# Qam constellation problem

Hi All,

I am plotting the constellation plot for qam8 before and after the root
raised cosine filter. Why I didn’t get clear 8 dots constellation before
the
root raised cosine filter (plotted from the data saved by
chunks2symbols.dat)? For both constellation before (data saved by
chunks2symbols.dat) and after (data saved by rrc_filter.dat) the filter,
I
got 8 fuzzy dots. Can anyone tell me why is that? However, for dbpsk, I
got
clear two points before the filter, but fuzzy two points after the
filter.

Brook

View this message in context:
http://old.nabble.com/qam-constellation-problem-tp31632653p31632653.html
Sent from the GnuRadio mailing list archive at Nabble.com.

BTW, I also got more fuzzy 16 points save by chunks2symbols.dat before
the
filter. Could anyone help me explain it?
Thanks,
Brook

Brook L. wrote:

filter.

Brook

View this message in context:
http://old.nabble.com/qam-constellation-problem-tp31632653p31632766.html
Sent from the GnuRadio mailing list archive at Nabble.com.

On Mon, May 16, 2011 at 9:23 PM, Brook L. [email protected] wrote:

clear two points before the filter, but fuzzy two points after the filter.

Brook

I’m assuming you are just talking about in the transmitter? I so, then
what
you are seeing is ISI. A root raised cosine filter is NOT a Nyquist
filter,
but a raised cosine filter is (or approximately so, at least). We
basically
split the filtering responsibilities between the transmitter and
When you pass the signal through two root raised cosines, you end up
with a
raised cosine. So the filter in the receiver removes the ISI introduced
by
the first filtering operation introduced in the transmitter.

So the transmitter filters to reduce out of band transmission while the
receiver uses a matched filter to remove the ISI and detect the symbol.

If you pass the “fuzzy” signal through another RRC filter, it should
clear
things up.

Tom

On Tue, May 17, 2011 at 5:36 PM, Brook L. [email protected] wrote:

the data saved by chunks2symbols.dat, which is before the rrc filter,
Thanks so much,
Brook

http://old.nabble.com/file/p31639677/dqpsk_before_filter.png

Oh, I see. It’s because you are using the constellation plot widget.
This
tries to do some phase/freq lock on the constellation, but I’m not sure
if
it can properly lock to a QAM signal (almost certainly not; especially a
non-square one).

Try using a time scope and set it to use dots instead of lines.

Tom

Hi Tom,

Thanks for your reply. Yes, you are right. Currently, I am seeing the
transmitter side.

Now, I understand that the constellation after the root raised cosine
filter
is a kind of “fuzzy” due to the ISI. But if I observe the constellation
from
the data saved by chunks2symbols.dat, which is before the rrc filter,
shouldn’t I see clear 8 dots without ISI? Because when I observe the
constellation for dqpsk before the rrc filter, there are four clear
points,
which is attached along with this message. Do I understand right that
there
is no ISI at the stage self.chunks2symbols =
gr.chunks_to_symbols_bc(rotated_const) but there is ISI at the stage
self.rrc_filter = gr.interp_fir_filter_ccf(self._samples_per_symbol,
clear points for dqpsk but “fuzzy” 8 points for qam8 before rrc filter
and
after self.chunks2symbols?

Thanks so much,
Brook

http://old.nabble.com/file/p31639677/dqpsk_before_filter.png

Tom R. wrote:

chunks2symbols.dat)? For both constellation before (data saved by
http://old.nabble.com/file/p31632653/qam8_before_filter.png
split the filtering responsibilities between the transmitter and receiver.

View this message in context:
http://old.nabble.com/qam-constellation-problem-tp31632653p31639677.html
Sent from the GnuRadio mailing list archive at Nabble.com.