Well, if you are going to try two different clocks, they don't have to be the exact same frequency. Maybe try a CCHD-957 at 49.152MHz, and one of the 27MHz or 50MHz standard clocks. Its a valid experiment although its not the experiment you initially had in mind.
Also, series resistors for most CMOS outputs should be 33R. That's because the IC output impedance is roughly 22R. Thus, the two impedances in series add up to 50R.
That would be a problem if the differences in sound were down around the threshold levels. For gross differences anyone can tell if there is a gross change in the sound. First thing I would do in this case is just try the cap and see if any perceived difference is gross or subtle.
Also, get other people to compare the sound too. Have them tell you what they are hearing that sounds different to them (just don't accidently bias them before you ask them to describe the difference). Ask them one at a time in private. Ask them not to discuss what they heard with the other listeners. Now we have some independent observers, which if they mostly describe the same differences are likely hearing something real. After all, if they are all imagining something, what are the chances of them all imagining the exact same thing. Its not a conclusive test, but its enough to suggest whether there is enough likelihood of some real difference to investigate further.
Of course, for subtle differences level matching is critical. For gross differences it may not make much or any difference. If there is loud buzz or something really obvious its still going to be there at most volume levels. Same with some other differences.
Now there are some caveats to look out for. If your test subjects describe the sound difference in vague terms like: "a little more dynamic," "little more bass or less bass," "punchier bass," those are exactly the types of differences people will describe if there are small level mismatches. If you are hearing descriptions like that then it would make sense to check levels carefully.
If you would like to know more about some kinds of real differences to check for, its not usually things having to do with FR. Its just because FR is so level dependent.
OTOH, things less commonly confused by small level mismatches IME might include: sound stage dimensionality in terms of width and depth, spatial location precision between and behind the speakers of virtual sources, degree of openness (clear separation between the sound of different instruments with some "black space" between them), if virtual sound stage is partially or entirely in front of the speakers or all between and behind the speakers, naturalness of instrument and vocals tones and textures, degree of presence or absence of low level room echos and reverberations in a symphony hall recording, etc.
Of course, all the above is IME and IMHO only. I can only suggest to try such things and try to learn from any mistakes made along the way as experience is gained.
Last edited:
Why do you think 49.152MHz would work better? Also why do you insist on using Crystek CCHD-957? If you wish to use an audio frequency clock NZ2520SDA 45.158MHz is available from Digikey, costs 10x less and has lower phase noise.
All those are impacted by even small level mismatches. And recording, speaker, room, listener position or even head movement have much higher impact on most of those than DACs or amps. So far no studies or tests exist that show well-implement DACs or amps making much of a difference on any of those. Of course it is another story if the goal is not to replicate what exists on the recording as closely as possible.
I have a similar question about clock for 9039q2m in i2s async slave mode. If I usually listen to 48K content, is it reasonable to use native clock for this frequency, such as 49.152Mhz or 24.576Mhz ?
As was already discussed with ASRC it may be beneficial to use a non-related clock frequency. But if there are no odd noises you should be ok with related clock frequency.
I didn't say that's what I think. It just happens to be a frequency where CCHD-957 is available. Its actually too high of a frequency to get very low close-in phase noise. Unfortunately, ESS dacs run async mode require a minimum clock frequency in order to play back certain sample rate content. Maybe good use the lowest clock frequency that meets the user's needs.
Joko Homo told me in detail about what it takes to get stable phase noise measurements out of those suckers. The slightest air currents make them drift too much so they had to be buried under blocks of Styrofoam in the most stable part of the room. Even the best selected low-measured-phase-noise ones I got from Jocko to try sounded bad in normal open room air.
Last edited:
In case it wasn't clear, I think 27MHz would be a better frequency than 49.152MHz, if all other things were equal. Lower frequencies tend to be better.
But not all other things were equal. I think CCHD-957 with MU caps will probably sound better than a "standard" clock.
Thus, I think CCHD-957 at 49.152MHz likely would sound better than "standard" at 27MHz.
But not all other things were equal. I think CCHD-957 with MU caps will probably sound better than a "standard" clock.
Thus, I think CCHD-957 at 49.152MHz likely would sound better than "standard" at 27MHz.
As I'm not interested in your opinion why don't you just let OP answer my question which was clearly meant for him if you can read.
I dont insist on using a special type, CCHD-957 was recommended above. I am just looking for a clock frequency and crystal type which is suited best for my usecase (and available at larger distributors). The source is Tidal (via toslink), so 49.152MHz would probably work great for 48, 96 and 192 kHz sample rate - of course there is still a lot material in 44.1kHz where the perfomance is questionable. There seems to be no common ground in the best suited clock for my application so I thought in implementing two which are switchable.
I am curios about what kind of listening (not measurement) equipment we are talking here where these differences can be observed so clearly.
So you think 24.576MHz would be better than 49.152MHz and probably 22.5792MHz would even be better for 44,1KHz if it is all from the same clock series?
As you are using ASRC the clock frequency does not have to match with the sample rate. So no need to use 48k multiple clock. And as discussed before non-related clock frequency may work better (e.g. 27MHz or 50MHz).
Very low distortion at all SPLs, speakers are Sound Lab large panel electrostatic speakers (645 frame size) which are a big part of what makes it possible to listen back up into the sound of the electronics. That and very careful room treatment. Slight changes in damping and or diffraction at certain points in the room can make surprising differences in the ability to reproduce a very precise sound stage. Also, frequencies below the 30Hz or so lower limit of the Sound Labs are augmented by dual REL subs. Website tends to be slow, but you can kind of get some idea about the speakers at: https://www.soundlabspeakers.com/majestic/
Power amps are kind of rare dual mono block Marantz MA9-S2. They have a sound, but its surprisingly consistent across all listening levels and frequencies, including for complex music signals. They are quite tolerant of the Sound Lab speakers' difficult to drive impedance curves.
Preamp is still under NDA, so can't say much about that. It has a sound too, however. Still, its quite easy to hear back up into the dac and what its doing.
In theory, phase noise increases by 6dB with each doubling of clock frequency.
In practice the first thing to do is to read the datasheet. Now what might be the minimum MCK frequency in ASYNC mode?
For HW# 16 it says 130*FS ≤ MCLK ≤ 49.152MHz. So minimum MCLK for 192kHz (max samplerate for me) should be 24.96MHz. Does this also mean a 50MHz clock would not work?