We are running FreePBX 13.0.197.21 and Asterisk 13.30.0 with spandsp 0.0.6, IAXmodem, Hylafax+ 7.0.0 and Faxpro 13.0.44 on CentOS 6. Long time running machine that does well on iaxmodem/hylafax.
The client wants to eliminate POTS and so added fax boards to their Canon Imagerunner MFP. Setup seems easy enough and it will send a few pages fine but when it comes to 30 page documents, it’s rare to get past 20. I’ve checked the bandwidth, latency and jitter which all well within specs required. Things are so bad I installed a SPA112 and while it does better, it trains down to 12000 baud before it reliably sends data. note the canon fax board is pretty bad and trains down to 4800 most of the time.
I’ve changed the error correction from redundant, fec, and adjusted maxdatagrams trying to find a difference and/or sweet spot, no luck. The vendor got a Canon tech online and proceeded to spend 8 hours “testing” faxes in and out. They found the same thing I’ve been telling them all along. The SPA appears to work better than the Canon Fax Board. Big help!
The last test I performed was to connect an SPA112 to the switch the FreePBX box is running on and viola, it ran perfect with 14400 connects and fast page transfers. So it appears the switch network is faulty somewhere, somehow. I don’t manage that part but they tell me they have prioritized for VoIP… it seems pretty obvious to me.
My question is when you have a “perfect” path, latency under 1ms, zero jitter and 124Mb bandwidth (iperf), can the switch interfere that much? There are four hops between the FreePBX and the problem printer.
When you run a fax through the spa112 does it actually show a t.38 call or is it doing g711? Whats the t.38 redudancy level set to on the spa112. As well as the jitter buffer adjust and level settings.
What are your ping and jitter levels. As well as like a max ping vs min. Try running WinMtr for 20 minutes straight to replicate a 30 page fax.
4 hops? Meaning 4 switches or do you mean routers?
Honestly I run spa’s all the time over public Internet without issues.
fax show sessions shows the call is in gateway mode:
Channel Tech FAXID Type Operation State File(s)
SIP/242-0000022c Spandsp 135 none gateway Active
Here are the changes for the SPA
Network Jitter Level: Very High
Call Waiting Serv: No
Three Way Conf Serv: No
Three Way Call Serv: No
Prefered Codec: G711u
FAX Enable T38: Yes
FAX T38 Redundancy: 1
FAX Passtru method: ReINVITE
Silence Supp: Disable
Echo Cancel: Disable
FAX Enable T38: Yes
FAX Tone Detect Mode: caller or callee
Here’s the output from mtr run for 20 minutes in report mode…
Hop 1 is a Cisco switch, hop 2 is a centos 5 linux router, hop3 is a pfsense box and hop 4 is the spa112.
dicko, we don’t have a provider. just trying to get these canon t38 fax boards to talk. The SPA was installed to test a known good device for comparison. It works better but still not reliable.
Any type of packet loss like that will screw up a G711 call. If your udptl ports arent open through the firewall t38 will not work only g711. What speed is the fax machine set to? T38 only works with 14400 and lower.
Yes, the packet loss is a problem and I wondered if t.38 was working at all. In the logs I see T.30 packets followed by T.38 but as this is all new to me I’m not sure what I’m looking for. I’d expect the fax stat to show t.38 calls yet only g.711 is listed. The funny thing is BOTH the canon fax board and the ata do the same thing, no t.38 session is listed in fax stat.
udptl_custom.conf is setup with udptlstart=4000 and udptlend=4999. No firewall in the way