Thanks, I now seem to have a decent soft decission demod and it works
pretty well over the air.
Error correction happens and works really well for the expected 'obvious
case': A pretty decent channel with the occasional bits being dropped.

On a terrible channel the end result is still junk as expected, and on a
good channel the parity bits are just there for the ride ;)


Regards,
Jeroen

On 06/16/2020 01:52 AM, Bill Cowley wrote:
>
> Hi Jeroen,
>
> Good work on generating a code for your application!  In theory the
> LLRs can have any value but since it's a log ratio, they are normally
> small.  You have probably seen David's use of the standard (BPSK)
> formula for getting LLRs from SDs which scales in proportion to the
> SNR. Zero indicates the bit is equally likely to be 1 or 0 and larger
> values indicated increasing certainly.  LLRs are often clipped to some
> magnitude (eg not allowed to be > 10 for "1" or <-10 for "0") as large
> values almost always cause numerical problems.  
>
> Cheers, Bill
>
>
> On 16/6/20 8:48 am, Jeroen Vreeken wrote:
>> Thanks Bill, I'll definitly dive into that.
>>
>> After figuring out what the current octave code is and isn't doing I
>> generated a 2/3 rate code (696 code bits, 434 data, 232 parity) and
>> got the mode 6000 unittests working with it.
>> I actually used a LDPC code generator I wrote a few years back when I
>> was looking into them the first time. (But never got around to
>> actually finishing it). It probably is not the best code, but at
>> least I can start testing on air again and if needed replace the code
>> later.
>>
>> As for making the modem softdecission what exactly is the range
>> run_ldpc_decoder() expects.
>> I see that the 'fixed' bits use -10.0. Is the value of the bits
>> important or is it completly relative?
>>
>> Regards,
>> Jeroen
>>
>> On 06/13/2020 07:31 AM, Bill Cowley wrote:
>>> Hi Jeroen, 
>>>
>>> Further to David's reply, you might like to look at the AFF3CT
>>> sitehttps://aff3ct.github.io/  which provides comprehensive info on
>>> channel coding.  David's codes have been Repeat Accumulate (RA) LDPC
>>> codes - we used them initially as encoding is very easy (eg well
>>> suited to a microController).   Of course you probably don't need
>>> this - so you might get slightly better performance with other
>>> options.   I haven't played with the AFF3CT software much, but it
>>> does cover RA.  
>>>
>>> Cheers, Bill 
>>>  
>>>
>>>     Cc: 
>>>     Bcc: 
>>>     Date: Thu, 11 Jun 2020 23:30:58 +0200
>>>     Subject: [Freetel-codec2] Generating LDPC code for 6000 mode
>>>     Hi,
>>>
>>>     The experiments with mode 6000 are going pretty well.
>>>     Another local HAM (PE1ITR) even managed to work our repeater
>>>     with just a
>>>     HackRF.
>>>     Compared to analog FM it seems to perform pretty well.
>>>
>>>     But I would like to take it a bit further and try to introduce
>>>     some FEC.
>>>     Since the currently used codec2 mode only needs 3200bps there is
>>>     still
>>>     plenty of room.
>>>     I was thinking about a rate 2/3 code (687 code bits for 458 data
>>>     bits),
>>>     but I am a bit puzzled with the ldpc octave code.
>>>     How exactly do I generate the needed matrix (and ofcourse the C and
>>>     header files)?
>>>
>>>     Regards,
>>>     Jeroen
>>>
>>>
>>>
>>>
>>>
>>>
>>>     ---------- Forwarded message ----------
>>>     From: David Rowe <[email protected] <mailto:[email protected]>>
>>>     To: [email protected]
>>>     <mailto:[email protected]>
>>>     Cc: 
>>>     Bcc: 
>>>     Date: Fri, 12 Jun 2020 07:29:34 +0930
>>>     Subject: Re: [Freetel-codec2] Generating LDPC code for 6000 mode
>>>     Hi Jeroen,
>>>
>>>     Hey that's great work with working the new mode on a HackRF, is that
>>>     about 50mW max?
>>>
>>>     Yes, FEC is a fine idea.
>>>
>>>     A friend has been kind enough to generate the codes for me,
>>>     however it's
>>>     not a process that is automated or straight forward.  Couple of
>>>     options:
>>>
>>>     1/ Use an existing code.
>>>     2/ Use a code with more data bits than you require, and fill
>>>     them with
>>>     known bits at the encoder and decoder that you don't transmit.  This
>>>     effectively lowers the code rate a little.  I do that for FreeDV
>>>     2020.
>>>     3/ Peruse the CML library codes and see if they have one that
>>>     fits your
>>>     use case, e.g. ldpcut.m
>>>
>>>     There is an Octave utility to generate .[ch] files and run the
>>>     codes in C.
>>>
>>>     Cheers,
>>>     David
>>>
>>>     On 12/6/20 7:00 am, Jeroen Vreeken wrote:
>>>     > Hi,
>>>     >
>>>     > The experiments with mode 6000 are going pretty well.
>>>     > Another local HAM (PE1ITR) even managed to work our repeater
>>>     with just a
>>>     > HackRF.
>>>     > Compared to analog FM it seems to perform pretty well.
>>>     >
>>>     > But I would like to take it a bit further and try to introduce
>>>     some FEC.
>>>     > Since the currently used codec2 mode only needs 3200bps there
>>>     is still
>>>     > plenty of room.
>>>     > I was thinking about a rate 2/3 code (687 code bits for 458
>>>     data bits),
>>>     > but I am a bit puzzled with the ldpc octave code.
>>>     > How exactly do I generate the needed matrix (and ofcourse the
>>>     C and
>>>     > header files)?
>>>     >
>>>     > Regards,
>>>     > Jeroen
>>>     >
>>>     >
>>>     >
>>>     > _______________________________________________
>>>     > Freetel-codec2 mailing list
>>>     > [email protected]
>>>     <mailto:[email protected]>
>>>     > https://lists.sourceforge.net/lists/listinfo/freetel-codec2
>>>     >
>>>
>>>
>>>     _______________________________________________
>>>     Freetel-codec2 mailing list
>>>     [email protected]
>>>     <mailto:[email protected]>
>>>     https://lists.sourceforge.net/lists/listinfo/freetel-codec2
>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> Freetel-codec2 mailing list
>>> [email protected]
>>> https://lists.sourceforge.net/lists/listinfo/freetel-codec2
>>
>>
>>
>> _______________________________________________
>> Freetel-codec2 mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/freetel-codec2
>
>
>
>
> _______________________________________________
> Freetel-codec2 mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/freetel-codec2

_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2

Reply via email to