You are here

Sampling Basics, Part 3

Tips & Techniques By Paul White
Published March 1996

Paul White concludes this series with a look at phrase and loop sampling, the use of keygroups and mapping, and the use of the modern sampler as a synthesizer. This is the last article in a three‑part series.

Over the past couple of months, I've introduced the basic concepts of sampling and, hopefully, made you aware of any hidden expenses such as buying more memory or adding a SCSI drive. Like any other area of music making and recording, sampling can be used either very simply or in great depth, but no matter how much you know already, there are always new tricks to learn.

When samplers were first devised, their relatively short sampling times and small memory capacity meant they were useful mainly for sampling individual instruments. Once looped, this allowed sustained sounds such as string pads to be generated. With the increase in the memory capacity of current samplers, it may be possible to record several minutes of either mono or stereo audio, making the sampler a convenient alternative to the analogue 2‑track recorder for 'spinning in' vocal choruses or other sections that occur more than once within a song. Samplers are also used extensively for sampling complete rhythmic phrases, especially drum patterns. All these applications require different things of a sampler — and of the user.

Looping Samples

In part one of this series, I commented that there's a world of difference between looping samples and sampling loops. Looping a sample is the art of sustaining the sound indefinitely by taking a part of it (usually during the decay part of the original sound), then playing it back repeatedly to create the illusion of a continuous sound. How well the start and end of the looped section match up determines the smoothness of the loop. One pitfall to be aware of is that stereo samples can be more difficult to loop than mono ones, as the ideal loop point on the left channel might not coincide with the ideal loop point on the right. The usual outcome is that one side loops smoothly, while the other has an audible glitch. You can't loop the samples at different places, otherwise the left and right channels would move out of phase with each other — so if you can't find a satisfactory loop point that works for both channels, you usually have to take your best shot and then use a crossfade to cover your tracks. It's often better to have a short crossfade than a long one, because during the crossfade region, the phase differences between the signals can cause a mild chorus or flange effect, which will detract from the natural character of the sound, and may even cause a noticeable dip in level.

In theory, the waveform being looped should be exactly in phase at the point where the loop joins up. If the optimum loop point has been chosen it will be, but unless the signal is absolutely constant in frequency, phase differences will occur the further you move from the loop point. The shorter the crossfade region, the less chance there is of running into phase problems.

Sampling Loops

Sampling loops is quite different from looping samples because, perversely, it's seldom necessary to do any looping at all. Sampling a 'loop' is actually the technique of sampling a musical phrase, usually a rhythmic one, that can be retriggered every time it ends, to create the illusion of a continuous musical part. You could do this by looping the sample itself, but used within a sequence, you'd almost certainly find the timing of your sample drifting with respect to the rest of the mix. Whenever you trigger a sample, it's important to realise that the only part of the sound you can guarantee to be in sync is the beginning. After that, you rely on the sampler's internal clock and the sequencer's clock not to drift apart. As they both utilise digital, crystal‑locked clocks, they should stay tightly in sync for at least the length of a typical song, but unless your loop length is matched precisely to the tempo of your sequencer, the timing will still wander out.

A far more satisfactory approach to is to retrigger your sample from the sequencer rather than looping it. If your sampled loop is four bars long, you simply play a new note every four bars, and make sure this note is quantised to the first beat of the bar. Now, if the tempo of the sequencer isn't spot on, the result will be a small timing discrepancy at the end of each sampled loop rather than a cumulative drift in timing, and if the error is kept small (by adjusting the sequencer's tempo as closely as possible to that of the looped section), nobody will hear it anyway. At the end of the song, you only need to allow the loop to play though to its natural end rather than retriggering it, and if you can pick a loop that actually has an ending (rather than simply being chopped off), it makes it easier for you to engineer a convincing finish to your piece. If all else fails, there's always the fade‑out to fall back on.

Many drum loop sample CDs include samples of individual drum sounds as well as the loops, and these samples can be used to play new parts over the original loops, to personalise them and add variety. To do this, you simply need to use the sampler in multitimbral mode, with the loop program assigned to one part and the drum sound program assigned to another. You can also use these sounds to build your own rhythms, fills and intros.

Anyone seriously involved in using rhythmic looped samples should check out Steinberg's ReCycle software, which makes the precise looping of patterns easy. It also enables you to split a single pattern into individual beats, so you can move, remove or substitute drum sounds to create new rhythms.

Phrase Sampling

Phrase sampling is similar to using loops, in that you record complete chunks of music rather than single sounds, but in normal circumstances, you wouldn't loop these sounds or retrigger them to play continuously. For example, if you have only one perfect backing vocal take for a chorus, you can sample it and then use it in every chorus in the song. A tip here is to sample in mono if you need to save memory: the sound will widen up when you add reverb during the mix.

Another use of phrase sampling is to use vocal or instrumental snatches in your mix. Ethnic vocal samples can often be broken down into phrases, and then retriggered to fit in with your own compositions — and the same is true of things like flute solos. There are also sample CDs featuring brass sections, which enable you to use powerful brass riffs in your work. The more you explore the world of sample CDs, the more inspirational sounds you come across.

Inevitably, the way you write music can be dictated to some extent by any sampled phrases you wish to use. For example, you'll probably want to stay in the same key as the sample you've chosen, rather than using pitch correction and all the side effects that process inevitably entails. Pitch‑shifting usually compromises the sound quality, unless the degree of pitch change is very small, whereas playing back the sample faster or slower by hitting a different key will not only change the tempo, but also the tonal character of the sound. Sometimes this can be used creatively — at others, it can't!

Timestretch‑type treatments may also be used to change the tempo of a section without altering the pitch, but once again, be aware that audible side effects will become evident if you change the tempo by more than a few percent. One option is to use external pitch/time processing such as Steinberg's Time Bandit software, in which case the degree of available change is rather greater before the sound quality is compromised.

It's worth mentioning at this stage that both looping and phrase sampling take up a lot of memory, especially if you are working multitimbrally or using long chunks of audio. In some cases, it may be worth considering moving up to an 'Audio plus MIDI' sequencer, so that longer sections can be recorded directly to disk to save on expensive RAM (for more on 'Audio plus MIDI' sequencers, see Mike Collins' overview of the major Mac‑based packages in SOS December '95). In my own case, I use Logic Audio, which allows drum parts to be looped in much the same way as you might do with a conventional sampler. Its benefits are even greater when you want to record long sections of vocals, and the visual interface is useful when it comes to cutting and pasting audio.

Pitch Invasion

Whether you take your sounds from CD‑ROMs or sample CDs, you may find that the multisampled area is only a couple of octaves wide, the reason being that the instrument or voice being sampled can only cover a limited range of pitches. However, the highest and lowest samples may still be mapped right out to the ends of the keyboards, which will allow you to use the sounds outside their natural range, providing you don't mind the unnatural sound. For this reason, a well‑sampled choir might sound perfectly natural in the middle of the keyboard, but if you go right up to the top, you get a sound akin to your local chipmunk close harmony singing group, and if you go down to the bottom, it can sound as though your angelic choir has descended into the bowels of Beelzebub. Figure 1 (see page 80) shows a typical keymap for a choir.

In many instances, a single sample assigned to each key is all that's needed, but if you want to emulate the tonal characteristics of a real instrument, you may need more. The majority of samplers allow you to assign two samples to a key, so that low note velocities trigger the 'quiet' sample, and higher velocities trigger the 'loud' sample. Only one sample plays at any one time, so you don't lose any polyphony, but the transition from 'quiet' to 'loud' sound will be abrupt, even though velocity still controls level in the usual way. Careful use of filters can help — if you set the filter so that it opens up more with higher velocities, the sound will get brighter as it gets louder, which is what most real instruments do.

Another option which, at first glance, looks attractive is velocity crossfading. Here, you have two samples, just like before, but this time they both play at once, only at different levels depending on the note velocity. At low velocities the 'quiet' sample predominates, while at high velocities, the 'loud' sample takes over. In the middle, you get a mix of both. This provides a much smoother transition from low to high, but because two notes are always playing, the available polyphony of your sampler is halved. You also hear a mixture of two samples, which isn't the same thing as one, intermediate velocity sample. Depending on the instrument, you might hear a phasing or chorus effect, which gives away the fact that you're not hearing a single sound. Different users will have different preferences, but on balance, I prefer velocity switching for most applications.

More About Maps

Once you start to explore sampling, you'll find you can combine phrase sampling and conventional simple note sampling in fascinating ways. For example, we've all heard those wonderful (and sadly overused) shakuhachi solos that end in breathy pitch drops and flourishes, but as you soon discover, you can't actually create those effects on single note samples, no matter how much you juggle with the bend and mod wheels.

I've found that one way of tackling the problem is to map out a couple of octaves of single note samples, played straight with no vibrato, then take a handful of useful phrases in the key of your song, and map these to single keys. By doing this, you can jump between playing a conventional solo line and triggering one of the phrases. Figure 2 shows a possible keymap for exploring this technique (see page 82).

Another way to do the same job is to use velocity cross‑switching, so that normal playing velocity gives you the conventional single note samples, and higher velocities trigger a phrase. Once you appreciate what can be be done in this way, you'll come up with your own way of mapping samples to suit your own playing style. In all cases, the single phrases should be mapped so that they play back at a fixed pitch regardless of the key to which they are assigned. In practice, it is possible to create a number of different‑sounding programs based on the same set of samples, simply by using different mapping options, different filter settings and different amplitude envelopes.


I'm aware that this series has only scratched the surface of what you can do with a sampler, but I hope that it has given those who don't own a sampler an idea of what they're missing, and existing sampler owners some new ideas for avenues of exploration. A good sampler is a very versatile tool, and though collecting and cataloguing a working selection of sounds takes time and effort, it is time well spent. A sampler can emulate conventional instruments, it can be a phrase recorder/player, or it can be a synthesizer in its own right. That makes it one of the most powerful music tools on the planet, so use it with respect and you'll reap the rewards.

Samplers As Synths

Now that most modern samplers are equipped with resonant filters, simple waveforms can be sampled and treated, to create the digital equivalent of an analogue synth. In theory, you could just load up your sampler with square, pulse, triangle, sawtooth and sine waves, then start synthesizing, but it's often more productive to start off by sampling the raw waveforms from a real instrument. The reason for this is that virtually all analogue synths colour or distort the waveforms to some extent, so if you can capture these benign colorations, you may end up with some of the character of the original instrument in your new sound.

When sampling a synth, it's usually best to set its low‑pass filter cutoff frequency at maximum and the resonance at minimum, so as to remove the effect of the filter altogether. This lets you re‑create the necessary filter sweep using the filter and filter envelope controls in your sampler. Similarly, turn off all LFO modulation, so as to get as static a sound as is possible. Again, you can put all this back using your sampler's own controls.

If the waveform is constant, you only need to sample a very short section — indeed, a single cycle will do the trick, and because the sample is very short, any auto‑looping facility you have should make light work of finding the best loop points for you. Sounds that evolve have to be treated differently, because the waveform is changing over time. If you have enough sample memory, you can sample a phase sync sound or a ring modulation effect, but unless you multisample, you'll find that the effect speeds up drastically towards the top of the keyboard, and slows to a near crawl at the other. Similarly, LFO pulse width modulation will change noticeably in speed, even if you do manage to loop the sample at a point that coincides with one full cycle of the modulation frequency. All in all, it's probably best to fake this particular effect by sampling an unmodulated sound, then either adding chorus, or layering the sound with a slightly detuned version of itself.

A typical modern sampler has good modulation facilities, portamento, nice‑sounding filters and powerful envelope shaping capability, so when you think about it, you've got everything you need in a synth except the basic waveforms. What's more, you can mix analogue waveforms with sampled waveforms, to create hybrid string patches and so on. This aspect of the sampler is probably the most underrated, but I suspect that over the next couple of years we shall see the fine line between synths and samplers dissolve altogether.

CD‑Roms: How Much??

Compared to straightforward sample CDs, CD‑ROMs seem hugely expensive (£200 is a typical CD‑ROM price, compared to more like £60 for a sample CD) — but when you consider that they can contain 600Mb or more of ready‑mapped samples, they still make good commercial sense — as long as enough of what's on the disk is likely to be useful to you. Indeed, there's little point in bootlegging samples from CD‑ROMs (which is illegal anyway), because the removable disks you need to store the data on tend to cost more per megabyte than the CD‑ROMs cost in the first place.

One of the limitations of the CD‑ROM is that the samples provided can't be edited unless you save the edited result to a writable drive of some kind. So if you want to change loop points or do other processing tricks, you'll still need an external SCSI drive. However, most samplers will allow you to save only the program information (key mapping, envelopes, filter settings and so on) to a floppy, without the need to save the samples themselves. That means you can make up program disks that allow you to use the existing samples in different ways, using inexpensive floppies.

Having made the case for CD‑ROMs, the truth is that you still end up paying for some samples that you'll never use, and I don't think it's any secret that the major sample providers are looking at ways of selling individual samples, possibly via the Internet. Whatever direction they take, as more people start to buy samples, the more the price will eventually come down — which has to be good news for all of us.

Sample CDS

Sample CDs are a lot cheaper than CD‑ROMs, and usually contain the same samples as their CD‑ROM counterparts (though occasionally fewer of them). The trade‑off is that you have to load all the samples into your sampler, map them into keygroups and create the programs yourself.

You'll find the key ranges for each sample listed on the accompanying sleeve or booklet, so the procedure is more tedious than it is difficult, but if any of the samples need looping, it's down to you to find the best loop points. This can be very time‑consuming, and requires a certain amount of experience to get right. If it's something you're going to be doing a lot of, then a software sample editing package such as Alchemy is well worth considering. I'm still using Akai's bundled Mesa software, which does the job well enough once you've cracked its cryptic routines, but its main weakness is that that it can't perform crossfade looping — you have to go back to the sampler front panel for that.

When it comes to mapping your samples, you don't always have to go along with what the CD literature suggests. For example, instruments such as strings and flutes can be transposed quite a long way before they start to sound wrong, so you could get away with just one or two keygroups, if most of your playing is near the middle of the keyboard. Even instruments that do suffer when transposed outside their normal range can be abused to create special effects: check out the Art of Noise's early material to hear excellent examples of this.

Human voices don't take kindly to being moved too far from their native pitch, and the same is true of acoustic instruments with strong resonant formants, such as acoustic guitars and pianos. Here, you may need to have a new sample every three or four semitones to maintain a natural sound. In this case, you should arrange the original pitch of your instrument to fall in the centre of the keygroup, then listen carefully for differences in tone as you move from one keygroup to another.

Whenever you play back a sample at a higher pitch, the whole harmonic structure of the sound is moved up, with the result that the highest sound in one keygroup is likely to sound brighter than the lowest note in the next keygroup up. If the keygroups are narrow enough, this effect may be trivial enough to ignore, but if you're using only one or two keygroups per octave, it may be advantageous to use the tracking filters set up with a negative value, so that as you move up a keygroup, the top end is tamed slightly. Some samplers allow you to crossfade from one keygroup to another, but this reduces polyphony and may produce unnatural chorus effects on solo instruments.