Could you further elaborate on this?
What is the advantage of using a FPGA over a standard sigma-delta modulator? In my understanding, both are able to provide a DSD-like stream. Aren't DSD ADCs basically analog-driven sigma-delta modulators, after all?
I don't see different implications for what concerns the filtering stage. What prevents you from using a "simple low pass filter" after a SDM?
To your first question "why FPGA", the advantage of an FPGA solution is that it has the potential to side-step the need for a reconstruction filter. The ever increasing number of PCM reconstruction filters is a good sign that an optimal filter design is still an unsolved problem. It's very interesting how many different paths have been taken in recent years to avoid the need for that filter. Examples include what Rob Watts is doing at Chord with his enormous filter tap lengths (eg the M-Scaler), Jim Kinne and team at Exogal (poor them, now in a corner with the obsolete FPGAs in the Comet), and of course all the different implementations of R-2R ladders. Of all these approaches, it seems that Chord's has the greatest potential. Rob is now saying that 44/16 audio files processed by his algorithm using one million plus coefficients (filter taps) sounds better than 192/24. That's fascinating, that he can now reconstruct complex analog waveforms in real time so accurately.
To your second question "using sigma delta to convert to DSD", that would require the use of a reconstruction filter for error correction of the sigma delta approximations, so what you're converting to DSD has already been compromised ("contaminated") by the reconstruction filter you've chosen.
To your third question, the answer is the same as the paragraph above.
T+A has an interesting dual solution in their DAC 8 DSD, where as you know they've implemented two independent circuits, one sigma delta decoder for PCM input, and a simple high pass filter for DSD input. It's too bad they chose to use a hi shelf resonant filter class for their PCM Bezier curve option, perhaps to differentiate themselves in the market but at the cost of that HF bump before the roll-off.
Re my comment about releasing half baked products, I didn't mean to be disrespectful of anyone but I do wonder sometimes whether Paul and his team understood how they were painting themselves deeper into corners with each of the three products I mentioned in my post, and whether they'll later regret that they compromised product quality and even their own reputations in doing that.
- cheers, Tom
PS - In retrospect, “side-step the need for a reconstruction filter” was a poor choice of words, eg “to allow for a less stringent or severe filter” would have been better. With delta-sig filters we’re still choosing between optimal freq response (brick wall) at the expense of temporal response (pre-ringing), or optimal temporal response at the expense of the HF attenuation slope. Maybe DSD is the best solution for now with its linear filter?
