[Asterisk-Users] T.38 fax summary

[Lee Howard]

Hello Steve.

It's an excellent read.

In two places you mention that V.34-Fax is 28,800 bps.  Actually, 
V.34-Fax has speeds ranging from 2400 baud to 33600 baud all using 
V.34.  And, while most V.34 connections are going to not probably be 
more than 28,800 bps, I have seen sustained analog V.34 fax connections 
at 33600 baud.

I think that it may be fair to mention ECM protocol a bit, in the 
fax-over-VoIP sections.  In the example you give of 20 ms of audio 
missing from the image data portion of the call, *if* the receiving 
modem doesn't declare end-of-page and actually is able to ride-over the 
audio loss, then yes, if the fax session used 2D-MR or 1D-MH 
compression and did not negotiate ECM protocol, then there will likely 
just appear a corrupt horizontal stripe or a few more.  However, if 
2D-MMR image compression is used, no image data corruption can be 
tolerated at all, or the image will be truncated at the point of 
corruption.  This is why usage of 2D-MMR image data compression 
requires usage of the ECM protocol, which will, in effect, cause the 
receiver to request that the sender retransmit the corrupted image data 
portions.  ECM protocol tends to make faxing a bit more resiliant, 
anyway, in practice, because the ECM protocol has methods of "flow 
control" that the non-ECM fax protocol lacked, which allow both the 
sender and the receiver to essentially pause the session at key moments 
when a slow-printing or slow-decoding or slow-framing machine may need 
extra time.  Consequently, most newer fax machines are going to use ECM 
with all image compression methods, including those that do not require 
it (MH and MR).  Plus, I think that it is attractive for fax 
manufacturers to limit copy-quality complaints by standardizing on ECM 
usage.

Sometimes I tend to believe that ECM is what's making fax-over-VoIP 
work at all "reliable" for a lot of people using it (granted, some can 
do fax-over-VoIP reliably because they have an IP network path with low 
jitter/loss/latency/etc).  I think that ECM will "mask" the real 
trouble of the "line conditions" from the users unless they actually 
are able to go in and look at a log of the fax session or if they 
analyze the time that is required to send/receive faxes over VoIP.  Of 
course, this is good I guess, to make faxing work sometimes over VoIP - 
however, I think that the downsides are that it hides the truth of the 
"line condition" problem from the end users, and it significantly 
increases the time required to communicate fax images.  So there may be 
no savings whatsoever in faxing-over-VoIP if you are paying $0.02 per 
minute but it takes you 3 times as long to complete the call.

I'v had some experience working with newer Panasonic Panafax fax 
machines over the last four years.  These machines do have an RJ-45 
jack in them, and do connect to a TCP/IP network.  They do have a 
feature that they call "internet fax", although it is actually neither 
T.37 or T.38.  It is closer to T.37, as it pretty much works like a 
scan-to-email system.  I think that there are some HP systems that do 
something similar.  So just because a fax machine has an RJ-45 jack and 
says something about "internet fax" in the features does not 
necessarily mean that T.38/T.37 is being used... or even just an 
adulterated version of them, either.

I concur with your statement, "T.37 is the sane way to handle FAX at 
this time."  My personal opinion for those businesses that are using 
VoIP exclusively and yet still need to support fax is to do whatever 
can be done (T.37 comes to mind) to get all of the faxes from all of 
the business to a common fax gateway system and then provide that 
system with the requisite PSTN capacity to push out the peak fax load.  
I also think that those businesses should be using DID for receiving 
faxes on that common gateway and routing them internally.  If DID isn't 
an option, then a trusted "office receptionist" can be used instead.  
Now, doing this is going to require additional hardware, most likely, 
in the form of TDM ports and perhaps some additional wiring, but I 
think that it's the necessary evil of the time for businesses 
transitioning into VoIP before FoIP is better supported.

For my clients I use HylaFAX, of course, to serve as that "gateway" fax 
device, but I suspect that spandsp+Asterisk can be used in that 
capacity with perhaps about thesame amount of "glue" work.

Anyway, it was a good read.

Thanks,

Lee.


On 2005.02.27 00:14 Steve Underwood wrote:
> Hi,
> 
> Questions keep comming up about this, so I started writing something 
> at http://www.soft-switch.org/foip.html . I think I covered the FAX 
> over VoIP issues fairly completely. T.37 is pretty simple to explain. 
> There is rather more to say about T.38, but at least this is a start. 
> If anyone wants to suggest corrections or additions, just blurt them 
> out.
> 
> Regards,
> Steve