Sampling Theory fun

Discussion in 'General Electronics Chat' started by Distort10n, Aug 9, 2007.

  1. Distort10n

    Thread Starter Active Member

    Dec 25, 2006
    429
    1
    Setup:

    TLV320AIC23B codec
    MCLK: 3.1457 MHz
    Input frequency: 700 Hz
    Base sampling rate F(s): 8.192kHz
    Oversampling: 384 * F(s)
    Do not want an anti-aliasing filter on the outputs because full bandwidth audio is needed (20 Hz to 20 kHz).

    Problem:

    ~8kHz frequency heard on the microphone output and lineout outputs.

    Question:

    Where do you think this person is going wrong?
     
  2. nanovate

    Distinguished Member

    May 7, 2007
    665
    1
    Is this device specified for a 3.1457 MHz clock?
    With a sampling rate of ~8 kHz you will not get the full audio bandwidth of 20 Hz to 20 kHz unless you are upsampling it with some kind of interpolation filter.
    I am not an expert in these matters and som maybe I am off base.
     
  3. Distort10n

    Thread Starter Active Member

    Dec 25, 2006
    429
    1
    The clock settings are appropriate. You do bring up a good point about the base sampling frequency of 8.192 kHz not being high enough for full bandwidth audio; however, the oversampling rate is 384 * F(s) or 3.145 MHz. So he should be ok here.

    The problem that I see, is that there is no anti-aliasing filter on the outputs. Even if you oversample, you still need to low pass filter the output. The TLV320AIC23B has internal digital filters where I am sure interpolation and decimation occur. One of the whole points of oversampling is to spread out the quantanization noise, and relax the requirements for the anti-aliasing filter; e.g., 2 pole rather than say 6 pole.

    Spectral components (the images) from every single Nyquist Zone excluding the first zone will fold back into the first Nyquist Zone. That is why he must be hearing the 8kHz sampling frequency.

    I think anyway.
     
  4. nanovate

    Distinguished Member

    May 7, 2007
    665
    1
    Can he boost the sampling rate since the output is going to full bandwidth? You'd lose some on the oversampling (maybe a little on the digital filtering also) but this might push the noise out past the audible range.
     
  5. Dave

    Retired Moderator

    Nov 17, 2003
    6,960
    144
    Although the purists may disagree, what you are essentially looking at here is an absence of reconstruction filters at the outputs.

    Dave
     
  6. Distort10n

    Thread Starter Active Member

    Dec 25, 2006
    429
    1
    Actually, I made a boo-boo. The AAF actually goes on the input to bandlimit the signal. I have been doing more reading on a delta sigma to gain a deeper understanding of the modulator.
    A work in progress.
     
  7. Dave

    Retired Moderator

    Nov 17, 2003
    6,960
    144
Loading...