I know four months have passed since the last post in this thread but I think the information in this topic is valuable and have some observations to add of my own.
I've also been struggling to get reliable operation of a Pi Zero + camera setup as a security camera using RPi Cam Web Interface. The system would only run for part of a day with various crashes/watchdog resets. It was a kernel oops that finally led me to this topic.
I had eliminated PSU problems as a potential cause as the Pi is powered from a 12V->5V 3A switch-mode regulator located right next to the PCB. I added a 220uF low-Z capacitor in parallel with the 5V O/P just for good measure. My scope confirms a 5.05V supply at the DC input with well under 50mV of noise.
While the OP in this topic refers to a Pi Zero W, mine is without on-board WiFi but I also re-located the WiFi USB dongle to the outside of the metal security camera case as I previously had it behind the front window near the camera and wondered if EMI might be an issue.
Despite this the frequent lock-ups persisted. But having seen this topic, I followed the suggestion to increase core voltage and force turbo on and the system has now behaved perfectly
for several days so far. Even if my luck breaks, the changes clearly have a significant bearing on the problem
The over_voltage parameter for the Zero seems to cause a bit of confusion as it is documented
that it defaults to 6. This default aready results in a standard supply of 1.2V + 0.025V * 6 = 1.35V and the maximum the voltage regulator subsytem can provide is 1.4V therefore:
- applies the maximum. Higher values just silently clip to 1.4V
However, I'm not convinced that upping the core voltage is what is making the difference to the camera issues.
As I understand the relevant history here, the CPU fab was enhanced to permit significant over-clocking of the Pi Zero and the default core voltage is increased to be already in place for this. I'm thinking that the instability could instead be due to the dynamic changes in clock speed when turbo mode is not forced on
To test this I've restored the default core voltage and just left:
Could it be that things go wrong with the camera interface when a dynamic clock speed changes takes place? Some more time with this setting in force might tell.