Effective use of output signal of a watchdog timer

Question

Subsequent to my previous question regarding watchdog timers, I am interested in deciding the RESULT of such a watchdog:

Assume I add a generic external watchdog timer to my circuit. If not toggled by my microcontroller within the watchdog timeout period, the watchdog will assert a Reset pin LOW for a few-millisecond period, then deasserts the signal.

Now, with this watchdog output signal, I have two options:

• Option A: I could apply it to the Reset pin of microcontroller
• Option B: I could connect it to the Enable-pin of my Voltage regulator which, in this case, thus resets the power for my entire circuit

My question: Are there any possible concerns with Option B? I have never seen a watchdog timer / reset-IC connected to a regulator in this fashion, whether in a watchdog timer datasheet or elsewhere. However, I find it desirable because that way, I would be able to "restart" all parts/ICs on my board, instead of just the microcontroller.

Answer 1

Using an external watchdog circuit to cycle the power to all your on-board circuity is a rather good idea. It helps to cover cases where some stray external event has caused an upset on some silicon chip circuit. Sometimes these events, whether they be high voltage spikes, reverse bias on signal inputs or even alpha particles can cause localized latchup on-board the chip and the only realistic way to clear such fault is to cycle the power.

Using the scheme to connect the WDT to the enable of the voltage regulator does require some due diligence on your part.

a) Make sure the WDT circuit is not powered by the same voltage rails as the disabled regulator.

b) Make sure that the length of time that the regulator is disabled is long enough. Bypass and filter capacitors on the output of the regulator need to discharge to the greatest degree possible in order to qualify as a power cycle to the silicon. In many instances a millisecond or so of disable time will only allow a partial discharge of the output rails.

Answer 2

With option B, you'll need a separate supply for the watchdog. Plus, interrupting power to the entire circuit is drastic step that will make other things (such as debugging) that much more difficult.

It would be better to design the system so that the microcontroller has the ability to reset/restart all of the other functional blocks under firmware control, either individually or as a group. Then the watchdog's only duty is to make sure the microcontroller itself is operating properly, which means that option A is the proper way to use it. Whenever the microcontroller gets reset, it then resets everything else as needed.

Answer 3

I read the people in favor of A and B solution. I would tell my story. My regulator LT1962 and the uProcesso ARM7) are exposed to RF signals (near two way radios antenna). In many cases, the regulator was tack without providing the 3.2 V output. No way to reset the uP because no power available. I connected a external WDT TPS3823 and when the uP always looking for the WDI oscillator. From this moment, the Regulator never was stack again.

OK: Looks that I was not clear in my answer. I have a Voltage Regulator LT1962 that operates in high level of Radio Frequency (RF) supplying voltage to a uP ARM7 from TI. Due to RF, the regulator was stack (was in flotaing state providing 0 V). The uP was OFF and I was unable to reset the regulator. I added a external WDT TPS3823 to reset the regulator when the uP was unable to provide the oscilating signal to the WDT. From this moment, I solved the problem of no power output when RF is strong.