Sampling Classic Electro Sounds
Nothing seems to be more demanding of a sampler than the sound of the bass drum on a TR808 drum machine. Why should a piece of primitive retro-tech pose problems for so many respectable and apparently ultimate quality machines? Wilf Smarties explains.
The simpler the signal, the harder it is to disguise the digitisation and aliasing that are part and parcel of the sampling process, and not many drum machines produce sounds lacking in harmonic content to the extent of the good old 808. (Incidentally, simple sounds also show up deficiencies in analogue equipment. A 1 kHz sine test tone recorded and played back on a poorly maintained tape recorder will sound unbelievably bad, even when the machine seems to work fine with music.)
Only the very best samplers, like the best recorders, pass the test of reproducing simple waveforms with flying colours, and only equipment of this quality will be able to directly replace analogue synthesizers and drum machines without impairing the mix.
There are many reasons for wanting to sample electronic instruments. The most obvious one is simply to extend your sound palette beyond the confines of your inventory, but there may be good reasons for sampling even toys in your possession. A monophonic (as opposed to monaural) synth may be sampled in order that you can play it polyphonically. A simple source can be layered to provide a thicker sound. A monaural source can be synthesised into stereo in various ways, and you can 'store' particularly juicy edits for those instruments which don't have their own patch memory facility.
Drum machines are prime candidates for sampling, since seldom (if ever) do the samples have to be looped. Most drum machines don't really 'do' much (sequencing apart), and so long as sound quality is maintained you don't lose much by using samples rather than the machine itself.
Be it a saxophone or whatever, a sampled instrument needs to be played in the manner of the original if a fair stab at realism is to be made, and drum machines and synths are no different. For example, I have a set of samples from the Boss Dr. Rhythm (DR55). Treat them with disrespect and the results sound like just another bunch of samples. Program constant hi-hat patterns, and use two fixed levels of dynamic (corresponding to the 'accent' voice on the drum machine), mix the sound in mono from one output (maintaining the relative dynamics of the original's voices), and there you are: step back in time to your second ever electro music experience, (What do you mean 'what was the first?' You mean you never owned a Casio VL-Tone?)
Early drum machines used synthetic sounds (ie. generated by simple synth circuitry) rather than sampled drums, and these provide perhaps the most interesting selection of drum voices for sampling. In many cases several voices shared the same source, and a passable set of 808 tom-toms and congas can be generated from a single well chosen sample, subsequently given a range of tunings and pannings, thus saving sampler memory.
While purists have the option of sticking rigidly to the performance capability of the original instrument, unlimited combinations of drum voices are obviously possible, and therefore you need to organise your samples in such a way as to be able to find the perfect combination of kick drum, snare, hi-hats etc. for a particular song. Rather than limit your options by using drum maps corresponding to the original instrument, there is much to be gained by keeping, for example, library files of kick drums, snares, and tom-tom and cymbal sets, assigned to usual keys, which can then be used to construct a drum kit in the mix.
Let's start with good old analogue. The first, simplest, and (some would say) best type of synthesis around. Whereas in some ways the 'simplicity factor' can ask deep (literally - low frequencies are especially troublesome) and meaningful questions about the sonic integrity of a sampler, it does open up some convenient options. I will begin by quickly recapping on some of the virtues and operational characteristics common to most analogue synthesizers.
Analogue synthesis is based on simple waveforms. Apart from 'noise', these are superb candidates for sampling, and in many cases a well chosen single cycle can be looped to regenerate an exact replica of the original waveform. (Noise can also be successfully sampled, but you need to take a relatively long sample, and loop it carefully, preferably with an alternating direction algorithm.) Only in the case of an LFO pulse-width modified wave need a longer sample be taken, to accommodate one cycle of the LFO frequency.
The second stage in the process of analogue synthesis involves the application of amplitude and filter envelopes to the waveform(s). (Some analogue synths have only a single envelope, which can modulate both amplitude and filter.) If you are lucky your sampler may have a decent filter of its own, but the chances are that it won't sound much like that of the original instrument. Indeed it is the filter which tends to be the prime factor defining the characteristic sound of a particular synth. When sampling a synth, therefore, it is not enough to sample only raw waveforms, although these can be invaluable. A good selection of complete voices is also required.
A third stage involves utilising an LFO to modulate the sound. This can readily be applied by the sampler itself and (apart from PWM) should be removed at the source.
When sampling single cycle waveforms, it is essential that only one oscillator at a time is sampled, and that the effect of the filter is removed. Looping such a sample is painlessly simple, especially if graphic editing is available. It does not really matter if you catch a single, or group of, cycle(s) in the loop: the results will be sonically indistinguishable. Many synth voices use more than one oscillator. Similarly, within the sampler you can stack wave samples to create richer sounds: if two or more oscillators are marginally detuned, you can create long-range chorus effects which belie their origins in very short samples. The sampler's own filters can then be applied to sampled waveforms to provide a range of voices in the manner of a true synthesizer.
Complete synth voices are much more expensive in terms of sample memory used than simple waveforms, firstly because many sounds have to be played until they fade out or stabilise sufficiently to permit successful looping of the sample's tail, and secondly because multi-sampling is almost always required to get a decent usable key width. In general, the more resonant the sound, the higher the number of samples required. Where a voice is dependent on a volume rather than a filter envelope (typically strings), try removing that envelope, sampling, say, C5 for a couple of seconds, looking for a good loop (around half a second upwards) and then re-applying the envelope with the sampler. A sampled envelope will be subject to 'key follow' (it will be shorter or longer depending on what pitch you replay it at), seldom a desirable result with looped samples.
Moving on to FM synthesizers, it turns out that these are excellent candidates for sampling. There is not much point in sampling raw waveforms here since, unless you are using a Synclavier, there is no way that a sampler will be able to combine these waveforms in remotely the same way as an FM instrument. However, the rules for sampling analogue voices apply here, the main difference being that, because digital oscillators tend to be absolutely stable, unlike their analogue counterparts, looping is actually much easier in many cases. For example, whereas an analogue synth string ensemble voice will have a vaguely random chorus running throughout, a similar sounding patch on an FM synth will constitute a complex but precise waveform which periodically repeats itself, offering the possibility of a perfect loop crossover point. No smoothing required.
Beyond FM and analogue synthesis, with contemporary instruments we move into a world where the keyboards are more or less samplers themselves, and the process of sampling these 'synthesizers' is really no different from that for real instruments, and is therefore beyond the remit of this article. In many ways the hardest part about retro-sampling is getting your hands on the original instrument and, even if you do, understanding it well enough to get it to deliver the best it has to offer in the time you have available to sample it!
Feature by Wilf Smarties
Previous article in this issue:
Next article in this issue:
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!