Re: Constellation Modulator Block Output is incorrect

[Ahmad Oweis]

Hi Marcus,

Yes the sps is set at 4.

I totally agree with you about the debugability of the block. It does not allow you plot the constellations to begin with.

Thanks for your reply.

Ahmad 

On Mon, 20 Feb 2023, 17:15 Marcus Müller, <mueller@kit.edu> wrote:

Hi Ahmad,

interesting; haven't tried it here, but I do get correctly that you set
a samples per symbol of > 1, right?

Generally, it's not my favourite block in the world: while combining the
constellation mapper and the pulse shaping in the same block does, in
fact, make sense, the chosen abstraction is kind of detrimental to
debugability. You really just a "Chunks to Symbols" (or use a "Multiply
const, const=2" follow by "Add const, const=-1") followed by an "Root
Raised Cosine Filter" with an interpolation == desired samples per
symbol to achieve the same.

That has no bearing on whether we should fix bugs with the Constellation
Modulator, though; we probably should, and we might maybe want to
consider deprecating it, if it's not that useful, but that would warrant
a wider discussion.

Best regards,
Marcus

On 2/20/23 10:08, Ahmad Oweis wrote:
> Hi everyone,
>
> My understanding is that the constellation modulator block should
> output an RRC pulse in the real and imaginary domains according to the
> constellation points. The pulse amplitude can either be positive or
> negative.
>
> However, if you build a flow chart on GRC and input a vector of zeros
> to the block, and use non-differential BPSK (so that symbol 0 is
> mapped to constellation point -1+j0, and symbol 1 to point 1+j0), the
> output will be a constant DC value of -1 without any pulse. If the
> input is 1, the output will have a positive amplitude pulse.
>
> This was reported as an issue on GitHub
> (https://github.com/gnuradio/gnuradio/issues/5449) but the answer was
> vague "It's the total I/Q signal that matters, since they are not
> transmitted or processed separately". This did not make a lot of sense
> to me, I still believe symbol zero should produce a pulse not a DC -1
> signal.
>
> Also, if you review issue
> (https://github.com/gnuradio/gnuradio/issues/4162), you will see that
> the one who posted it is probably having the same issue (go down to
> the last comment in the thread and compare the two graphs.)
>
> What do you think? Is the constellation modulator behaving as expected?
>
> Thanks!