"Jay - atldigi" wrote in message


In article , "Arny Krueger"
wrote:

"Chris Hornbeck" wrote: steps.

Dither is just noise, but noise has a special property in this case.
Although it can't smooth out differences, it can remove errors.

Dither doesn't remove the errors, it just makes them more palatable
to the ear.

It can remove distortion but the error still exists as broadband
noise.

Well, we call it noise but in fact it's 100% deterministic given that we
created the randomizing signal so we should know what it is.

I think Chris probably knew this but chose the wrong word.

Could be. But the point needs to be clearly made.

IMO, there's a lot of etymological weirdness in this area. Quantization
error is often called quantization noise. Spectral shaping of quantization
error is commonly called "noise shaping". Quantization error is noisy, but
it's noisy in the sense that loud neighbors are *noisy*. It's not noise in
the sense of random noise, because quantization error is 100% predictable.

I became impressed with how deterministic quantization error can be when I
started looking at measurements of a series of samples of the same sound
cards and saw how consistent they are. IME pure analog gear is not this
consistent. The consistency comes from the fact that the quantizers in the
converters were the major source of noise, and the quantizers were all
digital and therefore extremely similar.