OwenDuffy.net 


Modem reset mod

The problem

I have experience problems with modems 'locking up' in certain situations. The symptoms are that the modem is off hook but no longer passing data, and does not respond to input on the V.24 interface. It not possible to get into command mode, and the modem does not hang up when DTR is dropped (configured with AT&D2). It would seem that the CPU in the modem has got into a loop and is not scanning the V.24 signals and other inputs.

It has not been possible to reset the modem other than by removing the power. This requirement for physical intervention is often inconvenient and expensive in lost availability, travel and time.

The challenge is to find a way to enhance the reliability and availability of the modem.

The following describes a watchdog that monitors the V.24 /DTR and /OH leads (at TTL/CMOS levels) and when conditions indicate that the modem is locked up, it pulls the modem /RESET line low to force reset of the modem (similar to the reset at power-up).

Disclaimer

Such a modification may invalidate modem warranty, and modem type approvals, you may even damage the modem. The changes are external to any analogue signal paths and only serve to disconnect a modem that has failed in the off-hook condition,

The modification was developed and has been successfully deployed on five different modems and has worked as expected.

I do not warrant the modification in any way, you do any such mods entirely at your own risk, even if I supply you with a chip and / or information.

Algorithm

If /DTR has been low for at least 2.5S and goes high for at least 255mS, if /OH is low, it will pull /RESET low (via a 100 ohm resistor) for 20mS.

If you hardwire the /OH lead from the modemrst board to GND, the mod will force hardware reset at the end of every 'call', whereas if you connect it to the /OH line, it will only activate in the failure scenario.

Implementation

The watchdog is implemented in a Microchip 12C508A 8 bit microcontroller (MCU).

The only components required are the 12C508A, a 0.1uF filter capacitor for Vdd, and a 100 ohm resistor in series with the RESET connection.

The MCU uses the internal RC clock running at 4Mhz.

Board

This is a view of the underside of a peice of Veroboard showing the track cuts. Note that the track is cut between holes rather than drilling out the track around a hole. mdmrstboard.jpg (10018 bytes)

Fig 1: Veroboard track detail

View of a finished board. You can see the resistor and capacitor on the left side of the chip. The chip is socketed in this prototype.

The whole assembly is in shrink plastic tube to protect and insulate it.

modemrstassy.jpg (25982 bytes)

Fig 2: Finished assembly

Modem specifics

This section contains some details of connection points for three modem models on which I have tested the mdmrst chip. Whilst I have used other brands of modems, the lock up problem that I have described has been consistently observed on the three modem models below. I would expect that other models and brands built on the same chip set and software base may well exhibit the same problems.

ME071

VSS is available at the cathode of the PWR LED. VDD is available from the end of R24 nearest to the edge of the PCB. /RESET can be picked up from the via between C35 and the chip socket, /DTR is available from the via near pin 22 of U9.

me071mod.jpg (74461 bytes)

Fig 3: ME071 connection points

ME090

me090.jpg (31030 bytes)

Fig 4: ME090 connection points

ME100

me100.jpg (46949 bytes)

Fig 5: ME100 connection points

Conclusion

After two years of operation, the two modems fitted with this modification have not required any manual intervention due to lock-ups, whereas without it such intervention was necessary on average a few times per year.

The mdmrst mod has been successful in significantly improving the availability of the modem ports.


© Copyright: Owen Duffy 1995, 2021. All rights reserved. Disclaimer.