Re: [Discuss-gnuradio] Questions about synchronization and modulation in benchmark

[Timothy Lawrence Sitorus]

Hi Tom,

Thanks for answering all my questions.  Recently I am reading the source of gr_mpsk_receiver module and starting to understand how the algorithm consumes sps > 1 and produces one output.  

Thanks for the information about dqpsk2.py. I didn't now about it. I will test it soon.
Btw  ,can you tell me why the phase ambiguity issues occur? is it because gr_mpsk_receiver do not divide clock recovery  and frequency offset separetely? 
I checked dqpsk2.py source and saw that you are not using gr_mpsk_receiver anymore. 

Thanks,

Tim

On Sat, Feb 6, 2010 at 2:46 AM, Tom Rondeau <address@hidden> wrote:

On Fri, Jan 29, 2010 at 7:06 AM, Timothy Lawrence Sitorus

<address@hidden> wrote:

> Hello All,
>
> I have problems in demodulating code from benchmark qpsk. please help me
> with answering my question.
>
> 1. Why there are no bpsk, qpsk or 8psk in the modulation option ?there are
> only dbpsk , dqpsk, d8psk.

We don't use non-differential modulation because we didn't want to
work into the code the phase ambiguity issues involved. It's a simple
matter of removing the differential encoder in the modulator and
decoder in the demodulator, but then you'll have to add your own
method of handling the phase rotation.

> 2. Is there any different synchronization besides gr_mpsk_receiver_cc?

You can use the clock_recovery_mm_cc and costas_loop_cc blocks
separately. In the latest master in the Git repo, I have just added
dbpsk2 and dqpsk2 that use a whole different system of synchronization
that has proven much more robust. It still needs a little work, which
is why we haven't replaced the old dbpsk and dqpsk modules, although
this should eventual obsolete them.

In the new way of synchronization in these blocks, I used a band-edge
filter frequency lock loop for frequency acquisition, a polyphase
filterbank clock recovery algorithm, and the Costas loop for phase
correction.

> 3.I don't know about costas loop in detail, but in my knowledge, this module
> using modified Muller for the algorithms and I see that this method is just
> using 1 sample per symbol in recovering symbol timing with mu as a base for
> correction? am i right? if yes, may i know why they use just 1 sample per
> symbol?

No, the block takes in sps > 1 (2 or 4 samples generally) and produces
one output.

> 3. Is this costas loop just for differential modulation ? I have a problem
> in decoding the qpsk if frequency offset occurs.
>    * In the transmitter side , I just took of the diff dec . and in the
> receiver side, I made diff_phasor as a delay detector and recovered it into
> qpsk based on preamble.
>    (it works while i am using bencharmark loopback without frequency offset
> and time error)

No, the costas loop is not specific to differential modulation and
will work with non-differential. It doesn't have a wide bandwidth,
though, so this is probably your problem. It's also why I switched the
frequency tracking method in the dXpsk2 modules.


Tom

> 4. if no 3 is yes, please help me in solving this problem. I have no idea.
>
> Best Regards,
>
>
> Tim