Re: [Discuss-gnuradio] Decoding constellation (0, 1-1) using gnuradio

[abhinav narain]

Hi Marcus, others,

Actually things are not working on my side. I would be thankful if you could give some suggestion.

The points on the constellation diagram come to be at max in an elipse of half major axis length = 0.02,

while my constellation object takes the decision on 1+0j, -1+0j, 0+0j - which worked in simulation well.

When I change the decision boundary from 1,-1 to 0.02,-0.02 for the actual channel, I still don't get any decoded bits.

Is it possible to please suggest something for me ? 

Right now, I am using the Polyphase synch block followed by constellation decoder as the only two blocks on receive side, instead of using the FLL-BandEdge block.

Thanks,

Abhinav

On Mon, Oct 12, 2015 at 2:21 PM, abhinav narain <address@hidden> wrote:

Hi Marcus,

I messed that up! I am not sure if formatting is correct in the copied message, but yes I have emailed it to the discuss list now!

Thanks,

Abhinav

On Mon, Oct 12, 2015 at 11:16 AM, Marcus Müller <address@hidden> wrote:

Hi Abhinav,
I currently don't have the time to properly respond -- generally, it's a very good idea to reply to the mailing list rather than individuals. Could you do this for this mail?
Thanks,
Marcus

On 10/12/2015 08:07 PM, abhinav narain wrote:

Hi Marcus,

 

the problem is that you're not doing BPSK, really. It's an Amplitude shift keying, if you want so.

That is partially true - in ASK absence of pulse is bit 0, but in my case bit 0 is inverted pulse and absence of pulse is a third symbol. 

I want to send across two different bits sparsely spaced in time using gnuradio hardware.

Two things:
* you might want to consider what AGC2 does while you're transmitting zeros -- it will increase amplification until noise scales up to signal power. What you're getting after that will more or less be useless, unless your AGC is really set to be slow (I don't think so, according to the attack rate of 6/100).

Okay, I used it as this block is used in generic_mod_demod.py with these parameters

  

* you're throwing a FLL at your signal -- which would be fine, if there was definitely a carrier. I don't see how that should work for a signal that's zero most of the time.

Yes, I guess I need to do upconversion but I am right now just doing things in baseband. Even if I turn off the FLL Band-Edge block, I did not see any improvement in the output.

* Assuming the polyphase clock synth did work with such a signal, the resulting signal would still contain a symbol that was '0', and hence had random phase (the phase of the additive noise at the sampling time). Hence, the costas loop can't do anything reasonable about that.

I see, I am vaguely familiar with costas loop, should I not use costas loop then?

If I give order ==3 ( constObj.arity()) which is the size of constellation 0,1,-1, it doesn't work at all, hence I gave the parameter value as 2, and I get the previous output.

After following your advice of increasing rolloff in RRC filter and sps, I get the following stable output, although it is scaled to -2,2.

Please see the following image - (http://postimg.org/image/6kenur0ix/) constellation_after_channel_and_after_costas_loop.png

If I understand you correctly, however, you *actually* want to do BPSK, and the 0 symbol is just a "filler" in between symbols, right?

Yes! Exactly!

 

That might actually work, if you just increase the rolloff duration/samples per symbol of your pulse shaping filter.

Okay, I have increased the symbols, but the output of the decoder is wrong! Please see the image - (http://postimg.org/image/ngar9353z/) constellation_decoder_ouput.png.

It should be equally spaced values of 1, 0 with a lot of 2s(representing (0+0j)s transmitted) in betweeen.

I doubt the way I am using constellation decoder is the right way. 

I want the decoder to decode as the image attached - ( http://postimg.org/image/k0awzlq0h/ ) decoder_wanted.JPG 

So, I think it might be best to explain what kind of system you're trying to build, so that we can understand that better!

I am trying to build a covert system where I transmit very rarely so that it is difficult for someone to find out I am on the channel.

Thanks for your response,

Abhinav

Best regards,
Marcus

On 10/12/2015 06:00 PM, abhinav narain wrote:

Hi,

 I am trying to transmit -1,0,1 { [1]+ 100*[0]+[-1] }, basically BPSK lot of 0s  filled in between at certain frequency. I have few questions. "Constellation object" has 3 symbols ([1+0j,-1+0j,0+0j]) mapped to symbol map-> [0,1,2]  respectively. I agree that constellation with size 3 is weird but that is what I want. I want the decision boundaries to be at -0.5, 0.5 for bits and anything between that to be given symbol=2 which i can discard later on. I am currently doing this in baseband, but I will eventually trasmit at a some frequency X.

 I have the following baseband setup grc (http://postimg.org/image/8k7g1o375/  , grc file is attached).

 I have few problems which I am unable to fix since yesterday -

 1) The constellation after channel is stable (as expected) but it goes weird

 after the polyphase clock sync and costas loop ( http://postimg.org/image/pyg8h2vlh/  , attached image- constellation_before_after_costas.png)  and hence I don't get the expected output of equally spaced 1, 0 in the decoder  output ( http://postimg.org/image/a58cc35cj/  , attached image- custom_decoder_0_1_2_output.png)

 2) If I set the order of costas loop as 3 (size of constellation) my flowgraph doesn't work at all. I have used the parameters after reading some example code

 in gnuradio folders and web and i guess they are correct. Please correct me if  I am wrong.

Thanks,

Abhinav

_______________________________________________
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

clock_4_receiver.grc
Description: application/gnuradio-grc