So this means you've a system that can't play back any 48k-family sample rates?
Another question seeing as I've not come across such an arrangement as yours before - are the reclocked signals sent twice through the ABT574? If so, that would be a way to mitigate metastability.
Another question seeing as I've not come across such an arrangement as yours before - are the reclocked signals sent twice through the ABT574? If so, that would be a way to mitigate metastability.
No it's not like that at all. It's not heard as such. It can be described as more digital sounding, which is fatiguing to listen to for a long time. There is no hum, hiss or anything like that.
Do you mean synchronisation of a flip flop with an incoming stream? There's one more thing that should be taken into account: the first stage of synchronous reclocking is done in the converter. As soon the stream goes into the DAC from converter it is processed by a digital filter IC: SM5842
Ok, another dead end. I have no idea what that means. But if it is not random by nature then it probably is not related to random data errors or metastability.
I can only describe it as lower quality sound which is not so interesting to listen to. That's all. There are no clear artifacts in the sound.
If that were a problem then I think I would ocassionally hear drops in the sound but there are no drops. So this tells me that clock distribution is arranged properly.
I'm not clear what 'the converter' means here. In your described locking system there's no taking account of (for example) the propagation delays down the S/PDIF interconnects which means the phase of the 33.8MHz originating clock won't get preserved at the soundcard. Hence no telling where the data transitions will be vis-a-vis the 33.8MHz clock when the signal arrives back.
As this is a DIY forum I'd strongly suggest modding your DAC to use two flip-flops in series in your resynchronizer to reduce the likelihood of metastability. The ABT574 has 8 flip-flips so probably you won't even need to add an extra IC.
Please find attached the schematics of the converter.
Attachments
Do you mean that a signal must come through 2 flip flops istead of one flip flop before arriving to the DAC IC?
I use Ian Canada's FIFO II which, as I understand it, shifts the data in using the source clock and shifts it out using a new TCXO clock, thereby isolating any jitter on the input from the output. This is probably why I cannot hear any difference between the coax and the optical connections from my CD player.
A Dir9001 provides the I2S data to the FIFO II which then feeds the I2S to PCM converter for my dual TDA1541 dac.
"An asynchronous I2S FIFO is a kind of logic device which can buffer the digital audio stream, allowing the audio data
to pass through while isolating the original clock and replacing it with a new secondary clock. If the new clock has
less phase noise than the old one, the digital audio stream after the FIFO will have less jitter and that will make the
DAC or other digital audio device playing the stream sound better
Moreover, the sound quality of the playback will be independent from the digital audio source.
So, together with clock technology, the I2S FIFO is firmly believed to be one of the most effective solutions to deal with jitter."
If you wish to reduce the chance of a metastable FF output reaching your PCM63s* to a minimum, yes.
*Actually your SM5842, which sits prior to the PCM63s.
Last edited:
IMO this more or less leaves only RF noise pickup that is dependent on the I2S wiring. This also should be visible with scope.
I was thinking along the same lines. But why does in my case wrapping balanced i2s pairs in copper tape create more noise and distortion despite the fact that some Hi End vendors sell specially made i2s HDMI cables ? I.e. in my case additional screening of balanced i2s causes distortion and degradation of sound quality but in the case of a Hi End i2s cable with special screening it does not create distortion.
Last edited:
Ground loops and current return flow..
Once you get above about 100Khz (and definitely above 1MHz), the return current no longer follows the direct path but follows the path of least impedance.. typically under the signal trace itself, and if you have a signal coax.. then you may want to check how that is connected to allow that return path without making go through an alternative and noise+jitter creating path..
Like I said.. current loops and noise will cause more impact through the power supplies if not thought about. And you can't simply decide not to deal with it. If you have a isolator, then that will need a return path, separate on each side.. and if you have a signal coax, that shield will act like a return current path.
Once you get above about 100Khz (and definitely above 1MHz), the return current no longer follows the direct path but follows the path of least impedance.. typically under the signal trace itself, and if you have a signal coax.. then you may want to check how that is connected to allow that return path without making go through an alternative and noise+jitter creating path..
Like I said.. current loops and noise will cause more impact through the power supplies if not thought about. And you can't simply decide not to deal with it. If you have a isolator, then that will need a return path, separate on each side.. and if you have a signal coax, that shield will act like a return current path.
You know after more thinking about your advice on the use of 2 flip flops prior to PCM-63 I came to conclusion that my DAC and converter indeed uses exactly this set up. It's just that the first flip flop is in the converter right after DIR9001, then the stream goes to the DAC wherein it is processed through SM5842, an isolator and then a flip flop again and PCM-63. So I have in total 2 flip flops on the way of the sgnal.
I have slowly come to the opinion that the near idiot-proof and complete isolation afforded by optical connections are worth the jitter penalty here.
From the point of view of mitigating metastability, it isn't exactly the same set up. Its not the total number of FFs that's important, rather the use of two in quick succession with the same clock.