I'm quite sure I won't hear any difference between two reasonable DACs. I'm equally sure there are people on this forum who can hear a lot better than I can.
Like Mark wrote, sigma-delta DACs can produce lots of weird low-level artefacts, such as non-stationary noise and idle tones that get frequency-modulated by the desired signal. Whether any of those are audible to anyone, I don't know. If I had to guess, I would probably guess that they are inaudible, but I have guessed wrong before. Hence my vote for "Not sure".
Last edited:
The worst peak-peak ripple in the pre- and post-echo test was 0.2011647148 dB peak-peak, so just over +/- 0.1 dB. That seems to have been audible to participant 2, except on one recording he considered to be of poor quality.
In the first test, he even had the test with 0.030025949 dB peak-peak ripple correct on all but that one recording. Lucky shot or a good day? I don't know.
True. At least we can agree on that.
However, it costs more to make a Ferrari car than it costs to make a Yugo car. Someone may claim that a Yugo car can go the legal speed limit so therefore according to measurements it is just as good as a Ferrari. Most people would understand how much that person would be missing by taking such a position.
When it comes to dacs, people understand much less about costs and differences. I can't agree with your oversimplified assessment of what dac measurements matter. In my view you make too many assumptions that a dac is just like a line amp, or power amp (e.g. a few simple PSS measurements will tell the whole story). I don't know why a person who takes time to understand sigma delta dacs would take such an oversimplified view. It makes no sense to me.
However, I could imagine someone who doesn't bother to learn the quirks and technology of sigma delta dacs to ignorantly assume what measurements describe it accurately to be nothing different from a line amp or power amp.
Are you even familiar with the test for dacs that measures the noise at each DC offset? Do you know that absent dither that noise is a repeating pattern of the modulator output. Dither disguises that fact, but it doesn't mean that noise is like benign thermal resistor noise.
Do you know that the infamous ESS Hump Distortion was actually a deterministic signal about 20dB below the audio signal? You think that can't be audible to any human on earth? If so, you would be wrong.
Here is a measurement taken by ESS and posted by Scott Wurcer of hump distortion in early ESS dacs:
It sure doesn't look like random noise that would be inaudible to me. It also doesn't look like any distortion that occurs in a line amp or power amp. ESS didn't think so either, since they asked Mr. Wurcer to help them understand it.
Things like that don't happen in line amps or power amps. Sigma Delta dacs are very different, but you keep assuming they are more or less same in terms of standard measurements. That just isn't born out by the facts.
Last edited:
Like most car analogies for audio not accurate. Nor is your characterization of me. Many people think they hear so much that is not measurable. I think most of the time nearly all the time it is the reverse. So much is measurable that isn't audible. Hard for some to accept that other than in some cases power amps and very little else nothing makes audible differences beyond speakers.
@Markw4 What happened to your Deltawave ABx comparator tests of the PCM2DSD v3 recording vs PCM2DSD v4 recording vs original?
I know you were at least looking in to it as you asked questions about the comparator.
However, when I try to go to those posts I get this. I guess you deleted them?
I would love to know the results of your test.
I also thought I heard easily identifiable differences between v3 and v4. However, when I put it to an ABx test I couldn't distinguish them. I think this is a great test as we have quantification of how good the null is between the v3 and v4 recordings (-100 dB rms) and between the recordings and the original (-69 dB rms). We also have the tools to look at the exact waveform differences at points in time where you hear differences.
If you can accurately distinguish these files in an ABx test I would absolutely believe you can distinguish other minute differences between DACs. If you can't distinguish them, despite claiming large audible differences, it would call in to question your other opinions based on sighted listening tests.
Given you haven't posted the results, I can only assume your test results did not turn out the way you hoped but would love to know more.
Michael
I know you were at least looking in to it as you asked questions about the comparator.
However, when I try to go to those posts I get this. I guess you deleted them?
I would love to know the results of your test.
I also thought I heard easily identifiable differences between v3 and v4. However, when I put it to an ABx test I couldn't distinguish them. I think this is a great test as we have quantification of how good the null is between the v3 and v4 recordings (-100 dB rms) and between the recordings and the original (-69 dB rms). We also have the tools to look at the exact waveform differences at points in time where you hear differences.
If you can accurately distinguish these files in an ABx test I would absolutely believe you can distinguish other minute differences between DACs. If you can't distinguish them, despite claiming large audible differences, it would call in to question your other opinions based on sighted listening tests.
Given you haven't posted the results, I can only assume your test results did not turn out the way you hoped but would love to know more.
Michael
Why use ABX when there are better protocols? The problems with ABX are well understood in the professional sensory science field. Would you like to tell me what they are to show that you know about that?
Why were you even looking in to using the comparator then?
I once thought you were just very passionate about this stuff, now it seems like you are acting in bad faith. Oh well.
Michael
I once thought you were just very passionate about this stuff, now it seems like you are acting in bad faith. Oh well.
Michael
Also regarding ABX, a very interesting post on the subject at: https://www.diyaudio.com/community/threads/return-to-zero-shift-register-firdac.379406/post-7682348
No, my interest was and is to find the truth. A rigged test known to be biased towards false negatives is not the path to finding the honest truth.
If you truly know anything about ABX, you should know how it can be used properly. You show no such knowledge.
I always jest that subjectivists hear things that can’t be measured and objectivists measure things that can’t be heard. But it’s a reductive statement designed to entertain.
My 20-year old ears could hear more things than my 53-year old ears can. And it’s possible to measure and identify flaws that would sound horrible to all ears.
Let’s just say that being 53 erases a lot of the difference, if not all of the difference, between a well-made $1,000 DAC and a well-made $10,000 DAC.
I am happy to report that bad speakers still sound bad, however. And I can find those at all sorts of prices.
My 20-year old ears could hear more things than my 53-year old ears can. And it’s possible to measure and identify flaws that would sound horrible to all ears.
Let’s just say that being 53 erases a lot of the difference, if not all of the difference, between a well-made $1,000 DAC and a well-made $10,000 DAC.
I am happy to report that bad speakers still sound bad, however. And I can find those at all sorts of prices.
I also think professional reviewers should be obligated to publish their yearly audiogram for their audience to review. It would be very useful.
What protocols are better? Can you describe them please. Or link to some other post where you described them.
Last edited:
I have a lot of references, but some of the best summaries of general perceptual testing techniques are in areas other than audio. However, they are still quite applicable to audio or any other type of sensory perception.
Also, since training is a common part of proper perceptual testing, another good summary of general techniques is attached.
For general academic study of the subject, a text by Jian Bi is well regarded: Sensory Discrimination Tests and Measurements_ Sensometrics in Sensory Evaluation (2015, John Wiley & Sons)
The attached two papers are well worth some study. There is no perfect one-size-fits-all measurement technique. There are pro and cons that have be evaluated based on the research question or questions at hand. As can be seen, ABX appears in one table in the first reference. Some information is given in the "Limitations" column.
That said, there is evidence ABX can work well given sufficient training of test subjects. When such training is not readily available, then another protocol may be a better choice for achieving accurate results. I would probably suggest referring to Jian Bi's book for more information on that.
Also, since training is a common part of proper perceptual testing, another good summary of general techniques is attached.
For general academic study of the subject, a text by Jian Bi is well regarded: Sensory Discrimination Tests and Measurements_ Sensometrics in Sensory Evaluation (2015, John Wiley & Sons)
The attached two papers are well worth some study. There is no perfect one-size-fits-all measurement technique. There are pro and cons that have be evaluated based on the research question or questions at hand. As can be seen, ABX appears in one table in the first reference. Some information is given in the "Limitations" column.
That said, there is evidence ABX can work well given sufficient training of test subjects. When such training is not readily available, then another protocol may be a better choice for achieving accurate results. I would probably suggest referring to Jian Bi's book for more information on that.
Last edited:
Well looking thru the first document they discuss duo-trio (my favorite) and 2afc which is another I like. I think one of the best for the person trying a perceptual test is the up-down tests. I do think experience with the test however it is handled is very helpful. ABX can be very discriminating, but having used it until you are comfortable makes a big difference. As does doing some testing where you hear differences to learn what that is like. There is value in listener training and test training (or familiarization).
Last edited:
I want to echo the importance of listener ‘training’, which Mark has mentioned several times. My age means that my auditory sensory acuity (the parametric performance of my hearing) has diminished from when I was younger. My self-training over the years as a listener, however, has actually improved with age. By training, I mean, auditory observational skill. This is more a psychoacoustic skill.
For example, at one time, my wife didn’t notice the ‘audiophile’ characteristics of a decent audio system. Now, years later, she readily and unprompted, observes how a system images, its ambient stage, musical/tonal balance, etc. No, she’s not a golden ear, she simply now knows from experience what to listen for. It was always there, she had just never noticed it as a casual listener.
For example, at one time, my wife didn’t notice the ‘audiophile’ characteristics of a decent audio system. Now, years later, she readily and unprompted, observes how a system images, its ambient stage, musical/tonal balance, etc. No, she’s not a golden ear, she simply now knows from experience what to listen for. It was always there, she had just never noticed it as a casual listener.
Last edited:
Maybe the brain, as everything else in the body, likes homeostasis. I remember loving a specific set of headphones I had used for years. Then I bought something more expensive, but I was disappointed by it initially. Yet, I used the more expensive headphones for a couple of years. Then I figured I would return to my old time favorites for a bit, and I discovered to my surprise I no longer liked them at all. 🤷♂️
That would also explain the legendary burn-in effect.
That would also explain the legendary burn-in effect.
It would except there is more you don't know. However, what you suggest could be true in some cases. In other cases people have gone to considerable lengths to see if burn-in is real. That includes building multiple identical devices to see if each one individually follows a burn-in pattern. The observation has been that its more like a bathtub curve of initial wear more than anything else.
Thus, for example, maybe some electrolytic caps sound a little different having a bit of initial wear on their clock. For electrolytic caps that could mean a reduction in leakage currents and 1/f noise as the electrolyte chemically stabilizes under normal in-circuit bias conditions.
Last edited:
If I understand it correctly, his criticism boils down to small ABX tests being unsuitable for barely audible differences. If so, I have to agree. You see an example in the retest of the echo test: one poor recording sufficed to make the result insignificant at the 5 % level.
I have to explain why the answer is yes. Let's say I don't have to take a DAC as an example, but any modern operational amplifier and the simplest circuit, a non-inverting line preamp, I've made a lot of them. All those OPAs measure not well but fantastically well, and yet in the same assembly it is easy to notice that it sounds different by replacing the OPA. There is no need for blind tests, equalization of loudness and the like, it is immediately evident even to an inexperienced listener that there are some differences. I'm not saying that there are big differences, just that they are noticeable. What I haven't tried so far, OPA1612, 1622, 1656, 1642, LME49720, Burson V5, V6, V7, Sparkos SS3602, Muses02, some Chinese discrete opamps etc. It's the same with DACs. My DACs are all without oversampling and without any filters. So the measurements are for sure much worse than those Topping D50 and the like, and the sound is incomparably better. Topping D10, D50 is boring, regardless of any measurement. We changed their power supply, opamps, things improved a bit, but boredom remains. The question from the topic is just an occasion for a little discussion that brings nothing.
Last edited:
- Status
- Not open for further replies.