Have you looked at those tracks with something like Audacity? It has a digital clipping detection tool. Many times the track is limited just below clipping. I got a demo track that was severely clipped but had been processed down so the final version was not digitally overloaded. It still sounded awful. I sent it back to the creator and had him make one that was not clipped. He also sent one with "only" 5 dB of overdrive. The unclipped track sounded quite good. The processed track was still awful. This was synthesizer stuff and it was still really obvious. You don't really know what it would be like unclipped until you hear the unclipped version.
Modern production values unfortunately dictate small dynamic range and loud levels. Both really unnecessary with 24 bit systems.
I have long wondered how you can really test for the audibility of ultrasonics when its so difficult to capture them and reproduce them. And how accurate do you need to be in the reproduction of the ultrasonics to verify any thesis? The same would hold true for distortion as well.
Mr. Simon,
It is hard to fathom a 5 degree phase shift in the electronics over the entire audible range as being that great a problem when looking at the phase shift in a common crossover network or speaker that can be greater than 180 degrees at times. I do believe that this has much to do with sound quality and keeping phase shift and group delay to a minimum, but 5 degrees seems to be bordering on non detectable except in measurements before ever reaching a common loudspeaker device. How would you ever test for this in a double blind listening test without a playback system with less than this degree of error inherent in the test setup?
It is hard to fathom a 5 degree phase shift in the electronics over the entire audible range as being that great a problem when looking at the phase shift in a common crossover network or speaker that can be greater than 180 degrees at times. I do believe that this has much to do with sound quality and keeping phase shift and group delay to a minimum, but 5 degrees seems to be bordering on non detectable except in measurements before ever reaching a common loudspeaker device. How would you ever test for this in a double blind listening test without a playback system with less than this degree of error inherent in the test setup?
Regarding the last sentence above, since the phase shift varies with frequency then it seems clear that actual distortion would result, simply because the Fourier components would no longer all be correctly aligned in time. With phase shift that increased with frequency (everywhere), the faster an edge's rise-time or fall-time was, the more distorted the result would be. With a square wave test signal, the result would look like ringing. But all amplitude transitions containing frequencies above where the phase starts changing would suffer some distortion.
f_max = 1 / ( π · t_rise)
Last edited:
This morning I listened to an hour talk by Keith Johnson, on digital limits, with slides, that he did for a conference. I go along with Keith, not with Scott.
I think the EMI radiation and sensitivity is the next big issue and a place where significant improvements can be had. 25 years ago the noise from a digital display on a CD player was obvious and annoying. Today there is so much emi that the display is ignored. I suspect its every bit as much of a problem but swamped by all the other stuff. Getting a really emi free space is difficult and while CE has specs for immunity and radiation they are only triggered when an obvious failure is detected, not when the sound is degraded.
Oohashi et al. did brain imaging tests _and_ psychoacoustic preference tests (conducted double blind) as well. And for the statistics of listening test result they used Scheffe´s method to guard against false positives which is one of the most conservative ones.
I wonder if there exists a misunderstanding about the "non-auditory sources" because in a later publication they researched the role of bone conduction which was still related to the auditiory system.
Afaik no one else tried to replicate their experiments; as stated before, results of other studies were inconclusive, which is imho obviously related to the difficulties to keep the test parameters seperated (see for example our past discussion about possible IM products falling into the audio band) .
But to reduce all the efforts to "anecdotal evidence" is quite unfair.
I have to heartily concur. (And I think that would also be a much better subject of discussion that the Nyquist-related stuff.) I worry most about subtle, non-obvious degradations of audio output quality, from EMI, which could span almost from DC to daylight, since the obvious, "large" effects should be relatively easy to discover and mitigate; things like changes in transistor circuits' DC operating points, somewhere deep inside a chip, due to RF infiltration and rectification.
With a different EME (Electro-Magnetic Environment), suddenly a high-end system you designed might become significantly less-so, but not be so much worse that a user would realize it was not as designed and probably wouldn't dream there was some external cause, and thus your system's reputation could be injured, while, also, the user would not be able to experience the system's full glory. They wouldn't realize it should be better and you wouldn't realize their experience was degraded. Not a good situation.
The most-obvious remedies and mitigations seem to mainly include physical layout, shielding, and filtering, not counting layouts inside the chips themselves. "Minimize enclosed loop areas. Don't make antennas."
"Pay particular attention to inputs. Everything is an input, for RF."
Last edited:
Its possible that the effects of emi are contributing to the "sound" of some well respected high end systems.
I would say there is no doubt about that. And every location's electromagnetic environment is different. So different users probably each hear a slightly different "sound", from the same system. Then they try, remotely, to compare their experiences, or maybe worse, the results of their various tweaks.
It's a complex issue. Some examples, with the magnitude of measured effects, would be useful, because, as usual, I, at least, don't know how significant the effects might be.
It also affects the answers to questions about where to apply resources, during development. e.g. Would the resources expended in getting the last xx% improvement in theoretical sound quality be better spent on mitigation of EMI effects. It would, if the EMI effects could possibly swamp the last xx% improvement gained from other means. Or should both be done? That's an over-simplified example and it depends on the type of project. But there are usually trade-spaces that must be navigated.
Last edited:
and Keith just so happens to be in the business of selling hires audio recordings ... not to discredit the man, thats not my intention at all. it just seems that we are not talking about freshly broken ground here.
for the record I use hi-res when I can, but I think bit depth is more important than sample rate for solving most of the problems Keith brings up. such as dynamic headroom etc. things start to go downhill again in most cases after 96khz or max 192. though my dacs are capable of 32/384
Gordon appeared to be squirming a bit there in the background (as was I in some parts)... i'm wondering where these small bits of information are disappearing in these stages down the foodchain hes talking about? who records at 44.1 these days or even back then?
for the record I use hi-res when I can, but I think bit depth is more important than sample rate for solving most of the problems Keith brings up. such as dynamic headroom etc. things start to go downhill again in most cases after 96khz or max 192. though my dacs are capable of 32/384
Gordon appeared to be squirming a bit there in the background (as was I in some parts)... i'm wondering where these small bits of information are disappearing in these stages down the foodchain hes talking about? who records at 44.1 these days or even back then?
Last edited:
Keith Johnson has been known to me for the last 45 years. He did fundamental work on tape recording and solid state in the middle '60's that exceeds 90% of any audio research done here by anyone at any time. I keep an eye out for any input he gives on audio research.
He and I happen to be competitors, and I don't think that he thinks as well about my work as I do about his research. So be it, because for me, info is info, so long as the source is competent.
Quisp, please do something useful, rather than attempt to find fault in others.
He and I happen to be competitors, and I don't think that he thinks as well about my work as I do about his research. So be it, because for me, info is info, so long as the source is competent.
Quisp, please do something useful, rather than attempt to find fault in others.
If you took what I said as "all efforts" I apologize, I was speaking in the context of listening tests of music through speakers. I am unaware of any conclusive results and several to the contrary done under controlled conditions by highly regarded professionals (I'm not refering to E. Brad Meyer).
Is that the one where he said we need 256 bits?(or was that Doug Sax) 🙄 We could count all the particles in the universe in the mean time.
Oohashi got null results when switching to headphones, that I take is the reason for the renewed efforts.
- Status
- Not open for further replies.