Strange, we have got good result even with a Pierce and AT-Cut crystal, this is a Pierce at 6.144MHz
Attachments
Possibly air currents,, Xtal quality, I don't know enough about your oscillator to comment.
Again, I don't doubt it is good. I do wonder how good is good enough. If you are improving something, you want a baseline to compare to. You have not done that.
The music tests are merely an opinion and basically useless as proof of anything. Yes, I read the darned evaluations. You needed to correlate some measurements to actual tests to show reduced defects in the sound. Music is pretty useless for that purpose. It is good to make others feel good I guess, but falls far short of anything concrete that you should use to sell your oscillator.
I don't see you ever making measurements to see what the average oscillators performance is, or what makes a difference. You will leave that to others I guess. So I will devote my time to something that is worthwhile. Believe it or not, I was trying to help you. The information I was asking of you would have been useful to you and others. You just don't see it.
-Chris
Again, I don't doubt it is good. I do wonder how good is good enough. If you are improving something, you want a baseline to compare to. You have not done that.
The music tests are merely an opinion and basically useless as proof of anything. Yes, I read the darned evaluations. You needed to correlate some measurements to actual tests to show reduced defects in the sound. Music is pretty useless for that purpose. It is good to make others feel good I guess, but falls far short of anything concrete that you should use to sell your oscillator.
I don't see you ever making measurements to see what the average oscillators performance is, or what makes a difference. You will leave that to others I guess. So I will devote my time to something that is worthwhile. Believe it or not, I was trying to help you. The information I was asking of you would have been useful to you and others. You just don't see it.
-Chris
30dB better than the Crystek CCHD-957 at 10 Hz from the carrier?
Attachments
Thanks Chris - Glad to see some finality here.
I'm sure I will see you around on other threads. In the meantime the community can get back to subjective and objective measurements and testing.
Good luck in your ventures.
Sorry Marcel, I don't. I have to admit I have no idea what you are talking about here, and I am pretty sure you don't understand my point of view as well.
The process gain of a FFT is defined as 10log(N/2) where N is the number of lines in the FFT (here, N=8,000,000 something) which is 66dB process gain. Now, add this process gain to the FFT noise floor and one gets the SNR of the ADC.
You may want to review this reference https://www.analog.com/media/en/training-seminars/tutorials/MT-001.pdf pp. 7.
Next time when trying to engage, please stop using one line dismissals and try to explain yourself (or provide a reference if you don't have the time). I know we collided several times in the past on different topics, if you hold any grudges that's your problem, not mine.
Last edited:
I think I understood enough then... it does prove what is relevant, measured improvements are possible from improved clocking, even against one of the best measuring DACs on the market currently.
Not like it will be a surprise if a controlled version of the test shows similar improvements.
Except it doesn't. It just shows one DAC is better than another at that objective parameter. Correlation is not causality. Could be anything or dozens of things as everything gets critical as you chase the last few dB.
It could be anything and everything, but all related to clocking if we are looking at the effects on jitter... no?
Maybe, but then I don't understand the dismissive comment in https://www.diyaudio.com/forums/dig...itter-crystal-oscillator-438.html#post6711058 when it was obvious the process gain is wrong.
It should theoretically go significantly down. So it's not the result of the down conversion dividing the frequency deviation.
For me, it seems, correlates like as described by John here:The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator
Up, I think, as the ratio from the clock to the output frequency gets smaller. Anyway, it's clear there is something else going on, something that's audio frequency independent.
The signal power is actually -4.091 dB on Joseph's plots from post #4311, so the discrepancy is 4.091 dB worse than I wrote earlier. That doesn't change anything, of course.
I was not familiar with the term process gain. That is, I use it all the time, but I didn't know it is called process gain.
I understand that a sigma-delta modulate with lots of ultrasonic noise is more susceptible to far-off phase noise, because far-off phase noise can modulate that out-of-band noise into the band of interest, but I don't see why it would matter for close-in phase noise.
That is, the output signal (before filtering) of a single-bit sigma-delta DAC spectrally consists of the desired signal, a small amount of in-band noise and lots of out-of-band noise. When jitter randomly shifts the signal transitions back and forth, all frequency components get phase modulated. As the shift in time is the same for all components, the shift expressed in radians is larger for higher frequencies. The same holds for a many-bit DAC without noise shaping, but without the lots of out-of-band noise.
Close-in phase noise produces close-in skirts around all spectral components, but the only ones that end up in the audio band are those around the desired signal, which is the same for the single-bit and many-bit case. Far-off phase noise causes far-off sidebands around all spectral components, potentially converting out-of-band noise into the audio band in the single-bit case.
Obviously this whole line of reasoning is either wrong or incomplete, as Joseph measures much stronger sidebands than expected and also sidebands that don't depend on the audio frequency. I'm just wondering where the error or incompleteness is.
- Status
- Not open for further replies.