I wanted to come back to this post because it seems there are two issues being described here:
- 705,600 and 768,000 have a lowest common multiple of 112,896,000, whereas 44100 and 48000 have a lowest common multiple of 7,056,000, yet numbers "near" to 104,000,000 are not multiples of 7,056,000. The closest are 98,784,000 and 105,084,000, and then the next multiple is, ta-da, 112,896,000. So the "~104MHz" you use in the DACs and ADC force you to use a fractional interpolation.
Is the ~104MHz clock a side-effect of having to use an off the shelf crystal? Presumably there is no crystal at 112.896MHz?
You appear to imply that 44.1KHz or multiples (e.g. 705.6KHz) are exact divisors of the ~104MHz clock. e.g. a clock of 103,723,300 is 705.6KHz x 147. So fractional interpolation is not required in this case, only for multiples of 48KHz. Is that right?
- Studios require that a clock that's either a multiple of 44.1KHz or 48KHz is sourced externally to the ADC, so you can't rely upon a crystal inside the ADC? The external clock causes noise problems and jitter.
So this would appear to imply that 44.1KHz (or 705.6KHz) output suffers solely with problem 2, but 48KHz (or 768KHz) suffers from both problems 1 and 2. Is that right?
But, in this post:
https://www.head-fi.org/threads/watts-up.800264/page-104#post-15312499
you say:
Which implies some multiple of 7,056,000 doesn't it? If that were true then 705.6KHz and 768KHz would both be integer factors of 112.896MHz, but not "~104MHz". Which would then imply that problem 1 disappears entirely.
But you say that you are using a fractional interpolation filter, which implies problem 1 exists for 48KHz and all multiples up to 768KHz (since you describe the problem in terms of 768KHz output).
So I can't resolve the apparent contradiction. And, honestly, I'm curious why 112.896MHz isn't used instead, unless there simply isn't a crystal.