• State
  • Locked Locked
  • Replies 6 replies
  • Subscribers 31 subscribers
  • Views 1413 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

PCM2912A, sampling clock accuracy

Roland Welte
Intellectual 450 points

Other Parts Discussed in Thread: PCM2912A

I need to implement a soundcard with an USB interface, so I came across the PCM2912A. One of the requirements for my design is that sampling frequency of audio input and audio output needs to stay within 10 ppm accuracy. This means that the sampling frequency must be derived from an TCXO oscillator connected to the PCM2912A and must not be derived from USB transfer timing, or in other words, USB transfer timing must be locked to the oscillator driving the PCM2912A.

I have learned that the only way to achieve this is by using USB transfer mode Isochronous-Asynchronous. The PCM2912A datasheet states "Adaptive Isochronous Transfer for Playback" and "Asynchronous-Isochronous Transfer for Record". Based on this, do I understand correctly that on audio input (ADC) the sampling clock is indeed derived from the PCM2912A, but on audio output (DAC) the sampling clock is derived from the USB data flow due to the Isochronous-Adaptive transfer mode being used?

The datasheet also states "The PCM2912A employs SpAct™ architecture, TI’s unique system that recovers the audio clock from USB packet data."What does that mean exactly. The whole wide web does not know anything about SpAct™ architecture.

Any help with these questions is very much appreciated,

Roland

  • 0
    TI__Guru** 114395 points
    Hi Roland,

    I think your thoughts are correct. ADC transfer is derived from PCM2912A and DAC transfer is derived from the USB data.
    SpAct™ architecture refers to the specific technique used to perform the Adaptive Isochronous transfer. I think this is mainly related to the PLL but there is no detailed information available for sharing.

    Best regards,
    -Ivan Salazar
    Texas Instruments
  • 0 in reply to Ivan Salazar
    Intellectual 450 points

    Hi Ivan,

    Many thanks for your information. This is a real pity. If both, ADC and DAC, sampling could be locked to the PCM2912A's clock, then one could build a nice accurate single-chip sound card solution. Apparently, it doesn't look that way. I will have to resort now to a 2-chip solution with a USB Streaming Controller and a Codec  :-( . Or would you have any other suggestions?

    Have a nice weekend,

    Roland

  • 0 in reply to Roland Welte
    TI__Guru** 114395 points
    Hi Roland,

    All PCM29xx family use this architecture for adaptive isochronous transfer mode, the purpose of this difference between playback and record sampling rates is to make these paths independent. I think a two chip solution seems to be the best way to build a system that is able to handle the sampling rates the way you need it. However, PCM2912A and all PCM29xx family are very good devices with nice sound quality. If it is possible, you could evaluate PCM2912A performance to see if it is good for your system requirements.

    Best regards,
    -Ivan Salazar
    Texas Instruments
  • 0 in reply to Ivan Salazar
    Intellectual 450 points

    Hi Ivan,

    I believe you when you say that the PC29xx family provide nice sound quality. As a matter of fact, I have ordered 2 evaluation kits of the PCM2912A. However, I need a mandatory sampling clock accuracy of 10 ppm in both directions (audio band modem design).

    Best regards,

    Roland

  • 0 in reply to Ivan Salazar
    Prodigy 50 points

    Hi Ivan, Hi Roland,

    I'm working here on something similar to Roland, but i only need the asynchronous isochronous transfer for the output. The 2912A would then be wonderful...but:

    I ordered an evalkit, and after a few measurement ( with an external ocxo/gps clock generating the clock and the recorded signal) noticed that once recorded on an iphone through the PCM, the sine period has an error of 10ppm (typically the iphone quartz error). 

    To me, this really looks like an synchronous transfer. 

    Any hint?

    Best regards,

    Yann

  • 0 in reply to Yann Mossier
    Intellectual 450 points

    Hi Yann,

    As I have written above, the PCM2912A datasheet states "Adaptive Isochronous Transfer for Playback" and "Asynchronous-Isochronous Transfer for Record".

    If you want accurate sampling on output (playback) then the PCM2912A will not do it because it works with Adaptive Isochronous Transfer on output (determined by USB timing, not your clock oscillator).

    Best Regards,

    Roland