Fuel level sensors

[veebs]

Hi all,

I had a ‘failure to proceed’ event the other day - engine died with a engine system fault (or similar) on the dash, and refused to restart.

Got it to the Indy today, who diagnosed a bone dry fuel tank.

No warnings, and the gauge still read half full (well, slightly under half).

But here is the weird part: adding 20l of fuel, raised the fuel level on the gauge slightly.

So, this told us the level sensor is seeing a change at the very bottom of the tank (ie, if it was jammed halfway, it shouldn’t be able to detect any tiny amount of fuel).

I filled up on the way home, with something like 77 litres of fuel (the balance of what they put in, and the tank size) meaning the tank itself isn’t deformed.

There are, apparently two level sensors, on account of the long shape of the tank, and if i was guessing, I would say their readings get averaged to show the level on the gauge.

One is stuck at full, the other is correct, meaning we have between 100% and 50% showing on the gauge (ie, both full, or one erroneously full, the other empty for a 50% average)

Does this sound plausible? Anyone else had this?

No long range tanks - all stock.

[BradC]

Yowza! I'd go a buy a lotto ticket!

Checking the wiring diagram, both sensors are fed to the BCM separately, so any averaging would be done in the software. Do you have any live values for fuel level sensors in the GAP tool? (I've never looked)

[Tombie]

Very plausable - I have a similar issue to a lesser extent as I crunched my tank about 11 years ago and damaged one of them.

I havent replaced the sender yet (too slack), you can order them and do it though.

[veebs]

Yowza! I'd go a buy a lotto ticket!

Checking the wiring diagram, both sensors are fed to the BCM separately, so any averaging would be done in the software. Do you have any live values for fuel level sensors in the GAP tool? (I've never looked)

I was thinking the same re IID - I'll have a dig around later...

[veebs]

Fuel sensor levels are reportable using the gap tool - as a voltage from the variable resistor.

But both are working, and at what look to be consistent values. I tested using a high tech method: They are equal when parked on level ground, one is higher than the other on a slope with the nose facing uphill, and the values swap when i turn around to face downhill.

The sensors do appear to share a ground, which if faulty/loose could impact the sensors equally maybe?

I did find a service bulletin that covers my VIN, which, well, could explain everything... 184287

[BradC]

Yup. An issue with the common ground would cause all sorts of issues. The TSB shows it's an easy enough fix if you drop the tank.

In the mean time, back to watching the trip meter to estimate remaining fuel I suppose.

[veebs]

The plot thickens

I did two quick drives, logging the values of the fuel sensors. Awesome amount of data these cars create, but i digress.

As expected, the sensor voltages mirrored each other - when one was up, the other was down, to a maximum of 3.5v, minimum around 2.2v

I took the logs, and ran three 'fuel level' comparisons on the dataset:

1 - The average of the gauge voltages, expressed as a percentage of 3.5v, the assumed 'full' value
2 - The reported fuel level from the car ECU
3 - A burn-rate driven value, using an assumed 11 l/100km, and km driven since filling up.

I did 2 runs, and (1) and (3) above were fairly close, at about 90% (i've driven 80km)

The ECU however, was weird - it reported 95% on run 1, changing slightly every few seconds when I assume it recalculates. Extrapolating, this would yield the 50% showing at zero fuel i experienced.

On run 2, the ECU reported 100% the entire time.

Now, my voltage level assumption is premised on 0v being an empty tank, so there is potential for error here, though for the ECU to read 100%, when the levels are clearly moving, suggests there may be something else at play, within the software...?

184289

[BradC]

Challenging. The voltage you see will be from a resistive divider between a fixed resistor in the ECU and the tank sensor. It would be unusual to see as low as 0V as that'd require the tank resistor to be effectively a short. Usually on a 5V scale you'd see somewhere between 1 & 4V allowing room at the top and bottom for the ECU to reliably detect a short and open condition. Your best bet now is to get someone who has a D4 to give you some values to compare against.

I could give you some D3 values, but I know the tank and internals was "upgraded" between models, so I don't know what sort of validity that might have.

Edit : Re-reading the TSB you attached, resistance values between 46 ohms and ~1k you might see voltages lower than a volt. Still, won't be down at zero.
The spec in the D3 manual is 51.2 Ohms to 992.11 Ohms, so not that far from the mark. Interestingly it says the front sensor will see from 51 to 798 ohms (empty to full) and the rear will see 75.5 to 675 (it must be shallower).

The D3 fuel sensors are wired to the cluster, so that's different to the D4 for starters.

[BradC]

I did a run in the D3. Nearly full tank the sensors were ~3.5V. Under heavy braking the front sensor went up to about 3.8V and the rear down as low as 1.8V but generally around 2.2-2.5V. Seems to be a bit of sloshing.

[veebs]

Thanks BradC

I’m getting similar numbers, and the ecu has suddenly started calculating fuel again (NFI is it’s accurate)

All very weird.

[veebs]

So, for anyone playing at home, the problem has been fixed…

Turns out, the auxiliary battery had dropped a cell, and was being held up by the primary, via the traxide kit.
Somehow, this caused interference, as removing from the circuit meant all went back to normal.

Put the failed battery back in, and the problem returned.

So add this to the long list of “weird stuff that happens when the battery is weak/old”