newbie question about synchronization

Hello,

Currently I am trying to implement a transmitter-receiver unit on a DSP
board. Transmitter and Receiver will be implemented on seperate boards.

I have only one carrier with a fixed frequency and send my symbols for one
period. I guess that since there is only one carrier, I don`t need any
frequency synchronization. 

However I am not sure if I should apply symbol synchronization or would I
be fine without it?

any pointers would be appreciated.

thanks alot
Rifo

On Jul 19, 8:46 am, "rifo" <rtur...@yahoo.de> wrote:
> Hello,
>
> Currently I am trying to implement a transmitter-receiver unit on a DSP
> board. Transmitter and Receiver will be implemented on seperate boards.
>
> I have only one carrier with a fixed frequency and send my symbols for one
> period. I guess that since there is only one carrier, I don`t need any
> frequency synchronization.
>
> However I am not sure if I should apply symbol synchronization or would I
> be fine without it?
>
> any pointers would be appreciated.
>
> thanks alot
> Rifo

This question is not really meaningful without a more complete
specification.

If there's any clock mismatch between the Tx and Rx, it will
effectively induce frequency offset.  So what's the clock
mismatch?  How do you plan to synchronize the operation
of the Tx and Rx?

Julius

Assuming you might be having independent crystals on Tx board and Rx
board generating the
RF carrier and the sampling clocks, you may need to do:
(a) carrier frequency offset error estimation and compensation
(b) sampling clock offset estimation and compensation
(c) Symbol boundary estimation

>> I guess that since there is only one carrier, I don`t need any frequency synchronization.
Did not understand.

HTH,
Krishna
http://dsplog.blogspot.com

Hello thank you for your inputs,
I try to give more info about the system


>If there's any clock mismatch between the Tx and Rx, it will
>effectively induce frequency offset. So what's the clock
>mismatch? How do you plan to synchronize the operation
>of the Tx and Rx?

   I am using two motorola 56303 development kits hence I will have       
        independent crystals on board for Tx and Rx. I don't know about the
clock mismatch, I will check the datasheets and report back about this but
frankly I don't also have any plans for synchronization of Tx and Rx
operations.


>Assuming you might be having independent crystals on Tx board and Rx
>board generating the
>RF carrier and the sampling clocks, you may need to do:
>(a) carrier frequency offset error estimation and compensation
>(b) sampling clock offset estimation and compensation
>(c) Symbol boundary estimation
  
  Do I need to implement all three of these. I would also appreciate if
you 
point out any resources about the implementation of these algorithms. 

 
>>> I guess that since there is only one carrier, I don`t need any
frequency synchronization.
>Did not understand.
>

    I was going through the book of Sklar, "Digital Communications" and in
the synchronization chapter it said that;

"Take an AM radio, if there is only a central carrier and many receivers,
then in this system synchronization can be avoided"

    This led me to think that, since I only generate a single carrier with
a fixed frequency I may at least skip carrier synchronization (since I
would have coded the receiver DSP with the knowledge of this carrier
frequency) But now I see that I have a misinterpretation of the above
phrase.

I would be happy to hear your further advices
Thanks a lot
Rifo

On Jul 20, 6:10 am, "rifo" <rtur...@yahoo.de> wrote:

>    I am using two motorola 56303 development kits hence I will have
>         independent crystals on board for Tx and Rx. I don't know about the
> clock mismatch, I will check the datasheets and report back about this but
> frankly I don't also have any plans for synchronization of Tx and Rx
> operations.

Please do.

> >Assuming you might be having independent crystals on Tx board and Rx
> >board generating the
> >RF carrier and the sampling clocks, you may need to do:
> >(a) carrier frequency offset error estimation and compensation
> >(b) sampling clock offset estimation and compensation
> >(c) Symbol boundary estimation
>
>   Do I need to implement all three of these. I would also appreciate if
> you
> point out any resources about the implementation of these algorithms.

The dry treatment is available in most textbooks, like Proakis or
Sklar or Messerschmitt/Lee.

But if you want to truly optimize the parameters, I am a big fan of
d'Andrea and Mengali's text, and also the Meyr/Moeneclaey/Fechtel
text.  If you do a google search for "digital receiver
synchronization"
you can find several tutorial articles, but they do not go in much
depth.

If your goal is to just send from point A to point B and your SNR is
very high, then there is much fewer things to optimize, so the
tutorials
may be enough.

Regardless, for sure you need some sort of method for synchronizing.

>     I was going through the book of Sklar, "Digital Communications" and in
> the synchronization chapter it said that;
>
> "Take an AM radio, if there is only a central carrier and many receivers,
> then in this system synchronization can be avoided"
>
>     This led me to think that, since I only generate a single carrier with
> a fixed frequency I may at least skip carrier synchronization (since I
> would have coded the receiver DSP with the knowledge of this carrier
> frequency) But now I see that I have a misinterpretation of the above
> phrase.

No way :-).  Are you actually implementing a digital communication
system or an analog one?  If you are doing just AM transmission,
then Sklar is correct.  I had always assumed that you were doing
digital stuff.

> I would be happy to hear your further advices
> Thanks a lot
> Rifo

If your mixers are implemented in digital, you can simplify some parts
from the above list.

Julius

 
  I hope I have a correct interpretation of clock mismatch concept :) so
here it goes...

  Both development boards have the same codec (CS4125) which is driven by
a 24.576 Mhz crystal. As far as I remember crystals should have a good
precision...
  Okey I admit I don't know how to find the clock mismatch value :/


  By the way Sklar included a big "ANALOG" in the phrase I quoted above
and somehow I didn't see it :)
  My system is digital though and I create the high freq carrier signal in
the dsp. I will now check the tutorials you have pointed out. 

 Thanks a lot for your help.
 Rifo

PS: Julius, can you also tell me which parts I can simplify since my
mixers are implemented digitally

rifo wrote:
>  
>   I hope I have a correct interpretation of clock mismatch concept :) so
> here it goes...
> 
>   Both development boards have the same codec (CS4125) which is driven by
> a 24.576 Mhz crystal. As far as I remember crystals should have a good
> precision...
>   Okey I admit I don't know how to find the clock mismatch value :/

You can usually count on cheap run-of-the-mill crystals to be within 
1E-7 of nominal. That's worst-case mismatch of 2E-7.

   ...

Jerry
-- 
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

On Jul 20, 1:33 pm, "rifo" <rtur...@yahoo.de> wrote:
>   I hope I have a correct interpretation of clock mismatch concept :) so
> here it goes...
>
>   Both development boards have the same codec (CS4125) which is driven by
> a 24.576 Mhz crystal. As far as I remember crystals should have a good
> precision...
>   Okey I admit I don't know how to find the clock mismatch value :/

Two crystals, even manufactured in the same place at the same time,
will not be perfectly accurate.  This is specified by the PPM (parts
per
million) accuracy. Please read up on this.

>   By the way Sklar included a big "ANALOG" in the phrase I quoted above
> and somehow I didn't see it :)
>   My system is digital though and I create the high freq carrier signal in
> the dsp. I will now check the tutorials you have pointed out.

Please stop comparing apples with durians [1].  This is not good
practice :-).

>  Thanks a lot for your help.
>  Rifo
>
> PS: Julius, can you also tell me which parts I can simplify since my
> mixers are implemented digitally

It is possible to rely only on timing adjustment, without phase
correction,
assuming that you oversample the signal in the first place.

Sorry, but this is as far as I can help you for now.

Julius

[1]  http://en.wikipedia.org/wiki/Durian

Jerry Avins wrote:

> You can usually count on cheap run-of-the-mill crystals to be within 
> 1E-7 of nominal.

"Misunderestimate" (c) W Bush.

The microprocessor grade crystal from a pile is usually accurate to 
about 1e-5 of nominal.


Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

Vladimir Vassilevsky wrote:
> 
> 
> Jerry Avins wrote:
> 
>> You can usually count on cheap run-of-the-mill crystals to be within 
>> 1E-7 of nominal.
> 
> "Misunderestimate" (c) W Bush.
> 
> The microprocessor grade crystal from a pile is usually accurate to 
> about 1e-5 of nominal.

The world is going to pot. Back in my day, Sonny, ....

Jerry
-- 
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯