[Asterisk-Users] T.38 fax summary

[Steve Underwood]

Mark,

In the time it took to write all that you could probably have read up 
enough about T.38 to realise you were talking complete rubbish :-)

Regards,
Steve


Mark Eissler wrote:

>
> On Feb 25, 2005, at 10:20 AM, Lee Howard wrote:
>
>> In a traditional analog fax you have modulated audio data, that is, 
>> the data stream is converted into an audio representation by the 
>> transmitter, and the receiver demodulates the audio stream to produce 
>> the data stream.  A lot of data gets packed into very small portions 
>> of audio, which is why fax over VoIP (T.38 is not VoIP, it is FoIP) 
>> is unreliable - any jitter will likely cause data loss.
>>
>> There are no modulators in T.38.  So take the fax procedure, but 
>> instead remove the data modulation/demodulation part.  T.38 devices 
>> communicate raw data through the IP network, and the IP network is as 
>> good at communicating data as the PSTN is as good at communicating 
>> audio.  So if you could have a full T.38 delivery route from fax 
>> sender to fax receiver, the data never once gets converted into an 
>> audio signal - it doesn't need to be.
>>
>
> Sort of...but no. Fax requires a codec that supports the frequency 
> spectrum of a POTS audio channel. Currently, that means that anything 
> other than g.711 won't work since the other popular codecs achieve 
> their efficiency by dumping frequencies humans can't hear (just like 
> mp3). The problem isn't typically g.711 because that's the codec that 
> is generally used by the digital telco world. A common problem when 
> discussing g.711 often is packet size vs bandwidth limitations. T.38 
> can alleviate this problem because it doesn't rely on a codec.
>
> The bigger problem with faxing over VOIP is related to lost packets 
> and timing issues (jitter). Lost packets are the death knell for fax 
> because it isn't very tolerant of missing data. How do you complete an 
> image with missing data??? AFAIK T.38 can't do anything to recover 
> from packet loss...the fax machine needs to be tolerant of it. 
> Ironically, ECM was introduced to recover from information loss when 
> transmitting faxes over analog lines but ECM can actually cause 
> problems when used with T.38. If you can turn ECM off that's the best 
> thing to do when using T.38. Besides lost packets though if you have 
> to consider packets arriving at weird timing intervals (jitter).
>
> The fax machine needs to get its data in a steady stream. This is 
> supposed to be a realtime transmission after all. While T.38 can 
> absorb some of the problems triggered by latency and jitter, when the 
> problem becomes too excessive it tanks just as quickly as faxing 
> without T.38.
>
> So with those barriers out of the way what is it that T.38 tries to 
> accomplish? Instead of sending a fax over VOIP as a stream of sampled 
> audio, the protocol intercepts the audio at the endpoints and 
> packetizes it as blocks of data instead. The receiving gateway must 
> know how to handle the data stream so it can convert the fax back into 
> a T.30 fax data stream for POTS. During the session, progress is faked 
> so that the two fax machines don't think the transmission has 
> stopped...that's a crucial step because it takes time to convert and 
> send/receive the fax reliably.
>
> I think the best arsenal for faxing over VOIP today is to have a good 
> broadband connection, g.711, and a fax machine where YOU can set the 
> max transmission speed. Sadly, the last part seems to be missing quite 
> often. I've noticed that HP actually mentions faxing over VOIP in the 
> documentation for their 7410 all in one machine and, more importantly, 
> they include support for changing transmission speeds. Way to go HP!