Samples vs. Synthesis

Some customers ask which is better for LinnStrument's expressive control-- samples or synthesis?


The advantage of synthesis is that the sound is fully malleable, so it's easy to control note loudness from finger pressure, pitch from X-axis movement, and timbre (or whatever you want) from Y-axis movement.


If you want traditional instrument sounds, samples are more accurate. However, samples are fixed recordings at a specific pitch, loudness and timbre and are therefore not very malleable under LinnStrument's expressive control. For example:

* Natural velocity response requires switching between different samples recorded at different levels of force,

* Bending the Pitch of a sample more than a few semitones sounds unnatural because you're merely playing the sample faster or slower.

* Y-axis control of timbre doesn't work well because there's not much you can do to alter the timbre of a sample other than filtering it.

When playing LinnStrument, samples work well for piano or plucked-string sounds because there is no need for pressure or (usually) Y-axis control. In fact, all of the guitar sounds in my (Roger's) sounds file for Logic or MainStage are samples from the built-in Logci/MainStage sample libraries, but optimized for LinnStrument's expressive control.

But for sounds like solo winds or solo bowed strings that require continuous gestural control of pitch, loudness and timbre, samples don't work so well. Why? Because you need a different sample for each pitch range (perhaps every 3 semitones), and for each pitch sample, you need multiple velocity and pressure samples. And if you want continuous timbre control, you'd need multiple timbre samples for each of the velocity/pressure samples of each pitch range sample. That adds up to a lot of samples, and the transitions are usually noticeable. That said, if you only want ensemble winds or bowed strings, that usually don't require much if any expressive control, so sample libraries work pretty well for ensemble play.

Also, sample libraries tend to assume you're playing from a conventional MIDI keyboard's on/off switches, so the sample recordings usually include the performance gestures like vibrato, pitch slides, fast/slow attacks, etc, that you presumably bought your LinnStrument to perform in real time with gestures. For example, a violin sample recording may include performed vibrato or a slow bow attack, which you don't want when playing LinnStrument; you want to perform these gestures.

The best of both worlds

There's a third choice; physical modeling, led by the Audio Modeling company. Their bowed strings are mathematical models of bowed string instruments, and are so perfect that you'd swear they are samples. But unlike samples, they respond naturally to performed gestures for simultaneously control of each note's continuous loudness, pitch and a variety of timbral variations like bow position. And same for their wind instruments, though some of them use a hybrid of physical modeling and smart transitions between many samples. It doesn't really matter what's under the hood because they all sound perfect and respond naturally and authentically to LinnStrument's gestural control.

The Sample Modeling company is also very good. They use samples and run inside Kontakt, but they have a special technology that more seamlessly transitions between the different samples for pitch and loudness level.