I'm stumped. Help me fix my Trooper.

Here's the situation:

1991 Isuzu Trooper with a GM 3.4 V6 with TBI

The computer was reading -40 C temperature with the engine cold on a warm day, and 0ish when the engine was warmed up.

"Okay, I thought, it's a bad temp sensor." So, I changed it. No difference.

Then I changed it again, no difference.

Then I changed it a third time, no difference.

Then, I hooked up two temp sensors in parallel, the second just sitting on the intake manifold. Now the ECU measures nearly-correct temperature with the engine warm, but never goes below something like 60 C.

The wiring is fine, the ECU outputs the correct 5 volts and receives 4.93 volts back from the temp circuit, and all the sensors test in the correct range. There are no coolant bubbles. The engine is actually warming up. I've confirmed all of this with a datalogger.

I've tried two ECUs, one of them being brand new. They test fine, too.

WTF is happening here? And how do I hook up an arduino to fool the ECU into operating in the correct range? I think I know how to use an arduino to measure a thermistor (voltage dividing circuit), but I'm not sure how I can make the arduino emulate a thermistor in the eyes of the GM ECU. Will I need to hook up a bunch of different resistors and have it step between them as temperature changes?

"Insanity: doing the same thing over and over again and expecting different results."
- Albert Einstein

These temp sensors are known to be defective, which is why I tried 3.

Who has input on this: http://www.diyautotune.com/catalog/diypnp-delphi-56pin-unassembled-kit-p-389.html

Is it IMPOSSIBLE that you keep buying bad sensors? Sounds like everything is working as intended but it's not being READ that way.

I bought three bad brake calipers in a row and a brand new fuel pump was defective in the box.

Edit: looks like I was one post late !

OK. After reading the FULL post...

The voltage drops as the temperature rises, so the tandem sensors are reading near correct, so there is more resistance.

Figure out the formula and add the resistors in. Why you suddenly need resistors should be the next question.

Are all the sensors of the same brand? OEM?

It's impossible, I measured all the sensors and they have the correct voltages at each and every point on the Isuzu temp/ resistance chart.

The tandem sensors are reading less resistance than just one, they're hooked up in parallel.

Does this mean my wiring has too high of resistance? It measured fine, but maybe it's got a slight break that only shows up under load?

I tried just hooking up a normal resistor in parallel, but without one that scales (like the second thermistor) I could only get a correct reading in a tiny portion of the temperature range.

Have you visually inspected every inch of the wiring and traced back to where it grounds?

Err, no. It's all part of the main engine harness.

I'm going to try running wires directly from the ECU to the thermistor tomorrow, and see if that fixes it.

Oh, and whoever figures out this problem first gets $5 in PayPal moneys from me.

Racking my brain here. You have two sensors in parallel reading less resistance than just one. Hrm. Yea. I'm thinking backwards here, aren't I... Is it your average harness wiring that you are using? I have a hard time believing a small fault would cause such an effect.

Check for resistance in the line. That should give you a definitive answer to that question.

If the sensors are known good, and the ECU is known good, something must be fouling up their conversation?

Can you get at the pins needed so that you can install a sensor, measure it in-situ, then measure it again where it meets the ECU? There isn't much load going on in reading the sensor, but I still think the wiring is the most likely culprit. It's kinda gotta be sensor, ECU, wires, or connectors, right?

EDIT: Hrm. Guess you already kinda covered this with the ECU output and return voltages, huh?

What do you mean "the ECUs test good"?

If the ECU is sending out the correct voltage, and reading the correct voltage, but reporting the wrong temperature...

Could there be some other funky issue in the wiring that's confusing the ECU? I'm just chucking out random ideas at this point, but wondering whether a problem on a different sensor (or somewhere else) might be dragging the ECU's internal reading a funny direction...

Bad ground! It's always a bad ground, right?! But which one...

In reply to Tom Suddard: I know you probably don't have another Trooper available, but the testing faulty parts by swapping method really only works when you take the suspect component to a known good machine. If the part is faulty the problem should follow the part, if it doesn't the fault lies somewhere else.

My understanding of coolant temp sensors is that the resistance decreases as they warm up. So if something is causing an open circuit or an increase in resistance the ecm will interpret that as a very cold engine.

See if you can get a replacement connector end for the coolant sensor, and solder it in. Also if you haven't already, physically inspect and clean every connector related to that circuit from the sensor back to the ecm. Just because the ohm readings checked out doesn't mean that the circuit is good under operational load.

In reply to Ransom:

Bad ground is worth checking...

Hmm, that's an interesting thought.

I didn't test the ECUs myself, I took it to a shop and they said that both ECUs checked out fine. It's a standard wiring harness, and I don't measure any resistance in the wires or any continuity to ground in the wires.

I must have done something wrong–the supply voltage is great, the ECU is great, the sensor is great. It must be the wiring. Tomorrow, I'll run two wires directly from the ECU to the sensor, and see what happens.

You mean all four sensors are great.

Yeah.... smartass.

Ground?

Check your grounds, as others have said. Also check your battery cables to make sure they're all in good shape. My old Chevy truck had weird, random electrical issues that were caused by a bad negative battery cable. The cable looked fine, but with years of high heat from being so close to the engine, it just broke down.

What's the resistance range of the sensors? If they're pretty low values, then questionable wiring/grounds could certainly be the problem.

If you're really hooking up the extra sensor in parallel then you're lowering the total resistance of the circuit. If you hook them up in series then you're increasing the total resistance.

They're high resistance ranges, in the hundreds of thousands of ohms. Here's the chart and the wiring diagram:

I'm sure that I'm hooking them up in parallel.

Sorry Tom the EE is coming out in me. 3.4K is 3400 ohms. Not hundreds of thousands of ohms.

Parallel circuits allow for more paths of current flow. The more paths the lower the resistance. In series the current has to flow through each resistor one after the other. There is a voltage drop at each resistor in a series circuit. In a parallel circuit the voltage is constant.

I misread the chart above. I stand corrected the first number really is in 250000 ohms.

I'd bench test the resistance of the sensor at several different temperatures. Set the sensor in a pot on your stove and raise the temp. See what happens to the resistance. Compare it to the chart that you have above. If that works out then it's likely the wiring. It looks to me that you have an issue in your wiring harness. Adding the second sensor in parallel reduced the resistance and gave you a better reading.

Yeah Tom, the resistance drops pretty low once you get up near operating temps. I'm betting it's either a dirty/bad connector in the harness, or a dirty ground. I expect once you jumper straight from the sensor to the ECU it'll more or less go back to normal.