is the Viterbi decode for the AFS provided by Sylvain right?
Hi ,List: search some materials, find that the decode method of AFS convolutional code is different from the EFS`, it use RSC, and need SOVA(soft output viterbi algorithm). am i right? -- View this message in context: http://baseband-devel.722152.n3.nabble.com/is-the-Viterbi-decode-for-the-AFS... Sent from the baseband-devel mailing list archive at Nabble.com.
search some materials, find that the decode method of AFS convolutional code is different from the EFS`, it use RSC, and need SOVA(soft output viterbi algorithm). am i right?
No. It _does_ use RSC (and the viterbi code in libosmocore supports RSC). But that has nothing to do with SOVA and SOVA is _not_ required at all. AFAIK, the AMR code I sent a while is correct. Just to be sure I attached the latest version I have. I haven't tested all the modes but at least 12.2 and 10.2 should be correct. Cheers, Sylvain
Hi, Sylvain I use your code to decode the AMR in my network, but have no positive result.I have some questions. 1.when I have decode the TCH, and ready write to file, how to set the last argument of the function. length = osmo_ubit2pbit_ext(voice, 0, EFRAMR, 0, 8 + gsm690_12_2_len, 0/1);is it 0 or 1? I use your burst_id branch. 2.are you test your code relate to TCH AMR? have you ever succeed to listen the voice coded by AMR, in your clarifications about 27C3 GSM Sniff Talk, I find you hadn`t completed the AMR decode.you said:" - ... But since I'd like it to support AMR and viterbi soft output before that happens, it could take some time." how about your current process? 3.in your SDCCH sniffer demo, you use snr of the channel to decide a over of a channel, but in the speech of amr, because of DTX, it can not be used, how to settle the problem because I can`t implement the full stack. Look forwarding to communicate with you！ -- View this message in context: http://baseband-devel.722152.n3.nabble.com/is-the-Viterbi-decode-for-the-AFS... Sent from the baseband-devel mailing list archive at Nabble.com.
Bob i will suggest u to try implementing amr for mobile app then test whether it works properly then start playing with burst ind Or is it possible for u to share your code i will try to implement it in mobile app Sent from Android On Jul 27, 2012 6:56 AM, "bob" <avwiseav@gmail.com> wrote:
Hi, Sylvain
I use your code to decode the AMR in my network, but have no positive result.I have some questions.
1.when I have decode the TCH, and ready write to file, how to set the last argument of the function. length = osmo_ubit2pbit_ext(voice, 0, EFRAMR, 0, 8 + gsm690_12_2_len, 0/1);is it 0 or 1? I use your burst_id branch.
2.are you test your code relate to TCH AMR? have you ever succeed to listen the voice coded by AMR, in your clarifications about 27C3 GSM Sniff Talk, I find you hadn`t completed the AMR decode.you said:" - ... But since I'd like it to support AMR and viterbi soft output before that happens, it could take some time." how about your current process?
3.in your SDCCH sniffer demo, you use snr of the channel to decide a over of a channel, but in the speech of amr, because of DTX, it can not be used, how to settle the problem because I can`t implement the full stack.
Look forwarding to communicate with you！
-- View this message in context: http://baseband-devel.722152.n3.nabble.com/is-the-Viterbi-decode-for-the-AFS... Sent from the baseband-devel mailing list archive at Nabble.com.
Bob i will suggest u to try implementing amr for mobile app then test whether it works properly then start playing with burst ind Or is it possible for u to share your code i will try to implement it in mobile app
Those are two _very_ different tasks. Decoding AMR is a fairly easy task compared to implementing it in mobile ... Cheers, Sylvain
1.when I have decode the TCH, and ready write to file, how to set the last argument of the function. length = osmo_ubit2pbit_ext(voice, 0, EFRAMR, 0, 8 + gsm690_12_2_len, 0/1);is it 0 or 1? I use your burst_id branch.
Most likely 0
2.are you test your code relate to TCH AMR? have you ever succeed to listen the voice coded by AMR, in your clarifications about 27C3 GSM Sniff Talk, I find you hadn`t completed the AMR decode.you said:" - ... But since I'd like it to support AMR and viterbi soft output before that happens, it could take some time." how about your current process?
The code isn't ready yet and since I'm not actively working on it ATM, it's not really progressing.
3.in your SDCCH sniffer demo, you use snr of the channel to decide a over of a channel, but in the speech of amr, because of DTX, it can not be used, how to settle the problem because I can`t implement the full stack.
The SAACH associated with TCH doesn't have DTX so you can check if the channel is active by only looking for power during the SAACH burts. Cheers, Sylvain
Dear Sylvain In code provided above there is no *conv_tch_afs_XX_XX_next_term_state* &* conv_tch_afs_10_2_next_term_state*struct for some of bitrate so its giving wrong output for convolution decode. is it possible to give complete conv_tch_afs.c decode. On Thu, Jul 26, 2012 at 12:23 PM, Sylvain Munaut <246tnt@gmail.com> wrote:
search some materials, find that the decode method of AFS convolutional code is different from the EFS`, it use RSC, and need SOVA(soft output viterbi algorithm). am i right?
No.
It _does_ use RSC (and the viterbi code in libosmocore supports RSC). But that has nothing to do with SOVA and SOVA is _not_ required at all.
AFAIK, the AMR code I sent a while is correct. Just to be sure I attached the latest version I have. I haven't tested all the modes but at least 12.2 and 10.2 should be correct.
Cheers,
Sylvain
-- Akib Sayyed Matrix-Shell akibsayyed@gmail.com akibsayyed@matrixshell.com Mob:- +91-966-514-2243
participants (4)
-
Akib Sayyed -
bob -
Peter Stuge -
Sylvain Munaut