Saturday, February 27, 2016

A lesson on known and unkown unknowns

My Land Rover only just fits in my shop with the plow on!

This post, which is copied from our farm blog, is a lesson in binary logic troubleshooting, how not to use a multi-meter, and in battery technology.

My Land Rover starter has been giving trouble all winter, and I've been trouble-shooting it as we went along.

She always did start hard when cold, and often needed a jump and even a block heater when the mercury was below zero. This year the old girl was starting just fine up through the fall, but when the temperatures dropped below 10 F, we began to have the same kind of trouble and then yet more trouble. A jump start would almost always work in moderate cold, but once when we were well below zero, even a jump wouldn't turn the very tight, newly rebuilt engine block over, even after the block heater had been on for several hours.

Naturally, I suspected the starter. The battery was comparatively new, only three years old, while the starter was forty-five years old, and the engine was of course very tight after the otherwise very successful rebuild. But the new part was expensive so I waited. Eventually I had a a little extra cash left after paying the bills and shelled out nearly $300 for the replacement I needed.

Unlike most other spares for the series Rover, starters are expensive. On the recommendation of my Rover parts guy, a jovial Brit called Trevor who lives in Arkansas but used to live in Maine and operates the "Rovah Farm" parts house, I went for a brand new replacement "gear" starter. These are devices designed to give more and faster turns of the engine, especially in cold weather, which would probably make life a lot easier in winter. A sun-and-planet gear is used along with a modern motor to provide high rpm and greater turning power. A new old stock or reconditioned original Lucas-type starter was equally expensive, so it would be silly to opt for inferior but original equipment.

The new one is the shiny bright thing on the left. The nasty old black Lucas starter is on the right.

Unfortunately, once installed and despite the shine, the new starter wouldn't work. I pulled it out again and bench-tested it with our spare shop battery, a brand new truck-size unit, and it worked fine, so I put it back and tested the cables one-by-one.

It was while testing the cables that I finally noticed my multi-meter was over-reading. There was one point that the voltage available at one particular connection was 17 volts. This is essentially impossible for a nominal twelve volt lead-acid battery using six two-volt cells to provide. The battery voltage all along, since the fall, had been reading 15 volts, which is very high but not completely impossible, the kind of reading you might get with an exceptionally good twelve volt battery, and perhaps a voltmeter that was a little "off" but not completely unserviceable. This was why I hadn't suspected the battery. I immediately tested the voltmeter with a known voltage, a new AAA battery.

A new AAA should read around 1.6 V, and this one read 2.25 V.

Here's the archive shot of this particular experiment.

My meter well and truly faulted, I ran to the parts store to get a new one. Then I compared the two against the AAA. Actually, it read 22.5 V, and I had missed the fact all along that the decimal point was in the wrong place.

That 15 V reading that made me think I had such a great battery? It was actually reading 150 V, but my mind had inserted a decimal point in the "right place". (Later, after moving the meter inside and warming it up, the point went back to the right place, which makes me suspect the 9 volt battery in the multi-meter. I changed it, and the meter worked fine again, even when calibrated against the new one. But that's another story.)

This led me back immediately to fault the battery, which I switched out for the good new one I had in the shop. I also replaced the cables for good measure.

Now everything works just fine and my Rover is a lot happier starting. It should be, because almost every component in the starter circuit is brand new!

But what a great lesson in binary logic troubleshooting, which, by the way is AKA "the scientific method."

If you are troubleshooting a machine or other engineered system, or undertaking a scientific or social scientific inquiry, your main problem is to structure your inquiry so you test only one subsystem or one hypothesis at a time.

My first failed experiment in this particular inquiry began in the fall when the Land Rover first started hard. I had connected the multi-meter probes to the battery and gotten a relatively high reading, which made me fault the starter subsystem (starter, starter solenoid, starter switch or cables) rather than the battery-and-charging subsystem (battery, alternator, voltage regulator, associated cables). If a twelve volt car battery reads high, usually around 12.8 V, it's usually good. There are some exceptions but few enough that this is a good basic test to use. This one read high, because of the faulty meter, and so I called it good. From that point on, I was working with the wrong hypothesis.

Lets be clear here. The experiment failed because it didn't produce a good enough test of which of the two subsystems was at fault. If the experiment had discovered that the battery was bad, that would have been a good experiment.

Here I had multiple overlapping faults, but primarily the battery and the multi-meter were both bad, in such a way as to mask each others' symptoms. (The tight engine and weak starter contributed too, but secondarily.)

What I should have done was test the multi-meter before use against a known voltage, such as a new AAA battery. This kind of double checking of equipment is standard in science, and I should have known to do it. Instead, I've been jump-starting my Land Rover in the cold weather all winter.

I don't begrudge the money I spent on the new starter. It really does crank well, and will be a boon when snow-plowing during the winter next year and for as long as it lasts.

Because no car really likes to start when it's below zero, let alone one that is 45 years old. That's not science. That's experience.

No comments: