Multiplexing temperature sensor

My engine has a temperature sensor on the right bank for the dash gauge, and a different one on the left bank for the electronic fuel injection system. Both sensors are just temperature sensitive resistors of some kind. I would like a temperature reading for both banks. Rather than drilling and tapping a hole for yet another sensor for the left bank I was thinking about multiplexing the EFI sensor. Is there an simple circuit that I could cobble together to do this, perhaps based on an IC? The circuit would probably also have to modify the output to match an available temperature guage.

Any ideas?

TIA

Ed

Here is the EFI sensor resistance curve: Deg C Ohms 10 3700 20 2500 30 1700 40 1180 50 840 60 600 70 435 80 325 90 250 100 190

Here is the the curve for the sender to the dash temperature guage. If the output of the multiplexer circuit produced something close to this I could switch manually back and forth between the two, which would be fine: 'C ohms 0 1117 10 871 20 680 30 530 40 412 50 319 60 245 70 187 80 141 88 112 90 105 92 99 100 76 110 54

I used to have access to a Web page with the ECU internals, but it seems to be down at the moment, or perhaps has gone away. I'll continue looking for it, but maybe in the meanwhile the above will be sufficient to offer suggestions, of course with caveats.

TIA

Ed

I'd be very careful of messing withthe EFI circuit. You should read the voltage across the sensor with a high impedance circuit, so as not to disturb the EFI operation. So, the actual resistance curve of the sensor is not the issue, it's the voltage output at whatever current the EFI wants to drive it with. This may not be a constant current, since the EFI could be turning it on only during measurement cycles at whatever update rate it requires... maybe not too often, given the thermal time constant of the big hunk of metal it's attached to. So your first order of business would be to look at the voltage across the sensor while the engine is running, preferably with a scope. If it is constant (other than varying with temperature!) you can proceed to curve-fitting to match the other sensor. If it pulses at ahigh-enough rate, you may be able to filter it (after buffering, to not disturb the EFI) to get a suitable signal.

But curve fitting by analog means can be a bear. If you want to go whole-hog, you can use table look-up with a small processor chip. But my guess is that installing a separate sensor may be a lot easier in the long run.

Best regards,

Bob Masta dqatechATdaqartaDOTcom D A Q A R T A Data AcQuisition And Real-Time Analysis

Home of DaqGen, the FREEWARE signal generator

Thanks, Bob.

This is a very early EFI system, namely the Jaguar adaption of the Bosch D-Jetronic system, circa late 1970s. The ECU is an analog implementation with discrete components. It is of course driven by a pulse stream from the engine, picked up from a trigger board in the distributor. The primary inputs are engine speed (derived from the pulse stream), manifold pressure, coolant temperature, and air temperature. It is described at

if you really want to get into it.

Also, I have now found the site at which the Porsch 914 implementation of D-Jetronic is fully reverse engineered. It is at:

From that site you can navigate to which shows the portion of the ECU schematic where the cylinder head temperature sensor (since the

914 engine is air cooled). It looks like the sensor feeds into the base circuit of a transistor. I have read that this is eseentially the same as the Jaguar/Lucas implementation, although there are some differences.

The coolant temperature sensor (CTS) is a thermistor with resistance curce as given previously. One side is grounded and the other connects to the ECU. I just put a scope on the the ungrounded pin of the CTS in situ with the engine idling, warm. What I see is a very clean 2.2 volts, varying only by a negative 0.2 volt blip every 37 ms.

So, are you saying I might be able to just connect a gauge to the CTS through high impedence, no multiplexing required?

Thangs again for your interest.

Ed

That should work fine, as far as not disturbing the EFI. You can then use nonlinear circuits to linearize the voltage response for a digital meter, or you can make a custom face for an analog meter. (I'm assuming that you won't have an analog meter that is calibrated to match the EFI curve.) Or you can digitize and use lookup tables.

Best regards,

Bob Masta dqatechATdaqartaDOTcom D A Q A R T A Data AcQuisition And Real-Time Analysis

Home of DaqGen, the FREEWARE signal generator

Sounds easy! I'm wondering, though, about how to get the proper relationship between voltage at the sensor and temperature. Direct calibration is just about impossible for two reasons. First, it would require fitting another temperature measurement of the coolant at the same location in the system, which is what I'm trying to avoid. Second, covering the range at the high end would require overheating the engine. There must be some way to relate the known resistance vs. temperature to the voltage at the ungrounded terminal while it functions in the EFI system. Would this require details such as resistances etc. in the ECU where the CTS connects?

Ed

Answering my own question, I would think I could just measure the resistance between where the CTS pin and the battery voltage pin at the ECU, then view that resistance and the CTS as a "voltage divider."

Ed

if you can characterise and simulate the probe pulses sent to the sensor you can remove the sensor and heat it in a pot of oil with a thermometer while feeding it simulated pulses...

if the EFI sends different pulses to the sensor depending on the engine speed or temperature (etc) you have a tougher task.

--

Bye.
   Jasen

Measuring the ECU "resistance" probably won't work. (It's probably a current source, not a simple resistor.)

As I recall, you already have the resistance curve for the CTS. So shut off the engine and quickly measure the CTS resistance and that will tell you its temperature at some arbitrary reference point. The voltage across it just before you shut it off will allow you to find the ECU driving current from I = V/R. Now assuming that the ECU uses the same current all the time, you can apply that to the resistance curve to convert it to a voltage curve.

Best regards,

voltage Bob Masta dqatechATdaqartaDOTcom D A Q A R T A Data AcQuisition And Real-Time Analysis

Home of DaqGen, the FREEWARE signal generator

Thanks, Bob. Here is the relevant part of the ECU schematic. Could you take a look and tell me if it is a constant current source?

Thanks, Ed

Not constant current, just a plain old series resistor (plus a diode). But I'd avoid trying to measure the resistance of the ECU, just because it is a questionable practice in general, and also because that diode is in there to mess up your measurement. But since it isn't a current source, using a voltage divider approach is probably best. Just measure the CTS resistance (disconnected from the ECU, of course), as well as the voltage when it is connected, and you will have the current at that temperature. The effective ECU resistance will be 12 minus the CTS voltage divide by the current. You can probably assume that effective resistance over the whole range, and just use your CTS curve to derive the voltage curve.

Best regards,

Bob Masta dqatechATdaqartaDOTcom D A Q A R T A Data AcQuisition And Real-Time Analysis

Home of DaqGen, the FREEWARE signal generator

Thanks, Bob.

Ed

I remember forom analog computer days a device called a diode function generator (DFG). As I recall these were basically a set of diodes and resistors. Is there an IC that does this sort of thing? Or, what is the processor chip you are referring to? Microcontroller?

When I do a Google on "diode function generator" it's like a march back in time... pulls up all sorts of old papers by analog computer programmers. None of mine though... sigh.

Ed

I have never really used the traditional diode function generator... they are a morass of interacting bias resistances, not suitable for hand-tuning. Nowadays, I imagine that modeling software or even a special diode function generator program might make this more straightforward, though. An "improved" diode function generator uses op-amps to isolate each breakpoint, so they are somewhat easier to adjust, but they tend to have sharp corners at each breakpoint... not what you want here.

If you want to pursue this, I'd first take a long hard look at the correction you need to apply. You might get lucky and be able to use some simple circuit to do this.

The processor approach is not specific to any chip. You read the voltage with an A/D, and use that value as an address into a lookup table to get the corresponding temperature. If your chip has a lot of memory, you can have an entry for every degree. If not, you can segment the required curve into segments, and take the high N bits of the A/D value to select the segment, and apply linear interpolation to get the degree reading. Since temperature changes so slowly, even a slow chip should be plenty fast enough to do a lot of calculations. Note that I'm no expert at this, and have no specific chip recommendations here If you are not already comfortable with embedded development then you can either use this as an excuse to go off on a long, involved learning quest, or just fall back to creating a face for an analog meter.

Best regards,

Bob Masta dqatechATdaqartaDOTcom D A Q A R T A Data AcQuisition And Real-Time Analysis

Home of DaqGen, the FREEWARE signal generator