[Asterisk-Users] Major problems with TDM400 and specifictelephones: suggestions?

[Rich Adamson]

For the record & archives, there are apparently about eight different
revisions of the TDM card (observed via pci id revisions in driver code),
and this mod only impacts one of apparently multiple problems. Since
digium is very tight lipped about problems, there is a high probability
that other issues abound on early versions of the card.

It should also be noted the digium TDM card has a two year warranty,
therefore take advantage of that warranty to address the design problems
if problems continue to impact your system. The warranty will expire on
the initially shipped TDM cards within about twelve months or so.

Rich
------------------------
> I did the modification that Rich explains in his email on March 23rd below. 
> I believe it works for me, because before this mod I was getting "Ouch, part 
> reset..." errors at least once a week, rendering * unsuitable for production 
> systems. After this mod, the system is running flawlessly for almost a month 
> now.
> 
> The closest capacitor value I was able to find was 100nF though, but it 
> seems ok. And I had empty module slots, so I did not have to solder 
> anything, I just inserted the pins firmly to the slot (capacitors usually 
> have long legs). Very simple.
> 
> Now I am quite happy with TDM400, and I recommend Rich's mod to everyone 
> having such problems.
> 
> Thanks Rich...
> 
> ----- Original Message ----- 
> From: "Rich Adamson" <radamson at routers.com>
> To: "Asterisk Users Mailing List - Non-Commercial Discussion" 
> <asterisk-users at lists.digium.com>; <jnovack at stromberg-carlson.org>
> Sent: Wednesday, March 23, 2005 1:32 AM
> Subject: Re: [Asterisk-Users] Major problems with TDM400 and 
> specifictelephones: suggestions?
> 
> 
> >
> >> >I've improved the stability of my card by adding a capacitor on the
> >> >reset line. Hasn't taken a hit in over two weeks.
> >> >
> >> >
> >> Is this the E/F or revised H card? Where and what cap did you install?
> >
> > My card reports as E/F; only have one, so not sure what the differences
> > are between the various revisions.
> >
> > Adding the capacitor seems to have corrected the "my TDM card goes
> > out to lunch about every two weeks, and the only way to correct it
> > is to reload the drivers or reboot the machine" problem.
> >
> > Trying the following will likely void any digium warranty.
> >
> > Remove the TDM card from the PC and look closely at the pins associated
> > with the four fxs/fxo modules. The pins are labeled on the modules as
> > 1, 2, 19, and 20. The "reset" line is pin #2 while ground is pin #20.
> > Carefully solder a .02 ufd capacitor between pin #2 and #20 on one
> > module. Solder it onto a "single" module; no need to add one for each
> > module. Install the card and boot up. Nothing more to it.
> >
> >> Also, when the driver is loaded, my system reports an E/F card, but the
> >> board clearly says "H". Anyone know for sure which is correct.
> >
> > Best guess is the driver reports the E/F based on the pci controller
> > ID, which may or may not have been changed when revisions to the
> > physical card were made. (That guess can be verified by checking
> > the code; I remember seeing it, but don't remember which file.)
> >
> >> As illustrated by the problems this card has with PCI slots. Even some
> >> motherboards which clearly are PCI 2.2 can't see the card in ANY slot.
> >
> > Yes, but the flacky pci bus issue is a motherboard problem that
> > really has nothing to do with the digium card design. (The pci bus
> > issue is fairly well understood by those involved with heavy audio
> > apps. It just so happens to impact how the TDM card is used as well.)
> >
> >> The FXS module is configured as a ground start device to provide dial
> >> tone to an EM switch, as well as an inward path. Multiple FXS modules
> >> would allow multiple connections, and GS is normally used to prevent
> >> GLARE, or head on collisions on the outside chance that several calls in
> >> and out occur at the same time.
> >> Digium support person #1 has stated that GS does not work  on this
> >> module, and support person #2 says " it should work"  In fact it does
> >> provide a GS trunk that works well for outgoing calls. On incoming
> >> calls, the module does not behave properly, in that before ringing
> >> begins, Tip should be grounded, and stay that way throughout ringing and
> >> answer. In fact, Tip seems to float somewhere during the ring cycle, and
> >> providing an external ground causes ring voltage to cease but not trip
> >> ringing.
> >
> > That's kind of weird as the Ground Start pin on the chipset isn't
> > wired to anything whatsoever. The SI chipset apparently supports GS,
> > but the circuit board traces don't. Guess it might be possible to
> > _emulate_ GS through software, but its obvious the emulation isn't
> > the same as the real thing.
> >
> >
> >> Also dial pulse detection seems very narrow, and different dials that
> >> work fine with much other equipment is not so with this card.
> >
> > The dial pulse sensing would be something done in the drivers, so
> > sounds like that routine has the same narrow operating margins
> > the echo canceller has.
> >
> >> >Trying to follow the code path for a functional TDM card is not
> >> >to be taken lightly. Code is scattered across multiple drivers
> >> >and buried in asterisk modules. Even those that consider themselves
> >> >good asterisk developers stay way from this one.
> >> >
> >> >
> >> That doesn't bode well for any corrections, does it.
> >
> > Nope.