Godot Posted March 18, 2021 Posted March 18, 2021 1 minute ago, Eggmeister said: Were you blindfolded ? Doubly! 2
Ittaku Posted March 18, 2021 Posted March 18, 2021 10 hours ago, rmpfyf said: Random interrupt performance is worse than periodic performance. This is OS related jitter, not a line on output spectra. Is this documented or measured somewhere showing this effect and it having an audible effect or the result of your own experience/opinion/theory?
Ittaku Posted March 18, 2021 Posted March 18, 2021 2 minutes ago, rmpfyf said: My own and most RTOS Devs would concur - used to stick tracers all over my kernel builds for this sort of stuff But the audible noise effect would go against what normally happens with audio related noise issues. Anyway you're basically saying you're postulating it based on what you know about interrupt noise and making some pretty large leaps between that and making for audible issues. Can you not see why we're all struggling to see eye to eye? 1
Ittaku Posted March 18, 2021 Posted March 18, 2021 Unfortunately to me, Swenson is not high on my list of trustworthy sources. 3
davewantsmoore Posted March 18, 2021 Posted March 18, 2021 2 hours ago, rmpfyf said: First is the notion that AM or anyone else 'proves' anything. That's why I'm asking questions.... move on. 2 hours ago, rmpfyf said: There is nothing given here and as such all we can reasonably deduce is that for whatever the limits of the method or sensitivities involved, what was observed in testing did not exceed these. Indeed. So - does that mean that if you're using a bogo standard DAC/ADC loop like he is ...... then "the cables don't make any difference" is a fair conclusion?... or no? 2 hours ago, rmpfyf said: Second is that the J-test is worst case. Well.... The J-test is completely inappropriate for examining Ethernet, which uses packets. (Strange nobody has picked that up). 2 hours ago, rmpfyf said: Again, it's not. It is intended to give a result that is simplest to visualize, so you get 'shows up as worst case for that magnitude of jitter input for a digital signal' but that's not the same as 'this is the worst jitter you will experience'. If you fed a source a wide-bandwidth signal (e.g. music) you could reasonably expect there to be some temporal distortion of all freqeuncy components within the signal Limiting the discussion to only data based jitter, which is what Archmagio was doing..... then if we are using SPDIF (which he wasn't.... but assuming we were) ...... then I think you are wrong. The J-test, DOES, result in the worst inter-symbol interference possible...... assuming that the DAC doesn't have some other major data related quirk (which is usually a fair assumption, to my understanding). What other data based jitter could there be which is possibly worse?! (Dunn is quite explicit that it is the worst case. What has he missed?) 2 hours ago, rmpfyf said: To wit an earlier comment you made the total distortion is not xx dB at yy Hz, you need to add the entire distortion above the noise floor Yes. 2 hours ago, rmpfyf said: In practice I would think it'd still be small. That was my point.... I just looked at the major spectral components, and see that adding the others to it, isn't going to make it substantially higher. 2 hours ago, rmpfyf said: As it is typically randomised across channels, my experience is that the magnitude change needs to be very significant before you'd pick a tonal difference - but that a percieved difference in the accuracy of the stereo image Are you saying that you think that small amount of distortion correlated into one stereo channel are significantly more audible.... than say listening for distortion in one channel/speaker (ie. mono). I have seen quite a bit of direct and indirect research which agrees with that. Have you seen this? https://www.audiovero.de/en/acouratecleaner-xlr.php Encode into mid-side stereo.... DAC .... then an analogue circuit to return the MS-stereo to regular stereo.
davewantsmoore Posted March 18, 2021 Posted March 18, 2021 3 hours ago, rmpfyf said: but that a percieved difference in the accuracy of the stereo image (be it 'holography' or whatever other descriptor we might use) is noticed well before. This is consistent with audibility limits. I still do not think the what was shown in the archmagio charts. (ie. the extra components being so low..... except for what appears as "spread"... which is identical for each test - ie. it's in the baseline) .... shows anything which is going to be audible under any circumstance.
GaryT Posted March 18, 2021 Posted March 18, 2021 (edited) If I take a step back from all the frequency and time domain analysis of the the signal coming from the analogue outputs (which are subject to all sorts of random fluctuations when you are looking at the -120db levels) and just capture a PCM bitstream output from a player comparing a stream from a server via ethernet (using commodity cable and high end cable) along with a capture of a locally sourced file from a hard disk or flash drive, or a stream of the same exact file played from a remote server via the internet and I perform a bitwise comparison on all of those captures, would you see a difference? Edited March 18, 2021 by GaryT
Volunteer sir sanders zingmore Posted March 18, 2021 Author Volunteer Posted March 18, 2021 3 hours ago, rmpfyf said: A good experiment has a solid undersanding of the limits and sensitivites of the equipment and methods used. that’s a fair point in general. However it seems inconsistent to level it as a criticism when no limits and sensitivities are given here: 14 hours ago, rmpfyf said: ....my ears are more resolute than the test equipment I have and the time I can afford.
Ittaku Posted March 18, 2021 Posted March 18, 2021 16 minutes ago, GaryT said: If I take a step back from all the frequency and time domain analysis of the the signal coming from the analogue outputs (which are subject to all sorts of random fluctuations when you are looking at the -120db levels) and just capture a PCM bitstream output from a player comparing a stream from a server via ethernet (using commodity cable and high end cable) along with a capture of a locally sourced file from a hard disk or flash drive, or a stream of the same exact file played from a remote server via the internet and I perform a bitwise comparison on all of those captures, would you see a difference? No 1
GaryT Posted March 18, 2021 Posted March 18, 2021 Bingo. Anyone interested in a service where you send me your audio files and I transfer them back and forward over a fancy ethernet cable a few times then send them back on a gold plated flash drive? 1
allthumbs Posted March 18, 2021 Posted March 18, 2021 In regards to the Jitter test of 30 years ago it was interesting that in Amir's video addressing that study, he pointed out how they got rid of the cloth eared test candidates and took some time to train the others so that they could identify the "artifact" of jitter in order to prove audibility.
BugPowderDust Posted March 18, 2021 Posted March 18, 2021 7 minutes ago, GaryT said: Bingo. Anyone interested in a service where you send me your audio files and I transfer them back and forward over a fancy ethernet cable a few times then send them back on a gold plated flash drive? I’ve actually heard people claim that bitperfect music sounds different on disparate media types (eg ssd vs flash). There really is no end to this slippery slope 1
davewantsmoore Posted March 18, 2021 Posted March 18, 2021 10 minutes ago, GaryT said: Anyone interested in a service where you send me your audio files and I transfer them back and forward over a fancy ethernet cable a few times then send them back on a gold plated flash drive? This is a flawed way to look at the issue (all clocking, noise, etc. is stripped from the digital audio when it becomes stationary data). It might have been part of the joke.... but it's hard to be sure. 3 minutes ago, BugPowderDust said: I’ve actually heard people claim that bitperfect music sounds different on disparate media types (eg ssd vs flash). There really is no end to this slippery slope It's possible that it could...... but this would imply a pretty poorly designed system.
GaryT Posted March 18, 2021 Posted March 18, 2021 definitely a joke, but a well designed streamer buffers enough of the data (and crc/parity checked) to make it identical to a static file when you look at it bit by bit. 1
Guest Eggcup the Dafter Posted March 19, 2021 Posted March 19, 2021 2 hours ago, davewantsmoore said: It's possible that it could...... but this would imply a pretty poorly designed system. My Oppo 105 did just that - the USB-A input for flash drives was considerably brighter. I thought I was imagining it until (I think it was someone at ASR, can't go checking as I'm at work) someone measured the different inputs on a 105 and got a bad result for that input.. I haven't tried it with an SSD, or with a lot of different flash drives... but I could imagine that it might sound different depending on the cause of the issue. The thing to note of course is that any result on the Oppo 105 (or a bad batch of them which is possible) can't be extended to all properly designed streamer USB inputs. It only takes one popular model with a fault to turn something like this into solid audiophile lore
Guest Eggcup the Dafter Posted March 19, 2021 Posted March 19, 2021 2 hours ago, GaryT said: Confirmation Bias is a powerful thing. Prove it. I do believe the listening process is a flawed tool. But is confirmation bias the only thing that leads to a "wrong" result, or even the most important? If I'm going to be synical or distrusting of system related reasons for reported changes, I'm damn well going to be cynical about audiophile lore as it relates to my brain, or yours.
frednork Posted March 19, 2021 Posted March 19, 2021 6 hours ago, rmpfyf said: As it is typically randomised across channels, my experience is that the magnitude change needs to be very significant before you'd pick a tonal difference - but that a percieved difference in the accuracy of the stereo image (be it 'holography' or whatever other descriptor we might use) is noticed well before. This is consistent with audibility limits. For me this is the gold and correlates with what I perceive. And no tests are done on interchannel variation in most testing regimes so not picked up. For me the fragile phantom image created by two separate sound sources is the thing that changes most with cables or clocks. 3 hours ago, davewantsmoore said: Have you seen this? https://www.audiovero.de/en/acouratecleaner-xlr.php Encode into mid-side stereo.... DAC .... then an analogue circuit to return the MS-stereo to regular stereo. Have seen this before and wondered whether it might fix things but then forgotten amongst the haste to waste money on unnecessary cables and such. I think it could be a very elegant way to remove the interchannel variation and hopefully it's own flaws are less audible. Really didn't think this thread would provide anything but trolling opportunities and angst (although I liked the parrot bit). The "test" performed 6 years ago was done with stuff lying around and I would be surprised if I would hear a difference with the samples he chose So not the best planning. It's often said that reviewers are suspect due to their income streams. I think the same can be said of Amir and (Archimago) and frankly he is in too deep to flip 180 degrees now from his entrenched position. His audience need more "proof" that audio is mostly a rip off so he supplies it and takes the money.
Recommended Posts