With respect to the digital volume control James, I would expect that with a 16 bit/channel input, and 24 bit dacs -- assuming an 8 bit volume index (1 -- 256 --- yes, not 0 - 255), and a 8x24=32 bit multiplier, and a shifter, there would be no signal loss at all in the digital domain.
out = (in * volume) >> 8; // >> 8 is the same as dividing by 256.
In the case where the volume is at max (256) the output is the same as the input. Where the volume is the minimum -- 1, it is the same as shifting the bit pattern right by 8 bits -- the bits are actually left entirely intact -- they're just shifted over -- much like taking a base 10 number and dividing by 1000 -- the digits would be moved over 3 places, but would otherwise be unchanged. Assuming the source is 16 bits, it just shifts from the high 16 bits to the low 16 bits of the 24 bit word. All other intermediate volumes will be as correct as they can be in 24 bits.
So with the (quite phenominal) S/N ratio you have for this DAC I would expect that a digital volume control (computed as described above) would not cause any bit loss in the signal. So long as the operation takes constant time regardless of the input bits, there should be no additional jitter either. My rough, back-of the envelope math gives a noise floor of -92db below the original unscaled input signal when at the lowest possible volume setting (assuming -140db noise floor as published). Once the volume setting hits 1/32 or more, you're well out of the signal loss zone for a 16 bit input.
In any case with 16 bit/channel (or even 20 bit/ch) inputs, having 24 bits of DAC precision offers *plenty* of guard bits -- enough to ensure that any errors in volume computations are well below the precision of the input -- in fact even with 24 bit inputs any volume computation errors would almost certainly be somewhere very close to that ridiculously low noise floor you've managed to achieve. (I mean come ON -- the least significant bit of a 24 bit sample should be down at around -144db or so, and here you are at -140 S/N? How the heck did you guys DO that? I don't think I've ever seen someone make a power supply and dac combination that actually achieves that close to the theoretical precision of a 24 bit/sample signal -- at least not for a price within a few factors of 10 of what you have done.)
