In a university chemistry department, one works on a wide range of equipment, ranging from technologically recent equipment to old equipment, older equipment, and finally the "designers-have-been-grave-stoned for twenty years" equipment. One day a call came in from the Mass-Spec Lab: A quadrupole mass spectrometer, made in the UK, was alarming in standby mode but not in operate mode.
Fortunately, we had a set of schematics for the instrument, but the folks across the pond draw things a little differently. Fortunately, deciphering the drawings didn't take long using my super secret decoder ring (a.k.a., paper, pencil, and pencil sharpener).
The alarm was coming from a relay board that distributed 24Vdc power to three off-board circuits. On this particular machine, power for each circuit goes through a dedicated set of two (one operate and one interlock) SPDT relays rated to 5A. Both relays are energized to pass power in the operate mode and de-energized in standby mode. Any other relay combinations sounds an alarm. The NC contacts on each relay go to alarm logic, where they are XORed to each other. When the "operate" relay is de-energized, 24V (applied to the COM) passes through a 100kΩ to the NO contact, providing 24V low current signal to the second relay's COM. This is needed at the second relay so when it is de-energized there is a "high" logic signal to the alarm logic through the NC contact.
A visual inspection while switching "standby" to "operate" showed that all of the relays were toggling. Using a low-power binocular microscope, the relay contacts where inspected. As expected, the NO contacts showed pitting for current switching. The NC contacts appeared clean and pristine, also as expected since they switched very low current.
Measurements with an ohmmeter showed some relays having NC contact resistance ranging from half an ohm to tens of ohms on repetitive relay cycling using a bench power supply. One relay randomly measured higher resistance to megaohms. But continued cycling slowly reduced the resistance. After ordering an identical relay and replacing it, the alarm issue went away.
But a few months later the problem was back. This time another relay had the resistance issue. We ordered several identical relays. But the real quandary began after we installed one of the new relays. The alarm issue persisted and was traced back to the new relay. Several of these new relays were tested and a few showed the same contact issue. A "good" relay was selected and installed, getting the instrument back in operation. But what was going on?
After some hypothesis, analysis, meditation, perusing of datasheets on different kinds of relays, a headache, shot o'wiskey, dinner, more shots o'whiskey, sleep, hangover, breakfast, I began to formulate a theory. The relay uses silver contact material, which is correct for switching moderate power currents. But the COM-NC circuit is a microamp (small-signal) circuit. Small-signal relay contacts always use gold plating. Why? Because gold doesn't oxidize.
Silver is more durable and withstands arcing better than gold, but silver oxidizes. The arcing cuts through this oxidation. Microamp current at 24V does not arc and, therefore, does not clear the oxidation. Over time the oxidation builds up, causing contact resistance problems. Obviously a design goof or oversight by engineers on the big island. (This might be one of the reasons why in 1620 a small group of them left and started a colony called Plymouth.)
So what to do? One solution was to add a load resistor to the NC side to pull more current, thereby causing arcing to clear the oxidation. But this would not work on the second relay due to the 100k resistor. What we needed was a relay with silver on the COM-NO side and gold on the COM-NC side and also fit in place of the original relay. Right!
What else? Re-designing the circuit was not an option due to down-time and budget constraints. Hum! What followed almost defies commonsense, but we are academia and the rules do not apply. Take a relay, remove the cover then disassemble (see picture). Go to the junk drawer and get an edge-card connector with gold contacts and remove a contact. Un-curl it. Use a rounded-end punch and piece of hardwood to make two bowl-like dimples. Cut out the dimpled areas forming small pieces for soldering over the NC contacts. File the NC contacts down to make room for the gold contacts. Tin the NC contacts.
Using the old-school silicon heat-sink compound and wooden tooth-pick, carefully put heat-sink on the convex side of the dimpled contact. This will keep solder from flowing onto that surface (Ssshhh! This is a secret technique so don't tell anyone.) The heat-sink will also hold the small contact to the toothpick so it can be flipped over. Tin the concave side. Heat the relay contact to melt the tinning, then carefully place the concave side of the gold contact over the relay contact and wait for it to heat. Remove the heat and carefully hold until the solder cools.
Repeat process for other relay contact. Clean off heat sink. Reassemble relay but leave the cover off. Use bench power supply to cycle the relay and verify the NC contact arm flexes slightly, bending the NC arm as needed. This insures there is contact pressure. Put on cover and store this "special" relay. Breath deeply, have a cup of coffee, and stretch.
Move onto the next project -- plastic membrane switch/label eaten off the hot plate... chemists!
About two years later, the instrument failed again, alarming in stand-by. Testing showed one "bad" relay and another following suit. Repeat above process on another relay. We installed these "special" relays and tested. No alarms! Yesss! It has been three months now, and only time will tell if these relays fail. The modification for one takes about an hour of delicate work. A second identical mass-spec has since arrived in the department, improving odds on opportunities for us to mod more relays. But, hey, America's got talent!
The Frankenstein's Fix has just come to an end. Stay tuned to read the submissions and see what kind of difficult job of judging we have ahead of us! Submission details and full contest rules here.