of Microprocessors

DIRECT COMPUTER SYNTHESIS METHODS 115
sound complexity and other factors. Thus, the universal approach is to
compute the samples at whatever rate the program runs and save rhem on a
mass storage device capable of holding at least a couple million samples. An
IBM-type computer tape drive is probably the most economically suitable
device. A single $20 reel of tape can easily hold in excess of 10 million 16-bit
samples, giving an uninterrupted running time of over 3 min at 50 ks/s.
Large computer centers can also be expected to have enough disk capacity to
hold a sizable number of samples, which offers several advantages over tape.
Even the floppy disks commonly used in microcomputers can be a practical
storage medium in experimental systems, although lower sample rates must
be used and a single diskette can only hold 15-30 sec of sound.
After the samples have been computed and stored away, they must be
passed through the DAC at the final intended sample rate. The obvious way
to do this is to read the samples back from the mass storage device and
transfer them at a constant rate to the DAC. For high fidelity, however, the
sample rate must be rock steady; even a few nanoseconds jitter will increase
the background noise level substantially. Thus, the DAC would typically
have its own crystal clock and at least one sample buffer. The computer
servicing the DAC for playback must be able to provide the next sample
before the current one is finished. What this all means is that the computer
must usually be totally dedicated to the task of reading samples and sending
them to the DAC. Unfortunately, the operating systems of most large computers
are unable to insure uninterrupted execution of a single program so
music playback must be done in a stand-alone mode.
Often the expense of monopolizing the resources of a large computer
cannot be borne so the disk or tape containing the samples is played through
a DAC connected to a minicomputer. In the past, it was fairly common
practice to build a specialized hardware device for reading sample tapes. The
design of such a device was complicated by the fact that data on computer
tape is grouped into records of perhaps 3,000 samples each with gaps of
several hundred samples equivalent between. The playback device thus required
a substantial buffer memory to insure the uninterrupted flow of data
to the DAC. Of course, mini- or microcomputer installations for direct
synthesis should not experience any of these problems.
Sometlmes the tape or disk dnves available on smaller systems may not
be fast enough to provide the desired sample rate. This limitation may be
compensated for by running the DAC at half of the desired sample rate and
operating the audio tape recorder at half of the intended tape speed. Then
when the audio tape is played at full speed, the desired sample rate will have
been attained. One difficulty with this approach is that the high-frequency
equalization of the audio recorder is switched according to tape speed, which
would be expected to alter the high-frequency response somewhat. A more
severe problem is that during recording bass frequencies down to 10 Hz will
be generated but may not record. When played back at double speed, a