Title says most of it. After some extended testing and cycling power to the Arduino connected the Pixy, the camera lost all memory of trained sources. No physical damage to the hardware and when the saved .prm file was uploaded to the camera it was back to business as usual. Any suggestions on how to prevent this?
Hello,
I got the same issue. I made a setup that disconnects the power connection to the pixy with a relay. The delay between each reset is random from 500 ms to 3000 ms. The bug is recognized when the pixy does not react to i2c requests because it was reset to its default parameters. It takes around 1000 power cycles until the parameter data is lost.
I used the firmware from the homepage and build the master branch from github. Both had the issue.
The power supply is well regulated and the pixy is powered through the white power connector.
Right now I am trying to uncomment random things in the firmware to find the issue, but it takes a lot of time to test each iteration.
It would be nice if someone has a hint what I could do better 
The other threads that talked about the issue did not help me very much.
regards Alex
I managed to get the error every time now. The on-time needs to be around 975ms ±25ms and the off-time is not crucial. This needs to be repeated 2 times or more. It is also very important to add enough capacitors (300 uF) in parallel to the input of the power connector of the pixy. This guarantees that the turn-off is not too fast, so the timing does not need to be so precise. It works only with Voltages from 6V to 8V but with different timing it should be possible with higher voltages.
Right now the only fix I can see is hardcoding the parameters into the firmware. But that should not be the solution.
I hope someone on the Dev Team will have a look at this. If you need more information on how to replicate the bug, feel free to contact me.
Alex
Hello Alex,
Are you using the most recent firmware version?
http://cmucam.org/projects/cmucam5/wiki/Latest_release
This is usually caused by a “brown-out” of the power supply that Pixy is receiving. That is, if Pixy receives an undervoltage, 4.0 volts or less. How are you powering your Pixy? Are there batteries involved?
Edward
Hi Edward,
thanks for the reply. I tested the master branch on GitHub and version 2.0.19 from your link.
For testing, I power the pixy with a laboratory power supply at 4 A and 8 V. The “brown-out” is generated by turning the pixy off at the right moment with the relay. I made this setup to replicate the issue I was experiencing occasionally with my battery powered robot. It sometimes loses all its configuration when turning it on and off. This is a big issue for me, because the robot needs to perfom on competitions and losing configuration means losing a game.
Alex