Connect with us

Audio phase lock loop design info

Discussion in 'Electronic Design' started by Jorgito, Mar 21, 2008.

Scroll to continue with content
  1. Jorgito

    Jorgito Guest

    Hi All,

    Looking for basic design reference info for audio phase lock loops. For
    example, given a specified analog input S/N, what acquisition time is needed
    to achieve 1 degree phase resolution?


  2. MooseFET

    MooseFET Guest

    Acquisition time and S/N doesn't lead directly to phase accuracy. A
    great deal depends on the type of phase detector you are using. The
    initial locking process is very nonlinear. Once you get past that,
    the situation is often linear enough that a simple linear model of the
    situation works. The noise near the signal ends up mattering more
    than that far away.

    What roughly is your SNR?
    What sort of phase detector are you planning on using?

    It is a fairly common practice to make the filter part of the PLL
    change so that you get a quick lock and initial settling and then
    later enough low passing to get the needed noise rejection.
  3. Jorgito.

    Jorgito. Guest

    Hello, thanks for replying.

    My signal of interest is a continuous 20 kHz analog tone. The S/N will be >
    50 dB. I have several seconds or more to lock. I'd like to be able to
    measure the phase to better than one degree. Can you recommend an approach?

    Thanks again,
  4. MooseFET

    MooseFET Guest

    50dB -> 1:316

    arcsin(1/316) = 0.2 degrees

    Life is good. I'm not even sure you need the whole PLL.

    Your 1 degree measurement suggests that you have a reference signal
    and a signal to be measured. Is this the case or are you looking to
    detect a 1 degree modulation of a signal.

    The flip-flop phase detector in a 4046 will settle to 1 degree of
    error quite quickly. For small phase errors, the detector plus
    resistor can be simplified to be considered as a current source in the
    math. For small on times, the effective value of the resistance is
    very high so the math gets a lot easier.
  5. Tam

    Tam Guest

    If I did this right, 1 degree at 20 KHz is about 140 ns. A PLL should easily
    have a ststic error less than 1/10 of that.

  6. MooseFET

    MooseFET Guest

    Yes, at least for a HC4046.

    1/(20K * 360) = 138.888nS

    Also tan(1) = 0.017

    So you need a cut off frequency of about

    20K / 0.017 = 1.2MHz so fairly normal op-amps will do.
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day