The Analogue Sampler
Ever wished you could get some of those rich old analogue sounds working comfortably with your MIDI system? Greg Truckell samples the sounds of the past.
IF YOU'VE NEVER SAMPLED AN OLD ANALOGUE SYNTH, YOU'VE BEEN MISSING AN EASY WAY TO EXPAND YOUR SOUND LIBRARY AND TAKE THOSE OLD SOUNDS INTO THE '9OS.
SUBTRACTIVE SYNTHESISED SOUNDS (from analogue monosynths) are based on relatively simple waveform structures, typically derived from a handful of waveforms - sawtooth, square, triangle, and variable pulse. Sounds with simple waveforms are particularly easy to loop, and a very short loop of one or two wavecycles is often enough to recreate the sound of the original (thereby saving sample memory). If your sampler lets you detune a sample against itself, then some reasonably fat sounds can very quickly be at your disposal, even from a short loop.
The whole approach of sampling leans towards the idea of sounds consisting of an interesting and complex attack transient decaying into a sustained note with a more or less steady waveform. Most of the effort that went into programming interesting analogue synthesiser sounds went into creating a sound which provided a convincing sustain, but which had an attack transient that bore more than a passing resemblance to a bow scrape, flute chiff or mallet strike. Given the limited features of the average monosynth, you hardly need reminding how poorly they fared here. Monosynths came from an era when synthesisers were supposed to sound like synthesisers.
"TURN ANY SOUND INTO AN 'INSTRUMENT' BY ADDING AN IN-TUNE SYNTHESISED LOOP - IT'S CHEAPER THAN A D50 OR KORG M1, WHICH USE THE SAME PRINCIPLE."
Of course, you could easily synthesise a really unusual "event" in the context of an attack transient, sample something a little longer for the decay and loop, then mix the two samples together. Part of the problem with the analogue attack transient is to do with the relatively sluggish envelopes to be found on many monosynths. A DX7 envelope set to its fastest attack rate gives a good impression of a Batman sound effect - the attack rates often have to be decreased for sounds that you might ordinarily think of as percussive. In contrast, the fastest attack time on an analogue instrument can still have quite a lazy feel to it. However, for every octave you transpose your sample up, you halve the sample playback time - and therefore double the attack rate. Take this feature to extremes of three or four octaves, and you can turn a woolly analogue synth torn drum into a sparkling woodblock, claves, or weird tiny afro logdrum.
It may be the case that, when carefully analysed, the attack transient that you want to create requires features that your analogue synth is unable to provide. Perhaps there is just a chiff of noise right at the start, but there is plenty more happening before the decay phase of the sound. An extra envelope generator would come in really handy to attenuate only the noise generator, but we're missing the power of the sampler here - sample the noise transient first, then the oscillators, and then combine the two samples.
Alternatively, remember that you have a spare hand when you're sampling, and that it could be pressed into service turning down the noise generator (or whatever other edit or modulation you require).
Hang on a bit though - isn't this getting a bit spontaneous? I mean, sampling's difficult and it takes ages, right? Wrong: at least part of the point of sampling is to be able to easily recreate something which would otherwise be impractical, like expecting you to be able to tweak something exactly the same way every time. Don't simply think in terms of what sounds your monosynth can make that you could sample. Think well beyond your monosynth's ordinary performance, in terms of monstrously unlikely and unrepeatable events.
As an alternative to synthesising the attack transient, you could even sample something else entirely (nip into the kitchen with a hammer and a microphone), and turn it into an instrument by adding an in-tune synthesised loop. This approach can be considerably faster than trying to get a loop out of acoustic source material, and is often just as effective. It's also much cheaper than buying a Roland D50 or Korg M1, which use just the same principle - sampled attack transients, synthesised sustain sections. More importantly though, it can be immensely creative and original - now when did you last hear that said about sampling? If your analogue system has external signal processing features, then you could sample something which has been put through the analogue signal processors, and so has had filters, resonance and envelopes imposed on it, before you even have to employ your sampler's processing power. And it goes without saying that analogue synthesisers are not the only means of analogue processing which can provide worthwhile sample fodder - guitar combo overdrives, wah pedals, and good old thick analogue chorus come to mind.
Of course, it's difficult to get a good loop out of the kitchen - none of the stuff in there was built to have music performed on it. While you may not be able to get a loop out of whacking the wok with a pasta spoon, it is almost second nature to the seasoned (pun) samplist to be able to imagine what they might sound like if they could sustain a note. Imagination, analysis, synthesis; and the analogue monosynth's area of excellence is in sustained sounds (more later).
We can approach the above technique in a number of ways; it is quite possible to sample an attack transient without any thoughts as to its pitch, and to subsequently sample something intended to provide a wonderful hybrid sound, only to realise you have no idea how to get your second sample in tune with the first (assuming that you want to merge, rather than layer, the samples). The obvious solutions are (1) to document your archive of samples that might come in handy someday (I know, I was kidding), or (2) to find the source and calculate the pitch and sampling frequency at which it was made from the relative tuning and unity playback values (I said I was kidding!). The only alternative is making our samples during the same session - by far the simplest approach...
Tune your monosynth to the acoustic source, and you are ready to sample. That leaves us with the problem of how to balance the relative volumes, and how to line up the attack transients of the two samples. The most effective method I have come across so far was suggested to me by a bassist who had never before used a sampler, and to whom I am indebted (thanks Blackie); you get your hammer in one hand, put the other hand on the synth, and go "one, two, three, bong". Uncannily, the technique is even able to effectively synchronise more than one instrumentalist.
"IN FIVE MINUTES YOU CAN CREATE SOUNDS THAT PUT EVEN THE LIKES OF A TX816 OR MATRIX 12 TO SHAME - WITHOUT THE PROGRAMMING SABBATICAL."
Suppose your monosynth has two oscillators, two envelope generators, and one low-frequency oscillator - sound familiar? How would you like a huge modular system with ten oscillators, ten envelope generators, five filters, five LFOs and a handful of ring modulators or what have you? No problem. Analyse the sound you might create with such a monster, and break it down into, say, five parts. Synthesise the sustain - sample and loop it. Now synthesise and sample the part that lasts furthest into the decay section, and be sure to fade it out before it's as long as your first sample up to the loop start. Now add the samples together.
Since you saved your first sample to disk, if you get things wrong you can easily recover things. Proceed with whatever elements you need to add, right up to the attack transient, remembering to fade them out short of the loop start, otherwise you might find your loop difficult to perfect. Add high harmonics wailing out of tune and vanishing; sub harmonics growling and stuttering into silence. If you find that using your sampler's onboard editing or a computer-based sample editor gives an 'unnatural" edge to the fadeouts, then try shortening the synth's envelope generators. Some monosynths sport exponential rather than linear envelope stages, which have a lovely natural feel to them.
In five or ten minutes you can create awesome, complex sounds that put even the likes of a TX816 or Matrix 12 to shame - without the programming sabbatical, the computer-based visual editor, or even the huge financial outlay. The speed with which these sounds can be created is itself quite inspiring. Too many keyboard players these days shy away from the thought of creating new sounds, but lock them up with a sampler and a monosynth and they'll be sending out for a couple of boxes of blank floppy disks before very long.
SO FAR WE'VE covered short loops, complex attack transients, and extended attack/decay phases. Part of the power of the analogue monosynth was the ability to lean on any parameter during performance. The "meaty sync-bend", almost forgotten on contemporary synthesisers, stands out in my mind as the performance effect most likely to raise the hairs on the back of your neck. How can samplers recreate such drastic and complete changes in timbre, executed so smoothly on monosynths? A modulation wheel-controlled crossfade between two extremes is hardly close enough - but what about a velocity switch?
Sample the performance effect itself, as well as an unmodulated event. Remembering that monosynths lacked touch sensitivity, disable amplitude modulation from velocity to the sampler's envelopes. The top of the MIDI velocity range can now be reserved for the huge sync-bend sample, enabling it as a velocity switched performance effect.
We said we'd return to sustained sounds, and here we are. Even with a limited number of modulators, a sustained analogue monosynth is the standard against which the depth of movement of any synth patch is judged. Even if your sampler supports oscillator detuning, you're not going to be able to create lush, moving sounds with short loops. Analogue oscillators and filters are always drifting around their control settings, and the amounts of any modulation drift similarly. The drift is not so large as to be obtrusive or undesirable - quite the contrary in fact.
It has been said that there are only two kinds of good loop; short and long - who said size wasn't important? The catch with long loops is quite simple; they don't half take up a lot of memory. The usual rules apply (sort of) when sampling any modulated sound. Remember that if you transpose the sample up an octave then you halve the period of any modulation. This can actually be used creatively; LFO modulation of pulse width, filter cutoff, or even pitch, when sampled and replayed polyphonically, will effectively create independent LFOs for each voice. Be sure to keep the modulation amounts subtle though, as extremes don't transpose well. Oscillator beating effects will also behave like this when sampled.
Suppose though, that you want to sample a sustained synth texture with slow and not-too-subtle tonal modulation; slow pulse-width modulation for example. So long as the effect is cyclic, there should be no problem, as long as you set up the sound to be sampled properly. There are certain things to be borne in mind. Most obviously, if you're creating a long loop, then it has to be during the sustain phase of the sound. The loop can't start during the decay phase, as that would give rise to ramp-shaped amplitude or timbral modulation over the loop, caused by either the amplitude envelope or the filter envelope. If you are short of sample memory and have set your heart on a particular sampling rate, then you may find that you can't have both the length of loop you want and the attack/decay characteristics of your original sound. In this case you'll need to reduce the length of the attack/decay phases, and later use the sampler's envelopes to recreate the original. This will free more of the sample time for the loop. You might even be able to recreate the original envelopes entirely on the sampler, discarding any material before the loop.
"SAMPLING ANALOGUE SYNTHESISERS IS EASY; IT CAN ALSO BE VERY SPONTANEOUS. IT WILL ALMOST CERTAINLY FILL HUGE GAPS IN YOUR SAMPLE LIBRARY."
If you've ever used sample editing software on a computer, then you probably know how cute sampled synthesiser waveforms look on a screen (see diagram 1). Even apart from the educational value of such software, it is worth its weight in RAM when it comes to finding a good loop. The effects of cyclic modulation can be seen on screen, even with simple 2D software. Having allocated your memory and compressed the initial phases of the synth sound, it then becomes a simple matter to tweak the LFO frequency until one cycle fits into the memory allocated to the loop. One cycle will give better results than two, three or more, and with a one-directional, forward-only loop. Repetition of a loop pattern, in this context, is considerably more offensive to the ear than a simple loop, which will be perceived as modulation.
You will have noticed that for a long and modulating loop, I didn't suggest crossfade looping. There is good reason for this; analogue modulation is typically smooth and continuous. Crossfade looping may be smooth, but it is still a change in the relative balance of two discrete timbres, not a smooth modulation from one timbre to another, covering all the intermediate timbres. That rather nicely encapsulates the difference, by definition, between analogue and digital systems. Bearing in mind the way in which only analogue fills the gaps that digital cannot reach will help when it comes to recreating the sound with a sampler.
WE DISCUSSED FILTERS in some detail without discussing filter resonance. In the good old days, analogue filter resonance was created by feeding back some of the filtered signal into the filter, thereby boosting harmonics around the filter cutoff frequency, and giving rise to a far fuller sound than is obtained by contemporary filters, which instead create resonance by attenuating frequencies away from the cutoff frequency. The technicalities may sound the same, but the analogue method introduced more distortion, and therefore was a warmer sound.
Analogue waveforms without resonance can be transposed over the full five octaves of your sampler without much concern for the usual munchkinisation effects. This is because the pure waveforms contain no formants, or enharmonics for that matter. I'll qualify that right now; oscillator sync, cross-modulation and ring modulation can all introduce enharmonics (partials which have a frequency which is not an integer multiple of the fundamental frequency). Enharmonics generally differ from formants in that enharmonics will maintain their tuning relative to the fundamental over the full range of the instrument, whereas formants (which we'll be discussing in more detail in a future article) have their frequency more or less fixed regardless of the pitch at which the instrument is played.
There are a couple of ways in which analogue synths can create formants. One is to dedicate an oscillator to the creation of formants by disabling keyboard tracking for that oscillator. The other way, much more commonly employed, is to use filter resonance with little or no filter keyboard tracking or filter envelope contouring. It's probably fair to say that resonant, formant-based sounds have been used by many a synthesist who wouldn't know a formant if it leapt up and bit him on his modulator.
As many of you will know, formants are responsible to a great extent for sample munchkinisation, as they do not remain at their original frequency when the sample is transposed. Getting round the problem usually involves extensive equalisation before and after sampling, as well as considerable multisampling. Why on earth did I bother to mention them on analogue synths? Easy; now that you know about filter resonance, you will understand why your sampler always seems to create the effect of unity keyboard tracking (one octave per octave) - this is the only sort of filter keyboard tracking that your sampler will be able to reproduce without multisampling (you could create similar effects using the sampler's filter, but you will lose the character of the original). Of course, the effect of formant shifting can be used creatively to excess to create analogue synth sounds which sound very sampled without sounding like badly sampled acoustic stuff.
Another fairly common filter trick is envelope polarity inversion. If you invert the modulation polarity of a standard ADSR filter envelope, then the filter will start wide open, close to a minimum over the attack time, open over the decay time to a sustain level, and open even further on key-release. This effect requires what is usually termed an after-envelope. That is to say, in order to create the effect, you need more control after key-release than a simple release time. The reason for this is that a non-zero level is involved after key-release; envelope polarity inversion is the simple way to address the problem. Sampling the effect has its own problems. Unless your sampler allows you to jump to and play from a stage later in the sample than the loop on key-release, then you will not be able to accurately recreate the effect.
Or will you? Try making two samples; sample 1 with the filter attack and decay characteristics you require, but with a zero filter (and probably amplitude) sustain level. If you are being faithful to the original model, then the attack time should also be zero - but feel free to experiment. The second sample should have a slow filter attack time rising to full filter sustain. Merge the two samples and create a loop during the sustain phase of sample 2; discard anything after the loop and use the sampler's amplitude envelope set to acoustic-piano type values, but with a release time equal to the decay time. This will give you the effect of unsustained filter-envelope modulation polarity inversion. Experiment with different filter envelopes on both samples - you could also try making sample 2 a percussive filter sweep, then reversing the sample before merging and looping.
Sampling analogue synthesisers is easy; it can also be very spontaneous. It will almost certainly fill huge gaps in your sample library - it is also such a pleasantly creative process that you might find yourself creating new samples for specific projects, rather than depending on your existing library. Hybrid procedures, combining conventional samples with analogue sounds, are just as much fun, and will create completely new and original sounds. If you own a sampler and a monosynth, then you really can't resist the challenge. If you own a sampler but don't own a monosynth, then buy one now before they become expensive again.
Feature by Greg Truckell
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!