ADAU 1701 - Output noise vs sampling frequency

Hi all, I'm using adau1701 for an active dual channel amplifier to be used in 2-ways loudspeakers.

The dsp is used to drive differentially each input channel of the class d amplifier using an active filter similar to the one described in the datasheet. The measured a-weighted differential noise (dac0 - dac1) is around 100uV rms.

The Fs I'm currently using is 48 KHz. Is there any advantage in increasing the sampling frequency? My idea is that in theory noise shaping would be better.

If yes is there a comparison of Fs vs. output noise vs. current consumption?

Thanks,

Giacomo

  • 0
    •  Analog Employees 
    on Apr 5, 2021 9:16 PM

    Hello Giacomo,

    No, we do not have tests that we have done. We certainly have not done output noise verses current consumption. That is a bit of an odd pair of parameters to compare. Yes, the current consumption will go up with increased sample rate but it will not be a large rise. In this core the instruction rate of the core is always fixed. Increasing the sample rate only means that you have fewer instructions per sample period. What does run faster is the ADC and DACs so they will use more power. 

    The converters we use are oversampling converters so the out of band energy is well above the 24kHz Nyquist frequency of a 48kHs fs. So using an active filter on the ADC since and on the DAC side will improve the aliasing in the ADC and noise floor in the DACs. 

    Dave T

  • 0
    •  Super User 
    on Apr 6, 2021 12:21 AM in reply to DaveThib

         Re current consumption:  I measured 105 mA at 48K, and 106 mA at 192K.

  • 0
    •  Analog Employees 
    on Apr 6, 2021 1:49 AM in reply to KJBob

    Thanks Bob! I knew it would not be much more at 192kHz but the change you measured is less than I expected. 

    Dave T

  • Hi Dave,

    Thanks for the reply. It is clear that for SD adc most the noise spectral density is shaped to high frequencies and well above the Nyquist. What is not clear to me at this point is the way the adcs are handled depending on the sampling frequency that is set for the dsp. Is it correct to say that the SD sampling frequency is unchanged depending on the fs selection and that what changes depending on fs are the SD stream decimator settings? 

    Anyway, neglecting the adc implementation details, I've tried to evaluate the benefits of increasing Fs for folded noise. My anti alias input filter is a poor first order LP and moving from 48KHz to 96KHz has a noticeable effect.

    At this point I'm facing an issue: some sigma studio blocks are not working if I move from 48K to 96K. In particular all delay blocks are not working (No output) even if I've tried to set minimum delay and the code is safely compiled and downloaded.

    In my setup I have a 12.288MHz xtal, pllmode set to 10, and x512 instructions for Fs=96k (x1024 for Fs=48k).

    I remember a couple of years ago I had no issues with that. Could be something related to known issues with the latest versions of sigma studio?

    Thanks,

    Giacomo

  • I'd like to add that I've tried to remove by chance some components that are not on the signal path from my design such as logic ports, signal generators or filters used for signal detection and amplifier management. With some combinations the dsp works and with other ones not (for Fs greater than 48k).

    My impression is that it is a sigma studio issue. From the download page only 4.5 and 4.6 versions are available. I'd like to try with older stable versions. Where can I download them?

    Thanks,

    Giacomo