There are a considerable number of samplers on the new and second-hand market — and they all come with potentially misleading techno-speak specifications. So what is a '16-bit sampler'?

Sampling instruments are almost always referred to by their 'number of bits'. 16 bits is about the limit of affordability for most of us at the moment. The last five or six years have seen a steady progression from early 8-bit machines like the Ensoniq Mirage and Fairlight Mark 1, through the middle range 12-bit models like the Akai S900 and Prophet 2002, and finally to the current 'CD-Spec' machines like the Akai S1100, which offer similar performance to CD players in terms of reproduction quality.

It is very unlikely that the next five years will see anything like the same doubling of resolution — in the long term we can probably expect a standardisation on 24-bit wide 'sample-words', but it will be some time before the actual audio quality approaches even 20 bits. This apparent contradiction is actually just a problem in the way that the 'number of bits' measurement is being used. The important thing to bear in mind here is the difference between how the sample is stored and how it is replayed. It seems to me that when people talk about '16-bit samplers', whether they know it or not they are referring to the data storage format, and not the audio quality.

You may be wondering why any distinction is needed. Doesn't a sample stored using 16 bits and sampled at 44.1kHz produce CD quality when you play it back? Not necessarily, because the playback process will affect the sound just as will the data storage resolution. Consider this case: a machine which stores the samples as 24-bit words, but which only has an 8-bit digital-to-analogue convertor (DAC). If we apply the storage-driven description then this is a 24-bit sampling instrument. But the 8-bit playback is going to sound rather like telephone or AM radio quality, and so surely calling it '24-bit' is rather misleading.

Consider also the stored sample itself. If we use a 4-bit analogue-to-digital convertor to record it, then we have a sample of telephone answering machine quality, but stored as a 24-bit sample. If, for the sake of argument, this is replayed via a 16-bit DAC, then 12 of the 16 bits used for the output are actually reproducing things which have nothing much to do with the original 4 bits!

The interchangeability of units between storage and audio quality may seem unusual. It is easy to confirm that samples are stored as 16-bit 'words', but what exactly is meant by '16-bit audio'? '16-bit audio' is a sort of shorthand way of defining the quality you would get from a perfect reproduction of an audio signal stored using 16 bits. All the significant parameters will be at the theoretical values, so the signal-to-noise ratio will be 96dB, and the distortion will be a fraction of a percent.

Unfortunately, whereas it is easy to store a sample at 16-bit resolution, getting 16 bits-worth of audio quality is not quite so simple. The actual performance of many of the components that are used to convert from the 16-bit digital representation of the signal may be close to perfect, but the performance of the whole system can be compromised by just one poor part, or just something 'mechanical' like the physical layout. Some digital-to-analogue convertors rely for their performance on things like the matching between specific high quality capacitors, or the proximity of decoupling capacitors to the power supply pins, or even on the quality of the op-amp used to convert from the DAC's current output to a voltage output. Factors that you might not expect to have an influence in the world of digital audio, such as humidity or temperature, even differences between batches of components, can affect the performance of these components.

Understanding all this techno-speak is not essential: the important point to grasp is that there is far more to providing 'CD-quality sound' than just providing 16-bit storage and filling it with samples. This problem of interpretation also applies to many other items of digital hi-tech equipment. Many effects units now mention refer to 'CD Specification' (usually meaning 16-bit conversion at 44.1kHz) and then go on to talk about the storage or 'internal representation' as being 24 or even 32 bits.

What actually matters is how good (or should that be bad) is the worst part of the whole system. Having gone to considerable trouble to perform all the analogue-to-digital conversions with 16-bit accuracy, it would be silly to store the resulting samples with only 8 bits in memory, since we would be throwing away a great deal of the signal. In mixing terms, this would be rather like only ever moving the sliders in a mixer just up off the infinity line, and then using lots of gain at the output to compensate: the quality and noise would suffer dreadfully! The total system cannot be better than this one 'worst' part — if the DAC only has 14-bit performance then we can only get 14-bit audio at best.

Witch hunts to eliminate the component with the lowest quality only reveal the next worst part. In musical terms, a whole performance can be ruined because the bass player plays out of tune and can't follow the drummer's tempo, but replacing the bass player then reveals that the keyboard player has problems with triplets and times other than 4/4... What the designer of a piece of digital audio equipment attempts to do is choose the best and cheapest parts which are consistent with the overall performance and final price that is required. This means that for 16-bit audio quality, the lowest quality component has to be slightly better than 16 bits in performance terms, but the rest probably do not need to be much better than about 17 or 18 bits.

An interesting analogy is to compare the process with shooting a film or video, or even recording an album — much of the material that is recorded is later discarded, and often a compromise has to be made because it is not possible (or affordable) to re-record something. The end result is the optimum that can be achieved within the limitations of time and money; this applies to samplers, music, and many other aspects of life.