There can't be many sample libraries that do not contain at least a modest selection of analogue synthesizer samples. They are often tagged with such optimistic titles as 'Moog-142' or 'Oberheim-27' and vary in quality from the truly awesome to the truly awful. Many of them are memory hungry and often do not transpose particularly well, perhaps due to oscillator detuning or vibrato captured from the original sound source. One alternative to putting up with such dodgy material is to create your own, but sampling analogue synths can be a very frustrating experience. Here I'll make some suggestions which may help you get better results from your efforts.
By their very nature, analogue synth sounds are at their best when gradually changing over time, maybe as a result of filter sweeps, deliberate oscillator detuning, or (more likely) somewhat less deliberate tuning instability! One solution might be to make a long sample (I'd consider a 'long' sample to constitute a minimum of five seconds) in an attempt to capture more of the character of the sound. This will also allow the possibility of making a long loop, which generally gives a better chance of disguising the loop points.
When taking a long sample of an analogue synth patch, there are a few simple precautions you can take to prevent problems later on. If the source sound utilises two or more audio oscillators, you might consider sampling each of the oscillators in isolation to avoid also sampling the 'beat' frequency (the cyclic effect caused by two or more oscillators playing back with slight tuning differences). This will not only make sample looping easier, but will also make playback transposition more successful. In many cases you may find that the oscillators are merely producing detuned versions of the same basic sound, so taking more than one sample might prove unnecessary. In most modern samplers, duplicating and detuning layers to achieve similar results takes a few moments work.
If you are intending to loop the sample, it is probably a good idea to ensure that the sustain level of the amplitude envelope on the source instrument is set to its maximum value. Although this will effectively remove the decay portion of the amplitude envelope, this is probably the easiest audio element to replicate by using the sampler's own envelope generator. Finding suitable loop points will become considerably easier, since you will not have to contend with constantly changing volume levels across the length of the decay stage. It will also help to make the most of your sampler's bit resolution, by keeping the input level consistently high. Similarly, it might also be a good idea to apply a compressor to the analogue synth sound prior to feeding it into the sampler. Careful use of compression will produce a signal with a higher average volume level that will help greatly in the war against noise, and a more consistent signal level that should prove easier to loop.
If the source sound has any cyclic variation during its sustain period -- such as chorus, vibrato, or filter modulation -- then make sure you sample a whole cycle of the effect, or finding a good loop point could prove impossible. If in doubt, make the sample far longer than you might otherwise deem necessary -- you can always discard the excess later in the editing process.
After having taken your sample(s), you're really at the mercy of the facilities available on your particular sampler (unless, perhaps, you have some clever sample editing computer software like Sound Designer or Alchemy).
When looping, try to find loop points where the sound is at its brightest, tonally speaking. I find that clicks or glitches are usually easier to hide when the sound has a lot of high frequency content. The best loops are to be found where the cyclic variation in the sound repeats itself. Ideally this will be around a single repetition, but sometimes you may need to loop around two or more to make the loop less obvious. Bi-directional loops might also provide a better result if the going gets really tough -- providing your sampler allows such things. When all else fails, a short crossfade can hide a multitude of sins!
Making use of long samples has its merits, especially for one-shot sounds or off-the-wall effects, but it can lead to problems. When played back higher up the keyboard, for example, a gentle envelope sweep becomes a fast blip and subtle vibrato mutates into an irritating warble. Lower down the keyboard, the attack stage seems to last forever and the vibrato just leads to sounds being out of tune for much of the time. Multi-sampling provides one way around these kinds of niggles. Even though 15 or more samples may be needed to produce a convincing four-octave range, if a song calls for the sampled synth sound to be played back over a narrow range of keys, two or three of your samples may be enough to produce perfectly acceptable results.
An alternative method is to take a short sample of a raw analogue waveform, preferably from a single oscillator. By making a very short loop of this sample, possibly even down to an individual waveform cycle, the sampler's own synthesis engine can be pressed into service to emulate the original sound. Most samplers feature envelopes, filters, and LFOs which should be sufficient to do a reasonable job. For a thicker, multi- oscillator sound, duplicating the layer or keygroup with some subtle detuning or vibrato is usually sufficient. Employing these relatively simple methods, it is possible to get extremely close to the character of the original analogue texture, with a considerable saving on sampler memory.
Once the raw waveforms are safely archived (on floppy and/or hard disk), you also have a set of building blocks from which to construct original patches, maybe far removed from the sound you were originally trying to emulate. The disadvantage is that you are very much constrained by the quality of the sampler's own sound processing architecture. Digital filters, in particular, rarely exhibit the warmth and complexity of their analogue counterparts. However, there is a way to capture more of the character of an analogue filter, whilst retaining the flexibility of shorter samples.
First of all, create a patch on the analogue synth with a single oscillator passing through a low-pass filter. Apply a reasonably noticeable amount of filter resonance (this will help to make the results more obvious for this example). Remove any vibrato or other modulation effects, including filter envelopes. Turn down the filter cutoff to the point where it is close to reducing the oscillator's output to a sine wave. Sample the resultant waveform and make a short loop of it -- the shorter, the better. Now increase the filter cutoff about a quarter of the way between its current position and its fully open position. Sample the same note as before and then loop it. Do this twice more, increasing the filter cutoff another quarter of the way to fully open each time.
You should now have four sampled waveforms taken at various stages in the filter's cutoff range. Create a program with each of the samples in a separate keygroup or layer. Each keygroup should be assigned to play across the whole keyboard range. Now here comes the clever bit...
Set the envelope of the keygroup containing the first sample to a one second attack, full sustain, and a long release of about 10 seconds or so. Set the second sample's keygroup to a two second attack, full sustain, and an eight second release. Edit the third and fourth keygroup envelopes similarly, adding a second to the attack and subtracting two seconds from the release each time. The following table should help to make this more clear:
What you are creating here is a limited form of 'wavetable synthesis'. As you play the sample program from a keyboard, the various static filter settings of the four individual samples will blend together to create an impression of the filter opening and closing over time.
Experiment with the attack and release times to alter the nature of the 'filter envelope'. Adding more layers of samples will obviously yield far more convincing results, but setting them all up will become more unwieldy too! Judicious use of the sampler's own filter envelopes can help to smooth out the effect a little more.
Besides playing with envelope times, if your sampler can delay the time at which a layer is triggered, the flexibility of this technique becomes even greater. Try setting envelope attack times to their fastest values, and use delays to successively switch in the sample layers, for a hybrid analogue/digital effect that would be extremely difficult to achieve by analogue means alone. Many other techniques can be applied to bring sample layers in and out for some truly strange results, perhaps utilising LFOs or performance devices such as modulation wheel or breath controller. The relative volumes of the samples will also play a great part in the overall effect. Try detuning the layers to produce an even thicker texture.
Eminently usable results can be achieved by using this method -- and for very little cost in sample memory too, since you will be using relatively short loops. There is a slight catch, however, since you will use up a voice of polyphony for each extra layer you introduce on playback. Whether this is a fair trade-off will depend largely on the polyphony and memory capacity of your particular sampler.
Some samplers (such as the Ensoniq EPS) allow loop start and end points to be simultaneously modulated. This can also be used to create a more realistic impression of an analogue synth's filter. Simply record a short sample of the analogue synth while moving its filter cutoff control from one extreme to the other. Find the shortest loop you can at the start of the sample -- a single waveform cycle if possible -- and then allow the loop points to be shifted with the modulation wheel. On playback, you can now move the mod wheel to open and close a 'virtual' analogue filter! This is essentially the basis of true wavetable synthesis and has been used to great success in instruments such as Korg's Wavestation. If only more samplers had this facility...
Whatever methods you employ to capture your analogue synth sounds, try stacking the samples for some truly monster patches! It's amazing how even two or three half-decent analogue synth samples can combine to produce textures that would give a wall full of Moog modular units a run for their money -- and your samples will stay in tune!
None of the techniques I've discussed above are going to turn your Akai S900 into a MiniMoog, but with a little care and patience you might just fool more people a little more of the time. Happy sampling!
The first step towards making analogue synth samples is to get hold of an analogue synth! If you are not lucky enough to own one, then it's amazing how much loan time a lager and lasagne can buy from someone who does!
Although it may be tempting to get hold of a wall of analogue modules in the belief that the results will be much better for it, in reality even a fairly humble monosynth (Pro-One, Prodigy, SH101 etc) will produce some excellent results and will certainly be much quicker to set up!
Alternatively, you could try hiring a synth for a few days from a pro audio hire company (such as FX Rentals, Music Lab, Hilton Sound etc) or book a visit to the UK's Museum of Synthesizer Technology (pictured above) -- tel: 01279 771619.
For anyone whose sampler lacks a decent filter, there is some good news. Several companies are now producing analogue filters [eg. Analogue Systems FB3 (reviewed in this issue); Peavey Spectrum Filter] which take an audio input and can be triggered by MIDI note messages or control voltages. These devices are capable of giving a startlingly close impression of a true analogue synth, despite the fact that they may be filtering a digital waveform.
Some analogue sounds can be perfectly usable without resorting to multi-sampling. These would include organ-type pads, deep bass drones, or even string pads that do not feature a great deal of detuning. Lead sounds, such as oboe or flute-type patches featuring a very fast attack transient, can sometimes actually improve when transposed over large pitch intervals.