I've been asked to post some things about my research/dev with the FV-1.
A little about myself. I've worked with delta sigma converters for the last 15 years. I've gotten some patents. I've built products for pro audio (recording), guitar, and stage/DJ/live performance, including custom stuff used on albums you've no doubt listened to. The FV-1 was my first attempt at DSP. Enough boring things, though. Let's get to the good stuff.
Because my FV-1 design is currently being used in a commercially available product, I can't share specifics with this forum. That means no schematics, no measurements, no Bode plots. I can point everyone in the right direction though. I hope you understand.
First, the thermal noise. I've seen people make comments about this. I think the published spec is -97dB typ? I can't remember. In my experience, If you're using a standard 78xx regulator in your design, it's going to be noisy. Consider using a modern LDO or an LM431 shunt, and split the supplies into separate VDD and VCC. Using split power supplies, I was able to get the noise floor so low that it's inaudible at max gain. this includes two opamps, one of which is a makeup gain stage. I built my power supply using some research that Philips EMC Labs in Eindhoven published some years back, so my results might differ from yours.
Second, I experimented with a variety of power supply levels in my initial attempts to control noise. I found that 3.3v offers the best sonic performance, and keeps the FV-1's internal stages from overdriving. For reasons I'm not sure of, it really opens up the mid-range response as well. Maybe this has to do with the internal decimation filter behavior?
Third, use the fully differential IO, otherwise you're robbing yourself. You can tie your inputs together, and sum the outputs into a differential opamp stage. The difference in sound between single pin IO and differential IO is pretty amazing.
Fourth, the watch crystal clock is awesome. Seriously. I tried all kinds of fancy approaches including third order, wide Q XO's with fancy divider networks. The watch crystal sounds the best IMO.
Fifth, use a four layer board. This is technically a mixed-signal design, so four layer boards are your friends. If you're trying to save money on your design, consider using JLC PCB. They'll do low-volume four layer boards with parts placed for super cheap.
Anyway, I hope this helps someone here.
. Because I've worked with mixed-signal designs for a number of years I simply used circuitry that I've come to rely on. I was pretty amazed at the difference in noise performance.