2nd order decision directed PLL

Hi,

I'm trying to implement a secod-order decision-directed phase-locked loop
as part of acoustic algorithm. I know my loop constants but what I'm not
sure about is how long do I have to integrate my phase error. It also
occured to me that I might have to unwrap the phase also.

Sorry if this sounds trivial - I just couldn't find any reference on this
in the literature.

Thanks,
-Momo 

On Apr 16, 7:47&#4294967295;am, "Momo Loulah" <mlou...@hotmail.com> wrote:
> Hi,
>
> I'm trying to implement a secod-order decision-directed phase-locked loop
> as part of acoustic algorithm. I know my loop constants but what I'm not
> sure about is how long do I have to integrate my phase error. It also
> occured to me that I might have to unwrap the phase also.
>
> Sorry if this sounds trivial - I just couldn't find any reference on this
> in the literature.
>
> Thanks,
> -Momo

Are you talking about integrating the phase error prior to feeding it
into the loop filter? If the loop filter has an integrator, you can
just feed the phase error in directly. If you put an integrate and
dump in front of the loop filter, you need to adjust the loop filter
coefficients to account for it. It can be beneficial in fixed-point
implementations to have the I&D because it allows the loop filter to
run at a lower sample rate, meaning the coefficient magnitudes are
larger for a given BW.

The lock range of the PLL will follow standard textbook predictions if
you don't unwrap the phase. The reason to unwrap is to extend the lock
range, but look at your requirements and consider other ways of
extending lock range too before jumping to that solution.

John

>Are you talking about integrating the phase error prior to feeding it
>into the loop filter? If the loop filter has an integrator, you can
>just feed the phase error in directly. If you put an integrate and
>dump in front of the loop filter, you need to adjust the loop filter
>coefficients to account for it. It can be beneficial in fixed-point
>implementations to have the I&D because it allows the loop filter to
>run at a lower sample rate, meaning the coefficient magnitudes are
>larger for a given BW.
Thanks for the clarification John.

>The lock range of the PLL will follow standard textbook predictions if
>you don't unwrap the phase. The reason to unwrap is to extend the lock
>range, but look at your requirements and consider other ways of
>extending lock range too before jumping to that solution.
I think you'r right. I probably shouldn't tamper with the phase before
exhausting all other possibilities. Don't know whether allocating more
fixed-point precision would help at all in this.

-Momo

> I dont know about the specific algorithm you are using, but if it
> dictates you to stop tracking after the acquisition, you can stop
> filtering operation as well. Still you need to keep track of the error
> signal so that it is kept within the limits. When the error signal
> exceeds the threshold, you will need to acquire again.

Just spotted this. Strange, for some reason, your post doesn't appear
in dsprelated. Anyway, yeah I am actually worried that if I keep the
loop on all the time, the phase accumulator would saturate. So I guess
you'r right; I ought to use thresholds to control the loop operation.

Thanks for the hint Ubaid.

Salam,
-Momo