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.

Excessive noise with PCM1804.

Other Parts Discussed in Thread: PCM1804, PCM4104

Hello,

In my system I'm seeing -54dBFS from PCM1804, which I believe to be excessive. I'm reading this level inside the DSP software, the noise disappears when the relevant serial port is disabled which narrows down the source of the noise to the PCM1804.

The PCM1804 is configured as a slave to the DSP, changing the PCM1804 to be the master improves the noise reading by 6db to -60dBFS.

I get good performance from my DAC so I don't suspect jitter.

Analog input circuity is largely as in the datasheet, 10nF at the input pins with appropriate feedback cap on the preceding op amps.

BCLK and LRCLK signals look good on the 'scope. Triggering on MCLK signal I obtain stable traces for BCLK and LRCLK which I believe indicates they are synchronous.

Decoupling is as recommended in the datasheet, with extra ferrites (supply side of decoupling caps).

I have tried replacing the device but get same results. It does generate some heat, more than the PCM4104 I have in the system, perhaps relevant?

Can anyone help me to track down the problem please?

Thank you.

  • Hi,

    Sorry that you are experiencing trouble with our device.

    Have you verified the power supplies and decoupling caps for internal references are all clean? Having noise on these terminals would explain the decreased performance that you are seeing.

    -d2
  • Hello, thanks.
    With scope on 5mv/div sensitivity I can see no noise on the internal reference pins, at analog power supply pin I can see a slight thickening on the trace, looks to be around .5mV peak. Does that seem sufficiently low?
    Have just added an extra 10nF COG cap to the supply rail which didn't change things.

    EDIT: Found a spare probe pin to replace the broken one so was able to take a better (differential) measurement, supply noise looks to be around 100uV.

  • Hi,

    Do you have the 0.01 uF caps across the inputs? You do you have inputs biased at VCOM via external circuitry?

    -d2
  • Yep and yep. Inputs biased at 2.5V, 100nF caps at VCOM pins.
    Admittedly I've use ceramic caps where the datasheet recommends tantalums but I'd be surprised if that's adding 40dB odd of noise.
  • Is it possible to get a scope shot posted here of the data output and clocks (capturing at least 2 frames)? Also what is the settings of the HW control pins?

    Justin
  • Certainly, is there a file upload function on this forum that I can't see?


    HW control pins are as follows:
    FMT0-high
    FMT1-low
    S/M-high
    OSR0-low
    OSR1-low
    OSR2-low
    BYPASS-high


    Master clock=12.288MHz
    DSP sample rate=48k
  • In order to upload, you have to Use rich formatting. there is a hyperlink on the right hand side of the screen that allows you to change the editor style when composing a post.

    -d2
  • Thanks Don.

    I've built up another board and now I get the noise only intermittently, if I turn power off and on again several times it often goes away! Below are the scope shots, one at a time though I'm afraid. Made with 10x setting, inch long ground clip, 60MHz bandwidth scope. Changing the series resistors doesn't noticeably effect the waveform so I'm not sure but I may be reaching the limitations of my probe/scope here. Many thanks.

    MCLK:

    LRCLK:

    BCLK:

    SDATA:

  • The reason I ask for scope shots is to see the 4 I2S signals together and get a sense of the clock rates, as well as where the data is aligned so that I can match it up with the settings. So the above shots at least give me an indication that the clocks are clean looking but none of the other information. You have the part set for normal I2S which is actually right justified, is this what your DSP is set up for?

    Justin
  • Hi Justin, unfortunately my 4 channel scope is out of action at the moment. I can do two traces at a time, can you work with that? (Having said that, I think I can access another 4 channel scope so no big deal if not).

    DSP is indeed set up for I2S (delayed by 1). Have I definitely got it set up for 48kHz operation? When it's noisy I can change settings in DSP to 96kHz which normally stops the noise, but conversely, when it's noisy and at 96kHz, changing to 48Hz also often the noise, so don't know that that bears any relevance.
  • Hi, got the scope working again. Could really do with some help here please.


    (Found it impossible to get edges of all 4 traces showing at the same time, as LRCLK is so much slower than MCLK, maybe one of these new scopes could do it but not mine. Traces are less clean and tidy due to ground clips etc...).

    So, MCLK, BCLK and SDATA when noise is low:

    MCLK, BCLK and SDATA when noise is high:

    Can see a slight change in SDATA, is more visible in the next two.


    BCLK, LRCLK and SDATA when noise is low:

    BCLK, LRCLK and SDATA when noise is high:

    This change in SDATA is consistent, does this imply this excess noise is being sent to the DSP as 'information' as opposed to signal corruption somewhere?

    Changing the ADC from slave to master stops this noise however these 'spurs' crop up. Would imagine the problem is related, do these 'spurs' help with things to look for? With a 1k input signal the spurs are at 15kHz and 17kHz, with a 500Hz input signal, the spurs are at 14.5kHz and 16.5kHz.

    Many thanks.

  • Hi,

    If you change the sampling rate when in master mode does this also move the spurs? What is providing the master clock, a PLL or crystal. It would be helpful if you could post the schematic for this portion of the system to get a better idea of the inputs and outputs. Are you able to hook up the outputs to a know good I2S receiver such as an audio tester?

    Justin
  • Ah! Silly me, I was changing from slave to master but not changing the oversampling configuration pins, with those set appropriately the spurs disappear.

    Would you mind if I e-mailed you my schematic and layout please, to see if anything obvious is amiss? I now have a work around but would like for this to not happen next time!

    Many thanks, again.