May I politely correct your statement: NOS has nothing to do with LRCK clocking out the audio waveform. It is depending on DAC chip (or type).
BTW these phase noise plots seem familiar. Are you sure you are not acting as a proxy to a banned member?
I will keep an open mind about this and investigate further, thank you for your efforts!
Please see page 15 of the attachment on phase noise.
The noise skirt measurement recently proffered shows the conflation of two types of noise as depicted in the frequency domain. How are the two sources being separated so we know how much of it is phase noise? Also, FFT is an integral transform. As such there is an averaging effect. Mere power spectra does not show us what phase noise is doing in the time domain exactly. Nor does it show us how phase noise is affecting the two channels that are required for production of the stereo illusion. Looking at conflated, averaged noise for one channel is meaningless relative to perception of imaging in the sound stage. Unless of course, someone would like to explain in detail how to correlate the two in a meaningful way.
Also, please see page 17 of Bob Adam's presentation. If we are talking about DSD vs PCM, then jitter/phase-noise has a different relative importance because of the step size.
Last edited:
Looking for some trouble shooting help - I got the RaspberryPi4b/FifoPiQ7/HdmiPiPro going into a Denafrips Pontus 2 DAC and can’t get any sound, using Ropieee on the RPi4B, tried the different I2S settings on the DAC and still no luck, the DAC lights up showing 44.1 signal coming in but no sound, I feel like I’m missing something obvious
Thank Mark, interesting reading. Talked to a few RF testers here and they also agree FFT measurements to figure out phase noise performance are not reliable. Amplitude noise will mask or cause errors. A dedicated phase noise analyser is best Unfortunately, none have audio frequency types that could be used to measure what is coming out from a DAC. All mainly MHz-GHz suitable for clocks and audio/if signals. The closest that will fit seems to be this gear:
https://www.keysight.com/us/en/product/N9042B/uxa-signal-analyzer-2-hz-50-ghz.html
I was thinking of 1khz carrier to test and see what comes out as phase noise.
Thoughts?
Hi,
1. What clock are you using? If accusilicons you need to lift pin 1
2. I use generic rpi i2s driver on ropieee. Using the audiophinics es90… driver doesn’t work on my setup.
Hope this helps
As I said before the noise shown in my graph contains both phase modulated and amplitude modulated noise. These together are referred to as phase noise which may be quite confusing. Yes, amplitude modulated noise will most probably be dominant but that was my point. Clock induced phase modulated noise has less impact and according to studies amplitude modulated noise is more offending
(https://www.diyaudio.com/community/threads/phase-noise-in-ds-dacs.387862/post-7066124).
I'm more than happy to see real measurements at dac output instead of measurements of clocks. What you hear is what comes out of the dac, not what comes out of the clock.
Last edited:
All this applies to clock phase noise measurements as well. However that is not even a measurement of what you actually hear when listening to dacs.
e.g. if Fs is 44.1k pcm the why is LRCK (Input Freq) showing 3MHz instead for this test?
I stand corrected on how this particular dac works. IIUC it does upsample so that LRCK is roughly 2.8/3MHz. That is frame rate at which the data word changes.
In principle phase noise of, say, a 1kHz test tone could be demodulated. However, a potential problem is that it usually takes a measurement system that is, say, maybe, 10x better than the DUT so that test system performance does not produce too much measurement error.
However some measurement technique might be developed given sufficient time, effort, and money. What if we did that? Then we would probably still be having arguments about whether humans can hear any difference or not. More time, effort, and money could presumably solve that problem too. Question is, is anyone willing to pay what it would cost? IMHO and IME so far nobody cares enough to pay tens of thousands of dollars, if not more, to produce publication quality research on the subject. Even if someone were willing to do that, some skeptics will still say one research paper doesn't prove anything. They will say there needs to be broad consensus among experts who study such things.
Simple thing to do if anyone is interested enough would be travel to a place where there are multiple measured clocks and a sufficiently good reproduction system. Some people may hear the effects right away, and other people may have to take time to develop some skill. There may also be some people who might be classified as localization-deaf. Don't have enough experience with listeners at this point to have much idea what the distribution would look like.
Last edited:
Even with phase noise demodulator you cannot escape the fact that amplitude modulated noise is most probably dominating. And I seriously doubt demodulated phase noise of dac output would show anything close to the huge differences seen in clock phase noise measurements.
Audibility of clock phase noise at dac output is certainly questionable with even run-of-the-mill clocks and the law of diminishing returns applies here as well. Regarding your current hypothesis about soundstage effects there are various other sources for those as well. And I doubt many recording studios have SOA clocks for AD stage so according to your hypothesis the soundstage is already ruined at recording phase. Besides very few recordings have a proper soundstage to begin with. E.g. in classical music recordings close miking is often used which results in very strange effects such as high and low strings of solo violin having different location in soundstage.
Audibility of clock phase noise at dac output is certainly questionable with even run-of-the-mill clocks and the law of diminishing returns applies here as well. Regarding your current hypothesis about soundstage effects there are various other sources for those as well. And I doubt many recording studios have SOA clocks for AD stage so according to your hypothesis the soundstage is already ruined at recording phase. Besides very few recordings have a proper soundstage to begin with. E.g. in classical music recordings close miking is often used which results in very strange effects such as high and low strings of solo violin having different location in soundstage.
I’m using the clocks that came with the board, I have tried both the generic abs audiophonics drivers and still no sound, also have tried both the enabler and disabled settings for the 16 -32 conversion