| Basically, a sampling keyboard generates its sounds by playing back digitally
recorded pieces of sound. The term "sampling" comes from the technology
used to digitally record the sound. An input waveform, from whatever you
want (piano, voice, Chinese gong, etc.), is fed into an analog to digital
converter which measures the amplitude of the waveform very frequently.
The "sampling rate" is how many times per second the A/D converter measures
the amplitude of the input waveform. The output of the A/D converter is stored
in digital memory, which can be stored off to floppy disk for storage.
The rule of thumb about sampling is that the sampling frequency should be
a minimum of twice the highest frequency of the waveform you are sampling.
For example, if you record an instrument which has a frequency range from
20Hz. to 10Khz., you should sample at a rate of at least 20Khz to be able
to accurately reproduce the sound when you play it back.
Playback is accomplished by putting the digitally encoded sound through a
digital to analog converter (D/A or DAC) and then through some filters to
smooth out some of the error in the conversion process.
As you can imagine, storing more than a second or so of information gets
expensive in terms of the amount of memory you need. Techniques such as
looping allow you to play back a sound which is longer than the sound you
recorded by taking a piece of the original sound and appending it on to the
end of the original.
By the way, my understanding is that to get different pitches from a basic
sound, the clock frequency of the D/A on the output is varied. To get a
higher pitch, use a higher clock speed; lower pitch, lower clock speed.
The strength of sampling keyboards is that they can produce very realistic
sounds, because they use real digital recordings. The real expensive ones
(Fairlight, Synclavier, Emulator II with Macintosh and extra software) allow
you to edit the sounds graphically and then play them back. By doing this,
you can generate some really off the wall sounds. This type of capability
is also getting added to some of the more reasonably priced synths by using
external computers like an Apple II.
The drawback, from what I can see, is that machines which have high sampling
frequency (which means in most cases better sound) and lots of memory and
editing capabilities are quite expensive. Keyboards like the Ensoniq Mirage
and the new Prophet 2000 are making sampling much more approachable for the
average guy like me.
Please, someone correct me if I messed up the details along the way.
Derek
|
| Quite correct. I would just add this:
Pitch variation is generally not done solely by clock-speed changes to the D/A.
The observed pitch can also be varied by skipping samples. This is done when
creating a higher pitch would require a clock rate beyond the hardware
capability. Remember, moving up one octave requires doubling the frequency.
Thus, a four octave shift up (from C below middle C for example) would require
16 times the original sampling rate. At 44Khz (the generally accepted high
quality rate) you'd be into 704,000 D/A conversions per second. Forget it.
Also, the high quality machines don't do any of this at all. The reason is that
real acoustic sounds do not display uniform harmonic content at all pitches.
The human voice is a prime example. If you sing into a sampler and play back
using a keyboard at different pitches, you hear an awful lot of munchkinization
or death-booming depending on whether you go up or down. The right way to do
sampling is to sample the waveform at many different pitches, reduce the
samples to their dynamic spectrums using Fourier analysis or LPC, plot
various spline curves that define the change in harmonic content versus
pitch, then use all this information to interpolate the correct spectrum
for any given pitch. Doing this in real time obviously requires some pretty
incredible hardware, thus the high prices. Having played with cheap (<$2000)
samplers myself, I would offer this suggestion - don't bother.
Ted
|