Frequency offset vs. Symbol Rate Problem

Hi Everybody,

I have a problem that I thought would be pretty common, but I haven't
found a good solution yet.  The problem is basically this:  I have to
accommodate a frequency offset of 1450Hz on a signal coming in at
1200sym/sec.  If I sample 6 times a symbol, I get the required
bandwidth to satisfy Nyquist's, but the correlator length is too much
for the dsp to handle.  Does anybody have any ideas or any suggestions
as to where to look?  Thanks in advance!

-Frank

"Frank Cassidy" <nemesis@lost-wanderer.com> wrote in message 
news:abf316e4.0411041343.5be7f78a@posting.google.com...
> Hi Everybody,
>
> I have a problem that I thought would be pretty common, but I haven't
> found a good solution yet.  The problem is basically this:  I have to
> accommodate a frequency offset of 1450Hz on a signal coming in at
> 1200sym/sec.  If I sample 6 times a symbol, I get the required
> bandwidth to satisfy Nyquist's, but the correlator length is too much
> for the dsp to handle.  Does anybody have any ideas or any suggestions
> as to where to look?  Thanks in advance!

Frank,

It's not clear that you *are* satisfying the Nyquist criterion.
You haven't revealed the modulation method for those symbols.
The bandwidth isn't necessarily connected to the symbol rate.
It's the modulation method that would be used to convert the symbol rate to 
the bandwidth.
For example, if you are using binary signalling in an 8-bit per symbol 
alphabet then the bandwidth would be in excess of 8 times the symbol rate.

Also, you say there's a "frequency offset" but don't describe what this 
really means.

So, if one were to guess it might sound like you have a situation where the 
symbol rate matches the bandwidth - as might be true for PAM so the sample 
rate is 6x1200=7200Hz?.  However, when there's a frequency offset then this 
might have to be included in the necessary bandwidth.  1200+1450=2650.  The 
ratio of 7200/2650 is 2.72 which *may* be adequate sampling in a practical 
situation.

Much depends on the details here.....

Fred 

nemesis@lost-wanderer.com (Frank Cassidy) wrote in message news:<abf316e4.0411041343.5be7f78a@posting.google.com>...
> Hi Everybody,
> 
> I have a problem that I thought would be pretty common, but I haven't
> found a good solution yet.  The problem is basically this:  I have to
> accommodate a frequency offset of 1450Hz on a signal coming in at
> 1200sym/sec.  If I sample 6 times a symbol, I get the required
> bandwidth to satisfy Nyquist's, but the correlator length is too much
> for the dsp to handle.  Does anybody have any ideas or any suggestions
> as to where to look?  Thanks in advance!
> 
> -Frank

One approach, if you have enough compute power, is to shift the signal
by a set of offsets between -1450 and +1450 Hz in steps small enough
to satisfy your detector and look for a detect at each offset.

Another approach is to generate an AFC signal with a discriminator
that is proportional to the offset, use the AFC to bring the signal
to near baseband, then run your detector. 

What modulation are you using?

John

On 4 Nov 2004 13:43:04 -0800, nemesis@lost-wanderer.com (Frank
Cassidy) wrote:

>Hi Everybody,
>
>I have a problem that I thought would be pretty common, but I haven't
>found a good solution yet.  The problem is basically this:  I have to
>accommodate a frequency offset of 1450Hz on a signal coming in at
>1200sym/sec.  If I sample 6 times a symbol, I get the required
>bandwidth to satisfy Nyquist's, but the correlator length is too much
>for the dsp to handle.  Does anybody have any ideas or any suggestions
>as to where to look?  Thanks in advance!
>
>-Frank

There are lots of tradeoffs to consider for this sort of thing.

The first question I'd ask is how quickly do you have to acquire the
signal?   If you can spend some time to find it you can make it pretty
easy.   If you have to acquire it quickly, then there are various
techniques that can be tried, and many of these trade off speed of
acquisition for probability of acquisition or something like that.

As others have mentioned, there are some other fundamentals that are
helpful, like what type of modulation, receiver architecture, etc.


Eric Jacobsen
Minister of Algorithms, Intel Corp.
My opinions may not be Intel's opinions.
http://www.ericjacobsen.org