The Almighty Sampler
“There are £250,000 worth of uncleared samples on my latest album. Get listening!”- Norman Cook (Fatboy Slim)
The introduction of the sampler and sampling in the 1980’s became responsible for one of the biggest technological and creative shifts in the history of music. However, the audio workstation and in particular digital audio editing have overshadowed what was originally viewed as a marvel of modern technology.
Since it is now possible to cut, slice, move, stretch, compress and edit audio directly in the workstations editor, what was once limited to a sampler has become commonplace within a workstation. What’s more, with all sample CD’s now in WAV and AIFF format, audio can be imported direct into an arrange page and sample-based players such as Native Instruments Kontakt catering for a vast range of exotic multi-sampled instruments, practical sampling has taken something of a backseat.
However, the humble sampler still has a lot more to offer than the basic editing functions that can be performed in a workstation. And with a growing number of artists beginning to rely on step sequencing samplers such as the Rhizome and the Swedish Elektron Octatrack, an understanding of samplers, how they operate and some of their creative uses is still as paramount today as it’s ever been.
A sampler can be viewed as the digital equivalent of an analogue tape recorder but rather than recording the audio signal onto a magnetic tape, it is recorded digitally into random access memory (RAM) or directly onto the computer’s hard disk drive. After any audio signal has been recorded, it can then be manipulated using a series of editing parameters similar to those on synthesizers. This includes amplitude and filter envelopes, LFOs and in some units, the ability to mix synthesizer style oscillator with the sample. In addition, any sampled sounds can also be played back at varying pitches. To accomplish this, the sampler artificially increases or decreases the original sample frequency in relation to the note that is depressed.
This means that if the sample’s speed is increased or decreased by a significant amount it will no longer sound anything like the original source. For instance, if a single key strike from a piano is sampled at C3 and is played back at this same pitch from a controller keyboard, the sample will play back perfectly. If, however, this same sample is played back at C4, the sampler would increase the frequency of the original sound by 12 semitones (from 130.81 Hz to 523.25 Hz), and the result would sound more like two spoons knocking together than the original recording.
These extreme pitch adjustments are not always a bad thing, particularly for a creative user, but to recreate a real instrument within a sampler, it must be sampled at every few notes. Indeed, most samplers only reproduce an acceptable instrument sound four or five keys from the original root key so if an original instrument is needed throughout the key-range samples should be taken every few notes of the original source instrument. For example, with a piano it is prudent to sample the keys at C0, E0, G0 and B0, then C1, E1, G1, B1 and so forth until the entire range has been sampled. This technique is known as ‘multi-sampling’ and has occurred on every sample-based instrument available.
Naturally, recording sounds in this way will equate to a large number of samples, particularly if they are also recorded at a number of different velocities and settings. This means that the more accurate the reproduction, the more samples that are required and the more that are taken, the larger the file becomes and the more memory the sampler must have available to play these back.
Memory is rarely a problem today inside a computer since most sample-based instruments will stream direct from a hard disc but in many of the step-based sequencers; memory can still be a concern. Because these samplers hold the sounds in their onboard RAM, the maximum sampling time is limited by the amount of available memory.
At full audio bandwidth, one minute of mono recording will use approximately 5 megabytes (MB) of RAM. In the case of sampling a keyboard instrument in its most basic form, this could equate to 80 MB of memory, and you can double this if you want it in stereo. Consequently, various techniques have been adopted to make the most of the available memory, the first of which is to loop the samples.
As the overall length of a sample determines the amount of RAM that is required, reducing the sample’s length means more samples will fit into the memory space. Because most sounds have a distinctive attack and decay period but the sustain element remains consistent, the sustain portion can be continually looped for as long as the key is held, moving to the release part after the key is released. This means that only a short burst of the sustain period must be sampled, helping to conserve memory.
This is more difficult than it may appear, however, and the difficulty arises from what may appear to be a consistent sustain period of a sound is rarely static due to slight yet continual changes in the harmonic structure. If only a small segment of this harmonic movement is looped, the results would sound unnatural.
Conversely, if too long a section is looped in an effort to capture the harmonic movements, the decay or some of the release period may also be captured, and again, when looped, the final sound will still be unusual. In addition, any looping points must start and end at the same phase and level during the waveform. If not, the difference in phase or volume could result in an audible click as the waveform reaches the end of its looped section and jumps back to the beginning.
Some samplers offer a work around to this and automatically locate the nearest zero crossing to the position you choose. While this increases the likelihood that a smoother crossover is achieved, if the waveform’s level is different at the two loop points, there will still be a glitch. These glitches can, however, be avoided with crossfading.
Using this, the operator can fade out the end of the looped section and overlap it with a fade in at the start of the loop. This creates a smooth crossover between the two looping points, reducing the possibility that glitches are introduced. Although this goes some way to resolve the problems associated with looped samples, this is not always the best solution because if the start and end of the looped points are at different frequencies there will be an apparent change in the overall timbre during the crossfade. Unfortunately, there is no quick fix for avoiding the pitfalls associated with successfully creating a looped segment, so success can only be accredited to patience, experimentation, and experience.
Excerpt from Dance Music Manual: Tools, Toys, and Techniques by Rick Snoman © 2013 Taylor & Francis Group. All Rights Reserved.