Re: [Discuss-gnuradio] Convolutional Code output does not match polynomial

[bob wole]

Hey Jan,

Thanks for your reply, you are correct about the bit shift being LSB. So, I reversed the bit order of polynomials 1101101 to 1011011 and 1001111 to 1111001 in MATLAB, Now the output of gnuradio and MATLAB and my dry run matches. I am wondering why they did implementation like this? Any advantages?

Yeah, I too was hoping for a reply from Tom and Nick on this.


--

Bob

On Tue, Sep 8, 2015 at 12:37 PM, Jan Krämer <address@hidden> wrote:

Also I think Tom and Nick Foster for sure know more about it than I do.

Cheers,

Jan

2015-09-08 8:44 GMT+02:00 Jan Krämer <address@hidden>:

Hey Bob,

I have the same issues with my SSE Viterbi, works with gr-trellis but not with cc_encoder in gr-fec. The output of the cc_encoder is legit though.

I talked to a couple of guys about this at last year's GRCON. It seems that in most implementations, the bit order of the polynomials is reversed when compared to the textbook approach.

Therefor also the bit shifting in the encoding process (and the viterbi chainback) is also reversed. I think textbook is MSB while cc_encoder is LSB but I always mix them up xD.

But you can check https://github.com/gnuradio/gnuradio/blob/master/gr-fec/lib/cc_encoder_impl.cc beginning from line 166.

Hope that helps,

Jan

2015-09-08 8:12 GMT+02:00 bob wole <address@hidden>:

On Tue, Sep 1, 2015 at 2:55 PM, bob wole <address@hidden> wrote:

I am trying channel coding in gnuradio. I am using convolutional encoder with K=7 and R=1/2 with polynomials [109, 79] (default). However I am not getting the expected result. I input a known bit sequence, single frame using head block, and observe the output of the encoder using file sink. The output I think is not correct according to the given polynomials. Input bit stream is

http://pastebin.com/thsYtLFG

I am getting this output

http://pastebin.com/P42QNjiE

However, I am expecting (in hex)

00    00    E3    32    63    D7    4B    FA    D3 ......

 I have attached the flowgraph I am using for this purpose. I did a dry run by drawing a diagram of the encoder and inputing  initial two bytes. Diagram of the encoder is also attached. I also used MATLAB script to test and found that my dry run output matches with MATLAB but not with Gnuradio output. I also played with the endianness but could not get the desired result. Maybe I am doing something wrong in configuring the encoder?

Also another thing I noticed is that in QA codes of gr-fec I found

import fec_swig as fec
import blocks_swig as blocks

I think this should be

from gnuradio import fec, blocks

Am I right? I am using gnuradio 3.7.8.

--

Bob

I am still waiting to get a reply on this. Anyone else verified the output of the convolutional encoder?


--

Bob

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