In the third part of our short series, we untangle the remaining alphabetti spaghetti on the side plate of sampling...

N is for Note-Fattening

In many machines, samples can be grouped and detuned with respect to one another to produce chorus effects. (Make sure you don't detune too far or you'll end up with a backing vocal that finishes the day after the main vocals.) Similarly, with looped samples, try double triggering them from your sequencer, delaying one of the triggers by around 3 milliseconds so that the sample is double tracking itself. This is particularly effective on vocal samples; the longer the delay the deeper the effect.

O Is for Oldcut

Coldcut were the first and the original. "We call ourselves Oldcut says Jonathan More. But the years have been kind to Matt & Jon, and to prove it they still have a wealth of tricks up their sleeves - like MIDI sample triggering...

"A nice way to drop vocals (on choruses or whatever) into exactly the right place on a tape mix, is to set the machine in sync with the sequencer via SMPTE, set the sampler up to receive a MIDI Note trigger, and assign this note to your sample. Say you want to move the vocals from chorus one to chorus two: go into record just before you want the first chorus sample to start, and trigger the sample by pressing the key on the keyboard or playing from a sequencer the note that corresponds to the chorus sample. Then move this trigger note to the place in the sequence where you want to put the next chorus and again trigger the sampler. This way you don't have to mess around trying to pointlessly edit those samples to the right length or position."

P Is for Phase Checking

Using phase-shift to compare timing on tape with timing on the sequencer playback of samples is a trick used by One World. "Sometimes you get people that are worried about getting a sampled verse perfectly in time, every time," says Tim Lever. "A quick way of checking this when a tape machine and sampler are synced together, is to play back the vocal track on tape (the first verse, say) along with the sampled verse you want to trigger on all the subsequent verses. Listen out for the phasing on the sample itself. Once you've got the sample's phasing cancelled perfectly you know that the new verse is in time."

Q is for Cubase and Quantisation Noise

Cubase is still (probably) the most popular software sequencer, but when will Steinberg get around to including a dedicated sample edit page?

Now that most machines offer 16-bit sampling, the quantisation noise associated with 8- and 12-bit machines is not so much a problem. Nevertheless, adding compression to vocals etc, will help to minimise noise/hiss and help preserve clarity.

R is for Realism

Even from the earliest Ensoniq Mirage machines it's been possible to bring realism to drum samples through the addition of suitable amplitude and filtering envelopes. Add a velocity-sensitive attack and you'll come much closer to imitating the way a real drum sounds (percussion instruments always sound brighter when hit harder).

By setting up a velocity-sensitive pitch envelope (see 'Z is for Zoom') so that the harder you hit the deeper the frequency sweep, you'll be able to recreate another important characteristic of percussive sound: pitchbend. The harder you hit, the more significant the pitch fluctuation. Listen closely to real congas, tom-toms and bongos and you'll hear what I mean.

S is for Stereo

Here's a little stereo trickery...

You can generate a stereo sample from a mono sample by copying it and then raising the pitch of one sample very slightly and lower the other, also very slightly. Pan hard left and hard right, respectively, for instant stereo - well, pseudo stereo, anyway. Alternatively, set a slightly different start time for each of the two versions, place both onto a single key or sequencer part and again pan them to give an even more distinctive stereo effect.

T is for Time-stretching and Theft

Time-stretching doesn't always have to be used for making 'pitched' samples fit into the right rhythmic space. It can be pressed into service to stretch and shrink samples to produce a multisample, for example. Let's say you've extracted a good brass sample from a sampling CD. Use a reduction factor of 50% to play it an octave lower for the same length of time. Then shrink this version by a further 50% to create a sample two octaves down from the original. Use a stretch of 200% to play the original one octave up and set a stretch of 200% and 400% to play two and three octaves up.

Avoid breaking the law when sampling copyrighted material: read the first article in this series (MT, April '93) on sample clearance. You could save a fortune in legal bills.

U is for Unsightly Timestretch Marks

Why can't I get a clean sample without those awful time-stretch marks? Simply put: because it's fat. The more disorganised the harmonic content of a sampled waveform the more difficult it's going to be to wave the timestretch wand. The same problem is also noticeable when you play full keyboard ranges of notes from just one (non-multisampled) snippet of human harmonies. Remember, keep the waveforms simple and you'll stand much more chance of timestretching without the marks.

V is for Vocal Sampling

When sampling vocalists, make the phrase as long as possible. Vocalists lose energy and expression when asked to sing one word at a time. Even when recording one-word shouts or stabs, get the vocalist to sing them over and over again in time to the music, recording the results to DAT. You'll find the third or fourth take is usually the best. Obviously you must give your vocalists a pitch and time reference before they sing so you'll need to use headphones - you don't want click tracks or music leaking onto your samples.

Use the best microphone you can lay your hands on; a limiter and some means of EQ may also be helpful. But don't record with reverb/FX if sampling time is at a premium. Once your sample is complete, you can begin the task of truncating. Break the vocal samples up into as many sections as possible so that you can trigger them from a sequencer, or by 'walking across the keyboard' if you prefer to preserve a degree of human feel. Try to trigger the entire verse as a single sample and you run the risk of it drifting slightly out of sync.

Vocals, just like a good bassline or snare hit, need allowances made for time shifting so that they sit nicely in a groove. And because the ear usually likes to hear things laid back as opposed to being ahead of the beat, you'll find leaving a little extra room at the front very useful. Truncating your samples right up to the start of the waveform can sometimes sound unnatural - particularly if you lose that all important 'breath' at the beginning of a phrase. It makes more sense to use the existing 'ambience' at the beginning of a phrase than to reintroduce this artificially using delays on a sequencer.

The final stage in preparing phrased samples is rehearsing with them until the timing feels right. You've probably been using click tracks and metronome numbers when recording your samples, but now it's time to throw them out, turn up your monitors and use your ears. Because of their expressive, human quality, vocal lines are rather more flexible when it comes to tempo. If you've recorded a track at 120bpm, for example, you'll probably find you can work between 117 to 125bpm without any audible vocal problems. And any major differences can be put right by judicious use of time stretch.

Remember: most ballads, being fairly laid-back rhythmically, tend to make the player trigger late. Your samples will therefore require very little room at the start of each waveform. On the other hand, high energy stuff will probably make you rush each trigger (particularly if you're inputting from a keyboard in real time), so be prepared to stretch out the start time.

W is for Wavering String Samples

LFOs (Low Frequency Oscillators) are excellent at hiding badly bodged loops quickly. Just add a slight hint of LFO to that wavering string or pad sample and you should find you can mask the effects of looping quite effectively.

X is for Cross Fade

Sometimes, you may have a selection of samples that have all the right ingredients as a composite whole but no single sample is exactly 'right'. For example, a rare groove track might have many bars, but none capable of providing a clean rhythmic loop. In this situation, crossfading can help by eliminating (say) an annoying "Yeah!" from the end of a break. This is done by sampling a bar with a clean beginning and one from a bar with a more usable, clean ending. Set a decaying ramp (amplitude), envelope on the first and a reverse ramp on the second and then simply trigger the two samples (accurately without any delay) on top of each other.

Y is for Yamaha

The TX16W was discontinued several years ago and at present Yamaha do not manufacture a dedicated sampler. Their top performance synths do accept samples in MIDI Sample Dump protocol, enabling the import and export of samples via MIDI. Whilst many samples have considerable processing power inside, the results of combining samples with Yamaha's proven synthesis algorithms can be much more interesting. Units to try this on are the SY99, SY85 and the TG500.

Z is for Zoom

Some samplers - including the Akai S1000 series - allow you to use the envelope generators for pitched and sci-fi effects. By assigning one of the generators to control the pitch of just one of the samples, you can create a slight pitch slide or 'zoom' in the sound that has a pleasing chorus effect - but only on the attack part of the sound. You'll need an instant attack, a fast(ish) decay and no sustain (release is unimportant).

Interestingly, this was a favourite effect on the old PPG synth - applying an upward sweep to just one of the oscillators - so if this is an instrument you'd like to recreate, try it out.

Music Technology gratefully acknowledges the time and contributions of the following people, without whose help this feature would not have been possible:

Karl Twigg - Producer, One Family
Jonathan More - Coldcut
Norman Cook - Producer/DJ
Mike McEvoy - Producer
Tim Lever - Producer, One World
Peter Peck - Yamaha (Kemble) UK Ltd
Steve Howell - Akai (UK) Ltd
Dig - Sound Technology plc
Chas Smith - Roland (UK) Ltd