It sounds like a timing issues. Maybe interchannel phase but I would devise a test with two pulses, one in each channel with a time difference and measure the difference along with its stability. I would use my 5370A for that. Overkill (20pS single shot resolution) since air currents will degrade that in acoustic space but otherwise designed for the task. Would also be interesting to see audio frequency jitter in analog output with it.
I remember a demo of a wireless audio solution that shifted on every track. Really embarrassing when we pointed it out. Didn't go into production as I remember.
I remember a demo of a wireless audio solution that shifted on every track. Really embarrassing when we pointed it out. Didn't go into production as I remember.
Not just timing. For a certain range of audio frequencies, it is ITD timing (at least for lateral localization). For depth it relates to other types of cues.
The wiki article isn't too bad: https://en.wikipedia.org/wiki/Sound_localization
Some other references:
An Overview of the Major Phenomena of the Localization of Sound Sourcesby Normal-Hearing, Hearing-Impaired,and Aided Listeners
DOI: 10.1177/2331216514560442t
Localization Uncertainty in Time-Intensity Stereophonic Reproduction - IEEE
http://go.diyaudio.com/?id=69111X15...ef284c93&abp=1&xjsf=other_click__auxclick [2]
Spontaneous head-movements improve sound localization in aging adults with hearing loss
| https://doi.org/10.3389/fnhum.2022.1026056
The Effect of Head Turning on Sound Localization in the Horizontal Plane
Norbert Kolotzek, Gabriel Gomez, Bernhard U. Seeber
Audio Information Processing, Technische Universität München, Arcisstraße 21, 80333 München
Mechanisms of Sound Localization in Mammals
https://pubmed.ncbi.nlm.nih.gov/20664077/
The wiki article isn't too bad: https://en.wikipedia.org/wiki/Sound_localization
Some other references:
An Overview of the Major Phenomena of the Localization of Sound Sourcesby Normal-Hearing, Hearing-Impaired,and Aided Listeners
DOI: 10.1177/2331216514560442t
Localization Uncertainty in Time-Intensity Stereophonic Reproduction - IEEE
http://go.diyaudio.com/?id=69111X15...ef284c93&abp=1&xjsf=other_click__auxclick [2]
Spontaneous head-movements improve sound localization in aging adults with hearing loss
| https://doi.org/10.3389/fnhum.2022.1026056
The Effect of Head Turning on Sound Localization in the Horizontal Plane
Norbert Kolotzek, Gabriel Gomez, Bernhard U. Seeber
Audio Information Processing, Technische Universität München, Arcisstraße 21, 80333 München
Mechanisms of Sound Localization in Mammals
https://pubmed.ncbi.nlm.nih.gov/20664077/
OK, a mix of timing, response changes relating to intra-aural transfer function and phase shifts affecting the timing and percieved frequency balance. All measurable, none easy and especially acoustically. (The room interactions are quite involved) The head movement thing has been identified with localization and head tracking implemented properly makes a huge difference in headphones. Its why sound from headphones without head tracking existed between your ears or in some cases behind your head.
It seems unlikely that a linear system would alter the relationships without showing obvious phase/timing errors. But maybe?
Impulse timing in a sampled data system where the samples are sparse for the impulse may alter timing in ways that could also affect the effective spectrum of a tone burst or impulse (snare drum hit?). Definately worth exploring.
It seems unlikely that a linear system would alter the relationships without showing obvious phase/timing errors. But maybe?
Impulse timing in a sampled data system where the samples are sparse for the impulse may alter timing in ways that could also affect the effective spectrum of a tone burst or impulse (snare drum hit?). Definately worth exploring.
Hi Mark,
I was out for a seminar all day yesterday. Today I look and see you've been busy doing exactly what you've been asked to refrain from.
I am in complete agreement with Tom. You are an expert at threadjacking and holding people to double standards.
So ... officially, please stop doing this or I may consider bin time. Enough already.
I made a measurement at Fs = 384kHz and still the S-D noise shaping elevation that was present with older DACs is negligible.
Another analysis made up to 10MHz (but with lower dynamic range) shows some HF content above 200kHz (elevated noise floor), but still below -100dBV in amplitude. I do not see any reason to care about possible dv/dt overload of the preamp or power amp input stage, moreover, they are almost always equipped with RC lowpass input filters. It seems to me there is a permanent effort to find new and new hypothesis, none of them supported by any proof.
I'm kinda reluctant to post this, but here goes. Anatech - do your thing if you think it's appropriate.
Several years ago some of my friends began to experiment with conversion clocks for DACs. They tried all sorts of things. They found that some clocks, all ostensibly the same, just sounded better to them than others did. This was across several vendors, several clock circuit designs, and many samples of each. They tried measuring the difference between the clocks, both in the converted audio band and at the RF frequencies where the clocks operated. Nuthin. They tried the same clocks across different DAC circuits and got similar perceived sonic results, but no valuable measurement results.
BTW, there was no correlation between price and observed - not measured - performance.
BTW2, these were commercial clock oscillators or "home made" clock oscillators, not fancy audiophile products.
Eventually, a couple of the guys got together and bought a system that would do cross correlation Allan variance measurements of the oscillators at their operating frequency. Not application of the clock in an audio DAC - just the basic oscillator in a heroic level test fixture with really pure DC, proper terminations, and so on.
What they found was that there seemed to be a pretty good correlation between the clocks that "sounded good" to them and the ones that had better very low frequency phase modulation. As in below a Hz or two.
One of the strong points of this experiment was that the clocks were first listened to in a system and were then tested for phase noise. So, there wasn't any possibility for listener bias and there certainly was no bias on the part of the test system.
Now, I know that this wasn't a certified double blind test, although nobody had any idea of what clock oscillator should "sound best". And, the testing in the audio band didn't find anything interesting.
But, I have always wondered whether the problem was that the testing in the audio band wasn't for the right thing, whatever that might be. (Low frequency phase modulation of the entire waveform? That guess is as good/dumb as any other.)
In general, it's really easy to dismiss observations people make on various subjects without really investigating them. Especially when there's already a body of work that suggests that the idea may be dumb. It's smart to be skeptical. But, then I'm reminded of Barry Marshall and Robin Warren. Sometimes it's worth taking a second look from a different angle. Or a 97th look.
I am kinda reluctant to conclude that the entire sonic experience of a sound system can be explained with tests showing the amplitude performance of the system, including spectral plots over a couple of sweeps. Animal auditory systems use both amplitude and timing information to work properly. Again, I don't have the answers to this, but I do have some questions that would seem answerable.
Several years ago some of my friends began to experiment with conversion clocks for DACs. They tried all sorts of things. They found that some clocks, all ostensibly the same, just sounded better to them than others did. This was across several vendors, several clock circuit designs, and many samples of each. They tried measuring the difference between the clocks, both in the converted audio band and at the RF frequencies where the clocks operated. Nuthin. They tried the same clocks across different DAC circuits and got similar perceived sonic results, but no valuable measurement results.
BTW, there was no correlation between price and observed - not measured - performance.
BTW2, these were commercial clock oscillators or "home made" clock oscillators, not fancy audiophile products.
Eventually, a couple of the guys got together and bought a system that would do cross correlation Allan variance measurements of the oscillators at their operating frequency. Not application of the clock in an audio DAC - just the basic oscillator in a heroic level test fixture with really pure DC, proper terminations, and so on.
What they found was that there seemed to be a pretty good correlation between the clocks that "sounded good" to them and the ones that had better very low frequency phase modulation. As in below a Hz or two.
One of the strong points of this experiment was that the clocks were first listened to in a system and were then tested for phase noise. So, there wasn't any possibility for listener bias and there certainly was no bias on the part of the test system.
Now, I know that this wasn't a certified double blind test, although nobody had any idea of what clock oscillator should "sound best". And, the testing in the audio band didn't find anything interesting.
But, I have always wondered whether the problem was that the testing in the audio band wasn't for the right thing, whatever that might be. (Low frequency phase modulation of the entire waveform? That guess is as good/dumb as any other.)
In general, it's really easy to dismiss observations people make on various subjects without really investigating them. Especially when there's already a body of work that suggests that the idea may be dumb. It's smart to be skeptical. But, then I'm reminded of Barry Marshall and Robin Warren. Sometimes it's worth taking a second look from a different angle. Or a 97th look.
I am kinda reluctant to conclude that the entire sonic experience of a sound system can be explained with tests showing the amplitude performance of the system, including spectral plots over a couple of sweeps. Animal auditory systems use both amplitude and timing information to work properly. Again, I don't have the answers to this, but I do have some questions that would seem answerable.
I think the question would be how the LF fluctuations affect the audio. You should see those as sidebands close in to a fixed tone. It would be pretty easy to take a reference clock with a voltage tweak (vricap tuning diode) and modulate it and se what comes out. This is stuff that would be very difficult to capture on a conventional FFT. Some of the skirt you see on tones is from the windowing making it hard to separate the origins. And if you are using a common clock any modulation will be largely cancelled out.
I should point out that a tape system has much worse time base errors (wow) not to mention a turntable with off center disks. And most of us have been oblivious to this in the analog era.
I should point out that a tape system has much worse time base errors (wow) not to mention a turntable with off center disks. And most of us have been oblivious to this in the analog era.
WRT HF output from DAC's. I would be surprised if there were much because that would run into FF stuff. However there probably is some HF radiation (its almost impossible to eliminate).
My concern was the HF energy going into the reconstruction filter's opamp. Even Scott Wurcer mentioned that issue. Its really challenging for a current output DAC.
The issue with identifying these issues is that all the measurement tools we have are based on the system being in a steady state with one or more tones either continuous or repeating. Both describe either post modern classical music or really boring. Not what most people listen to. (I'm not most people, being a fan of Philip Glass, Steve Reich and John Adams.)
My concern was the HF energy going into the reconstruction filter's opamp. Even Scott Wurcer mentioned that issue. Its really challenging for a current output DAC.
The issue with identifying these issues is that all the measurement tools we have are based on the system being in a steady state with one or more tones either continuous or repeating. Both describe either post modern classical music or really boring. Not what most people listen to. (I'm not most people, being a fan of Philip Glass, Steve Reich and John Adams.)
Depending on the DAC it may make no difference at all. Some DACs have clock-jitter cleaners on the clock input. Contrary to common belief, feeding a clock-jitter cleaner a cleaner reference clock does not necessarily make the output of the jitter cleaner any cleaner. Usually a PLL is used for the jitter cleaning. And recall that a PLL acts as a lowpass filter on the reference clock and a highpass filter on the VCO input. The cutoff frequency for those filter function is the PLL loop bandwidth. What this means is that the DAC performance will most likely be determined by the performance of the VCO that's built into the DAC, which renders the fancy femtosecond clocks rather useless ... at least from a technical perspective. They make great marketing bullet points.
Tom
That's very true. Every time I look at the test reports of a turntable system showing its wow and flutter performance, I shake my head in wonder.
The Plangent Process really seems to help in this regard for tape recordings. (Plangent Process for those not familiar) I certainly prefer the sound quality of a Plangent restored recording.
I also think that some people are just more sensitive to this than others. For example, Vandersteen loudspeakers are designed and built with an almost maniacal devotion to preserving the time domain performance of the acoustic waveform. Richard Vandersteen has said more than once that not everybody is sensitive to these time domain effects, or maybe don't care. Most loudspeakers aren't as focused on this detail, and people certainly buy and use them. So, it may be true of other time base errors, too.
I've been trying to find a software application that will let me phase modulate the entire waveform of a recording to see what I can see. That would be kind of similar to modulating the reference clock, but presumably with more predictability. So far, I'm still looking. Ideally, something would be visible in an FFT measurement of that, but who knows?
Edit: After thinking about tomchr's comment about PLL's in conversion chips, it makes me wonder if humans are able to filter out relatively gross imperfections in sound reproduction better than they can ignore more modest low frequency changes. Turntable wow may be easy to ignore, while subtle phase shifts may be tough. Like seasickness, in a way.
Last edited:
Nothing controversial about those results. Low close-in phase noise (or jitter) has been one of the goals of digital audio since its introduction. But in published studies the audibility threshold for jitter is surprisingly high. So as with THD or noise there is quite likely a limit to how much audible improvement is possible with lower phase noise clocks.
And while the test you referred to had mitigated listener biases most listening claims related to especially expensive audiophile clocks lack all controls so perceptual biases make those claims worthless. Another issue is that since measurements do not show any improvements it is typical that various non-measurable characteristics are associated with low phase noise clocks.
All very true.
But, what if (just a game here...) the effect that some people may be hearing is related to effects that are low enough in frequency so that the PLL's don't reject whatever it is in the effective reference clock?
Again, I don't know. And I don't want to sound like a denialist of science like so many people are these days. But, what if our existing tests are accurate but incomplete?
I do know that if a loudspeaker cabinet moves during playback, it does affect the sound quality. Could the same effective thing happen in the electrical domain driving the loudspeaker? Dunno. That's what I am asking - not claiming.
As I said, expensive audiophile clocks were not part of their experiment. I don't know for sure, but I'd guess that expensive clocks aren't always what they claim to be and the listening claims might be kinda dubious. Again, a guess by me.
Plus, the measurements that have been made to date are not everything that can be measured, are they? I am not the guy who will scream that "you can't measure everything - it's impossible!" Instead, I will say that I don't think that we do measure everything that can be measured today. That is my real point. If we measure some performance aspect and consistently find nothing, then either our test methodology is flawed or there's nothing there. You don't know until you try.
What if our listening tests are a complete joke from a scientific point of view? Should they then carry the same weight as measurements with calibrated equipment?
I can rattle off a good handful of psychological effects that would explain why someone would perceive a difference between two identical stimuli. I can list even more reasons why the commonly used listening tests are flawed. It is very difficult and expensive to set up a scientifically valid listening test.
The dynamic range and frequency resolution of modern test equipment by far exceeds that of the human ear. Is it possible that we aren't testing everything? Sure. As any lawyer will tell you, anything is possible. That does not mean that it is likely or even probable, though.
While I do think the topic of measurements vs perceived experience is an interesting one, I really think it would be better served in its own thread.
Tom
I second to this.
And, my experience with A/B tests with a listening panel was that there was often preference to distorted or noisier sample compared to original files. For similar reasons, I consider sighted tests to be pointless.
My retort is that this gear absolutely calibrated and proper, but the testing is not all encompassing. I'm not saying that listening tests are the be all and end all - they certainly are not - but they might provide some places to examine phenomena through the testing.
But, OK, I'll stop.
Absolutely. Olive & Toole did a bunch of that at Harman Kardon. You can find their results in various AES papers. There are many other papers on the topic too, but they're frequently used as toilet paper by subjectivists because they don't support the subjectivists' opinions.
I'm not seeing anything alarming here.
Tom
I like to consider myself to not be in any camp, subjectivist or otherwise. I don't have many opinions to support, other than the ones that I've proven to myself either through experimentation or really good academic research that also matches my experimentation and observations - even academia sometimes has axes to grind. Largely that experimentation is through what most people would consider objective measurements. (Full disclosure - I do not have a degree in either electrical engineering or psychology.)
But, you guys have convinced me that things in audio are already as good as they could possibly be, so I'm going to go pursue something else in life. Thanks for all the fish. 🙂
Just adding precision here: I don't think things in audio are as good as they could possibly be. I'm sure there will be improvements. Some with eke out another 0.5 dB improvement on some parameter. Maybe even more. Many thought we'd reached the absolute best when DACs crossed 120 dB THD. Then ESS went, "hold my beer" and now we're even further. I expect that to continue.
But pragmatically, I think audio has been the best it's ever been for quite a while now. What's left is the "numbers game". Some poo-poo that as just a numbers game. And there's some validity to that. It just so happens that I have a lot of fun chasing that last dB. Doing so allows me to use a broad spectrum of knowledge including from physics, material science, and engineering. And judging by the responses by quite a few members here I don't think I'm the only one who enjoys that. 🙂
Tom