It seems the entire history of stand-alone DACs in hi-fi has been subject to RF noise injection from electrical cables. Certainly with my DAC from the 1990s, optical (non TOSLINK) was the best input and many other DACs from that era also had an optical input that was considered the best.
I think it's fair to say as stand-alone DACs get better, the problems caused by RF become more obvious. Self-noise (whether due to an FPGA or other circuitry) then becomes more important and if there is no external source of noise (e.g. with an optical input and assuming that jitter is completely irrelevant) then self-noise is the entire problem.
(It's worth remembering that Rob's pulse array design is a key technique to reduce self-noise. The pulse array itself doesn't produce noise-modulation as it's a constant noise DAC. The higher the element count, the lower the self-noise, so 20 elements in DAVE is the best of all the Chord DACs.)
It also means it's hard to isolate how much of the sound quality benefit of Blu 2 is due to DAVE's FPGA mostly being "off", since WTA 1 filtering is "pass through, unmodified". Because WTA 1 takes most of the FPGA in DAVE, there is theoretically less noise inside DAVE when Blu 2 is feeding 16FS music. The sound quality benefits ascribed to Blu 2 are the same as those for reduced noise floor modulation, so that makes it harder to tell what improvement Blu 2 brings doing solely due to the extra taps.
We would need a DAC with no WTA 1 and then an upsampler that has two output modes: 16FS with 1 million taps and 16FS wtih 164,000 taps.
The closest we have is that Blu 2 has a "2/3 of a million taps" mode, where the output is 16FS (video mode). This is a non-WTA filter. It should sound worse than the WTA 1 filter in DAVE (because it's not WTA) but apparently people prefer it to DAVE stand-alone. That might indicate that DAVE's self-noise is a very serious problem.
A non-WTA filter should never sound better than a WTA filter, no matter how many taps it has, if WTA is the correct approach. But, the video mode has "0.1s" delay (about the same as DAVE):
which might imply that it incorporates DAVE's WTA filter: 82,000 taps from the future and then about 600,000 taps from the past, with 82,000 of those taps being DAVE taps. So that would make it a pseudo WTA filter: DAVE for about 0.05s into the future and DAVE++ from the past, 0.05s of DAVE WTA, then other taps. So then it's a question of whether the asymmetric nature of the filter degrades it to sound worse than DAVE or other pure WTA filters (which are trying to be a sinc filter, which is symmetric). If not, why is it that an asymmetric filter can sound better than a symmetric filter?
Regardless, the video mode versus WTA mode in Blu 2 is the closest we can get to discerning how much of a problem DAVE has from self-noise.
Ideally we would be able to listen to a pulse array DAC which only has a 2048FS input. The self-noise would be solely due to crosstalk between the channels and the element count. Then we would know what sound quality problems the FPGA in DAVE is causing.
It's possible to do a version of this experiment if you have Blu 2, DAVE and Hugo 2. Since DAVE has dual-data mode output and since Hugo 2 has 16FS input using its dual-data mode, you can listen to Hugo 2 with either 1 million taps upsampling from Blu 2 or 164,000 taps upsampling from DAVE. But you will need a lot of ferrites (until adding more makes no difference) on the cable going into Hugo 2, since DAVE's BNC outputs are much noisier than Blu 2's.
Malc, you have everything (except the cable for Hugo 2?) required to do this experiment, don't you? Obviously Hugo 2 has a lower quality of pulse array, so it won't represent DAVE, but it's still potentially an interesting experiment.
@ray-dude also has the electronics needed for this experiment, I believe. Qutest is the other option instead of Hugo 2.
The real problem is getting the data into Hugo 2/Qutest without adding so much RF noise that it swamps the comparison being sought.
Now playing: Max Cooper - Order from Chaos