Bartek Kania wrote: >-----BEGIN PGP SIGNED MESSAGE----- >Hash: SHA1 > >On Sat, 1 Apr 2006, Steve Underwood wrote: > > >>Max CtRiX Cetra wrote: >> >> >>>I am trying to make T38 termination work under openpbx. >>>I have used egnarf T38 trunk and it compiles cleanly with the latest >>>spandsp snapshot (commiting it to mainline could be useful). >>> >>> ><snipp> > > >>Here's a hint: "it doesn't work" is a completely useless report. :-) >>What does happen? What do the logs show? Does it negotiate T.38 at all? >> >> > >Firstly, >that branch is under development.. Or, it would be if it wasn't for a >lot of other things that I need to do and the fact that I haven't had >the time to read the T.38 specs yet =) > > The T.38 spec is not that illuminating when you read it. :-)
>I've been testing with RxFax and trying to receive a fax using UDPTL > >The problems so far are: >* chan_sip doesn't seem to detect the fax tones. I don't really see > why, but it seems the goertzel detector doesn't detect it. WHat I > found is that this check > if (!hit && (fax_energy >= FAX_THRESHOLD) && > (fax_energy >= DTMF_TO_TOTAL_ENERGY*s->energy) > the second line never gets to be true. If I remove it it seems to > detect the faxtones. But I don't understand what it does so I don't > really want to remove it. Steve, if you could look at that it would > be great! > > That sounds like the signal is disorted by the codec to the point where it lacks sufficient purity to be detected as a single tone. Which codec did you use for the audio? >* A really annoying bug in chan_sip/rtp/udptl/idontknowyet causes > openpbx to ignore incoming frames until it has sent the first frame > itself. So RxFax doesn't receive audio / udptl at all until it has > sent the first frame itself. I needed to add a PlayTones() before > the call to RxFax to hear the initial fax-tones in my test-phone and > via the modem. > > I think this is something to do with NAT handling, isn't it. I guess I could make an initial transmission, just to kick things off. >If anyone has any experience with T.38 and feels like looking at it, >please do so! I probably won't have the time for a while =/ > > You need to use a recent snapshot of spandsp. A couple of silly issues in the T.38 code were reported to me, and fixed - bits being packed in the wrong order at one point was something of a killer :-) People have reported sending faxes in both directions with a recent snapshot, but don't read too much into that. There are a lot of grey things in the T.38 spec, and further work to ensure interworking with a lot of kit that implements the T.38 spec in their own sweet way will be needed. Also, I don't expect the current code to be that robust. Its really not mature right now. The frame structure gained the sequence number of the frames a while ago. It isn't used to the fullest extent, though. For example, it should be used to define the sequence number of UDPTL frames going out, and it isn't. This is important, since to make UDPTL robust certain key frames should be repeated to minimise the risk of loss. This is not specified, but it is common and sane practice. Regards, Steve _______________________________________________ Openpbx-dev mailing list [email protected] http://lists.openpbx.org/mailman/listinfo/openpbx-dev
