[asterisk-users] spandsp (foip)

[Steve Underwood]

Lee Howard wrote:

> On Sun, Sep 24, 2006 at 01:58:21PM -0700, Christopher Corn wrote:
>
>>   A couple of faxing methods im confused about.
>>
>>   The pass through method, sending fax data over G711 codec
>>   versus
>>   Relay method, t30 to t38 conversion
>>
>>
>>   Can someone explain to me why the pass through method doesn't
>>   require t30 to t38 conversion ( or does it do it?)? i believe
>>   the conversion to t38 is so that it can be routed through a
>>   packet network and then back to t30 so that the fax machine can
>>   understand. why is it that if you use a pass through method, and
>>   your still passing through a packet network, you dont need to
>>   convert to t38 and t30?
>>
>
> Be careful about your wording.  People here generally refer to "pass 
> through" as T.38 pass-through and not G.711 pass-through.
>
> I think that if you understood how faxing works you would see that 
> your questions here don't really make sense.
>
> In traditional PSTN faxing you have a total of two endpoints 
> performing T.30 protocol.  In a simplified form, the sender takes 
> scanner image data and modulates it (into an audio waveform) and then 
> passes that audio over the PSTN to the receiver which demodulates it 
> (takes the audio and turns it into data again).  As long as the 
> demodulated data is identical to the original data, then everything 
> should be okay... for the most part.  However, if you start to 
> consider audio corruption on the PSTN, then that's where difficulties 
> start to ensue.  If you have some audio, modulated data, and then you 
> compress it or fracture it or otherwise corrupt it, then there's no 
> possible way that the demodulator is going to be able to come up with 
> the original data.
>
> Now introduce VoIP telephony... where a small amount of audio 
> corruption (jitter) is anticipated on the UDP channel... and mix it 
> with faxing and hopefully you can see how it just doesn't work well.  
> VoIP is packetized audio passed over an IP network.  Packetized audio 
> is nothing new.  ISDN circuits have had it for a long time now.  Those 
> circuits are digital - meaning the audio waveform is digitized at 8000 
> Hz... so the audio is represented with bytes and are packetized into 
> frames.  Those traditional digital circuits are designed to prevent 
> any loss of that data.  VoIP works similarly, except that the medium 
> is lossy UDP/IP networking.

ISDN doesn't packetize voice. ISDN is a strict circuit switched TDM system.

> Since VoIP works on *IP* networks, and since IP networks already 
> handle data communication very well, there really is no reason to 
> perform the modulation or the demodulation - just send the raw data 
> through.  So that's basically the punchline of T.38... it's fax 
> protocol without the traditional modems involved.  Then you have FoIP.
>
> However, these days the world is a hybrid of VoIP and PSTN 
> environments (mostly PSTN still), and thus anyone using T.38 will need 
> to have a "gateway" involved somewhere along the call path that can do 
> that traditional modulation/demodulation.  That is what the T.38 
> gateway is.  If a T.38 relay does not act as a gateway (i.e. no 
> modulators) then it performs only T.38 pass-through - meaning it only 
> is useful for situations where calls are end-to-end T.38 or where an 
> external FoIP service provider is used.
>
> Because of the way things work T.38 gateways will not only need to 
> have traditional modems (hard or soft) but will also need to perform 
> T.30.  So when faxing with T.38 and the call is not end-to-end T.38 
> then you have at least three points along the call path performing 
> T.30 (versus the traditional scenario of just two).
>
> So, to answer your questions...
>
> Why does using G.711 not require T.38?  Because from the viewpoint 
> that the question was given, G.711 and T.38 are competing approaches.  
> T.38 was designed to replace G.711.  You can packetize G.711 audio 
> just fine without converting it to anything else.  So when faxing with 
> G.711 T.38 is not involved because its basically mimicking the 
> old-style traditional PSTN faxing, except that the audio is passing 
> over a different (less-reliable) medium.
>
> So the reason that T.38 exists is because UDP/IP is lossy and is not 
> therefore reliable for the purposes of faxing with G.711 unless the 
> communication can be guaranteed to be nearly lossless.  For those that 
> work on lossy channels, G.711 will just not work reliably.

Lossless channels are only a part of it. If you look at 
http://www.soft-switch.org/foip-with-real-atas.html you will see 
examples of other problems that happen with a wide range of ATAs. Once 
that have FAX support modes, yet cann't possibly ever work with FAX.

Steve