowed, plucked, struck, scraped and so on. Consequently, without them the brain has difficulty in differentiating between, say, a cello and a bassoon or a real trumpet and a synthesised one.

Roland's LA Synthesis, as found on the D50, D70, MT32 and so on, takes advantage of this by combining acoustic instruments' sampled attacks with synthesised sustain tones or short steady-state sample loops. Quite a variety of instruments' characters can be replicated in this way while using minuscule amounts of RAM. Those familiar with the range of Roland's LA sounds will agree that it works, and that even LA's worst examples can still sound infinitely more convincing than imitative synthesis as performed on most "pure" synthesis systems.

An additional point - and one which forms the basis of this editing technique - is that the amount of attack transients, when considered in relation to the sustain portion of an acoustic instrument sound, gives us information as to how loud it is being played. Loud bowed strings are accompanied by more bow scrape on the way into a note, woodwinds and brass more lip, reed or breath noise.

Although the character of an instrument's attack changes drastically with dynamics, there are things that can be done to many of your existing samples that can add extra expression to them. All you need is a sampler that allows you to layer one sound upon another and increase the loudness of one as MIDI velocity increases.

This sample editing trick seems to work most effectively on bowed strings, brass and woodwind - sounds in which the attack transients are considerably softer than their sustained portions. I've tried this on plucked and struck strings and mallet percussion but the effect is not as pronounced. It's worth a try, though; you might just come up with something interesting.

EDITING

The technique is quite simple. All it involves is isolating the attack portion of a sample (maybe the first ¹/_10th to ¹/₄ second of the sound), saving this as a separate sample, then layering it with the original sample in a keyboard "map". With velocity response set up so that the louder you play, the more the attack transient sample comes through, you'll find additional expression can be had from otherwise meek or timid samples. Playing a keyboard hard will add more bite to the front end of a note.

To avoid any phasing problems, don't mess around with either sample's start points unless any changes made are identical to both. Of course, since your attack samples will be "one shot", make sure loops are turned off.

PROFESSIONAL TOUCHES

Those of you with the appropriate facilities will want to maximise attack samples' amplitude to their fullest for the best signal-to-noise ratio and fade their endings so that there isn't an abrupt change in level as your attack samples reach their end (although you might decide you like this effect - try it).

With attack samples maximised, you'll probably not want them to be louder than 25% the volume of the "straight" sample when played at full MIDI velocity, but let your ears be the judge as to the ratio most suited to your needs.

When dealing with multisamples, keep in mind that lower notes generally take longer to reach their steady state than higher notes, and this must be taken into consideration when making your attack samples - lower-pitched samples' attacks will probably need to be a bit longer than the higher ones.

The only pre-requisite to this simple but effective and little-used technique is that something has to be there to work with in the first place. For example, many bowed string library samples are little more than sustain loops with no bowed attacks anywhere in sight. Instead they rely upon amplitude envelopes to fade them in from nothing and die away in the same manner. But for those samples that do have attacks present, using this technique you'll find the additional chuff of flutes and piccolos, "burps" of bassoons, oboes, clarinets and saxophone reeds, and lip "blurts" of brass instruments really will help these samples come across as more punchy and poppy in your music the louder you play. Check it out.