PCM2902 XTI and external clocksources

[hifi]

I seem to recall having read somewhere that PCM2902 or maybe it was the PCM2702..or some other family member.. Have trouble accepting an externaly generated clock on XTI?


Can somebody confirm / deny this?


It would be fun trying this:

http://rocky.digikey.com/WebLib/CTS/Web Data/CB3,-LV Series.pdf

And the ideas presented here:

http://www.wenzel.com/documents/finesse.html

 

[John Swenson]

I've done an external clock to a 2706 with no problem at all. I haven't tried a 2902 though.

My experience has been that the 270x parts sound better than the 290x parts, so if you have to choose between a 2902 and a 2702 I would go with the 2702.

John S.

 

[hifi]

Thanks, I´m aware of the preference for the 2702 over 2902 when it comes to analog outputs. But I´m just after the Spdif output of the 2902.

I´ll be ordering boards in a week or so 🙂

/ Mikael

 

[John Swenson]

There is one other very important difference, the 2902 runs at just 48KHz, the 2702 will run at 44.1. The 2902 says it will take 44.1, but it converts it to 48, the S/PDIF is always going at 48, no matter what your input format is. The 2702 will output what you put in.

On the clock isue my experience has been that feeding a very low jitter clock into a 2706 really does significantly improve the sound over using a crystal with the builtin oscillator.


John S.

 

[paologatto]

There is one other very important difference, the 2902 runs at just 48KHz, the 2702 will run at 44.1. The 2902 says it will take 44.1, but it converts it to 48, the S/PDIF is always going at 48, no matter what your input format is. The 2702 will output what you put in.

How you can say this? I made extensive tests with PCM2902/6 in loopback using MLS and analysing its impulse and frequency response. I found that the impulse is perfectly clean at any of the standard sample rates.. and in my experience this tells that there is NO resampling, so the internal clock of the codec changes accordly to the selected sample rate.

In other soundcards, known to work always at a fixed sample rate, the effect in the impulse response in loopback is evident!

On the clock isue my experience has been that feeding a very low jitter clock into a 2706 really does significantly improve the sound over using a crystal with the builtin oscillator.

John S. [/B]

But.. and the SpAct inside the PCM2902/6 (and 2706? don't remember)? Doesn't this override any other concern about external crystals? (maybe I'm wrong in this..)

 

[hifi]

I know the 2902 can´t be resampling to 48Khz.

The PLL in my dac locks on to the 64fs produced from a 44.1Khz source just fine but not 48Khz (The "source" being a 2902 feeding known data of 44,1/48khz)

 

[John Swenson]

Hmm, thats very interesting, I've never been able to get a 44.1 S/PDIF out of a 2902. No matter what driver, operating system, application I use it always puts out 48. I have both a 2902 and 2706 board and with the same OS/driver/app/files when I play a 44.1 file to the 2902 I get a 48 S/PDIF out and a 44.1 from the 2706. I wonder if they changed the chip somewhere along the line? Mine is about 2 years old now.

On the clock front, I'm not exactly sure why using a low jitter clock makes a difference. I didn't think it would either, but I decided to try it, and low and behold it does make a signficant difference.

The latest data sheets for the2706 say it runs in adaptive mode in playback, if this is true then the output clock would be generated from a PLL referenced to the system clock rather than from the SPACT itself, in this case the quality of the input clock would have an effect on the output clock. No it would NOT be as good as the input clock because it does still have to go through a PLL, but the input clock would make a difference.

John S.

 

[paologatto]

In my experiments the S/PDIF from PCM2902/6 is:

- at 32 kHz when a file at 32 kHz is played
See http://purebits.com/temp/pcm2906-32KHz.jpg


- at 44.1 kHz when a file at 44.1 kHz is played
See http://purebits.com/temp/pcm2906-441KHz.jpg


- at 48 kHz when a file at 48 kHz is played
See http://purebits.com/temp/pcm2906-48KHz.jpg


- at 48 kHz when a file at any other not standard sample rate is played
See http://purebits.com/temp/pcm2906-40KHz.jpg (40 kHz)
and http://purebits.com/temp/pcm2906-96KHz.jpg (96 kHz)


Tests are done playing files at different sample rates and connecting the S/PDIF output of the PCM2906 (via toslink) to the input of an RME Fireface800, that has a nice utility DIGICheck that analyzes incoming data streams.

 

[00940]

The latest data sheets for the2706 say it runs in adaptive mode in playback, if this is true then the output clock would be generated from a PLL referenced to the system clock rather than from the SPACT itself, in this case the quality of the input clock would have an effect on the output clock. No it would NOT be as good as the input clock because it does still have to go through a PLL, but the input clock would make a difference.

John S.

All TI's PCM290* and PCM270* are running in adaptive mode (which just means that the PCM2*** will adapt to the incoming flow). I'd be highly surprised if the USB receiver section in each of those chips was any different of the system developped originally for the PCM2702. Thus, the final system clock is generated by a clock generator using the signal sent by the DCO inside the SPACT. The input clock will only control the USB endpoint (little interest) and the phase error detector and DCO of the SPACT.

Just as shown in those diagrams made by the designer of the SPACT :

An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.

 

[paologatto]

The full story of the PCM2702 development is really interesting too:

http://www.planetanalog.com/showArticle.jhtml?articleID=12801995