A History of Sampling

Sampling has been with us more more than 10 years.Paul Wiffen takes a retrospective look at its development and how it has shaped the music of the last decade.

Fair but not light : The Fairlight sampler system

The first sampler I ever heard was on a 1980 record. The track Army Dreamers, from the Kate Bush album Never For Ever. Although my ears told me that I was hearing a cello there was no mention of a cellist, just a credit to one Geoff Downes on 'Fairlight'. Whatever this mysterious device was, I wanted one!

For many years the name Fairlight became synonymous with sampling (although it also boasted the first digital sequencer - Page R and a variety of synthesis methods). To own one was the unfulfilled wish of many, many musicians.
By today's standards, the original Fairlight was extremely limited. Its memory was a mere 32k, its resolution 8-bit and the average sample rate a mere 16kHz, giving results that would be sneered at by the average sampler owner today. Listening to that Kate Bush track again, I was amazed at how rough that cello sample sounded. But, the fact remains that it is still unmistakably a cello rather than a synthesised copy, and that's what made the Fairlight so desirable. It sounded like a bad recording, rather than a good imitation.

In 1981, the Fairlight was joined by the Em-u Systems Emulator, which concentrated on sampling (leaving out the synthesizer and sequencing functions to keep down costs). The Emulator widened the circle of musicians who had access to sampling technology, but it was still the domain of pop stars, large recording studios and highly paid session musicians.

These early samplers both had rudimentary looping (see Newslink Autumn 1990),but with only a few parameters available for adjustment. Finding a usable loop was often more a matter of luck than good judgement.
Similarly multi-sampling - the ability to place different samples on different keys - was extremely limited, and samples were only recognisable as versions of the sound source when played within a few semitones of the original pitch. It wasn't long before some artists like Peter Gabriel were using samples transposed down several octaves to produce long complex timbres, the like of which had never been heard before.

Roland S-770

8-bit resolution was the standard for many years. Sampling was originally linear : the numbers stored in memory were straightforward representations of the analogue signal fed in. Various schemes were devised to encode these values to achieve greater accuracy, such as Delta-modulation. This worked by measuring the difference between one sample and the next. The theory was that you would normally have much smaller changes in value between samples than the whole range available, so you could use the same bit resolution to define a smaller change mere accurately. The 1984 Emulator II was noticeably superior to its 8-bit linear predecessors thanks to this kind of recording system. Other systems of sample encoding include Roland's Differential Interpolation method (see David Marshall's article, Newslink Autumn 1990).

Sample memories grew too. The Emulator II, with over half a megabyte of memory, allowed more than 17 seconds of sampling at 26.6kHz unparalleled figures until, that is, New England Digital expanded the Synclavier's capabilities to include sampling. Never ones to do things by half, they introduced 16-bit resolution, a 50kHz sample rate and megabytes of memory, limited only by the size of your budget (provided your budget would buy you a large country house). There was one major drawback however (apart from price), and that was that, for several years, sampling on the Synclavier was monophonic.

At the (slightly) more affordable end of the market a new use appeared for sampling. This was to use it as a technique for encoding dedicated chips for drum machines. As far back as the original Linn LM-1, manufacturers had used sampling to get realistic sounds into their drum machines. Now both Linn and E-mu came out with drum machines that could sample, so you could record and play back your own drum sounds before, during or after creating the drum pattern that you wanted. Manufacturers of sampling keyboards, meanwhile, started to introduce analogue components of synthesis such as filters and envelopes.These allowed the sound to be reshaped in playback and had a variety of uses.

First, poor quality samples could be cleaned up to some extent (or the shortcomings could at least be disguised) by filtering and changing volumes. Second, the natural envelope of a sound could be reintroduced after a sound had been looped (it would otherwise remain at a constant volume). This was particularly useful in sampling piano and other percussive pitched sources, where previously large amounts of memory had been needed to allow the sound to decay naturally. Now decay could be added to a loop, thus saving memory.

The Fairlight was not the first instrument to use recordings of real sounds. That distinction belongs to a British keyboard, the Mellotron, which dates back to the late 605. The Mellotron was responsible for the strings on tracks by King Crimson, Yes, Genesis, The Moody Blues and Barclay James Harvest (rumour has it you couldn't be a progressive rock band unless you had a Mellotron), as well as the flutes on Led Zeppelin's Stairway to Heaven. The difference between the Mellotron and the Fairlight was that, although both replayed recordings of actual instruments, the Fairlight used digital memory to store its recordings, as opposed to the clumsy tape replay system of the Mellotron (which had to rewind after 9 seconds). Another major advantage was that, with the Fairlight, you could make your own samples, instead of being stuck with whatever sounds the manufacturer had decided to record. Mellotron

But the third use was the most creative: samples from the natural acoustic world could now be used as source waveforms (looped or unlooped) for standard synthesis techniques. Instead of the sawtooth, square or pulse waveforms which had been the basis of sound in analogue synthesis, an infinite variety of waveforms became available.

Roland U-220 : A Sampler,not a submarine

This was one of the major features of the much more affordable samplers that began to appear in 1985-1986. Whatever their bit rate (some still used 8-bit, although most had moved up to 12-bit), whatever their memory size (most had half a meg of memory), whatever their polyphony, they all began to feature filtering and enveloping very heavily. The result was machines that were much more versatile than their more expensive brethren which may have had sample rates of 100kHz but whose features for modifying the sound were still pretty limited.

Looping and multisampling became more sophisticated, even on the cheapest machines. It was not uncommon for more than 50 samples to be available across the keyboard (if you had the memory for them). Various techniques began to evolve to make the tedious business of looping less of a hit-and-miss affair.

The advantages of sampling percussion as against pitched instruments were obvious. Drum sounds needed less memory than strings or pianos, and they didn't need looping or multi-sampling. Sampling drum machines needed much less memory and were therefore much cheaper. Sampling drum machines were also the first area where 12-bit sampling was used. Although 8-bit companding systems were a significant improvement on 8-bit linear for sustained sounds, they weren't so hot on percussion sounds. The move to 12-bit meant cleaner, crisper drum sounds.

Machines were either able to search for zero-crossing points or other suitable places for click-free looping. Some could even create a loop of the same length as the sound frequency to give a sort of autoloop feature (the Roland S-10 was the first of these). Then crossfade looping arrived, whereby the sampler could recalculate the waveform around the loop point to smooth out any clicks or glitches.

The sampler became the subject of musical fashion and there was a flood of N-n-n-n- nineteen type records, in which the sampler was used more as an effect than as an instrument. Sampling features were built into digital delay pedals. like the Boss DSD-3, so that a guitarist performing live could play a phrase which would loop repetitively, whilst he played a different part over the top.

The expense of RAM memory has always been a major factor in the cost of samplers. In 1988 its price started to fall dramatically.
At the same time, digital techniques for filtering came into their own, and the analogue-based filter and envelope technology was replaced by more flexible digital equivalents. This overcame a major drawback, which was that separate outputs were, until then, forced to be monophonic because of the discrete hardware required for analogue filtering on each voice. Digital technology meant that samplers could finally have polyphonic separate outputs.

Boss DSD-3

Visual editing - always available on the top end systems which were essentially computers - became increasingly common on more affordable samplers. Roland led the way in this respect. Ever since the S-50, released in 1986, most Roland samplers have had a CRT output for connecting a monitor, giving greater ease and flexibility of looping, mapping and so on.

Samples began appearing in mainstream synths with the introduction of the Roland D-50. Nowadays, there are few synths on the market which do not have a wide variety of multi-sampled standard acoustic sounds. Some, like the Roland U-20, are not so much synthesizers, as sample players, using resynthesized PCM technology to give immediate access to a very accurate library. As a result, musicians are no longer forced to turn to samplers for realistic acoustic sounds. Piano, strings and brass are now to be found in the average synth's ROM.

This has had a major effect on the way samplers have been used in the 90s (as a listen to the radio will reveal). Those who still want to use them for recreating instruments have to look to the top end of the market to achieve anything that is not already available on sample-playing instruments. But the shift has been towards musical styles where the sampler is used to reproduce whole sections of music, be it vocal phrases, rhythmic grooves or other musical passages. These also need high fidelity, stereo sampling and long record times, as developed for the Roland S-770.

RolandS-550

Sampling has come a long way in 10 years, from primitive machines owned exclusively by the rich and famous, to a much more highly developed technology that is available to the average musician. Sound quality, facilities and programmability have increased as dramatically as prices have dropped. Today's memory capacities are beyond even the wildest dreams of the early sampler owners. So where does it go from here? Sampling has grown closer and closer to hard disk recording (to which it is already related by virtue of the digitization process). Roland's new DM-80 is an example of this. Memory on hard disk (or optical) will become available for real-time playback, and eventually the recording and performance applications will fuse into a creative whole under the direct control of the musician or producer. Roll on the next 10 years!

Roland S-50

Basic Synthesis 1 | Basic Synthesis 2 | Advanced Synthesis 1 | Advanced Synthesis 2| Sampling 1 | Sampling 2


MAIN INDEX

REFERENCE GUIDE

TRANSCRIPTS

SYNTH-GLOSSARY

Chaos Quantum Logic Cosmos Conscious Belief Elect. Art Chem. Maths


Roland Newslink Spring 1991 File Info: Created 19/8/2000 Updated 3/8/2012 Page Address: http://leebor2.100webspace.net/Zymic/synth7.html