Receiving symbols using non-coherent M-FSK

Hi There,

How does the demodulator of a non-coherent M-FSK system correctly time
the "reading" of the symbols?  In the book on Digital Comms by
Proakis, it is mentioned that a bank of 2M correlators can be used.
In the book there is also a model for non-coherent M-FSK in an AWGN
environment.  I have used the model to create a matrix of metrics and
done the decoding/demodulation from there.  However, in practise, how
would one "syncrohonise" the transmitter and receiver?  If the
receiver samples the correlators at the wrong time, you will run into
a lot of trouble.  Should one adapt a scheme such as the RS232 scheme
where the line is always high, followed by a start bit, the data and
one or more stop bits?  Someone also suggested that with RS232 you
should oversample the signal, say sample every bit 10 times.

Another question, with BPSK and QPSK (and QAM), if one wants to deploy
non-coherent demodulation, how does the receiver know what the phase
of the transmitter's carrier frequency is, which must be known for
decoding?

Your time, effort and suggestions will be greatly appreciated
Jaco

jaco.versfeld@gmail.com wrote:

> 
> How does the demodulator of a non-coherent M-FSK system correctly time
> the "reading" of the symbols? 
> In the book on Digital Comms by
> Proakis, it is mentioned that a bank of 2M correlators can be used.

There is 1001 of the good methods for symbol synchronization and 65536 
books about that. Unfortunately, Proakis is not among those books. Look 
for Gardner, Mueller and Muller, Viterbi, Van Trees.

> In the book there is also a model for non-coherent M-FSK in an AWGN
> environment.  I have used the model to create a matrix of metrics and
> done the decoding/demodulation from there.  However, in practise, how
> would one "syncrohonise" the transmitter and receiver?

It all depends. What hardware do you have, how fast it should acquire 
the lock, what is the channel, what is the signal, how optimal do you 
want to be, etc. etc.

>  If the
> receiver samples the correlators at the wrong time, you will run into
> a lot of trouble.  Should one adapt a scheme such as the RS232 scheme
> where the line is always high, followed by a start bit, the data and
> one or more stop bits?  Someone also suggested that with RS232 you
> should oversample the signal, say sample every bit 10 times.

:-) You have to start learning from the very beginning.

> Another question, with BPSK and QPSK (and QAM), if one wants to deploy
> non-coherent demodulation, how does the receiver know what the phase
> of the transmitter's carrier frequency is, which must be known for
> decoding?

The whole point of the non-coherent demodulation is that the receiver 
doesn't have to know the phase.

> Your time, effort and suggestions will be greatly appreciated

I like this figure of speech. How much exactly is the great appreciation?


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

On Nov 22, 5:59 am, jaco.versf...@gmail.com wrote:
> Hi There,
>
> How does the demodulator of a non-coherent M-FSK system correctly time
> the "reading" of the symbols?  In the book on Digital Comms by
> Proakis, it is mentioned that a bank of 2M correlators can be used.
> In the book there is also a model for non-coherent M-FSK in an AWGN
> environment.  I have used the model to create a matrix of metrics and
> done the decoding/demodulation from there.  However, in practise, how
> would one "syncrohonise" the transmitter and receiver?  If the
> receiver samples the correlators at the wrong time, you will run into
> a lot of trouble.  Should one adapt a scheme such as the RS232 scheme
> where the line is always high, followed by a start bit, the data and
> one or more stop bits?  Someone also suggested that with RS232 you
> should oversample the signal, say sample every bit 10 times.
>
> Another question, with BPSK and QPSK (and QAM), if one wants to deploy
> non-coherent demodulation, how does the receiver know what the phase
> of the transmitter's carrier frequency is, which must be known for
> decoding?
>
> Your time, effort and suggestions will be greatly appreciated
> Jaco

As has been mentioned already, non-coherent detection operates without
carrier phase information. Obviously, you still need to recover symbol
timing. There are many ways to do it. As an example, I recently
developed an FPGA-based CPFSK demodulator that used a fractionally-
spaced equalizer to recover the timing. The equalizer sampled the
signal at twice the symbol rate (hence the term "fractionally-spaced")
and used decision feedback to compensate for channel distortion and
provide sampling phase correction, thus no PLL was required.

<jaco.versfeld@gmail.com> wrote in message 
news:e7d2c66d-c711-44db-9ced-cfda809cb4c1@d4g2000prg.googlegroups.com...
> Hi There,
>
> How does the demodulator of a non-coherent M-FSK system 
> correctly time
> the "reading" of the symbols?

There is a really good paper by Louis Litwin that will get 
you started.  You can find it here:

http://rfdesign.com/images/archive/0901Litwin32.pdf

or here:

http://images.industryclick.com/files/4/0901Litwin32.pdf