symbol synchronisation

Hello People

I would like to develope a adaptive symbol synchronisation assuming I have
got perfect phase synchronisation.

Can you please suggest some techniques.? I have looked into some books but
the treatment is not enough.

Your guidance will be greatly appreciated.

Thanking you,

Chintan

On Thu, 31 Jul 2008 15:39:54 -0500, "cpshah99"
<cpshah99@rediffmail.com> wrote:

>Hello People
>
>I would like to develope a adaptive symbol synchronisation assuming I have
>got perfect phase synchronisation.
>
>Can you please suggest some techniques.? I have looked into some books but
>the treatment is not enough.
>
>Your guidance will be greatly appreciated.
>
>Thanking you,
>
>Chintan

Phase-locked loop with a Timing Error Detector.  Usually, though,
timing locks first and then phase.

Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.ericjacobsen.org

Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php

On Aug 1, 4:39&#4294967295;am, "cpshah99" <cpsha...@rediffmail.com> wrote:
> Hello People
>
> I would like to develope a adaptive symbol synchronisation assuming I have
> got perfect phase synchronisation.
>
> Can you please suggest some techniques.? I have looked into some books but
> the treatment is not enough.
>
> Your guidance will be greatly appreciated.

When you say "phase synchronisation" I assume you mean carrier phase
rather than symbol phase. Most symbol sync techniques don't actually
care too much about carrier sync., and are typically placed before it.

Learning how to do symbol sync sucks. Every comms book covers the
topic, but none really get to grips with it. They all describe simple
techniques that work in simple cases, but generally avoid the serious
real world cases, and how the various algorithms can fail.

You didn't say anything about the kind of symbols you are trying to
sync to. It matters a lot. If there is significant excess bandwidth in
the signal, and it is PSK, simple techniques like a Gardner timing
error detector (which if you've looked at any digital comms book you
must know about) work really well. If there is lots of excess
bandwidth, and you are able to use slow tracking, Gardner can even
work well with high order QAM. You just need to consider the average
of a number of Gardner tests, rather than considering each individual
test to be meaningful. However, simple techniques like this tend to
fall apart, or suffer from multiple minima (i.e. they can settle to
certain wrong places just as easily as to the right one) when the
excess bandwidth is small - i.e. the typical real world condition, for
anything moderately advanced.

Techniques based on cross correlating the two band edges takes more
compute power, but are very predictable. This is a common basis for
symbol sync in things like V.34 modems, where the bandwidth of the
channel is very close to the underlying bandwidth of the signal.

Regards,
Steve

On Jul 31, 4:39&#4294967295;pm, "cpshah99" <cpsha...@rediffmail.com> wrote:
> Hello People
>
> I would like to develope a adaptive symbol synchronisation assuming I have
> got perfect phase synchronisation.
>
> Can you please suggest some techniques.? I have looked into some books but
> the treatment is not enough.
>
> Your guidance will be greatly appreciated.
>
> Thanking you,
>
> Chintan

The best references for this are D'Andrea and Mengali's book,
and Meyr and Ascheid's book.  Meyr has a second book with
Moeneclaey that is very digitally-oriented, and is worth a look.

Like Eric said, typically symbol synchronization is done in two
parts:  a timing error detector/estimator, and a timing filter
to smooth out the estimates.  Then you need a timing
compensator to take the smoothed estimate, and use it to
correct the next (batch of) symbols.

There are several timing error detectors, with their pros and
cons.  The two books I mention above have this information.

Julius

cpshah99 wrote:

> Hello People

Hello pupil,

> I would like to develope a adaptive symbol synchronisation assuming I have
> got perfect phase synchronisation.

This is trivial and impractical situation. If you have the perfect 
carrier synchronization, you already have the symbol synchronization 
(unless there is a frequency shift in the channel). Carrier sync and 
symbol sync are directly related; the only thing to find out is the 
phase of the symbol sync wrt carrier.

> Can you please suggest some techniques.?

There are many. It largely depends on what is the signal and what are 
the other requirements: speed of tracking, processing power, algoritmic 
delay, bandwidth, SNR. For the simple cases, Gardner's algorithm works 
well.  For the complex cases, I like to extract the timing information 
from the equalizer.

> I have looked into some books but
> the treatment is not enough.

Viterbi & Omura "Digital Communication"

Heinrich Meyr; Marc Moeneclaey; Stefan Fechtel
"Digital Communication Receivers : Synchronization, Channel Estimation"

> Your guidance will be greatly appreciated.

How great exactly is the great appreciacion?

> Thanking you, 
> Chintan



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

Hello All

Thanks a lot for responses.

My signal is QPSK modulated at carrier frequency of 12 KHz and with
bandwidth of 8 KHz. Now as I am modeling doppler in my system, I also have
doppler correction at receiver which uses Linear Interpolation. That
algorithm works very well, in the sense that it estimates the delay from
the equaliser and I use that delay to correct symbol timing and to correct
the carrier phase.

But this technique doesn't help me with turbo decoding. So I was thinking
that when we use turbo decoding do we just correct the symbol timing once
and for all.

I also implemented joint equaliser and PLL given by Godard's algorithm.
But it doesnt work well when my doppler is more than 0.2 m/s.

Thanks a lot. 

To Vladimir: Sir, your help is greatly appreciated :-)

Chintan

cpshah99 wrote:

> Hello All
> 
> Thanks a lot for responses.

You'd better start with getting to work the basic Costas and Gardner 
before jumping into the complicated things.

> My signal is QPSK modulated at carrier frequency of 12 KHz and with
> bandwidth of 8 KHz. Now as I am modeling doppler in my system, I also have
> doppler correction at receiver which uses Linear Interpolation.

Come on. What kind of doppler can be at 12kHz ?

> That
> algorithm works very well, in the sense that it estimates the delay from
> the equaliser and I use that delay to correct symbol timing and to correct
> the carrier phase.

Incredible.

> But this technique doesn't help me with turbo decoding. So I was thinking
> that when we use turbo decoding do we just correct the symbol timing once
> and for all.

There is no relation between sync and decoding. Unless you are using 
some sort of the non-causal decision directed algorithm. Of course, sync 
should be good enough not to impair the performance of the decoder, 
which can be challenging at low SNR.

> I also implemented joint equaliser and PLL given by Godard's algorithm.

Wow.

> But it doesnt work well when my doppler is more than 0.2 m/s.

Which means that the system simply doesn't work.

> Thanks a lot. 

Go to hell.

> To Vladimir: Sir, your help is greatly appreciated :-)

Your appreciation is cheap and worseless.

> Chintan


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

>
>
>cpshah99 wrote:
>
>> Hello All
>> 
>> Thanks a lot for responses.
>
>You'd better start with getting to work the basic Costas and Gardner 
>before jumping into the complicated things.
>
>> My signal is QPSK modulated at carrier frequency of 12 KHz and with
>> bandwidth of 8 KHz. Now as I am modeling doppler in my system, I also
have
>> doppler correction at receiver which uses Linear Interpolation.
>
>Come on. What kind of doppler can be at 12kHz ?
>
>> That
>> algorithm works very well, in the sense that it estimates the delay
from
>> the equaliser and I use that delay to correct symbol timing and to
correct
>> the carrier phase.
>
>Incredible.
>
>> But this technique doesn't help me with turbo decoding. So I was
thinking
>> that when we use turbo decoding do we just correct the symbol timing
once
>> and for all.
>
>There is no relation between sync and decoding. Unless you are using 
>some sort of the non-causal decision directed algorithm. Of course, sync

>should be good enough not to impair the performance of the decoder, 
>which can be challenging at low SNR.
>
>> I also implemented joint equaliser and PLL given by Godard's
algorithm.
>
>Wow.
>
>> But it doesnt work well when my doppler is more than 0.2 m/s.
>
>Which means that the system simply doesn't work.
>
>> Thanks a lot. 
>
>Go to hell.
>
>> To Vladimir: Sir, your help is greatly appreciated :-)
>
>Your appreciation is cheap and worseless.
>
>> Chintan
>
>
>Vladimir Vassilevsky
>DSP and Mixed Signal Design Consultant
>http://www.abvolt.com
>
%%%%

Hello

Thanks again. 

Ok I will start again from basics.

But just wanted to ask, how does it relate doppler and carrier frquency. I
can introduce any sort of doppler. From what I have seen from real time
experiment that even doppler of 0.3 m/s can cuase hell lot of trouble.

Now that joint PLL of Godard does not work above certain cases, that
doesnt mean it doesnt work at all.

Just for the fun, I was laughing abt 'Go to hell'.

Few days back somebody asked me: do u believe in hell or heaven?
I said : Yes.

Then he asked: where will u go?
I said : Hell.

Thanks again.

Chintan

On Aug 1, 2:30&#4294967295;pm, "cpshah99" <cpsha...@rediffmail.com> wrote:
> >cpshah99 wrote:
>
> >> Hello All
>
> >> Thanks a lot for responses.
>
> >You'd better start with getting to work the basic Costas and Gardner
> >before jumping into the complicated things.
>
> >> My signal is QPSK modulated at carrier frequency of 12 KHz and with
> >> bandwidth of 8 KHz. Now as I am modeling doppler in my system, I also
> have
> >> doppler correction at receiver which uses Linear Interpolation.
>
> >Come on. What kind of doppler can be at 12kHz ?
>
> >> That
> >> algorithm works very well, in the sense that it estimates the delay
> from
> >> the equaliser and I use that delay to correct symbol timing and to
> correct
> >> the carrier phase.
>
> >Incredible.
>
> >> But this technique doesn't help me with turbo decoding. So I was
> thinking
> >> that when we use turbo decoding do we just correct the symbol timing
> once
> >> and for all.
>
> >There is no relation between sync and decoding. Unless you are using
> >some sort of the non-causal decision directed algorithm. Of course, sync
> >should be good enough not to impair the performance of the decoder,
> >which can be challenging at low SNR.
>
> >> I also implemented joint equaliser and PLL given by Godard's
> algorithm.
>
> >Wow.
>
> >> But it doesnt work well when my doppler is more than 0.2 m/s.
>
> >Which means that the system simply doesn't work.
>
> >> Thanks a lot.
>
> >Go to hell.
>
> >> To Vladimir: Sir, your help is greatly appreciated :-)
>
> >Your appreciation is cheap and worseless.
>
> >> Chintan
>
> >Vladimir Vassilevsky
> >DSP and Mixed Signal Design Consultant
> >http://www.abvolt.com
>
> %%%%
>
> Hello
>
> Thanks again.
>
> Ok I will start again from basics.
>
> But just wanted to ask, how does it relate doppler and carrier frquency. I
> can introduce any sort of doppler. From what I have seen from real time
> experiment that even doppler of 0.3 m/s can cuase hell lot of trouble.
>
> Now that joint PLL of Godard does not work above certain cases, that
> doesnt mean it doesnt work at all.
>
[snip]

Since you haven't told us yet what the specs are (jitter, drift,
excess
bandwidth, SNR, modulation, fc, fs, etc.) it's hard to guess what will
or will
not work, except trivial cases :-P.

Just like most things in real life, there are many solutions, but the
regimes
where each solution is reasonable often do not overlap.

>On Aug 1, 2:30=A0pm, "cpshah99" <cpsha...@rediffmail.com> wrote:
>> >cpshah99 wrote:
>>
>> >> Hello All
>>
>> >> Thanks a lot for responses.
>>
>> >You'd better start with getting to work the basic Costas and Gardner
>> >before jumping into the complicated things.
>>
>> >> My signal is QPSK modulated at carrier frequency of 12 KHz and with
>> >> bandwidth of 8 KHz. Now as I am modeling doppler in my system, I
also
>> have
>> >> doppler correction at receiver which uses Linear Interpolation.
>>
>> >Come on. What kind of doppler can be at 12kHz ?
>>
>> >> That
>> >> algorithm works very well, in the sense that it estimates the delay
>> from
>> >> the equaliser and I use that delay to correct symbol timing and to
>> correct
>> >> the carrier phase.
>>
>> >Incredible.
>>
>> >> But this technique doesn't help me with turbo decoding. So I was
>> thinking
>> >> that when we use turbo decoding do we just correct the symbol
timing
>> once
>> >> and for all.
>>
>> >There is no relation between sync and decoding. Unless you are using
>> >some sort of the non-causal decision directed algorithm. Of course,
sync
>> >should be good enough not to impair the performance of the decoder,
>> >which can be challenging at low SNR.
>>
>> >> I also implemented joint equaliser and PLL given by Godard's
>> algorithm.
>>
>> >Wow.
>>
>> >> But it doesnt work well when my doppler is more than 0.2 m/s.
>>
>> >Which means that the system simply doesn't work.
>>
>> >> Thanks a lot.
>>
>> >Go to hell.
>>
>> >> To Vladimir: Sir, your help is greatly appreciated :-)
>>
>> >Your appreciation is cheap and worseless.
>>
>> >> Chintan
>>
>> >Vladimir Vassilevsky
>> >DSP and Mixed Signal Design Consultant
>> >http://www.abvolt.com
>>
>> %%%%
>>
>> Hello
>>
>> Thanks again.
>>
>> Ok I will start again from basics.
>>
>> But just wanted to ask, how does it relate doppler and carrier
frquency. =
>I
>> can introduce any sort of doppler. From what I have seen from real
time
>> experiment that even doppler of 0.3 m/s can cuase hell lot of trouble.
>>
>> Now that joint PLL of Godard does not work above certain cases, that
>> doesnt mean it doesnt work at all.
>>
>[snip]
>
>Since you haven't told us yet what the specs are (jitter, drift,
>excess
>bandwidth, SNR, modulation, fc, fs, etc.) it's hard to guess what will
>or will
>not work, except trivial cases :-P.
>
>Just like most things in real life, there are many solutions, but the
>regimes
>where each solution is reasonable often do not overlap.
>

%%%%%%

Hi

I am doing this simulation for underwater comms, mt fc=12 KHz, fs= 48 KHz,
QPSK Modulation, signal bandwidth 8 KHz.

I will still look into basics again like Gardener or Costas loop.

Thanks again.

Chintan