That could be the simplest circuit using discrete OTS parts, and the capacitor in the reset circuit avoids the stuck-low problem. I'd add the diode+resistor in the feedback to speed up the discharge phase in case it has to repeat.
Peter, about the PUT. I had an article in a IEEE Circuits & Systems special issue on chaotic dynamics, with experimental verification using driven PUTs, an integrate and fire model. Prof. Leon Chua was editor of that issue, and he had a keen interest also in synthesis of 4-layer semiconductors using bipolar transistors, per jmg's comment. Also i had an article in Physica D, using cross-coupled PUTs as a number-theoretic model of coupled neurons . That was long ago when I was a grad student, before I forgot everything. And I guess it makes me a passé world expert on PUTs.
Also, thinking of tunnel diodes, i still have one of the original HealthKit tunnel-dippers.
Here is the really final circuit with only 6 components!
I changed the transistor to PNP type to eliminate the discharge diode.
The base capacitor was removed to enable the micro to set the timeout period and
also have the option to temporarily disable the watchdog. ( at the small risk of a lock-out )
Here's the prototype board which i plan to add the watchdog circuit.
There are still a few issues here and there, for example
a. I still don't know how to mount the speaker properly to the board ( now using super glue )
b. yellow audio cable looks out of place
William, that looks like a clever adaptation to me, using the transistor for dual function to reset via either rst\ or io\ . The base-emitter junction has very low leakage current, better than a typical 1N4148, and the 3.3V power supply is well below the zener threshold. One thing you might try would be to connect the transistor in inverted mode, that is, switch the collector and emitter leads. The base-collector junction is also low-leakage, not quite as good as base-emitter, but what you gain is a lower saturation voltage, so the capacitor can discharge to a few millivolts (given enough time and enough drive current) instead of bottoming out at some 600 millivolts. Another note: If the circuit does get locked in reset, the period will be a lot shorter than it will be when the capacitor starts at zero volts. The hysteresis band of the 1G14 is only about 0.75 volt . That could be a "feature".
Watchdog timers can be a real pain though. Especially during development--It has to be turned off!
Comments
Also, thinking of tunnel diodes, i still have one of the original HealthKit tunnel-dippers.
I changed the transistor to PNP type to eliminate the discharge diode.
The base capacitor was removed to enable the micro to set the timeout period and
also have the option to temporarily disable the watchdog. ( at the small risk of a lock-out )
There are still a few issues here and there, for example
a. I still don't know how to mount the speaker properly to the board ( now using super glue )
b. yellow audio cable looks out of place
Why is 90% of copper etch away?, pour should be used.
I still like to see where the tracks run. Pour is used on on the bottom layer.
Watchdog timers can be a real pain though. Especially during development--It has to be turned off!