Some initial research would need to be done to figure out what hardware would be required. I can tell you right now that DAC-3 has a sound, and is not perfectly transparent. Since transparent doesn't exist, figuring out how to test the top 10% of the 5% might be complicated, time-consuming, and expensive. That should not be surprising.
It isn't useful to specify particular equipment - what needs to be established is that for each test some internal controls are also produced which show that the test sensitivity is suitable (this includes the listener) - running a calibration (training session) will show what know difference in low level details the test is sensitive down to & a evaluation about how suitable the sensitivity is for the proposed test.
I'm not sure this is particularly widely accepted and certainly not by researchers into what we can hear! (BTW this made more sense before you edited it).
Let's just say that the current accepted threshold is 0.1%THD detection for the 95% (blah blah Gaussian distribution). So amongst the top 5% we might expect some to have a threshold 10x better (0.01%) and so the replay chain will need to be better than that. That doesn't sound like a hard target to meet?
Yes, that is correct. Present thinking about what the top bits of the top 5% can hear is wildly off. That's okay though, we don't have to argue about it now. Any researchers would have to go through a process of validating tests and test equipment. It might take an iterative process of seeking out people like mastering engineers that are respected by other mastering engineers for their hearing abilities, and then working with a small set of people to make a first cut on what is likely to be good enough. If researchers fail to do things like that and rely too much on research based on the 95% and conservative minds at diyaudio then they will fail badly. Again, no point in arguing. Only careful objective research will tell.
Last edited:
A good analysis from David Griesinger of the possible way phase coherence works in auditory perception. Although his focus is mainly hall acoustics & reflections leading to clarity & engagement, there's still lots of good stuff in there
One of the issues which has dogged auditory research in the past & becoming recognized as an issue now, is the historic use of tones, noise & simple test signals - the same issue with measurements, I would suggest - test signals are not revealing enough?
This was directed at your comment from a technical perspective (and John's blind cheerleading of anything that is in opposition, no matter the merit, good or bad, of certain folks positions).
In short, it doesn't add up as something major (if at all), and would show up as grossly muddying transient events that are sitting close to the noise floor far beyond a stationary signal at the same level. Or show up as a sidelobe in a nulling/bridge experiment, where incoherence lights up as a imperfect cancellation. E.g gain up a signal, divide it back down and feed the gained and straight signal into two sides of a bridge. Does anything show up in burst tones that looks like a modulated noise effect?
Can you otherwise propose an experiment that would show this specific effect and be able to manipulate it to be greater and smaller with a modified differential pair?
I guess I'm wondering why, when I make experiments with events barely poking out of the noise floor, why I don't see the effect you're describing? If it's real, lesson learned and I'll need to be more careful about what parts I pick. If it's real and sits so far into the noise floor, cool effect I don't need to sweat, and if it not real, then that's fine as well. I'm just incredulous, it's my natural state of being.
Last edited:
That’s not at all, even a tiny bit, what I was getting at. Zero, nada, zip, squat.
Sorry to interrupt - back to regular programming. I’m out.
Except surely by saying up front you HAVE to have a DAC3 for these tests you are totally failing to be objective from the get go?
Note, just to re-iterate. I have nothing against Benchmark and believe they produce very well engineered products. I just have questions over if that level of accuracy is needed or audible. I also find it interesting that I made a comment some days back on how using DAC3 and AHB2 you can run a very different gain structure from normal which some reports suggest is superior. No one commented despite that being an area where there might be some real stuff happening. Funny old world
I agree for sure but foobar does have setting to assure bit perfect 24 bit data to the DAC if the user makes sure to use them (and their hardware supports it). So Foobar does have these internal controls. The ABX feature is just a plug-in and does no file processing and I find the comments that Foobar ABX itself is "doing" something to the files is silly.
That being said I personally think this particular test would survive resampling and 16bits, but I would never ask you to accept that.
Perhaps I'm lost on what you're proposing? I thought I understood, and had a decent idea about coherence and what a gross change in coherence length would look like due to nonlinearities (at least at one point in my life, that's sadly a less exercised muscle). But I'm wondering why setting up some sort of bridge between two paths (a high and low, respectively, path) would not show where the two paths lose their stationary interference as an additional transient signal, no (in addition to harmonics from the nonlinearity).
I do disagree with your apology -- the coherence length of any discussion in this thread is <1 post, so carry on.
The exact meaning of phase coherence in this context needs to be better defined. I spent literally 100's of hours studying ADSL drivers both in simulation and in the real analog domain with huge QUAM constellations and there were never any surprises or ambiguities. I'll also note they were VERY class A/B sometimes 100:1 output current to bias current and used GOBS of feedback.
- Status
- Not open for further replies.