Re: Clock synchronization problem of modulated signal: How to set the parameters to make the synchronization effect better?

[Marcus Müller]

Hi
1. I currently have a modulation scheme (joint channel source coding). Its 
input is bit
data, and its output is two real numbers: a and b, where a is the Q signal and 
b is the I
signal, which is a+bi Symbols, and a and b are constantly changing, which means 
that the
constellation diagram will continue to change. Every time a bit codeword is 
input, the
output symbol is different.

2. This module was selected because it was used in "generic_mod.py" (I think
pfb_arb_resampler refers to this "polyphase arbitrary resampler"). The 
selection of
parameters is also set in "generic_mod.py". The filter parameters are as 
follows:
# pulse shaping filter
?0?2 ?0?2 ?0?2 ?0?2 nfilts = 32
?0?2 ?0?2 ?0?2 ?0?2 ntaps = nfilts * 11 * int(self._samples_per_symbol)?0?2 
?0?2 # make nfilts filters of
ntaps each
?0?2 ?0?2 ?0?2 ?0?2 self.rrc_taps = filter.firdes.root_raised_cosine(
?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 nfilts,?0?2 ?0?2 ?0?2 ?0?2 ?0?2 # gain
?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 nfilts,?0?2 ?0?2 ?0?2 ?0?2 ?0?2 # sampling rate 
based on 32 filters in resampler
?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 1.0,?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2# symbol 
rate
?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 self._excess_bw, # excess bandwidth (roll-off 
factor)
?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ntaps)
?0?2 ?0?2 ?0?2 ?0?2 self.rrc_filter = 
filter.pfb_arb_resampler_ccf(self._samples_per_symbol,
?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 
?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2 ?0?2self.rrc_taps)

3. My choice of constellation points is random and there is no rule. I just 
want to verify
whether the "polyphase clock sync" module has an effect on any constellation 
diagram.
Facts proved to be effective, but not very good.

4.My question seems very complicated. Just like the modulation method in which 
the
constellation diagram continues to change as I described above, can the 
transmitted *a and
b* be restored after channel interference? (Including clock offset, phase 
frequency
offset, etc.)

Best regards,
Thanks for your reply!




------------------ ???????? ------------------
*??????:* "Marcus M??ller" <mmueller@gnuradio.org>;
*????????:*?0?22021??6??3??(??????) ????6:12
*??????:*?0?2"discuss-gnuradio"<discuss-gnuradio@gnu.org>;
*????:*?0?2Re: Clock synchronization problem of modulated signal: How to set 
the parameters to
make the synchronization effect better?

Hi!
On 03.06.21 10:28, ???????? wrote:
> ?0?2 I am trying to modulate an unconventional signal. Its constellation 
> points are not as
> beautiful as QPSK or QAM.

So, what are they? This might be very relevant to synchronization aspects. 
Especially
important: is the average of all your symbols after source coding and channel 
coding still 0?


> then use "polyphase arbitrary resampler" for pulse shaping,

Unusual choice! You'd almost always want to use an interpolating FIR filter for 
this.

What is the pulse shape you're using? This defines which synchronization 
algorithms can work.


> because some constellation points are too close, the distinction is not very 
> obvious. I
> want to know how to set the parameters to achieve a better synchronization 
> effect?

This really depends on your constellation! You'll have to tell us more about it.

It's also an indication for why your constellation might not be a good idea, if 
this is
what you see at the receiver:
You're literally telling us "there's a high probability that the receiver 
mistakes one
symbol for another", which means that you're not making good use of the 
channel, i.e. the
mutual information between output and input of your channel is not as high as 
possible.

So, I assume you use that constellation for some other reason, some constraint 
that we
don't know: Often, such constraints do have influence on synchronization!
I think you'll want to tell us *why* you chose that constellation.


> In addition, I also need to add frequency offset and multipath effects later. 
> According
> to the CMA equalizer in the "QPSK Modulation and Demodulation Tutorial",it 
> only works on
> signals that have a constant amplitude, or modulus. What other existing ones 
> can be easily
> removed Any method or module for frequency phase shift and multipath effect 
> of any signal?

Well, equalization is very system-dependent; when simple statistical properties 
like
(e.g., constant envelope for CMA) don't work, then you need something specific 
to sense
your channel. We don't know enough about your system, and its motivation, to 
advise here.
For example, I wonder: Maybe pilot symbols are a good idea? But what about the 
constraints
that led to your special constellation, are they in conflict with that? Maybe 
FBMC or OFDM
solve the multipath problem more elegantly? ...


Best regards,
Marcus