22.09.2015 Views

of Microprocessors

Musical-Applications-of-Microprocessors-2ed-Chamberlin-H-1987

Musical-Applications-of-Microprocessors-2ed-Chamberlin-H-1987

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

110 MUSICAL ApPLICATIONS OF MICROPROCESSORS<br />

assumption that the wave did not change very much from sample to sample is<br />

not necessary.<br />

Of course, for sound and music a whole spectrum <strong>of</strong> frequencies must<br />

be reproducible, not just single sine waves. Fortunately, a sampled arbitrary<br />

waveform behaves just the same as a sampled sine wave. The spectrum <strong>of</strong> the<br />

waveform is reproduced unaltered and then symmetrical pairs <strong>of</strong> copies<br />

around each harmonic <strong>of</strong> the sampling frequency are also introduced. Figure<br />

4-3 shows the result <strong>of</strong> a full 20-Hz to 20-kHz audio spectrum sampled at a<br />

50-ks/s rate. Since there is no overlap between the desired spectrum and the<br />

copies, a low-pass filter can once again be used to eliminate the copies.<br />

The preceding argument is equally valid for conversion <strong>of</strong> sound<br />

waveforms into numbers. The waveform is first sampled with a balanced<br />

modulator or its equivalent and then the sample pulse amplitudes are measured<br />

with an analog-to-digital converter (ADC),which is nothing more than<br />

an ultrafast digital voltmeter. Each sample thus becomes a number that may<br />

then be ptocessed by the computer. No information about the curves and<br />

undulations <strong>of</strong> the waveform is lost provided that the spectrum <strong>of</strong> the<br />

waveform has no frequency components above one-half <strong>of</strong> the sampling rate.<br />

Dnfortunately, some natural sounds do have appreciable .energy beyond the<br />

audible range so a low-pass filter is needed to prevent these high frequencies<br />

from re~ching the sampler and ADC. Since these frequency components are<br />

beyond the audible range anyway, the filter does not audibly affect the<br />

sound. If signal frequencies higher than half the sample rate are allowed to<br />

enter the sampler, their <strong>of</strong>fspring in the copies will overlap the original<br />

spectrum and will cause distortion. Such a situation is termed aliasing and<br />

the resulting distortion is called alias distortion.<br />

Any Sound Can Be Synthesized<br />

Figure 4-4 shows a block diagram <strong>of</strong> a complete computerized audioto-digital<br />

and back to audio again conversion system. If the two low-pass<br />

filters are matched and have infinite attenuation beyond one-half <strong>of</strong> the<br />

sample rate, the waveforms at points A and B are exactly alike! Since the<br />

waveform at point A is not audibly different from the input, this is truly a<br />

high-fidelity system. The computer could also be programmed to supply its<br />

own stream <strong>of</strong> numbers to the DAC, and it therefore follows that the<br />

computer can produce any sound. Note also that the system has no lowfrequency<br />

limitation, thus allowing signals down to dc to be reproduced.<br />

ANY<br />

AUDIO<br />

SIGNAL<br />

LOW-PASS<br />

FILTER<br />

~<br />

20 kHz<br />

LOW-PASS<br />

~<br />

20 kHz<br />

POINT A<br />

POINT B<br />

Fig. 4-4. Computerized digital audio system.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!