FWIW we had the following issue: even before we went digital pressing the PTT button influenced the oil pressure reading. Especially COM2 sent it to zero. After installing the MVP 50 the shop reported that transmitting drove the device nuts. We never understood what the problem really was, but an improved ground wire to the engine did the trick. The shop also added a few ferrite ring chokes but that didn’t solve the root cause. In way over thousand hours the problem reappeared exactly once, shortly after installation and never again. After pressing PTT all analogue inputs of the MVP went crazy, the digital ones kept working. Resetting the unit resolved the problem and it never occurred again. My guess was therefore that somehow the AD or maybe the circuitry that provides the reference voltage got jammed. Maybe you want to check if your problem is somehow related to pressing PTT or any other ‚electrical event‘.
The TB20 had this problem. The oil pressure dropped to very low when transmitting on some VHF frequencies. A specific DME channel also triggered problems. The fix was a ferrite DB connector filter. But the effect was on analog circuitry, which recovered immediately upon cessation of the interference.
I don’t think this is the same thing though. It is probably some VHF/UHF radiation/conduction, and it may be triggering spurious CPU interrupts at a high speed. That is a vulnerability in many badly designed digital systems. Another form is where a misbehaving device on an ETH LAN is generating multicast packets at MHz speeds, which cause RX interrupts in every device attached 
It happens because all the smart arse devs like to have interrupt-driven RX on ETH.
Most designers today are not good with analog, and knocking up an instrument like this and getting it working on the bench is basically easy. It is quite a fun project, with a multichannel ADC and a big LCD with a graphics library for $500.
Peter wrote:
I don’t think this is the same thing though. It is probably some VHF/UHF radiation/conduction, and it may be triggering spurious CPU interrupts at a high speed.
I scratch my head about where interference could be coming from, though; given that the problem persists when every single CB in the airplane is pulled apart from the JPI.
Start to wonder what kind of clothing you guys are wearing…
Next step – naked diagnostics.
Peter wrote:
I don’t think this is the same thing though. It is probably some VHF/UHF radiation/conduction, and it may be triggering spurious CPU interrupts at a high speed.
I don’t think it was an ‘analogue’ problem in my case either. However, at least the main CPU kept on working. The failure wasn’t transient though, it persisted once it had occurred. Therefore I’m pretty sure a ‘digital’ part of the system was put into a state it couldn’t get out.
The thermocouples do not measure absolute temperature but the difference between the hot-junction (inside the sensor) and cold junction (inside the instrument) . You need some absolute reference inside the instrument in order to add to thermocouple voltage and obtain measured temp. Ref temp measurement inside the instrument thus affects all thermocouples with cold junctions within the same box equally As the instrument warms up the delta will vary. If it happens with engine off as the instrument warms up ( you can also warm it up carefully with a hair dryer, not a heat gun) then you know the problem is within the instrument wherever the “cold junction” is ( for JPI I think it is inside the display, but others have a box external to the display). If you tried two displays then either there is an external junction box you need to replace or some wiring issue (typically bad- or undue-ground or short) is wreaking havoc with your instrument’s reference temp.
When measuring probe wires make sure the connector on the instrument- or junction-box side is disconnected. Make sure to measure vs ground but also vs other wires in the vicinity.
That is very helpful Antonio, thank you; it gives me a much better understanding of how this system works.
A problem with the cold junction compensation would not subtract 360F from the reading, unless the cold junction was elevated by 360F.
I have designed many thermocouple (and PT100 etc) instruments. CJC is normally implemented by locating the CJ sensor at the terminal block at which the thermocouple wire is terminated. Typically it is a cheap silicon diode, but nowadays people place the ADC right next to the connector, and modern ADCs contin a temp sensor which is handy for this. I’ve just done a product with an ADS1118.
If it does it with literally all other equipment turned off then it must be a faulty instrument. Embarrassing But it is not unknown in avionics for mfgs to send out a sequence of boxes each of which contains some fault. The biggest names have done it…
What happens if you power up the instrument on the bench, with some sensors connected?
I would also just lean towards two faulty boxes. If one wanted to get esoteric one could fantasize about something installation-specific that destroys the boxes after x hours but Occam‘s razor says „bad quality“.
