i've had a persistent low-grade problem on my 89 civic which i now think i've cured and want to post for the archive. i think it's important because i had been diagnosing it as a sensor problem, and it would appear that others have too if the fact that in all my local junkyards, unless you get to a fresh vehicle very quickly, the sensor has been removed, presumably by someone chasing the same problem on their own vehicle. it appears to be an issue with most honda vehicles of late 80's, through to the mid-late 90's. i noticed the problem because i have a 90 japan-made crx, otherwise identical to my 89 civic, and it does not have the symptoms. prior to that, i thought my civic was "just that way" and was simply slow to warm up and uneconomical because it was aged. symptoms: i. poorer fuel economy. it's not awful, but it's 10%-20% lower than normal. ii. rapid discoloration of engine oil iii. slow engine warm-up and thus delayed torque converter lock-up on automatics. this is the most immediately noticeable. diagnostics: 1. check coolant system. i tried three different thermostats before deciding for sure this wasn't the problem. 2. check sensor. these can and do fail. when i replaced mine, the fault disappeared for a while showing i'd scratched the itch, but then it returned. 3. check the sensor wiring. analysis: the thermistor used by honda is much lower resistance than is used by many other manufacturers. only about 2.5kohm at 25C. [others use 10k - 50k.] that's fine for the electronics, but when the operation temp climbs, thermistor resistance is in the 100-200 ohm range and any connection problems or noise start to become a significant percentage of such a low value. on checking the connector plug, i found that the inside terminals were shockingly loose and would grip a test spade hardly at all. static electrical resistance was ok, but when hot or subject to vibration, i think electrical noise would become very significant. as the computer would appear to average readings to mitigate electrical noise on unshielded leads, it would seem to be using a false higher thermistor reading in its coolant temp sensor math than would otherwise be the case with a noise-free connection. solution: the problem then, how to remedy? the plug is "tamper proof" and hard to disassemble. it's also below the notoriously leaky honda distributor and covered in oil and filth. upon dissecting several junkyard donor plugs, i concluded that it was possible to remove the terminals from the plug since they have a snap plug-in feature like a lot of connector plugs - the problem is finding out how to push the release tab and pulling the lead so it comes out. also the rubber bung at the end of the plug is glued in place - great weather proofing, but a bitch for terminal removal. [also, later model connectors have the release tabs occluded, to removal of leads only would be pretty much impossible.] but, with terminals removed finally, it was possible to crimp the connectors back so they would grip the terminal spades tightly, reinsert with silicone grease to keep the weather out, and reconnect. http://www.flickr.com/photos/38636024@N00/4451349382/ http://www.flickr.com/photos/38636024@N00/4451349388/ http://www.flickr.com/photos/38636024@N00/4451349400/ http://www.flickr.com/photos/38636024@N00/4451402360/ upon reassembly, the connector offered a little slide resistance going back onto the sensor - just as it should! a quick test drive had the car warming up just as fast as the crx, with torque converter lockup just as it should. conclusion: you can recondition the connectors the hard way, as above, or you can just snip the plug and replace with a junkyard donor tested to connect better. i prefer the former because i'm a pedant, but the latter would have been MUCH quicker.
