Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu
The tone on/off conditions should be, minimum on 40ms and 50ms off. I recommend you to record a call with those double digit detections. If you can reproduce them yourself by dialing quickly it'd be even better. Once you have the recording you can run libteletone on the recorded audio (see libs/freetdm/src/detect_dtmf.c) and may be start tweaking it. You could also run spandsp dtmf detector over the recorded audio to see if does a better job (I'd think so). Another way to test if spandsp does better is to disable dtmf detection completely from freetdm (the disable_dtmf dial plan app does that) and then use spandsp_start_dtmf application on the dialplan (see mod_spandsp). I'd be interested in looking at the recording too, you can send it to my email if you want me to take a look. You can record all the read/write audio using: ftmod trace /path/prefix span [chan] That starts recording all incoming and outgoing data on the channel (if no channel is specified all channels on the span are recorded), the specified prefix will be used for each of the files. The recording starts immediately (although if there is no call then most likely no one is reading/writing to the channel), you must stop the recording when done with: ftmod notrace span [chan] Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com On Sat, Jul 3, 2010 at 4:22 AM, Jeroen C. van Gelderen jero...@thegreek.com wrote: If however you have this issue even with HW EC, call Sangoma tech support and we will be happy to look at the issue. I wish this was Sangoma hardware esp. with hardware ec :) My DTMF bouncing issues were indeed caused by echo but now that I've got OSLEC running (thanks for patching that) calls are echo-free and crystal clear. The issue now is that my users dial their transfers (-say- #44) reasonably fast and that often causes two successive DTMF digits to be detected as one (#4). In order to reliably reach the extension the clerk cannot dial more than one digit per second. That seems excessively slow. I suspect there must be a knob somewhere to tune that (in libteletone)? Other causes to look for? Cheers, -Slim ___ FreeSWITCH-dev mailing list FreeSWITCH-dev@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-dev UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-dev http://www.freeswitch.org ___ FreeSWITCH-dev mailing list FreeSWITCH-dev@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-dev UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-dev http://www.freeswitch.org
Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu
If however you have this issue even with HW EC, call Sangoma tech support and we will be happy to look at the issue. I wish this was Sangoma hardware esp. with hardware ec :) My DTMF bouncing issues were indeed caused by echo but now that I've got OSLEC running (thanks for patching that) calls are echo-free and crystal clear. The issue now is that my users dial their transfers (-say- #44) reasonably fast and that often causes two successive DTMF digits to be detected as one (#4). In order to reliably reach the extension the clerk cannot dial more than one digit per second. That seems excessively slow. I suspect there must be a knob somewhere to tune that (in libteletone)? Other causes to look for? Cheers, -Slim ___ FreeSWITCH-dev mailing list FreeSWITCH-dev@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-dev UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-dev http://www.freeswitch.org
Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu
Yes, CID works back again with HWEC disabled. However, I get echo without HWEC ;-) François On Wed, 30 Jun 2010 18:48:40 -0400, Moises Silva wrote: Just fyi, you can disable hw ec from the CLI so you don't have to start/stop the port for your tests. Start the card with HWEC enabled, then you can use wan_ec_client. # wan_ec_client wanpipe1 disable Then enable again: # wan_ec_client wanpipe1 enable You can verify it's enabled/disabled with # wan_ec_client wanpipe stats Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com [1] On Wed, Jun 30, 2010 at 6:29 PM, wrote: I'll give a try tomorrow with HWEC disabled. DE : freeswitch-dev-boun...@lists.freeswitch.org [3] [mailto:freeswitch-dev-boun...@lists.freeswitch.org [4]] DE LA PART DE Moises Silva ENVOYÉ : mercredi 30 juin 2010 23:15 À : freeswitch-dev@lists.freeswitch.org [5] OBJET : Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu Can you try 3.5.12 with and without hw ec enabled and check if cid is there? Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com [6] On Wed, Jun 30, 2010 at 2:58 PM, François Legal wrote: Hello, did try to upgrade to wanpipe 3.5.12 (from 3.5.11) and thought the DTMF problem seems to be fixed (did not had the opportunity to really test it thourougthly), I seem to have lost the CID feature in the upgrade. Roll back to 3.5.11 and CID is back there. François On Mon, 28 Jun 2010 18:54:19 -0400, Moises Silva wrote: Hello, I spent a few hours playing with DTMF stuff and analog cards and it seems there is 2 issues at hand. 1. Bleeding DTMF. 2. Echo DTMF. The first issue, for software DTMF, can be solved with Anthony's pre buffer size feature. However that introduces delay by design, and it will not work for large DTMFs (if the dtmf is larger than the buffer). For hardware DTMF a new driver was just released that includes a configuration to allow the EC chip to perform the dtmf tone removal which cuts down the bleeding to only 20ms (in my testing) there is no way a DTMF detector will consider that a valid DTMF and therefore the bleeding should be solved with no delay introduced. The option is HWEC_DTMF_REMOVAL = YES, must be added along with the usual TDMV_HW_DTMF = YES in wanpipex.conf Ideally the software DTMF detector (in this case teletone) should cut it at the same time that it detects it. I thought may be spandsp would help, but it seems spandsp does not have an option to squelch the dtmf tone. May be Steve can help with that. I pinged him on IRC and he said he may get some code working, but there is no date for that and also that would involve integrating spandsp into FreeTDM, any reason to not do this now that spandsp is LGPL? As for the echo dtmf. It seems sometimes an outgoing DTMF may be detected as incoming DTMF due to echo. There is not much we can do there if you don't have echo cancellation. If however you have this issue even with HW EC, call Sangoma tech support and we will be happy to look at the issue. In another note, I added a variable and application to disable DTMF. That will disable DTMF (either software or hardware) in the leg executing that app. If you want to disable in the outgoing leg (before a bridge), you must export a special variable: The DTMF is enabled automatically on each call, so there is no need to enable it for each call. But in case you need to enable it: Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com [8] On Wed, Jun 16, 2010 at 12:09 PM, Anthony Minessale wrote: openzap_pre_buffer_size is a variable you can set to specific number of MS 60 for example. it will pre_buffer the audio on the channel so when you detect dtmf it will completely drop the buffer so all of the original dtmf should be dropped as well. probably if the dtmf is too long then it will cause problems anyway. if that pre buffer does not fix anything it would point to echo or bleeding. We could make a variable to disable dtmf detection completely on a per-call basis possibly but you will still probably hear it bleeding. This type of problem was reported fixed with sangoma because of the echo canceler. I don't use dahdi or digium stuff much so I can't comment on what happens when you use it. On Wed, Jun 16, 2010 at 3:41 AM, François Legal wrote: Does this really fix it ? I wonder because the problem I see here is also that the FXS side detects the DTMF and then queues it on the FXO side for generation. That leads to the called party receiving twice the DTMF, the first one is the inband DTMF, the second is the one queued/generated by the FXO channel. For a clean
Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu
Hi, You can try the EC in speech recognition mode. By default, the ec is in normal mode. But you can set it to speech recognition mode with this: # wan_ec_client wanpipe1 msr channel msr = mode speech recognition Then verify the channel has the required mode: # wan_ec_client wanpipe1 stats channel You can take it back to normal mode (IIRC) with: # wan_ec_client wanpipe1 mn channel In any case I am interested in your problem with the ec in normal mode. I need to check with the driver developer to see if something changed in the ec settings. Please let me know if speech recognition mode helps. Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com On Thu, Jul 1, 2010 at 1:46 PM, François Legal de...@thom.fr.eu.org wrote: Yes, CID works back again with HWEC disabled. However, I get echo without HWEC ;-) François On Wed, 30 Jun 2010 18:48:40 -0400, Moises Silva moises.si...@gmail.com wrote: Just fyi, you can disable hw ec from the CLI so you don't have to start/stop the port for your tests. Start the card with HWEC enabled, then you can use wan_ec_client. # wan_ec_client wanpipe1 disable Then enable again: # wan_ec_client wanpipe1 enable You can verify it's enabled/disabled with # wan_ec_client wanpipe stats Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com On Wed, Jun 30, 2010 at 6:29 PM, de...@thom.fr.eu.org wrote: I’ll give a try tomorrow with HWEC disabled. *De :* freeswitch-dev-boun...@lists.freeswitch.org [mailto: freeswitch-dev-boun...@lists.freeswitch.org] *De la part de* Moises Silva *Envoyé :* mercredi 30 juin 2010 23:15 *À :* freeswitch-dev@lists.freeswitch.org *Objet :* Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu Can you try 3.5.12 with and without hw ec enabled and check if cid is there? Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com On Wed, Jun 30, 2010 at 2:58 PM, François Legal de...@thom.fr.eu.org wrote: Hello, did try to upgrade to wanpipe 3.5.12 (from 3.5.11) and thought the DTMF problem seems to be fixed (did not had the opportunity to really test it thourougthly), I seem to have lost the CID feature in the upgrade. Roll back to 3.5.11 and CID is back there. François On Mon, 28 Jun 2010 18:54:19 -0400, Moises Silva wrote: Hello, I spent a few hours playing with DTMF stuff and analog cards and it seems there is 2 issues at hand. 1. Bleeding DTMF. 2. Echo DTMF. The first issue, for software DTMF, can be solved with Anthony's pre buffer size feature. However that introduces delay by design, and it will not work for large DTMFs (if the dtmf is larger than the buffer). For hardware DTMF a new driver was just released that includes a configuration to allow the EC chip to perform the dtmf tone removal which cuts down the bleeding to only 20ms (in my testing) there is no way a DTMF detector will consider that a valid DTMF and therefore the bleeding should be solved with no delay introduced. The option is HWEC_DTMF_REMOVAL = YES, must be added along with the usual TDMV_HW_DTMF = YES in wanpipex.conf Ideally the software DTMF detector (in this case teletone) should cut it at the same time that it detects it. I thought may be spandsp would help, but it seems spandsp does not have an option to squelch the dtmf tone. May be Steve can help with that. I pinged him on IRC and he said he may get some code working, but there is no date for that and also that would involve integrating spandsp into FreeTDM, any reason to not do this now that spandsp is LGPL? As for the echo dtmf. It seems sometimes an outgoing DTMF may be detected as incoming DTMF due to echo. There is not much we can do there if you don't have echo cancellation. If however you have this issue even with HW EC, call Sangoma tech support and we will be happy to look at the issue. In another note, I added a variable and application to disable DTMF. That will disable DTMF (either software or hardware) in the leg executing that app. If you want to disable in the outgoing leg (before a bridge), you must export a special variable: The DTMF is enabled automatically on each call, so there is no need to enable it for each call. But in case you need to enable it: Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com On Wed, Jun 16, 2010 at 12:09 PM, Anthony Minessale anthony.miness...@gmail.com wrote: openzap_pre_buffer_size is a variable you can set to specific number of MS 60 for example. it will pre_buffer
Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu
Hello, did try to upgrade to wanpipe 3.5.12 (from 3.5.11) and thought the DTMF problem seems to be fixed (did not had the opportunity to really test it thourougthly), I seem to have lost the CID feature in the upgrade. Roll back to 3.5.11 and CID is back there. François On Mon, 28 Jun 2010 18:54:19 -0400, Moises Silva wrote: Hello, I spent a few hours playing with DTMF stuff and analog cards and it seems there is 2 issues at hand. 1. Bleeding DTMF. 2. Echo DTMF. The first issue, for software DTMF, can be solved with Anthony's pre buffer size feature. However that introduces delay by design, and it will not work for large DTMFs (if the dtmf is larger than the buffer). For hardware DTMF a new driver was just released that includes a configuration to allow the EC chip to perform the dtmf tone removal which cuts down the bleeding to only 20ms (in my testing) there is no way a DTMF detector will consider that a valid DTMF and therefore the bleeding should be solved with no delay introduced. The option is HWEC_DTMF_REMOVAL = YES, must be added along with the usual TDMV_HW_DTMF = YES in wanpipex.conf Ideally the software DTMF detector (in this case teletone) should cut it at the same time that it detects it. I thought may be spandsp would help, but it seems spandsp does not have an option to squelch the dtmf tone. May be Steve can help with that. I pinged him on IRC and he said he may get some code working, but there is no date for that and also that would involve integrating spandsp into FreeTDM, any reason to not do this now that spandsp is LGPL? As for the echo dtmf. It seems sometimes an outgoing DTMF may be detected as incoming DTMF due to echo. There is not much we can do there if you don't have echo cancellation. If however you have this issue even with HW EC, call Sangoma tech support and we will be happy to look at the issue. In another note, I added a variable and application to disable DTMF. That will disable DTMF (either software or hardware) in the leg executing that app. If you want to disable in the outgoing leg (before a bridge), you must export a special variable: The DTMF is enabled automatically on each call, so there is no need to enable it for each call. But in case you need to enable it: Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com [1] On Wed, Jun 16, 2010 at 12:09 PM, Anthony Minessale wrote: openzap_pre_buffer_size is a variable you can set to specific number of MS 60 for example. it will pre_buffer the audio on the channel so when you detect dtmf it will completely drop the buffer so all of the original dtmf should be dropped as well. probably if the dtmf is too long then it will cause problems anyway. if that pre buffer does not fix anything it would point to echo or bleeding. We could make a variable to disable dtmf detection completely on a per-call basis possibly but you will still probably hear it bleeding. This type of problem was reported fixed with sangoma because of the echo canceler. I don't use dahdi or digium stuff much so I can't comment on what happens when you use it. On Wed, Jun 16, 2010 at 3:41 AM, François Legal wrote: Does this really fix it ? I wonder because the problem I see here is also that the FXS side detects the DTMF and then queues it on the FXO side for generation. That leads to the called party receiving twice the DTMF, the first one is the inband DTMF, the second is the one queued/generated by the FXO channel. For a clean fix, I guess some kind of application should be created, that would prevent DTMF to be queued on the other channel. Such application would then be called before the bridge. Maybe there is a cleaner way to do this. François On Tue, 15 Jun 2010 18:23:13 -0500, Jeroen C. van Gelderen wrote: Hi everybody, I have a problem that is very similar to a problem reported by François [1] except for the fact that my FXS and FXO ports are on a Xorcom Astribank device instead of a Sangoma. To quote François: The problem is that each leg of the bridge is detecting the inband DTMF, and so [F]reeswitch sends each detected DTMF from one leg to the other, and so on and so forth (as each leg detects the DTMF again and again) HIS words but evidenced by the MY log :) The snippet below has DTMF coming in on the FXS port (1:1) and bouncing between it and the FXO port (3:1). The ports are simply bridged together with bridge(OpenZap/3/1/F) or bridge(OpenZap/3/1/w) I'm running: FreeSWITCH version: 1.0.head (git-01c0c69 2010-06-08 16-22-21 -0500) dahdi: Version: SVN-trunk-r8762 Output of lsdahdi at end of message. 8 2010-06-13 04:18:39.439252 [DEBUG] zap_io.c:2062 1:1 GENERATE DTMF [4] 2010-06-13 04:18:39.637249 [DEBUG] mod_openzap.c:721 queue DTMF [4] 2010-06-13 04:18:39.676248 [DEBUG] zap_io.c:2062 3:1 GENERATE DTMF [4]
Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu
Can you try 3.5.12 with and without hw ec enabled and check if cid is there? Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com On Wed, Jun 30, 2010 at 2:58 PM, François Legal de...@thom.fr.eu.orgwrote: Hello, did try to upgrade to wanpipe 3.5.12 (from 3.5.11) and thought the DTMF problem seems to be fixed (did not had the opportunity to really test it thourougthly), I seem to have lost the CID feature in the upgrade. Roll back to 3.5.11 and CID is back there. François On Mon, 28 Jun 2010 18:54:19 -0400, Moises Silva wrote: Hello, I spent a few hours playing with DTMF stuff and analog cards and it seems there is 2 issues at hand. 1. Bleeding DTMF. 2. Echo DTMF. The first issue, for software DTMF, can be solved with Anthony's pre buffer size feature. However that introduces delay by design, and it will not work for large DTMFs (if the dtmf is larger than the buffer). For hardware DTMF a new driver was just released that includes a configuration to allow the EC chip to perform the dtmf tone removal which cuts down the bleeding to only 20ms (in my testing) there is no way a DTMF detector will consider that a valid DTMF and therefore the bleeding should be solved with no delay introduced. The option is HWEC_DTMF_REMOVAL = YES, must be added along with the usual TDMV_HW_DTMF = YES in wanpipex.conf Ideally the software DTMF detector (in this case teletone) should cut it at the same time that it detects it. I thought may be spandsp would help, but it seems spandsp does not have an option to squelch the dtmf tone. May be Steve can help with that. I pinged him on IRC and he said he may get some code working, but there is no date for that and also that would involve integrating spandsp into FreeTDM, any reason to not do this now that spandsp is LGPL? As for the echo dtmf. It seems sometimes an outgoing DTMF may be detected as incoming DTMF due to echo. There is not much we can do there if you don't have echo cancellation. If however you have this issue even with HW EC, call Sangoma tech support and we will be happy to look at the issue. In another note, I added a variable and application to disable DTMF. That will disable DTMF (either software or hardware) in the leg executing that app. If you want to disable in the outgoing leg (before a bridge), you must export a special variable: The DTMF is enabled automatically on each call, so there is no need to enable it for each call. But in case you need to enable it: Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com On Wed, Jun 16, 2010 at 12:09 PM, Anthony Minessale anthony.miness...@gmail.com wrote: openzap_pre_buffer_size is a variable you can set to specific number of MS 60 for example. it will pre_buffer the audio on the channel so when you detect dtmf it will completely drop the buffer so all of the original dtmf should be dropped as well. probably if the dtmf is too long then it will cause problems anyway. if that pre buffer does not fix anything it would point to echo or bleeding. We could make a variable to disable dtmf detection completely on a per-call basis possibly but you will still probably hear it bleeding. This type of problem was reported fixed with sangoma because of the echo canceler. I don't use dahdi or digium stuff much so I can't comment on what happens when you use it. On Wed, Jun 16, 2010 at 3:41 AM, François Legal de...@thom.fr.eu.orgwrote: Does this really fix it ? I wonder because the problem I see here is also that the FXS side detects the DTMF and then queues it on the FXO side for generation. That leads to the called party receiving twice the DTMF, the first one is the inband DTMF, the second is the one queued/generated by the FXO channel. For a clean fix, I guess some kind of application should be created, that would prevent DTMF to be queued on the other channel. Such application would then be called before the bridge. Maybe there is a cleaner way to do this. François On Tue, 15 Jun 2010 18:23:13 -0500, Jeroen C. van Gelderen jero...@thegreek.com wrote: Hi everybody, I have a problem that is very similar to a problem reported by François [1] except for the fact that my FXS and FXO ports are on a Xorcom Astribank device instead of a Sangoma. To quote François: The problem is that each leg of the bridge is detecting the inband DTMF, and so [F]reeswitch sends each detected DTMF from one leg to the other, and so on and so forth (as each leg detects the DTMF again and again) HIS words but evidenced by the MY log :) The snippet below has DTMF coming in on the FXS port (1:1) and bouncing between it and the FXO port (3:1). The ports are simply bridged together
Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu
I’ll give a try tomorrow with HWEC disabled. De : freeswitch-dev-boun...@lists.freeswitch.org [mailto:freeswitch-dev-boun...@lists.freeswitch.org] De la part de Moises Silva Envoyé : mercredi 30 juin 2010 23:15 À : freeswitch-dev@lists.freeswitch.org Objet : Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu Can you try 3.5.12 with and without hw ec enabled and check if cid is there? Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com On Wed, Jun 30, 2010 at 2:58 PM, François Legal de...@thom.fr.eu.org wrote: Hello, did try to upgrade to wanpipe 3.5.12 (from 3.5.11) and thought the DTMF problem seems to be fixed (did not had the opportunity to really test it thourougthly), I seem to have lost the CID feature in the upgrade. Roll back to 3.5.11 and CID is back there. François On Mon, 28 Jun 2010 18:54:19 -0400, Moises Silva wrote: Hello, I spent a few hours playing with DTMF stuff and analog cards and it seems there is 2 issues at hand. 1. Bleeding DTMF. 2. Echo DTMF. The first issue, for software DTMF, can be solved with Anthony's pre buffer size feature. However that introduces delay by design, and it will not work for large DTMFs (if the dtmf is larger than the buffer). For hardware DTMF a new driver was just released that includes a configuration to allow the EC chip to perform the dtmf tone removal which cuts down the bleeding to only 20ms (in my testing) there is no way a DTMF detector will consider that a valid DTMF and therefore the bleeding should be solved with no delay introduced. The option is HWEC_DTMF_REMOVAL = YES, must be added along with the usual TDMV_HW_DTMF = YES in wanpipex.conf Ideally the software DTMF detector (in this case teletone) should cut it at the same time that it detects it. I thought may be spandsp would help, but it seems spandsp does not have an option to squelch the dtmf tone. May be Steve can help with that. I pinged him on IRC and he said he may get some code working, but there is no date for that and also that would involve integrating spandsp into FreeTDM, any reason to not do this now that spandsp is LGPL? As for the echo dtmf. It seems sometimes an outgoing DTMF may be detected as incoming DTMF due to echo. There is not much we can do there if you don't have echo cancellation. If however you have this issue even with HW EC, call Sangoma tech support and we will be happy to look at the issue. In another note, I added a variable and application to disable DTMF. That will disable DTMF (either software or hardware) in the leg executing that app. If you want to disable in the outgoing leg (before a bridge), you must export a special variable: The DTMF is enabled automatically on each call, so there is no need to enable it for each call. But in case you need to enable it: Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com On Wed, Jun 16, 2010 at 12:09 PM, Anthony Minessale anthony.miness...@gmail.com wrote: openzap_pre_buffer_size is a variable you can set to specific number of MS 60 for example. it will pre_buffer the audio on the channel so when you detect dtmf it will completely drop the buffer so all of the original dtmf should be dropped as well. probably if the dtmf is too long then it will cause problems anyway. if that pre buffer does not fix anything it would point to echo or bleeding. We could make a variable to disable dtmf detection completely on a per-call basis possibly but you will still probably hear it bleeding. This type of problem was reported fixed with sangoma because of the echo canceler. I don't use dahdi or digium stuff much so I can't comment on what happens when you use it. On Wed, Jun 16, 2010 at 3:41 AM, François Legal de...@thom.fr.eu.org wrote: Does this really fix it ? I wonder because the problem I see here is also that the FXS side detects the DTMF and then queues it on the FXO side for generation. That leads to the called party receiving twice the DTMF, the first one is the inband DTMF, the second is the one queued/generated by the FXO channel. For a clean fix, I guess some kind of application should be created, that would prevent DTMF to be queued on the other channel. Such application would then be called before the bridge. Maybe there is a cleaner way to do this. François On Tue, 15 Jun 2010 18:23:13 -0500, Jeroen C. van Gelderen jero...@thegreek.com wrote: Hi everybody, I have a problem that is very similar to a problem reported by François [1] except for the fact that my FXS and FXO ports are on a Xorcom Astribank device instead of a Sangoma. To quote François: The problem is that each leg of the bridge is detecting the inband DTMF, and so [F]reeswitch sends each detected
Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu
Just fyi, you can disable hw ec from the CLI so you don't have to start/stop the port for your tests. Start the card with HWEC enabled, then you can use wan_ec_client. # wan_ec_client wanpipe1 disable chan Then enable again: # wan_ec_client wanpipe1 enable chan You can verify it's enabled/disabled with # wan_ec_client wanpipe stats chan Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com On Wed, Jun 30, 2010 at 6:29 PM, de...@thom.fr.eu.org wrote: I’ll give a try tomorrow with HWEC disabled. *De :* freeswitch-dev-boun...@lists.freeswitch.org [mailto: freeswitch-dev-boun...@lists.freeswitch.org] *De la part de* Moises Silva *Envoyé :* mercredi 30 juin 2010 23:15 *À :* freeswitch-dev@lists.freeswitch.org *Objet :* Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu Can you try 3.5.12 with and without hw ec enabled and check if cid is there? Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com On Wed, Jun 30, 2010 at 2:58 PM, François Legal de...@thom.fr.eu.org wrote: Hello, did try to upgrade to wanpipe 3.5.12 (from 3.5.11) and thought the DTMF problem seems to be fixed (did not had the opportunity to really test it thourougthly), I seem to have lost the CID feature in the upgrade. Roll back to 3.5.11 and CID is back there. François On Mon, 28 Jun 2010 18:54:19 -0400, Moises Silva wrote: Hello, I spent a few hours playing with DTMF stuff and analog cards and it seems there is 2 issues at hand. 1. Bleeding DTMF. 2. Echo DTMF. The first issue, for software DTMF, can be solved with Anthony's pre buffer size feature. However that introduces delay by design, and it will not work for large DTMFs (if the dtmf is larger than the buffer). For hardware DTMF a new driver was just released that includes a configuration to allow the EC chip to perform the dtmf tone removal which cuts down the bleeding to only 20ms (in my testing) there is no way a DTMF detector will consider that a valid DTMF and therefore the bleeding should be solved with no delay introduced. The option is HWEC_DTMF_REMOVAL = YES, must be added along with the usual TDMV_HW_DTMF = YES in wanpipex.conf Ideally the software DTMF detector (in this case teletone) should cut it at the same time that it detects it. I thought may be spandsp would help, but it seems spandsp does not have an option to squelch the dtmf tone. May be Steve can help with that. I pinged him on IRC and he said he may get some code working, but there is no date for that and also that would involve integrating spandsp into FreeTDM, any reason to not do this now that spandsp is LGPL? As for the echo dtmf. It seems sometimes an outgoing DTMF may be detected as incoming DTMF due to echo. There is not much we can do there if you don't have echo cancellation. If however you have this issue even with HW EC, call Sangoma tech support and we will be happy to look at the issue. In another note, I added a variable and application to disable DTMF. That will disable DTMF (either software or hardware) in the leg executing that app. If you want to disable in the outgoing leg (before a bridge), you must export a special variable: The DTMF is enabled automatically on each call, so there is no need to enable it for each call. But in case you need to enable it: Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com On Wed, Jun 16, 2010 at 12:09 PM, Anthony Minessale anthony.miness...@gmail.com wrote: openzap_pre_buffer_size is a variable you can set to specific number of MS 60 for example. it will pre_buffer the audio on the channel so when you detect dtmf it will completely drop the buffer so all of the original dtmf should be dropped as well. probably if the dtmf is too long then it will cause problems anyway. if that pre buffer does not fix anything it would point to echo or bleeding. We could make a variable to disable dtmf detection completely on a per-call basis possibly but you will still probably hear it bleeding. This type of problem was reported fixed with sangoma because of the echo canceler. I don't use dahdi or digium stuff much so I can't comment on what happens when you use it. On Wed, Jun 16, 2010 at 3:41 AM, François Legal de...@thom.fr.eu.org wrote: Does this really fix it ? I wonder because the problem I see here is also that the FXS side detects the DTMF and then queues it on the FXO side for generation. That leads to the called party receiving twice the DTMF, the first one is the inband DTMF, the second is the one queued/generated by the FXO channel. For a clean fix, I guess
Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu
Hello, I spent a few hours playing with DTMF stuff and analog cards and it seems there is 2 issues at hand. 1. Bleeding DTMF. 2. Echo DTMF. The first issue, for software DTMF, can be solved with Anthony's pre buffer size feature. However that introduces delay by design, and it will not work for large DTMFs (if the dtmf is larger than the buffer). For hardware DTMF a new driver was just released that includes a configuration to allow the EC chip to perform the dtmf tone removal which cuts down the bleeding to only 20ms (in my testing) there is no way a DTMF detector will consider that a valid DTMF and therefore the bleeding should be solved with no delay introduced. The option is HWEC_DTMF_REMOVAL = YES, must be added along with the usual TDMV_HW_DTMF = YES in wanpipex.conf Ideally the software DTMF detector (in this case teletone) should cut it at the same time that it detects it. I thought may be spandsp would help, but it seems spandsp does not have an option to squelch the dtmf tone. May be Steve can help with that. I pinged him on IRC and he said he may get some code working, but there is no date for that and also that would involve integrating spandsp into FreeTDM, any reason to not do this now that spandsp is LGPL? As for the echo dtmf. It seems sometimes an outgoing DTMF may be detected as incoming DTMF due to echo. There is not much we can do there if you don't have echo cancellation. If however you have this issue even with HW EC, call Sangoma tech support and we will be happy to look at the issue. In another note, I added a variable and application to disable DTMF. action application=disable_dtmf/ That will disable DTMF (either software or hardware) in the leg executing that app. If you want to disable in the outgoing leg (before a bridge), you must export a special variable: action application=export data=freetdm_disable_dtmf=yes/ The DTMF is enabled automatically on each call, so there is no need to enable it for each call. But in case you need to enable it: action application=enable_dtmf/ Moises Silva Senior Software Engineer Sangoma Technologies Inc. | 50 McIntosh Drive, Suite 120, Markham ON L3R 9T3 Canada t. 1 905 474 1990 x 128 | e. m...@sangoma.com On Wed, Jun 16, 2010 at 12:09 PM, Anthony Minessale anthony.miness...@gmail.com wrote: openzap_pre_buffer_size is a variable you can set to specific number of MS 60 for example. it will pre_buffer the audio on the channel so when you detect dtmf it will completely drop the buffer so all of the original dtmf should be dropped as well. probably if the dtmf is too long then it will cause problems anyway. if that pre buffer does not fix anything it would point to echo or bleeding. We could make a variable to disable dtmf detection completely on a per-call basis possibly but you will still probably hear it bleeding. This type of problem was reported fixed with sangoma because of the echo canceler. I don't use dahdi or digium stuff much so I can't comment on what happens when you use it. On Wed, Jun 16, 2010 at 3:41 AM, François Legal de...@thom.fr.eu.orgwrote: Does this really fix it ? I wonder because the problem I see here is also that the FXS side detects the DTMF and then queues it on the FXO side for generation. That leads to the called party receiving twice the DTMF, the first one is the inband DTMF, the second is the one queued/generated by the FXO channel. For a clean fix, I guess some kind of application should be created, that would prevent DTMF to be queued on the other channel. Such application would then be called before the bridge. Maybe there is a cleaner way to do this. François On Tue, 15 Jun 2010 18:23:13 -0500, Jeroen C. van Gelderen jero...@thegreek.com wrote: Hi everybody, I have a problem that is very similar to a problem reported by François [1] except for the fact that my FXS and FXO ports are on a Xorcom Astribank device instead of a Sangoma. To quote François: The problem is that each leg of the bridge is detecting the inband DTMF, and so [F]reeswitch sends each detected DTMF from one leg to the other, and so on and so forth (as each leg detects the DTMF again and again) HIS words but evidenced by the MY log :) The snippet below has DTMF coming in on the FXS port (1:1) and bouncing between it and the FXO port (3:1). The ports are simply bridged together with bridge(OpenZap/3/1/F) or bridge(OpenZap/3/1/w) I'm running: FreeSWITCH version: 1.0.head (git-01c0c69 2010-06-08 16-22-21 -0500) dahdi: Version: SVN-trunk-r8762 Output of lsdahdi at end of message. 8888888 [...] 2010-06-13 04:18:39.217256 [DEBUG] mod_openzap.c:721 queue DTMF [4] 2010-06-13 04:18:39.256255 [DEBUG] zap_io.c:2062 3:1 GENERATE DTMF [4] 2010-06-13 04:18:39.397253 [DEBUG] mod_openzap.c:721 queue DTMF [4] 2010-06-13 04:18:39.439252 [DEBUG] mod_openzap.c:780 Dropping frame! (write not ready)
Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu
Does this really fix it ? I wonder because the problem I see here is also that the FXS side detects the DTMF and then queues it on the FXO side for generation. That leads to the called party receiving twice the DTMF, the first one is the inband DTMF, the second is the one queued/generated by the FXO channel. For a clean fix, I guess some kind of application should be created, that would prevent DTMF to be queued on the other channel. Such application would then be called before the bridge. Maybe there is a cleaner way to do this. François On Tue, 15 Jun 2010 18:23:13 -0500, Jeroen C. van Gelderen jero...@thegreek.com wrote: Hi everybody, I have a problem that is very similar to a problem reported by François [1] except for the fact that my FXS and FXO ports are on a Xorcom Astribank device instead of a Sangoma. To quote François: The problem is that each leg of the bridge is detecting the inband DTMF, and so [F]reeswitch sends each detected DTMF from one leg to the other, and so on and so forth (as each leg detects the DTMF again and again) HIS words but evidenced by the MY log :) The snippet below has DTMF coming in on the FXS port (1:1) and bouncing between it and the FXO port (3:1). The ports are simply bridged together with bridge(OpenZap/3/1/F) or bridge(OpenZap/3/1/w) I'm running: FreeSWITCH version: 1.0.head (git-01c0c69 2010-06-08 16-22-21 -0500) dahdi: Version: SVN-trunk-r8762 Output of lsdahdi at end of message. 8888888 [...] 2010-06-13 04:18:39.217256 [DEBUG] mod_openzap.c:721 queue DTMF [4] 2010-06-13 04:18:39.256255 [DEBUG] zap_io.c:2062 3:1 GENERATE DTMF [4] 2010-06-13 04:18:39.397253 [DEBUG] mod_openzap.c:721 queue DTMF [4] 2010-06-13 04:18:39.439252 [DEBUG] mod_openzap.c:780 Dropping frame! (write not ready) 2010-06-13 04:18:39.439252 [DEBUG] zap_io.c:2062 1:1 GENERATE DTMF [4] 2010-06-13 04:18:39.637249 [DEBUG] mod_openzap.c:721 queue DTMF [4] 2010-06-13 04:18:39.676248 [DEBUG] zap_io.c:2062 3:1 GENERATE DTMF [4] 2010-06-13 04:18:39.859244 [DEBUG] mod_openzap.c:780 Dropping frame! (write not ready) [...ad infinitum...] 8888888 Because I needed DTMF pass-through working now I applied an ugly HACK. This drops DTMF tones detected on spans 3 and 4 (which are my FXO spans). This is very WRONG but it does solve my immediate problem: 8888888 diff --git a/libs/openzap/mod_openzap/mod_openzap.c b/libs/openzap/mod_openzap/mod_openzap.c index 5aebfea..ff3b081 100644 --- a/libs/openzap/mod_openzap/mod_openzap.c +++ b/libs/openzap/mod_openzap/mod_openzap.c @@ -718,8 +718,12 @@ static switch_status_t channel_read_frame(switch_core_session_t *session, switch for (p = dtmf; p *p; p++) { if (is_dtmf(*p)) { _dtmf.digit = *p; - zap_log(ZAP_LOG_DEBUG, queue DTMF [%c]\n, *p); - switch_channel_queue_dtmf(channel, _dtmf); + if (tech_pvt-zchan-span_id == 3 || tech_pvt-zchan-span_id == 4) { + zap_log(ZAP_LOG_DEBUG, Ignoring DTMF [%c] on FXO port %d:%d\n, *p, tech_pvt-zchan-span_id, tech_pvt-zchan-chan_id) ; + } else { + zap_log(ZAP_LOG_DEBUG, queue DTMF [%c]\n, *p); + switch_channel_queue_dtmf(channel, _dtmf); + } } } } 8888888 Can someone point me in the right direction instead? Do I need to do this at the OpenZAP/DAHDI level by disabling some kind of DTMF detection like was done for the Sangoma driver? Any and all pointers appreciated. Cheers, -Slim [1] [Freeswitch-dev] FXS bridged on FXO ports and DTMF http://www.mail-archive.com/freeswitch-dev@lists.freeswitch.org/msg02830.htm l [Freeswitch-dev] Problem with sending DTMF on FXS port bridged to an FXO port http://lists.freeswitch.org/pipermail/freeswitch-dev/2010-April/003607.html 8888888 [r...@localhost freeswitch]# lsdahdi ### Span 1: XBUS-00/XPD-00 Xorcom XPD #00/00: FXS (MASTER) 1 FXSFXOKS 2 FXSFXOKS 3 FXSFXOKS 4 FXSFXOKS 5 FXSFXOKS 6 FXSFXOKS 7 FXSFXOKS 8 FXSFXOKS 9 Output FXOKS 10 Output FXOKS 11 Input FXOKS 12 Input FXOKS 13 Input FXOKS 14 Input FXOKS ### Span 2: XBUS-00/XPD-10 Xorcom XPD #00/10: FXS 15 FXSFXOKS 16 FXSFXOKS 17 FXSFXOKS 18 FXSFXOKS 19 FXSFXOKS 20 FXSFXOKS 21 FXSFXOKS 22 FXSFXOKS ### Span 3: XBUS-00/XPD-20 Xorcom XPD #00/20: FXO 23 FXOFXSKS
Re: [Freeswitch-dev] FXS bridged on FXO ports and DTMF - Deja Vu
openzap_pre_buffer_size is a variable you can set to specific number of MS 60 for example. it will pre_buffer the audio on the channel so when you detect dtmf it will completely drop the buffer so all of the original dtmf should be dropped as well. probably if the dtmf is too long then it will cause problems anyway. if that pre buffer does not fix anything it would point to echo or bleeding. We could make a variable to disable dtmf detection completely on a per-call basis possibly but you will still probably hear it bleeding. This type of problem was reported fixed with sangoma because of the echo canceler. I don't use dahdi or digium stuff much so I can't comment on what happens when you use it. On Wed, Jun 16, 2010 at 3:41 AM, François Legal de...@thom.fr.eu.orgwrote: Does this really fix it ? I wonder because the problem I see here is also that the FXS side detects the DTMF and then queues it on the FXO side for generation. That leads to the called party receiving twice the DTMF, the first one is the inband DTMF, the second is the one queued/generated by the FXO channel. For a clean fix, I guess some kind of application should be created, that would prevent DTMF to be queued on the other channel. Such application would then be called before the bridge. Maybe there is a cleaner way to do this. François On Tue, 15 Jun 2010 18:23:13 -0500, Jeroen C. van Gelderen jero...@thegreek.com wrote: Hi everybody, I have a problem that is very similar to a problem reported by François [1] except for the fact that my FXS and FXO ports are on a Xorcom Astribank device instead of a Sangoma. To quote François: The problem is that each leg of the bridge is detecting the inband DTMF, and so [F]reeswitch sends each detected DTMF from one leg to the other, and so on and so forth (as each leg detects the DTMF again and again) HIS words but evidenced by the MY log :) The snippet below has DTMF coming in on the FXS port (1:1) and bouncing between it and the FXO port (3:1). The ports are simply bridged together with bridge(OpenZap/3/1/F) or bridge(OpenZap/3/1/w) I'm running: FreeSWITCH version: 1.0.head (git-01c0c69 2010-06-08 16-22-21 -0500) dahdi: Version: SVN-trunk-r8762 Output of lsdahdi at end of message. 8888888 [...] 2010-06-13 04:18:39.217256 [DEBUG] mod_openzap.c:721 queue DTMF [4] 2010-06-13 04:18:39.256255 [DEBUG] zap_io.c:2062 3:1 GENERATE DTMF [4] 2010-06-13 04:18:39.397253 [DEBUG] mod_openzap.c:721 queue DTMF [4] 2010-06-13 04:18:39.439252 [DEBUG] mod_openzap.c:780 Dropping frame! (write not ready) 2010-06-13 04:18:39.439252 [DEBUG] zap_io.c:2062 1:1 GENERATE DTMF [4] 2010-06-13 04:18:39.637249 [DEBUG] mod_openzap.c:721 queue DTMF [4] 2010-06-13 04:18:39.676248 [DEBUG] zap_io.c:2062 3:1 GENERATE DTMF [4] 2010-06-13 04:18:39.859244 [DEBUG] mod_openzap.c:780 Dropping frame! (write not ready) [...ad infinitum...] 8888888 Because I needed DTMF pass-through working now I applied an ugly HACK. This drops DTMF tones detected on spans 3 and 4 (which are my FXO spans). This is very WRONG but it does solve my immediate problem: 8888888 diff --git a/libs/openzap/mod_openzap/mod_openzap.c b/libs/openzap/mod_openzap/mod_openzap.c index 5aebfea..ff3b081 100644 --- a/libs/openzap/mod_openzap/mod_openzap.c +++ b/libs/openzap/mod_openzap/mod_openzap.c @@ -718,8 +718,12 @@ static switch_status_t channel_read_frame(switch_core_session_t *session, switch for (p = dtmf; p *p; p++) { if (is_dtmf(*p)) { _dtmf.digit = *p; - zap_log(ZAP_LOG_DEBUG, queue DTMF [%c]\n, *p); - switch_channel_queue_dtmf(channel, _dtmf); + if (tech_pvt-zchan-span_id == 3 || tech_pvt-zchan-span_id == 4) { + zap_log(ZAP_LOG_DEBUG, Ignoring DTMF [%c] on FXO port %d:%d\n, *p, tech_pvt-zchan-span_id, tech_pvt-zchan-chan_id) ; + } else { + zap_log(ZAP_LOG_DEBUG, queue DTMF [%c]\n, *p); + switch_channel_queue_dtmf(channel, _dtmf); + } } } } 8888888 Can someone point me in the right direction instead? Do I need to do this at the OpenZAP/DAHDI level by disabling some kind of DTMF detection like was done for the Sangoma driver? Any and all pointers appreciated. Cheers, -Slim [1] [Freeswitch-dev] FXS bridged on FXO ports and DTMF http://www.mail-archive.com/freeswitch-dev@lists.freeswitch.org/msg02830.htm l [Freeswitch-dev] Problem with sending DTMF on FXS port bridged to an FXO port