[Asterisk-Users] Ouch, part reset, quickly

[Rich Adamson]

> > i've debugged the driver well enough and know that the Ouch message happens
> > when register 0x08 of the module returns >0, which indicates in most times
> > that digital loopback is enabled on the card. this register is set to
> > /disable/ digital loopback upon an init.
> 
> > the power alarm happens when the line feed (hookstate) of the module is not
> > in sync with a driver variable which tracks hookstate. the resetting bit
> > you see is just informational to let you know that the driver is setting
> > the on-module registers back to what the internal variable says it should
> > be.
> 
> Very interesting; thank you for sharing this.  In my experience the card 
> starts to act funny when I get ONE of these -- perhaps in some situations 
> there are more than one register that is going awry and the resetting code 
> doesn't reset them all?  I should hack in some debug code that dumps the 
> registers whenever it detects this "power" alarm.
> 
> I should also grab the datasheets and any erratta for the SLIC chipset and see 
> if anything interesting turns up.  Thanks for giving me a direction to start 
> in.
> 
> > i can explain what the driver does when these things happen, however, i'm
> > thinking that it's more of a hardware issue than anything else. based on my
> > (admittedly limited) reading of the Tiger320 ProSLIC datasheet, the
> > registers mentioned shouldnt go awry, yet they do.
> 
> I thought the TJ320 was a PCI bridge that provided an 8-bit parallel 
> interface, timer and a serial interface or two, and that there was a separate 
> SLIC chipset which did the actual interfacing to the phone line.

FYI, the tdm fxo chip set from Silicon Labs ( www.silabs.com ) uses
the 3019 for handling pstn line-side interface (and electrical isolation)
and the 3050 for PCM encode/decode, impedance, hybrid, near-end echo
cancellation, interrupts, etc.  The *.pdf's are rather hard to find on 
their site but very detailed (3050 has 110 pages). A quick check of SI's
revision history tends to suggest very few anomalies since released in
2003. The very first tdm fxo modules sold by digium used the rev-C 
and rev-D chips.

The Tiger320 handles, as you mentioned, the pci v2.2 bus interfacing.

Given the sophistication of the SI chip set, it would appear that at
least some functionality exists that has not been taken advantage of 
within the wctdm/zaptel drivers, etc. Part of that history is probably 
related to attempted reuse of code that was written for the x100p (in 
multiple * modules and drivers).