Pic of prototype Squarer Board:
It sounds very good, is somewhat tunable, should work with Andrea Mori and or Acko Labs clocks, and it can almost entirely replace the General Purpose Clock Board when used as shown in the pic (which is to say, no PLL is included on this new board). It is six-layers and probably more sophisticated than other squaring boards I have seen. Still have to implement the clock switching circuitry on this prototype, and then do at least one more rev to pick up on any small changes. However, unlike the other boards of this thread, the squaring board is probably not going to be open source. We'll have to see.
In other news, starting to get some early adopters building clock, interconnect, and reclocker boards. Reports are to the effect that its well worth the effort (although currently trying to work through a problem one builder is having which involves the PLL and the PCM2DSD board not always working reliably together; so far unable to duplicate reported problem here).
It sounds very good, is somewhat tunable, should work with Andrea Mori and or Acko Labs clocks, and it can almost entirely replace the General Purpose Clock Board when used as shown in the pic (which is to say, no PLL is included on this new board). It is six-layers and probably more sophisticated than other squaring boards I have seen. Still have to implement the clock switching circuitry on this prototype, and then do at least one more rev to pick up on any small changes. However, unlike the other boards of this thread, the squaring board is probably not going to be open source. We'll have to see.
In other news, starting to get some early adopters building clock, interconnect, and reclocker boards. Reports are to the effect that its well worth the effort (although currently trying to work through a problem one builder is having which involves the PLL and the PCM2DSD board not always working reliably together; so far unable to duplicate reported problem here).
Last edited:
Frequency switching works okay. Interesting result is that Andrea Mori clocks (11/12MHz, with one set of doublers) sound different from 22/24MHz AckoLabs clocks (both sets of clocks are using the same power supplies and the same squarer board, only the clocks are different).
To some extent the Squarer Board can be tuned to lessen the difference between clock brands, but not make them exactly the same. Andrea clocks are smoother and softer sounding; Acko clocks can be quite sharp and dynamic (even to the point of being a little hard to listen to for too long). Yet both clocks measure exceptionally well in terms of phase noise. Some experiments with tuning the squarer may give a few small hints, yet not are be fully satisfactory in the way of explanation for the different sounds. BTW, I noticed the same general characteristics of the clocks before when using other squaring circuits. At the moment I like the Acko clocks best but only if used with a particular squarer tuning.
So, what could be going on to explain the difference between the sound of different brand clock designs if they both measure exceptionally well in terms of phase noise?
This is something I have been thinking about for awhile. Aside from measurement errors (and aside from listening errors), so far I can only think of one basic physical explanation. What about anyone else, any ideas in terms of clock/oscillator physics?
To some extent the Squarer Board can be tuned to lessen the difference between clock brands, but not make them exactly the same. Andrea clocks are smoother and softer sounding; Acko clocks can be quite sharp and dynamic (even to the point of being a little hard to listen to for too long). Yet both clocks measure exceptionally well in terms of phase noise. Some experiments with tuning the squarer may give a few small hints, yet not are be fully satisfactory in the way of explanation for the different sounds. BTW, I noticed the same general characteristics of the clocks before when using other squaring circuits. At the moment I like the Acko clocks best but only if used with a particular squarer tuning.
So, what could be going on to explain the difference between the sound of different brand clock designs if they both measure exceptionally well in terms of phase noise?
This is something I have been thinking about for awhile. Aside from measurement errors (and aside from listening errors), so far I can only think of one basic physical explanation. What about anyone else, any ideas in terms of clock/oscillator physics?
decoupling around the crystals ? But here it is not iso layout, you have a doubler for one of the conf if I read rigth (i fastly readed on a smartphone), so you have your culpritt perhaps.
The crystals are on the oscillator boards. If decoupling in that area is a problem, then how and why would the phase noise measure exceptionally good? IOW, suppose the oscillator clocks are in black boxes (we don't know what is inside), we only know the published measured phase noise for each black box. If phase noise is super low for all of the black boxes, then shouldn't the clock oscillators sound about the same? If not, then what could be different about them that would account for the difference in sound?
Last edited:
I tried to sleep a night over this, but I must say, I am as puzzled as you are.
It reminds me of my listening tests with different clocks, especially the Accusilicons versus the Crytecs (I mean the XO versions, same specs, same price kind of comparable league). Same results, The Accus were smoother, the Crystecs were sharper in comparison between the two. By the way: Andrea Mori beats them all hands down (I use the 5,6/6,1 with two sets of doublers with an automatic switching between 44/48 and selecting one or two frequency-doublers in the clock chain by automatic detection of FS signal)
Beats me, no idea what that could cause this? Something we do not (know how to) measure yet? Anything real time like THD of the sinewave? any DC offset? Wild guessing appears...
It reminds me of my listening tests with different clocks, especially the Accusilicons versus the Crytecs (I mean the XO versions, same specs, same price kind of comparable league). Same results, The Accus were smoother, the Crystecs were sharper in comparison between the two. By the way: Andrea Mori beats them all hands down (I use the 5,6/6,1 with two sets of doublers with an automatic switching between 44/48 and selecting one or two frequency-doublers in the clock chain by automatic detection of FS signal)
Beats me, no idea what that could cause this? Something we do not (know how to) measure yet? Anything real time like THD of the sinewave? any DC offset? Wild guessing appears...
Have you considered ABX-testing to rule out confirmation bias? This is - as you say listening errors - the most likely explanation. A blinded confirmation of the difference in sound would be very interesting !
Actually, thinking about it overnight resulted in a short list of some possible electronics/physics explanations (no claim is made as to completeness of the list).
In no particular order, they are:
1. Some clocks may be more sensitive than others to power supply type/characteristics
2. Clocks may not be operating at their published specs
3. Squarer might be sensitive to incoming signal level (maybe if clock signal levels were found to be different)
4. The dac in use may be sensitive to doublers in the clock source (doubler use might show up as THD of the 22/24MHz sine waves, if we could measure THD at those frequencies)
5. Phase noise plots do not show peak time jitter.
Regarding #4, the dac in use is 4-tap DSD FIRDAC.
Regarding #5, Phase Noise can be approximately converted to RMS time jitter, but an estimate of RMS jitter still cannot be converted to peak jitter without first knowing the phase-deviation-versus-time waveform (which is what we would really like to know in the first place).
Also, would just mention it doesn't have to be only one factor causing a difference in sound. Could be some mix of factors.
In no particular order, they are:
1. Some clocks may be more sensitive than others to power supply type/characteristics
2. Clocks may not be operating at their published specs
3. Squarer might be sensitive to incoming signal level (maybe if clock signal levels were found to be different)
4. The dac in use may be sensitive to doublers in the clock source (doubler use might show up as THD of the 22/24MHz sine waves, if we could measure THD at those frequencies)
5. Phase noise plots do not show peak time jitter.
Regarding #4, the dac in use is 4-tap DSD FIRDAC.
Regarding #5, Phase Noise can be approximately converted to RMS time jitter, but an estimate of RMS jitter still cannot be converted to peak jitter without first knowing the phase-deviation-versus-time waveform (which is what we would really like to know in the first place).
Also, would just mention it doesn't have to be only one factor causing a difference in sound. Could be some mix of factors.
Last edited:
I'm not at all confident, it's about likelihood. I'm not trying to troll at all. I just think it is a shame that so often the way to rule out bias is overlooked. We all know that it is an important factor we have to deal with. I too have e.g. tweaked a room correction curve to 'perfection' that turned out to be in bypass mode🙄 . ABX testing is not hard, it just takes some time.
If the difference is real, that would be very interesting and a possible path forward to better sound.
If the difference is real, that would be very interesting and a possible path forward to better sound.
How can someone know or estimate the likelihood of a possible explanation in this particular case?
For example, if someone is comparing two sounds that differ slightly only in FR, isn't easier to make a mistake than if OTOH, one were comparing the sound of a bird chirping to the sound of a train whistle? I don't see why we would need to consider ABX in the latter case. IOW, the degree of sound difference combined with the type difference should inform whether ABX (or other equivalent protocol) is needed?
Part of the reason I ask is that some people tend to demand ABX "proof" from other people when the person making the demand is acting on their own bias at the moment, and not as a matter of general principle. We see that kind of bias a lot at some other audio website out there.
For example, if someone is comparing two sounds that differ slightly only in FR, isn't easier to make a mistake than if OTOH, one were comparing the sound of a bird chirping to the sound of a train whistle? I don't see why we would need to consider ABX in the latter case. IOW, the degree of sound difference combined with the type difference should inform whether ABX (or other equivalent protocol) is needed?
Part of the reason I ask is that some people tend to demand ABX "proof" from other people when the person making the demand is acting on their own bias at the moment, and not as a matter of general principle. We see that kind of bias a lot at some other audio website out there.
Last edited:
I like (informal but double-blind) ABX tests as well, but it's easy enough to make some methodological error that leads to inconclusive results. For example, the high-order dither test you find here https://www.diyaudio.com/forums/everything-else/313257-dither-listening-test.html was not entirely conclusive because I failed to take into account the possibility that the test persons might be able to hear differences between two realizations of the same random process.
The pre- and post-echo test from about three years ago that you can find here, https://www.diyaudio.com/community/...makes-nos-sound-different.371931/post-6761338 , seems to show that test person #2 could hear the difference except on one recording of which he found the sound quality inadequate, but that conclusion is somewhat speculative because the test was too small (see the report attached to the post I linked to for details).
In fact, you can only calculate how large an ABX test needs to be when you have a decent estimate of the probability of making a mistake if the effect is real. When you then calculate how many trials are needed to, for example, be 95 % confident that the effect shows up if it is real, you probably end up with an impractically large number of trials unless the effect is so obvious that the chance of making a mistake is very small.
I would be more worried about subharmonics, particularly the second, see posts 2830 to 2850 of the well tempered master clock thread, https://www.diyaudio.com/community/...jitter-crystal-oscillator.261651/post-6372426 The FIRDAC much reduces the effect, but doesn't quite eliminate it.
If a sine wave is distorted such the the duty cycle is not very close to 50%, or if successive sine wave cycles are of slightly alternating frequency then wouldn't that type of thing show up on a hi-res FFT? I mean, seems there should be correlation with more than one FFT bin frequency?
Actually, I have a question about that. After you found the error and took it out of bypass mode, did the sound change audibly? Basically just wondering if the room correction did anything at all for the sound?
If so, then did you then perform ABX on room correction enabled or bypassed to verify if you could hear a real difference? If so, how many trials and what was your score?
I do indeed do ABX test on changes I have my doubt about. You are correct that sometimes a difference is so obvious that a test would be unnecessary. But you said yourself that you were uncertain in this case.
I like to test to try to make it objectively true, you don't. To each their own.
I like to test to try to make it objectively true, you don't. To each their own.
Sure, but in your second case, when you take the average sine wave frequency to be the fundamental, the peaks at half that frequency and its odd multiples do not count as THD because they are not harmonics but subharmonics.
Okay. How about if I take the lower of the two frequencies as the fundamental, and call the higher frequency a non-integer harmonic? Its just a system that is more non-LTI than we may see in some other audio devices? IOW, it not the result of a slightly curved yet unchanging transfer function. That said, maybe it is PSS because we see it as a spur relative to our chosen fundamental?
Guess what I'm getting at is, a lot can depend on definitions.
Guess what I'm getting at is, a lot can depend on definitions.
I was uncertain about the underlying causal mechanism of dac and clock system behavior, but not uncertain as to whether there was an audible difference.
Anyway, when I have a doubt I don't typically use ABX. I use other skilled listeners and ask them to provide descriptive analysis. If they all independently give the same descriptive analysis, not only will I know there is an audible difference, but I will probably also know exactly what it is.
Very true, but when ABX-ing not a heterogeneous population, but on single person: me. Thus, the important source of interindividual noise disapears, which is fine because I am only interested if I can here a difference mysellf on my own equipment. I used to design and conduct clinical trials for a living, but I must confess I don't do power calculations before testing, Very bad practice indeed.
On the other hand, often there are moments to unblind a trial, to see if it is necessary or worth continuing. I guess that is why I do, do 20 or so tests, see if a difference is looming, and continue or not. It's just a hobby after all.
Last edited:
Just a reminder: my comment of post #192 relates to your remark
in post #188. I'm sure an imperfect doubler would show up in the clock spectrum, but the deviation from the ideal clock spectrum would not necessarily be THD.