[asterisk-users] PRI debugging ...

[Matthew Fredrickson]

Gordon Henderson wrote:
> On Fri, 16 May 2008, Gordon Henderson wrote:
> 
>> Have a problem with an ISDN30 line in the UK.
> 
> So following up my own post.. I've not solved this issue, but I think I 
> know what causes it.
> 
> This was my experiment to put 2 cards in one 1.3GHz system - a TDM400 with 
> 2 x FXO and 2 x FSX and a TE120P - E1 card.
> 
> The PRI card loses interrupts, so I'm guessing it loses a frame of data 
> when it loses an interrupt, and eventually it gives up and does a reset. 
> The TDM card was rock solid. The system is using oslec too FWIW.
> 
> When I unloaded the wctdm module the PRI performend flawlessly.
> 
> So I'm suspecting the 1.3GHz processor and underlying IO is marginal for 
> this application. The Mobo doesn't have an APIC, just old PIC hardware, 
> although both cards were on separate IRQs - the TDM card had the higher 
> priority IRQ though - didn't have time to test it with the cards swapped 
> over, but loading the modules in a differnt order didn't make any 
> difference. Turning off the USB hardware didn't help either.
> 
> The processor does seem to have a highish high-priority interrupt load (as 
> seen by top). I'll be trying a newer kernel when I get a chance though 
> (this is 2.6.18, compiled to match the motherboard exactly)
> 
> Making calls through the TDM card just made it worse.
> 
> However when it was working, it was working very well indeed, but the 
> occasional time when it dropped all calls (about once an hour) wasn't 
> good.

You might try turning off echo cancellation to see if your D-channel 
performance improves.  That would be a good test to tell if you should 
look into perhaps getting either a faster CPU or a hardware echo 
canceller.  It's possible that you may be saturating your poor 1.3 Ghz 
CPU by doing echo cancellation for too many channels on it.

-- 
Matthew Fredrickson
Software/Firmware Engineer
Digium, Inc.