mmerrill99--have you ever actually *done* an experiment in your life? I do them daily for my job and they cover a broad spectrum of biology, materials science and engineering. There are certainly different protocols for each experiment (duh) but a common mindset to how you build up a test. It's the same sort of mindset, with its field-specific caveats and adaptations you see in preference testing (the food guys seem way ahead of everyone else) or even clinical trials. It's also why when I read any technical paper I read the abstract followed by the methods section to figure out what exactly they tested versus what they claimed to test.
Given anything about coherent you've written about designs of experiment has essentially parroted or directly quoted someone else, I do wonder if you have any actual understanding. I haven't seen you suggest a design that would reduce the objections you have to present testing. And your position otherwise completely lacks nuance that would suggest you realize that any experiment you do has some sticky bits and non-idealities which are not dealbreakers (sometimes you do have dealbreaker non-idealities). EVERYTHING involved with experiments has some "you have to be careful" steps. Needing to double check whether you're running in 24 bit vs 16 bit or if a accidental dither/truncate ended up where you didn't want it to is a part of the "you have to be careful" business.
It's also why one doesn't take internet abx tests too seriously. But there's nothing necessarily wrong with ABX accepting its strengths and limitations (more degrees of freedom mean fewer chances for guessing at the cost of higher mental load lowering the overarching sensitivity, which may be partially mitigated by allowing for an individual to switch back and forth a number of times before choosing)
Given anything about coherent you've written about designs of experiment has essentially parroted or directly quoted someone else, I do wonder if you have any actual understanding. I haven't seen you suggest a design that would reduce the objections you have to present testing. And your position otherwise completely lacks nuance that would suggest you realize that any experiment you do has some sticky bits and non-idealities which are not dealbreakers (sometimes you do have dealbreaker non-idealities). EVERYTHING involved with experiments has some "you have to be careful" steps. Needing to double check whether you're running in 24 bit vs 16 bit or if a accidental dither/truncate ended up where you didn't want it to is a part of the "you have to be careful" business.
It's also why one doesn't take internet abx tests too seriously. But there's nothing necessarily wrong with ABX accepting its strengths and limitations (more degrees of freedom mean fewer chances for guessing at the cost of higher mental load lowering the overarching sensitivity, which may be partially mitigated by allowing for an individual to switch back and forth a number of times before choosing)
How do you prove a test is sensitive enough to tell two identical files apart?
diyAudio member MaxHeadroom worked on that for a while in September-October 2016.
Help Needed Please - Copy/Paste Batch File
Can I just clarify here. We are talking about the amplifier noise floor being modulated, not the noise floor of the music (which is often tens of dB above the amplifier noise floor).
A good amplifier noise floor should be low enough so that the noise so far into the weeds that I really cannot see a mechanism for audibility.
Let' talk about post rather than people, OK?
My playing devil's advocate with experimental setups is what makes them more robust.
You can disregard this analysis of experimental weaknesses ( as you seem to want to do constantly) - that's your prerogative but don't attack me for pointing them out.
I pointed out a flaw in your statement about how you find unbelievable this idea about noise modulation.
Again, I'm simply pointing out the weakness of home run ABX tests, Foobar in particular.
I would advise to try HQPlayer.
You can do your Poo-ging
with the various modulators and filters etc.https://www.signalyst.com/consumer.html
It takes control of OS and shuts everything that doesn't need to be running.
It is pretty much the best quality audio you can get out of a computer ATM - the rest is hardware dependent.
It has far superior digital filters, modulators and OP's whatever format you desire.
DSD up to 1024 x and PCM up to 32 bit / 1.5MHz.
I know people who have achieved really high quality sound with a fairly cheap DAC, for example, by using HQP
and converting to DSD256, as such most of the DAC's filters etc are bypassed. There are many possibilities.
T
Last edited:
I thought I've been trying to make some useful suggestions. I honestly was asking how one determines the low limit on the sensitivity of a test like this.
Richard may want to do other things besides run a player. Maybe speaker cross-overs, and corrective EQ, for example. Does HQPlayer do all that too?
Likely at least part of the reason Earl said new tests would have to be developed for the 5%.
Also, before deciding on files to compare there would also be the issue of what hardware would be needed. DAC-3 and AHB2? What else?
Phase coherence is a quality of the signal. Google has a zillion explanations.
I have no idea how an audio amplifier may or may not be phase coherent. A marketing term?
Then PMA is wasting his time, as it is folks opening foobar and plugging a headphone into their laptop are included.
Surely that's jumping the gun a bit and assuming that there is something special about the DAC3 that only the '5%' can hear? Given how much the likes of Geddes, Toole etc poo poo expensive electronics I would be interested how you would persuade them that this was needed over one costing 1/10 the price?
You said that users needed to be warned about Foobar use on Windows - possible resampling happening (not to mention the lowish quality internal soundcard). By this I assumed that you recognize any 'tests' run with resampling are not valid tests. Therefore if such a test returned a null, what is it telling you - nothing, it's not a valid test, Agreed?
I said that there were no internal controls in Foobar ABX to allow us ascertain if this resampling happened. Similarly nothing to tell us if the internal soundcard is suitably useable for the test. Similarly nothing to tell us other possible issues in the running of the test. In general, there are no internal controls for self-verification of the test. Without these basic controls, we cannot even get to the calibration necessary for sensitivity testing of Foobar ABX
I think we've been through the ideas about this before but if there's no internal controls as above, there's really no point in even discussing this as it's not verifiable that during the test the sensitivity that was established at calibration stage is still in operation during the test
Last edited:
No, not at all. Just says most dacs don't reproduce small details well. Small details have to be reproduced accurately before someone tries to see if they can hear them.
Last edited:
- Status
- Not open for further replies.