Motivation

Challenging the paradigm of taking sampled drum sounds for granted and using them as, if it is one of the most conventional practices that constitute the framework for creating “electronic music”.

Attempt at Some Historical Excourse

Since when had it became like that?

Sampling one-hit percussion, or, alternatively, recreating it with synthesis, was the primary direction in early electronic music, that made it into consumer sector, since the nature of battery instruments implies repetition, and those sounds are usually short, which made then good candidates to be utilized in first low-memory digital samplers. Since those instruments usually do not have a strongly defined pitch, so the pitch change is not used as a form of expression, when playing those. That eliminates the problem of handling changing pitch of an instrument, during synthesis or having to store a large number of pitches in sampler’s memory. Early rhythm-machines, were mostly built around borderline neglecting the timbral structure in favour of persistence of generated rhythmic pattern, e.g. a sequence of transients.

Description

Overall, the sound of this small library, can be described as something really lo-fi, but still moderately punchy and usable. To put those samples together, I’ve been digging through a whole lot of source material. This tutorial about recreating a snare drum sound with FM synthesis, and this SOS article about cymbal synthesis, were the most important bits of information that kickstarted the whole thing. Rest of this kit’s pieces were done intuitively. Part of those are very basic, like subtractive synthesis, other came out from more adventurous approaches, including resonant filter banks, and layering of found sounds.

Using this lib, I put together a simple gloomy rock tune, what would serve the demonstration purposes. But before we proceed, I feel like I need to put a disclaimer: this collection of sounds it not designed to be high-quality commercial grade product, but rather an exercise, showcasing the exciting possibilities that open up the moment you get off the beaten “pre-cooked drum samples” path.

Takeaways

Sometimes, rhythmic patterns and note length are really more important than the timbres themselves. This sets us back to the origins of rhythm-machines. This aspect of how humans perceive rythms is what made the concept of “programmed drums” wholesome.

So, while trying to nail down the timbres, I should have really been paying more attention to dynamics, for this whole sample patch to be more expressive. Well, there were reasons for that beyond my ignorance. Initially, using a free version of TX Software Sampler, I found a number of expression-related settings to be non-intuitive, while other to have UX and stability issues. No offence to the authors, it might be my DAW’s or my OS problem as well. Probably, my (kill to) cure from that would be trying to do the same workout on Yamaha sampler, so I’d realise how much of privileged “mouse-smudger” I am, using point-and-click software.

Cymbals naturally have their dynamic range very wide. And practically none of the widespread approximations (FM, samples, resonant filters) does cover the desired palette of hues on their own. So, ideally, when trying to approximate a cymbal sound, you would have to change the model/approach when reaching a certain velocity boundary. Like, one FM patch for the lower velocities, another one for the moderate ones, and when, probably some combination of pre-recorded sound and resonant. And since the change of timber could be dramatically obvious, you’ll probably need to apply to crossfading between them as well.

The latter sounds like a whole lot of manual work on any sampler, so, if I ever go down that path, that would be, I’ll have to think about automating the process of laying down the usual 32/64/128 levels of velocity, each from a separate sampled sound.