Issues with Teensy and Longevity

I'm having some issues with Teensy boards dying, seemingly at random. I'm using the board as an always on temperature controller and have had it running for about a year. In that time, I've gone through 3 boards. First a Teensy 3.0 and then 2 3.1's. I swapped the first two boards out and the third died yesterday. The setup runs wonderfully for months on end and then I'll look at it one day and find that the board is no longer functional (no longer running code and non-responsive over the serial connection). Setup is very simple, 9 output pins connected to a relay board and one pin for 1-wire communication to some DS18B20's. I cut the trace to separate VUSB and the board is powered from a 5V power supply in the enclosure. I'm going to order another couple of boards to get back up and running, but it seems like a board should last more than a few months. Anyone out there have any suggestions for things to check?

Two things,

IF the issue is power related (and it may not be), the 3.2 offers better input voltage protection than past 3.0 and 3.1 models. For example, depending on how the relay boards are set up, they may be sending voltage spikes into the 5V bus as relays unload. One thing I would try is a separate 5V power supply for the relays and the teensy.

Secondly, how are the outputs of the teensy connected to the relays? Some boards feature quality opto-isolators (making the above suggested power-supply separation easier), some may suck more power out of the teensy than is sustainable in the long run. I would have a look see and determine how the relays are powered and whether you may be exceeding the safe output levels that Paul has proposed in the past (9mA, max IIRC but design for 5mA)

I'd suspect noise or RF pickup on the 1-wire signal, and/or back-EMF from the relays.

If you have any way of logging data, you could have the Teesny monitor its own die temperature. Could things be getting toasty in the enclosure?

https://forum.pjrc.com/threads/25143-on-chip-temperature-demo

Is it possible there's a rare sequence of events that's causing the power drawn by the Teensy to spike? Say, normally just one or two relays are energized at a time, but maybe occasionally all of them switch simultaneously?

There is also this to monitor Teensy Voltage: https://forum.pjrc.com/threads/26117-Teensy-3-1-Voltage-sensing-and-low-battery-alert?highlight=getInputVoltage

@pictographer - I pulled that p#4 linked (p#1) code and ran the sketch on my T_3.2 - it was showing a temp warmer than my room 77-80°F. I put my finger on the MCU and the temp went DOWN 3°F or more and is now climbing back up 77°F? I put it in my hot laptop exhaust and it dropped under 70°F - in fact DOWN to 59°F and dropping . . . under 50°F before I moved it - the air is easily 100°F. Out of that air it is now back up to 74+.

Sounds like a calculation error. I used the code as the basis for some code in http://pictographer.com/didah and the values I got were reasonable. You could try posting the code you actually used. Maybe one of us will spot a sign error something.

There were a couple of threads on improving the accuracy of the realtime clock on the Teensy 3 by compensating for temperature using this mechanism. If one is willing to calibrate individual units, it's possible to achieve pretty extreme timekeeping accuracy. I can't imagine that would work if the temperature sensor wasn't functional.

Thanks for all the input everyone! Sounds like I have a few things to try. I'm guessing against heat issues, as this first happened at 45* ambient and there's not much in the enclosure that's producing heat. My relay board just uses transistors to drive the relays, so EMF could be the issue. As could noise on the 1-wire bus, as it's a fairly long run and I have noticed some transmission errors. Would I be safe introducing opto-isolators on all my signal lines (1-wire bus & relays)? Along with a second power supply dedicated to the Teensy?

pictographer said:

Sounds like a calculation error. I used the code as the basis for some code in http://pictographer.com/didah and the values I got were reasonable. You could try posting the code you actually used. Maybe one of us will spot a sign error something.

Click to expand...

I used the code direct from this forum post: https://forum.pjrc.com/threads/25143-on-chip-temperature-demo?p=42329&viewfull=1#post42329

Very difficult to do optoisolation on 1-wire. Seen the schematics in the past, it was involved.

I'd start with a opto-isolated relay board. You may or may not have to adjust the drop down resistors for 3.3v operation ( I had to on mine). Also separate the power supplies for relays and teensy.