Sampling Confidential - Lateral Thinking (Part 2)
Dom Foulsham continues our series on the digital ins and outs of sampling, and talks to some arch sampling professionals about the tricks of their trade. Discover how to get a convincing crash cymbal; how to break the rules creatively and stay legal; and 20 things to do with a redundant manual. Read it and weep, Jive Bunny
In the second of our series looking at the side issues of sampling, we poll the good and the great on how best to use, and occasionally abuse, the humble sampler...
"I'm not actually very good with samplers. They make them do far more than I ever need them to. There's also the fact that I have a religious aversion to ever reading a manual. I tend not to know what half the pages do." So says Norman Cook, sample pioneer and one-time Beats International protagonist.
Sampling has now been with us for more than a decade; the terminology associated with it has slipped quietly into the language with scarcely the raising of a technophobic eyebrow. But beyond the manual, what of the secret life of the sampler - the more unusual stories of sampler use and abuse. There's an alphabet of answers...
Finding the perfect loop can be a nightmare. The trick is to find a good point of equal amplitude and tone so that the loop is smooth and glitch-free. So cheat. The S1000/S1100 and S3000 series have an auto-loop function that will have a bash at finding the best points of equal amplitude, but this may not always guarantee a click-free loop. Try using crossfade auto-looping; this will crossfade a portion of the sound (before and after the loop point) with the loop to smooth things out.
It's particularly useful for those fat synth sounds with multiple detuned oscillators. The secret is to take a very long sample - maybe five or six seconds - and then look for suitable points in the waveform that look like they'll give a passable loop (auto-loop might help here). Set a 1-second loop. Then set an outrageously long cross-fade so that the whole sound is 'splodged' and cross-faded into one homogeneous waveform. This can have several results.
Firstly, it should produce a seamless loop (not always - it depends on the sound), but because parts of the sample from the beginning and the end are effectively being layered onto the looped portion, the sound becomes even fuller and fatter. You also tend to lose the 'cyclical' feel of looped samples because the crossfade evens everything out. Once completed, you can cut the portion before and after the loop, to leave a huge, fat, 1-second sample that can be used as the basis of a synth sound, with all the usual TVF and DCF programming that most samplers now provide.
This effect doesn't work as well with all groups of sounds, but the results are always interesting with brass, strings and vocals
If you own a DAT recorder, backup all your samples onto DAT tape. It's one of those obvious things people don't do until they lose samples deep in hidden menus or through accidental erasures.
Beefing up sampled cymbals is essential if you want to get away from those less than credible, short, 'memory-hungry' sounds that seem to have been cradle-snatched from an early Dr Rhythm. "Try taking a crash cymbal, and playing identical samples one octave below and one octave above on top of the original and then adding a 'bidirectional loop' equivalent of the same sample" suggests Karl Twigg of One Family. "The 'bi-directional' sample is one which, when finished, begins playing back towards the start point, whereupon it again changes direction and plays back the original way, then backwards again - and so on. By choosing an early looping point (short enough that the drop-in volume doesn't give you that cymbal-dropped-on-the-floor-and-rolling-around sound) you can then use your amplitude and filter envelopes to decay the loop to the required silence." Hey presto! - a huge crash cymbal.
Doubling up vocals is essential if you don't have a singer with a particularly strong voice. Unlike a harmoniser, vocal doubling on a sampler doesn't produce that characteristic 'watery' effect, but your machine will require plenty of memory...
Sample a whole verse, get it locked into the sequencer and then timestretch it to 99% and 101% of its original length to produce two new samples. Then, tune down the shorter one by approximately 0.8 of a semitone from the original, and the longer one up by the same amount. At this point all three samples should be the same length but pitched slightly differently. Played back together the samples should produce a much fatter vocal line with all the strength of a harmoniser but none of the watery effect: timestretch harmonising with no stretch marks!
With stereo samplers try changing the loop points to different settings on the left and right channels. (Some autolooping samplers do this by mistake!) The result is a waveform that slowly goes out of sync as it loops, producing a sound that gives a nice swirly effect in stereo.
Ever wondered why factory samples sound so great? The answer is that to get a true sound out of a sampler you don't always put a true sound in. Even the best samplers have a way of adding their own colour to a sample when played back. To avoid this try playing back a sample through an equaliser and altering the settings until you are totally happy with the new sound (...increase the bass, cut the mid, add a little parametric EQ to a snare to get that commercial 'pink' sound, for example).
Then, keeping the same settings, place the equaliser at the input of the sampler and re-sample the sound. This time, the EQ will be built into the sample giving you the sound you want and compensating in advance for for any colouration that the sampler will introduce.
"One of my favourite rim shots is just a bit of a bloke saying something" reveals Norman Cook, "It sounds just like an 808 rim shot, but a lot tougher - all by accident. With Beats International, one of the tracks on the album Three Foot Skank has a duck sample that makes an excellent snare drum. That said, I'm not one of these people that waves a DAT machine at every Safari animal they see, and then makes naff concept albums."
Masters At Work have developed an interesting hallmark through the use of bizarre, almost surreal delays like the Suzanne Vega 'Blood Makes Noise-oise-oise-oise-oise' or the famous 'Only Love Can Break Your Hah-Hah-Hah-Hah-Heart'. That sort of loop over a clashing drum beat can produce an excellent effect sometimes.
Tackhead do the same - looping the end of their drums, but adding parametric EQ so that a snare ends up soaring up the octaves. (A similar effect using analogue slap-back was used by Laura Brannigan's 'Self-Control'.)
With samplers like the Ensoniq ASR-10 you can use the sampler as a real-time effects processor, to add, say, distortion to the guitar, reverb to vocals - or whatever. With some machines its possible to actually use them as an effects processor whilst simultaneously replaying a sequence of samples. That's possible whether or not you want to actually sample that voice or guitar.
When you've got a bass drum and you're having problems trying to get enough of a slap or click to be heard in the mix, try adding the same bass drum sample an octave up. This may not be necessary for a club mix, but on radio the second kick helps the beat to cut through better.
"On the S1100, a lot of the bass drums and deep bass samples have a click at the very end, and so we simply set the DFs to close just before the end of that sample," says Tim Lever of One World Productions. "By the time the click occurs the filter is closed and you don't hear it." This doesn't happen with the S3000.
Triggering multiple MIDI channels at the same instant can cause problems, as events have to be sent sequentially from the MIDI Out port. The use of pre-delays when dealing with large numbers of simultaneous notes can avoid the data crush that produces a slight 'ripple' effect.
"With samplers, we like to leave as many things running live as possible - that way you can change things like drum rolls very easily in the mix," says Tim Lever. "But one of the big problems we have is timing. Running from something like Notator or Cubase, if you have everything quantised, you're not going to get all those sixteenth notes coming out bang on the beat. However, it is possible to delay events on both these programs; move them off the beat by factors of minus 1 or minus 2 etc. So we've developed samples that have short bits of silence added to them that are exactly the same lengths as these delays. So, if we're triggering a hi-hat, for example, the MIDI notes are programmed slightly ahead of the quantised beat, but the actual sound of the hi-hat occurs at exactly the right time."
Why oh why are there only jack sockets and no XLR inputs on the Ensoniq EPS series samplers? Answers on a postcard...
Don't randomly assign your samples across the keyboard. Use a method that makes good musical sense. If your vocals are sung in the key of C and follow a I-IV-V progression, for example, then they should be assigned to the notes C, F and G. If you have more than one vocalist try to ensure all the samples associated with each one are at a specific point on the keyboard - within a certain octave, for example. It's far less confusing later on.
Make sure, also, that you switch off any pitchbending, modulation and velocity sensitivity on the keyboard so that the sample plays back at the same pitch and volume every time. Pitchbend can be used to exaggerate vocal bends, but duration is then also affected, and any choral samples will sound pretty surreal as the chord's intervals play havoc with the melody. Aftertouch, if you have it, can also help add dynamic swells to the music; if you've sampled anything in stereo, try assigning positive aftertouch to the left sample and negative aftertouch to the right. The result? Orbiting voices across the stereo spectrum. Just wait till you try your first Roland Sound Space samples!
A lot of dance music is based around loops, but getting a repeating loop to synchronise with even just one other loop or track can be a problem. Adding a third makes life even more difficult. What you can do if you have, say, a 3-second loop is to divide it into eight separate segments, working out the different millisecond lengths of each, and assigning them to consecutive notes on your keyboard (between say C3 and C4).
You can then play these segments by hand - or trigger them on the sequencer - in time with the sample loop you wish them to be added to. The effect is to keep any timing problems to a minimum as the second loop is effectively retriggered and brought back onto the beat (in this case) eight times during a three second period. It keeps the timing really tight, and allows different loops to be edited without additional headaches.
The technique of breaking down sample loops into a series of repeated assigned notes is also useful where you want to get the best possible vocal chorus line. If you have several takes of a chorus by a vocalist and find that some parts of some takes are better than some of others, you can use onboard editing facilities to cross-fade a good beginning of one phrase with a good ending from another. Trying to achieve a convincing result by conventional splicing of the two samples would be near enough impossible - vocalists never sing phrases in the same way twice!
Record a chord or sequence of notes on a sequencer and use it to simultaneously trigger a combination of different samples each, perhaps, assigned to a different type of effect. These can then be recorded to DAT and finally resampled into your machine as a single 'mega sample'. Samplers like the Ensoniq ASR-10 can do all of this internally and in real time.