Posted By
TLC
on 2007-01-10
16:56:23
|  Re: SID-supported software ???
Bionic: you were right in that subject of sigma-delta dac's...
I confused two types of d/a conversion by each other... which are similar in principle to the point that at the heart, both are based on '1-bit' d/a conversion, but are different otherwise.
The method of d/a conversion I was referring to first was in fact "delta modulation". There, it's true that the output can be either increased or decreased by a single step in a cycle. The more the steps, the higher sampling frequency is needed to achieve high slew rates.
The delta-sigma dac is not like that... the output of this dac is a high frequency digital stream of +vref and -vref, whose "average" gives the signal. These (amongst other things) utilize digital filters to achieve "noise shaping", ie. reduce the quantization noise in the passband at the price of generating excessive noise above the passband (which is said to be easy to filter out by a simple rc filter).
...My nose still tells me not to use delta-sigma DACs. IMHO, the only, inherently linear _and_ inherently low-noise D/A conversion method would be the delta modulator way (if operated at a fast enough internal frequency). Delta-sigma modulation, at this sight, asks me the same question that originally multibit dac's were sweared at... For multi-bit dacs, it was said (amongst other things) that it's almost impossible to build a feasible reconstruction filter (ie. it's hard to build a filter that has good roll-off, so that everything above the passband can be filtered out, and has low phase-distortion + ringing). With sigma-delta dac's, it's said these are unneccessary... even a simple rc filter will do... But for me, it asks the same question. What would happen with the generated, high-level, high-frequency noise of the sigma-delta dac?... It's unlikely that these could be filtered out by a simple RC, if it's close to the passband and it's "strong" (as it's seen even on the diags of the manufacturers). ...Are these dac's smart enough to shift the noise spectra _so_ high that there's no (or low) noise even far beyond the passband?... I'd have a couple of other questions left, but that's a different story.
|
