In the past two days, I’ve lost two imp001s due to an assumed voltage spike. At this site, we’re using a Kele (kele.com) DCP-1.5-W power supply to convert a readily available 24VAC to 5VDC. Our device had been in service for 3 months without issue. The same configuration is in service on 2 other similar pieces of equipment without issue.
When our device went offline yesterday, I checked the incoming voltage and discovered 5.2VDC to the April board. With the imp001 installed, instead of 3.3VDC, I’m seeing 1.2VDC. Without it installed, I’m seeing 3.3VDC. Since the replacement device went offline last night, I believe this is the case now with both imp001 cards. I’ll be checking on the second one when the rain stops.
My assumption is that the voltage regulator on the DCP-1.5-W is either failing or allowing a spike from its supply to pass through as a >16VDC supply to the April card (above the buck regulator’s capability), thereby frying the imp001. Any thoughts?
I’m wondering if putting a 5.1V zener (Sparkfun 1N4733A) across the output of the power supply (reversed, of course) would make sense as long as the power supply is adjusted to be something like 4.9VDC). Again, thoughts?
Also, would EI have any interest in seeing the dead imp001 to do a post mortem?
At $30/imp001, I can’t afford to have more fail, so I’ll be considering different sources of power for these applications.
The problem with using a zener there is that they tend to destroy themselves when they clamp at high currents.
You could try using a varistor (eg at 12v), which are better for dealing with spikes. You can get some pretty meaty ones.
If the board is showing 3.3v with no imp in there, then I’d tend to assume the PSU is ok; have you tried another imp001 in there? Is it possible the imp saw a transient on an I/O pin vs in the power input? That could fry the imp and cause it to draw huge amounts of current which would explain the low voltage. Is the imp getting unusually warm?
The failed imp001 does get warm (I don’t notice a working imp001 getting warm at all during service).
The field wiring is simply two thermistors and 4 “dry contact” devices (pressure switches and current switches). Unless there’s a “new” induced voltage somewhere, a transient on the I/O would have come out of nowhere. Not saying this isn’t possible, but it is unlikely.
I’ll likely try a more standard form of power supply and see if I lose another. That would certainly point to field wiring.
How long are the wires to these sensors? You can get decent induced spikes on wiring, especially if it runs next to power wires that might have large current changes on them (eg a compressor turning on & off).
Some in-line resistance by the imp and a zener clamp would definitely help to improve robustness.
