Autumn starts in a few days bringing the promise of cooler days. So back to the 23cm project. I decided, based on what was in the junk box, to use a MAX2681 mixer. Reasonably cheap if you have to buy one I was able to salvage a couple from 900MHz data modems.
One of the unforeseen bonuses with this part is the input matching at 23cm.
The input impedance is 50-j133 ohms. So a 16nH inductor in series with the input pin will cancel the input reactance.
Too easy!
73's
So maybe things are starting to degrade?
The closing times, with the caveat the contact bouncing is not allowed for, look consistent:
Group 1 was using the house power supply (13.8V) where after it was shifted the the shed for two runs (groups 2 - 20). Then, with hot weather, the last run (groups 21-30) was back in the house where the temperature extremes are smoothed out by air conditioning.
Contact resistance also looks consistent though the presence of my phone near the test board when in the house would explain some of those outliers in groups 21-30.
However, opening times appear to have changed. There is a very tight distribution of opening times and an upward trend for groups 21-30. Groups 21-30 was when the R was halved from 1.35ohms to around 0.7 ohms. That means the current on contact closing has increased.
The capacitor internal resistance is probably now a factor and if I was being thorough I would measure the wave-forms with a digital oscilloscope to derive the current at contact close.
Suffice to say it looks to be the first evidence of degradation. It only took some 2.6 million cycles of increasing abuse. Another run "as-is" appears warranted to see if the trend of slower opening times continues. And some measurements of the current at contact close is planned.
73's
Still no deterioration.
We have a heat wave starting today (+35 for 7 days) so conditions are not suitable for leaving the test board running continuously in the workshop.
I'm going to increase the contact closing current by decreasing R from 1.35ohms to something much smaller depending on what I have in the parts bin.
Ok, so another 262,000 cycles. But this time with the closing contact current stressed to 10 Amps. I pulled the data onto the PC and grouped the opening items into 10 divisions (approx. 26,000 observations in each group).
All I saw was a variation in opening times which would correlate well with the shed temperature over the time it took to fill the memory chip:
Just out of idle curiosity I looked at the closing time. Not expecting anything I was surprised to see two things: Firstly, the time taken to close appears to have an inverse relationship to the time taken to open. As the time taken to open increases, presumably with temperature changes, the time taken to close decreases.
The second thing I noticed was that Group 6 had very little variation. I'm unsure why this is the case and contact bounce could be the reason for this.
And finally, the sum of contact resistance readings for each cycle:
Again, support for changing temperature over the duration of the test can be seen.
However, what I'm not seeing, is degradation. This relay has been subject to over 2 million cycles now. I'll run this test fixture through 260,000 cycles once more before I up the ante because I really want to know what failure looks like.
73's
I added the RC network discussed in the previous entry. Everything is working correctly and today I will shift the test board up to the workshop for an extended run, being the balance of the 262100 cycles the 32MB Flash chip holds. So far, after 30,816 cycles, there has been no obvious change.
While running in the house attached to a terminal I can monitor progress. The legend for what is shown below is:
\ Closing time
- ADC voltage reading
/ Opening time
So at a glance I can tell if anything has significantly changed since each symbol is printed in proportion to the reading.
PAGE : 1911 Cycles : 30576\\\\\\\-----////////////
PAGE : 1912 Cycles : 30592\\\\\\\-----////////////
PAGE : 1913 Cycles : 30608\\\\\\\-----////////////
PAGE : 1914 Cycles : 30624\\\\\\\-----////////////
PAGE : 1915 Cycles : 30640\\\\\\\-----////////////
PAGE : 1916 Cycles : 30656\\\\\\\-----////////////
PAGE : 1917 Cycles : 30672\\\\\\\-----////////////
PAGE : 1918 Cycles : 30688\\\\\\\-----////////////
PAGE : 1919 Cycles : 30704\\\\\\\-----////////////
PAGE : 1920 Cycles : 30720\\\\\\\-----////////////
PAGE : 1921 Cycles : 30736\\\\\\\-----////////////
PAGE : 1922 Cycles : 30752\\\\\\\-----////////////
PAGE : 1923 Cycles : 30768\\\\\\\-----////////////
PAGE : 1924 Cycles : 30784\\\\\\\-----////////////
PAGE : 1925 Cycles : 30800\\\\\\\-----////////////
PAGE : 1926 Cycles : 30816\\\\\\\-----////////////
73's
ps If you're not using STM8EForth you are missing out on a powerful tool for such exercises. See the Wiki
Somewhat frustrated that the relay was still clicking away, so mechanical failure had not been reached, I went of to research what electrical failure means. I had assumed it would manifest itself as a sharp rise in contact resistance but another test is contact opening times.
At 20% of the relays switching capacity it appears I was unlikely to reach an electrical failure.
Since I had two opening time observations from early in this exercise, 1820 uS each, and the opening times were now 1910 uS it seemed worthwhile to also log the opening times. Closing times are a low priority since I had no easy way to measure contact bounce with this test board.
In addition, I gave some thought to how I could increase the 2 Amp test current I was using. The next relay I wanted to test had a 2Amp rating so I didn't want to undertake the heavy modifications to the test board a 10 Amp current would require. I elected to place an RC circuit across the test contact to add 8 Amps across the contacts when initially closing.
Since the constant current generator had a no load voltage of 10.5V the mean R should be 1.31 ohms (10.5V / 8Amps). I plan to use two 2.7 ohm resistors in parallel for 1.35 ohms. The capacitor needs to be fully charged in about 400mS.
I am using a constant current generator, not a constant voltage, and a series R. At 2 Amps I lose 2.7V across R, so C charges to 7.8V. Then a constant voltage relationship takes over to finish charging. I will start with a 1000uF capacitor. It=CV so that should charge to 7.8V in 39mS. Then a further 68mS to reach about 10.5V.
Now to consider the extra heat the constant current generator has to dissipate when charging the capacitor. The additional 2 Amps for 39ms plus the tapering current for 68mS is extra heat, but this compares with the 2 Amps being forced through the relay contacts when closed for 50 ms.
I decided to reduce the on time to 21ms. ( Wait 5ms then take 8 readings at 2ms intervals ) A further pause of 1 second each time the buffer had finished writing to the external flash storage would let me disconnect by "ear" should the need arise and help with the heat dissipation.
Might as well store the closing time, even with no de-bounce checking, since I had bytes to spare. I was filling a 256byte buffer with 32 readings of 8 bytes. Now I was filling it with 16 readings of 8 bytes plus 2 bytes for each time measurement.
We're back in a heat wave now with daytime temperatures rising above 37C. So leaving anything running in the shed would have to wait a few days till the next cool spell. I'll make all these mods and see what happens.
In the interim this is where I got to:
My conclusion is there is no degradation in contact resistance after 1.83 million cycles when switching resistive loads at 20% of contact rating.
I'm disappointed there was no failure. Next time!