FPGA Central

hi all,
i am having a basic doubt on basis of current frequency,phase,
channel tracking algorithms. say we will take one example.. assume we
have

the first row below are the original values and the second row values
are my tracked values based on 1st order PLL (a simple loop filter)

original values 16429 16105 16021 16186 16278

tracked values 16125 16512 16573 16492 16381

and also please refer this pdf at

http://www.sal.tkk.fi/Opinnot/Mat-2....les/enyb03.pdf

even in this case also, the tracking is very near to original values
(please note, i said very near,SAME AS IN ABOVE CASE).

even in a simple loop filter, i need to ensure that the algorithm has
input of difference between current and previous estimate

even in phase locked loop, after the error has propogated it will take
a while (say few samples to compensate it)

my doubt is when we know the error why dont we compensate it
instantly..say in the above case of two rows, we keep on accumulating
the error and then compensate it..why dont we correct with the value
known ..instead we accumulate it and then try to correct it

thanks in advance for all answers..

please do let me know..am i missing any fundamental principle here...

thanks again
particle (filter) reddy

On Jan 25, 8:05 pm, "PARTICLEREDDY (STRAYDOG)"
<particlere...@gmail.com> wrote:
> hi all,
> i am having a basic doubt on basis of current frequency,phase,
> channel tracking algorithms. say we will take one example.. assume we
> have
>
> the first row below are the original values and the second row values
> are my tracked values based on 1st order PLL (a simple loop filter)
>
> original values 16429 16105 16021 16186 16278
>
> tracked values 16125 16512 16573 16492 16381
>
> and also please refer this pdf at
>
> http://www.sal.tkk.fi/Opinnot/Mat-2....les/enyb03.pdf
>
> even in this case also, the tracking is very near to original values
> (please note, i said very near,SAME AS IN ABOVE CASE).
>
> even in a simple loop filter, i need to ensure that the algorithm has
> input of difference between current and previous estimate
>
> even in phase locked loop, after the error has propogated it will take
> a while (say few samples to compensate it)
>
> my doubt is when we know the error why dont we compensate it
> instantly..say in the above case of two rows, we keep on accumulating
> the error and then compensate it..why dont we correct with the value
> known ..instead we accumulate it and then try to correct it
>
> thanks in advance for all answers..
>
> please do let me know..am i missing any fundamental principle here...
>
> thanks again
> particle (filter) reddy

The accumulator (integrator) is necessary to track a frequency offset.
Otherwise you are constantly chasing lock but never get there, because
the reference will always spin with respect to the data. Therefore you
are constantly correcting for the spin, and your system BER will
suffer because of phase noise. What you want to do is just what the
books say -- feed the error into a PI filter. This gives you a classic
second order system which will overshoot at first but then lock onto
the offset, so the reference and the data are in phase and no longer
spinning with respect to one another.

John

PARTICLEREDDY (STRAYDOG) wrote:

> even in phase locked loop, after the error has propogated it will take
> a while (say few samples to compensate it)
>
> my doubt is when we know the error why dont we compensate it
> instantly..say in the above case of two rows, we keep on accumulating
> the error and then compensate it..why dont we correct with the value
> known ..instead we accumulate it and then try to correct it

Indeed there are "feed forward" algorithms which estimate the phase and
the frequency in advance as opposed to the PLL which is a "feed back"
algorithm. This is kind of like FIR versus IIR filters. The reason for
using the PLL is that it is simple and it does the job.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

On Fri, 25 Jan 2008 17:05:16 -0800, PARTICLEREDDY (STRAYDOG) wrote:

> hi all,
> i am having a basic doubt on basis of current frequency,phase,
> channel tracking algorithms. say we will take one example.. assume we
> have
>
> the first row below are the original values and the second row values
> are my tracked values based on 1st order PLL (a simple loop filter)
>
> original values 16429 16105 16021 16186 16278
>
> tracked values 16125 16512 16573 16492 16381
>
> and also please refer this pdf at
>
> http://www.sal.tkk.fi/Opinnot/Mat-2....les/enyb03.pdf
>
> even in this case also, the tracking is very near to original values
> (please note, i said very near,SAME AS IN ABOVE CASE).
>
> even in a simple loop filter, i need to ensure that the algorithm has
> input of difference between current and previous estimate
>
> even in phase locked loop, after the error has propogated it will take a
> while (say few samples to compensate it)
>
> my doubt is when we know the error why dont we compensate it
> instantly..say in the above case of two rows, we keep on accumulating
> the error and then compensate it..why dont we correct with the value
> known ..instead we accumulate it and then try to correct it
>
> thanks in advance for all answers..
>
> please do let me know..am i missing any fundamental principle here...
>
> thanks again
> particle (filter) reddy

Bit clock and carrier signals are noisy, and depending on the modulation
scheme and filtering there is often incomplete information at each bit
period, or intersymbol interference (which appears as noise). The action
of the PLL is to average the carrier and clock phases over many bits,
which rejects noise and spans the stretches where there isn't solid bit
clock information.

If you are given bursts of data you can fit the bit clock and carrier to
it after the fact, then decode the data -- but you're still carrying out
the same task as the PLL does, just in a different way.

--
Tim Wescott
Control systems and communications consulting
http://www.wescottdesign.com

Need to learn how to apply control theory in your embedded system?
"Applied Control Theory for Embedded Systems" by Tim Wescott
Elsevier/Newnes, http://www.wescottdesign.com/actfes/actfes.html