Re: [Discuss-gnuradio] Constellations, EVM and SNR Questions

[Federico 'Larroca' La Rocca]

Hello again,

To export from GNU Radio to octave, you should first use a "File Sink" block and then one of the scripts in gr-utils/octave. I'm pretty sure they are compatible with matlab too.

About the SNR, that depends on what you want. If you want the power of noise that actually affects your decision system divided by the power of the symbols, then MER is what you are looking for (assuming your sampling and frequency correction works properly). If you want the more classic SNR (i.e. power of the useful signal divided by the total noise power in the band), then that's more involved. You may try the following: since you know the constellation points and the pulse, you may easily calculate the power of the useful signal (assuming your AGC works properly and that the channel is flat fading). The difference between this power and the actual received power is due to noise (assuming independence). 

However, what matters performance-wise is the amount of noise that gets into the decision system, and not the actual SNR. 

Hope it helps!

2018-06-07 10:11 GMT-03:00 Luis Felipe Albarracin Sanchez <address@hidden>:

Hello all,

Thanks for all the responses,

I will try the measurements you just told me Federico, and i will let you know. And yes  Ron  i used a  DPSK Modulator, now i know that it has a root raised cosine filter, and i will try those other blocks you told me.

Still, do you have any idea of how i can  export data file form GNURadio (the output  of a modulator block) and plot the costellation in matlab properly? and also repositories for SNR for different modulations?

Thanks again.

On Thu, Jun 7, 2018 at 2:32 PM, Federico 'Larroca' La Rocca <address@hidden> wrote:

Hi Luis,

For measuring EVM, you may use our gr-mer OOT module: https://github.com/git-artes/gr-mer. Among several measurements, it outputs the MER, which is very similar to the EVM (at least that's my understanding).

Now, to output actual constellation points, you should do as Ron suggests: use the block "chunks to symbols". For instance, the attached flowgraph image generates QPSK constellation points randomly and shows them on a constellation sink.

best

Federico

2018-06-07 7:58 GMT-03:00 Ron Economos <address@hidden>:

Are you using the "Constellation Modulator" block? It's really a hierarchical block with a root raised cosine filter. So you're not going to see the original constellation points.

If you just want to see constellation points with the QT GUI Constellation Sink block, you can try the DVB-S2X Modulator in the Digital Television component. It has a bunch of built-in constellations. Or you can use the "Chunks to Symbols" block and define your own constellation.

That's just a starting point. Actual EVM measurements is a complex topic.

Ron

On 06/07/2018 03:24 AM, Luis Felipe Albarracin Sanchez wrote:

Hello Everyone,

I am trying to run a model with different modulations, and to see different measurement values ( Constellation Diagram, EVM, and SNR). i have done the following but it does not seem to work:

1- I try plotting a DQPSK modulation with "QT GUI Constallation Sink" (directly form the modulating block and it shows a graph  like this:

As you can see i have dots that are not where they are supposed to be (in every corner), to try to solve this i placed a "sink file" block from the out of the "modulating block", and i tried to plotted in matlab to see if i had a better image. Unfortunately it went worse. In matlab i read the bin file it showed a row with  many numbers (none of them bigger than 255) so i assumed it came formatted every 8 bits .

Fist thing i separated what i have read is the real and the complex part, and i tried to plotted, and  it did not work....show many points none of them in the corners where they are supposed to be. After that i thought maybe the Most Significant Bit is the "sign", so i adjusted  the program in matlab and ran the plot, but it also did not work.

Below the plot i get:

I wanted this so i can design a program on Matlab to calculate EVM for future measurements, nevertheless i wanted to measure EVM on GNURadio, and also i am looking for a way to measure SNR with different modulations.

Given this, i would appreciate your help with the following:

1- How can i have a graph of the constellation for different modulations in GNURadio, that is accurate.

2- How can i export data file form GNURadio (the output  of a modulator block) and plot the costellation in matlab properly?

3- Is there any  git repository where i can find blocks for EVM and SNR for different modulations, if not is there a way or known resources, or  flowgraphs to do SNR measurements and EVM measurements?

Once again thanks for all  the help.

Kind regards.

_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio

_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio

_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio

--

Eng. Luis Felipe Albarracin
PMP
CCNA/CCDA/CCNP/CCDP/CCIP
ITIL v3 Foundation
Msc. Telematics / MBA
"Die Grenzen meiner Sprache bedeuten die Grenzen meiner Welt"