Friday, April 08, 2005

 

Advanced Troubleshooting

My current "beater" is a '91 Chevy Caprice wagon, representing a total investment of perhaps $1500. It's not pretty, but it's functional, and I just wanted something to commute in so I could keep the miles off my pickup and perhaps save a bit of fuel while I'm at it (20 MPG is great when you're used to 13 MPG).

But much to my dismay, while driving home in a fierce snowstorm a couple of months ago, it suddenly died. Driving along at 45 MPH, the engine just quit. Numerous attempts at restarting resulted in strong cranking, but it wouldn't catch fire - not even so much as a stumble or cough. A brief inspection while waiting for the tow truck revealed that the SES ("check engine" light) did not illuminate during the bulb check or with terminals A and B of the ALDL connector jumped (this should have resulted in a flashing SES). All of the fuses in the dash-mounted fuseblock (according to the owners manual, three of the fuses affect ECM function) passed a brief visual inspection, and there is no underhood fuseblock such as that used on GM vehicles starting in the mid-90s.

Unfortuantely, after getting a ride home behind a F-450 (there's something shameful about getting rescued by a Ford) and sitting in my driveway for a while, it started right up again. There went my evidence. Inspection of the underdash wiring revealed little (except for a starter wire mangled by the previous owner's alarm install, which was unrelated to this failure but another problem waiting to happen). I soaked down the ignition module and distributor cap with a spray bottle, but that didn't result in any abnormal behavior. With nothing left to do but drive it around and wait for it to happen again, I armed myself with a DMM, schematics, and a tool kit.

And it happened yesterday during my lunch break, just as I was completing a left-hand turn. The engine died, and the SES lamp was dark with the key in the Run position. I coasted into a parking lot and went about the process of pulling the ECM and checking the appropriate power and ground connections at the harness, suspecting that's where my troubles lied. Nope - I had +12V and ground wherever they were required. Drats - that leaves the ECM (GM ECMs are not known for failures, even after over a decade in service).

Back at the office, I rigged-up a temporary harness (using Molex CL-Grid terminals) that allowed me to test the basic function of the ECM. Hmm - it works on the bench. That leaves an intermittent failure of either the ECM or the wiring (probably the later) - not what I wanted to deal with.

My friend Josh, fellow B-body owner and electrical troubleshooting genius (he works in the lift-truck repair industry) showed up after work and brought me back to my stranded car. Swapping in the ECM from his car didn't fix the problem, and my ECM worked fine in his car, verifying what I had found on the bench - the ECM is good. Further investigation ruled out excessive resistance in the ground path. That left the +12V supply. The open-circuit voltage looked just fine, but what about its performance under load?

With a whole lot of wiggling around, I was able to back-probe the connectors with the harness plugged into the ECM - not an easy task when it's tucked up in the passenger's-side kickpanel.
This revealed a solid 12.4 V on the Run/Start/Bulbtest supply, but only 1.8 V on the Hot At All Times terminals. That explains a lot. But what of the root cause? According to the owner's manual and a Haynes manual, this circuit is feed by one of the fuses in the dash-mounted fuseblock, but some further investigation of other schematics that I had pulled off the 'net showed the terminals I was probing were actually fed by an underhood fuse.

Keep in mind that this vehicle is from the era before underhood Maxi-fuses, where fusible links (basically a wire that melts when exposed to excessive current) provided the feed from the battery to the interior fuse block (short-circuit protection being required between the battery and the firewall, in case a wire was cut during a collision or just through normal wear). But this car has something odd - a little fuseholder next to the fuel-pump relay, looking somewhat like a Weatherpac connector and holding a standard Mini-fuse. This apparently provides power to the ECM and the fuel pump, but doesn't get mention in the owner's manual, nor is it clearly called-out by Haynes. Inside was this 30-amp fuse:



Looks fine, right? It checks out good with a ohmmeter (never depend on visual inspection to tell you if a fuse is good or not).

Wiggling around the fuseholder resulted in a SES light that'd flash on and off - we'd found the general area of the problem. It didn't seem to be related to wiring strain, and thus was probably the fuseholder itself as twisting the fuse in the holder caused the problem. We performed a precision adjustment of the holder's terminals, plugged the fuse back in, and... nothing.

Take a closer look at the fuse. Instead of the legs and link being stamped from one piece of material (as Littelfuse typically does) this has two seperate legs connected by a link that's spotwelded in place. You can see bubbled melted plastic where those welds are located, presumably from heat caused by excessive resistance at the welds. Once the plastic got soft, it didn't support the fuse internals, and a crack in one of the welds probably developed.

Stealing a quality 30 A fuse from the defroster circuit has so far cured the problem.

The point? There really isn't one, except that electrical troubleshooting is a real bitch, and that one should never get hung-up on assumptions or "pet theories" while troubleshooting. The above represents at least 3 hours of work by an automotive electrical engineer and a experienced technician (not to mention the driving time required for the problem to reoccur). What are the odds that this would have been properly troubleshot by a repair shop?

Comments:
Awesome post!
 
I think it took about as long to create the post as it did to fix the car.
 
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?