[sdiy] RE: [AH] control voltage was/ Keyboard in space

[Czech Martin]

Most books say that quantisation produces some kind of hissing noise.
This is sometimes true. But if you quantize a narrow band signal
(all musical instruments in the decay phase) you can better
describe this as distortion.

This is a trap some people fall into some time.

I know some guys who wanted to design a RF interface with ADC.
The theoreticians thought about their books and figured out
that 9 bit resolution would be similar to most existing
analog solutions. Because all artefacts would drown
in the hissing noise.
NO! This is narrow band and so all kinds
of distortion and mixing products came up, sharp little
needles in the FFT diagram.
At first they tryed to blame the poor analog designer of the
ADC, but he made a very good job, the ADC made effectively 8.7 bits
or so. 
So they had to go for 12 bits, of course this causes a lot of
problems for fitting into an existing chip, and also the
distortion products vary a lot from sample to sample (chip).
This is because of component mismatch.

Still the performance is worse then it was in 30 years old
bipolar analog chips...

m.c.

>of slots. This is called "quantization" and produces quantization noise,
>also called zipper noise. Binary encoded systems also need to be "sampled"
>which is time limited, and produces "aliasing" noise. These additional noise
>sources can be minimized with proper design.