Hi, >> Hi Damjan! >> >> Thanks, this was really a good starting point. Do you know are there any >> papers/reports/personal experience about the robustness of T.38 against >> jitter/delay/loss? >> > > Audiocodes claims it can deal with up to 400 milliseconds of jitter and 12 > seconds round-trip delay (www.audiocodes.com/objects/FaxRelay.pdf). As for > packet loss, redundancy/FEC give you protection against a few lost packets > in a row, but in practice router tail-drops are bursty... > I wonder how they came up with that. To get beyond about 4 seconds of loop delay you need some serious spoofing tricks, but I've never seen an Audiocodes box try to do that. My T.38 gateway can only handle about 4 seconds. I haven't yet attempted to implement the tricks needed to go further, and it seems very few other implementors have. >> btw I wonder about the bandwidth consumption of T.38. Is it really just >> the fax bandwidth (e.g. 14400 bits/s) plus the IP/UDP header or is there >> something else? >> > > It's no bandwidth during V.21 preamble, 300 bits per second during the > V.21 HDLC, and up to 14400 bits per second during image transfer. Faxes > are always half-duplex. Redundancy/FEC increase these numbers, but I've > heard a T.38 fax still uses less bandwidth than a G.729 call. > The bandwidth used depends a lot on the amount of redundancy you have configured. Callweaver currently follows the commonest rule - quite a lot of redundancy for the 300bps control messages, and no redundancy for the image data. If the image data is send in reasonably large chunks, V.17 can take less than the small chunks of G.729 consume.
Regards, Steve _______________________________________________ Callweaver-users mailing list [email protected] http://lists.callweaver.org/mailman/listinfo/callweaver-users
