No degradation. No failure. No end in sight.
If nothing else the last run of ~520k cycles confirmed that putting the whole test board in a more stable RF environment reduce the number of outliers. (Groups 17-27 below)
Excel couldn't handle this many observations so I switched to SAS. I was a bit rusty but I got there.
Any guesses at how many cycles before the relay fails? Rated mechanical life is 10 million cycles. I'll gather another ~520k cycles of data and update in due course.
73's
Well I still feel a bit disappointed. No dramatic deterioration after 780,000 cycles.
At least I have worked out how to get Box and Whisker plots out of Excel to help assess changes over time. Trying to do anything with Excel for so many observations is about as slow as the time it takes to transfer the contents of the 32 Megabit flash at 9600 baud. Brings back fond memories of SAS on IBM mainframes. But I digress!
I divided the 782,224 observations into 12 groups. The first 4 groups are from 16 observations per cycle. The last 8 groups are from 8 observations per cycle, the sum of each 8 observations then scaled by 2.
The shift in the readings from group 5 onward is related to ambient temperature. Up until then we were in a strong heatwave. Thankfully it subsided and ambient temperatures dropped some 15-20 degrees. In addition, the air was circulating faster over the heat-sink of the constant current generator because the window was open and an ex tropical cyclone was responsible for a steady stream of air through the window.
Group 15 shows what happens when the phone is not in the office and there air is still. Still some outliers but I can't rule out family members walking past the office with a phone in their pocket.
With a few days of mild weather expected I'm moving the board up to the workshop. No RF, no strong air currents. Hopefully with those potential sources of error removed I will see a large deterioration in the contact resistance. At present the voltage drop across the relay pins gives and estimated contact resistance of 10 milli-ohms.
So I should hit 1.3 million cycles in a few days. Fingers crossed something wears out.
73's
Well I'm feeling disappointed. I had thought that by now some evidence of degradation would be visible in the contact resistance. But after operating the relay 262,000 times I don't see any strong evidence the end of life is near. The ADC reading for the voltage across the closed contact has been quite consistent over time. The variability could be due to fluctuations in ambient temperature and RF interference from cellular and wifi sources.
I'm going to modify the code so that 8 readings of the contact resistance are taken, instead of the present 16. That will allow me to record twice as many cycles before filling up the 32mb memory chip. It will also reduce the "on" time each second which will reduce the temperature of the LM317 used in the 2 Amp constant current generator. I suspect the LM317 gets hot enough that the temperature protection is causing the current generator to fold back to a lower figure.
But I am starting to wonder if switching 20% of the rated current means the life is governed by mechanical, not electrical, failure. That would be over a million cycles.
73's