I would not, but it would not be fair to assume that the audible effects of jitter are not masked by anything at all.
First of all, what audible effects of jitter? You haven’t determined there are any and Secondly, audio masking only happens in the ear and requires a similarly or higher level signal to mask it, both of which would be apparent in various measurements or more obviously in a null test.
If we see improvements in various areas, it's worthwhile to check whether this may have been a bottleneck in previous testing would it not?
Assuming there are any audible bottlenecks.
But posts in the realm of "Well I don't think it matters" or "No one else has shown a difference so why are you trying" are not constructive at all.
If any improvement is below the level of audibility then it doesn’t matter as far as listening is concerned. It will only matter to those who place importance on better inaudible specifications. I didn’t say and don’t believe that you shouldn’t try just because no one else has shown a difference. If you did really demonstrate a difference, I for one would find that interesting, although I wouldn’t know if that result had any practical implications until it was investigated formally and reproduced by others.
In regards to methodology, this is the setup:
I have a few issues/potential issues of varying concern.
Why are you converting to 705kHz? With such a relatively high data rate, there’s the potential concern of output/transfer errors. EG. Are you hearing differences between filters or errors due to an abnormal digital audio transfer rate, from your computer or DAC? This isn’t a big concern because even if there are errors they should be random and not bias the result but it is potentially a point of failure (of methodology).
The Holo May is also a slight concern. The R2R topology is one of those esoteric/pathological design topologies which can indeed be audible. It’s only a slight concern because the Holo May appears to perform virtually as well as standard topology DAC but again, why not eliminate this potential “point of failure”/criticism if you can?
A more serious point of concern, please state the test SPL reference. Obviously, one could ABX a small, quiet section (say the fade out), whack up the amplification by XdB and then pretty much any artefact at any level can be made audible.
In your second section, I think you should not combine tests, say a filter + dither test. Test for them individually, the idea being to try and make sure that the only single variable changing is the one you’re testing. Having 2 simultaneous variables introduces more potential points of failure and if you manage to falsify the null hypothesis, you don’t know why, which of the variables caused it.
Again, the SPL reference please. TDPF dither is trivially easy to hear (and distinguish) at even just a fairly modestly elevated level.
Also, very high and “stupid” number of taps, are you absolutely certain you’re actually testing the difference between filters and not the difference between artefacts caused by the software/computer while trying to calculate them?
G