Tracking what I know about undervoltages and power issues

[CamDavidsonPilon]

Some peripherals, like an air pump, will cause very short spikes in voltage that dip below the Rpi's alert threshold of 4.65V. Without an ossciliscope, you probably can't detect these (but the Pi can). We pass this alert to the user.

Does this happen with all Rpi models?

I only know that it has happened with my Rpi 3B, acting as a worker/leader, and a 5V 2.5A PSU.

Solutions

• Buy a larger PSU, something like 5V & 3amp.
• hardware solution? https://electronics.stackexchange.com/questions/264313/capacitor-sizing-for-small-motors-to-reduce-inductive-voltage-spikes
• Use a dedicated PSU for the PWM Amps.
• Hypothesis: use a smaller board: like a 3A+, or ZeroW
• Possible optimizations that seemed to work:
  • echo 0 | sudo tee /sys/class/leds/led1/brightness to remove power LED
  • echo '1-1' |sudo tee /sys/bus/usb/drivers/usb/unbind to remove USB
  • sudo /opt/vc/bin/tvservice -o to remove HDMI display (edit: this already is part of our installation)

[CamDavidsonPilon]

Another thing I've measured is the 5V rail during heavy CPU load. Ex: when I start up a bunch of jobs quickly (i.e. busy), my 5V rail drops to ~4.9V (for example), which also seems to reduce the output from the LED, relating it to #81.