Anyone interfaced to an engine IAT sensor?

[IDEngineer]

An engine's Intake Air Temperature (IAT) sensor is typically a thermistor. It has a two wire connector. One side is tied to ground, so all you have to work with is the "high side". The natural solution would seem to be a voltage divider and measure with an A/D.

However, to prevent self heating (and potential thermal runaway due to NTC) every IAT I've run across demands incredibly low max current - like on the order of 100uA max. In a 5V circuit (which every one describes as typical) that's an equivalent source impedance of 50K. That's a rather high impedance for a one-wire sensor at the far end of a wiring harness in a very electrically noisy environment. And their resistance ranges are typically 30K at 25C ambient dropping to a few/several hundred ohms at engine operating temperatures, so if you size the pullup to limit current at the thermistor's minimum you end up with a narrow dynamic range. Narrow dynamic range + noisy environment + high effective source impedance = not so great.

Can't run it differentially because one side is grounded close to the sensor (some actually integrate the ground into the sensor housing mount). I can drive it with a constant current source but that doesn't change the high impedance nature of things... I suspect 100uA would be swamped by the induced currents from all the noise/HV/etc. within the harness as well as nearby on the engine as the harness winds its way around, with plenty of length to serve as a very effective inductive pickup.

Online references I've found basically just wave their hands at a resistive divider "solution" without worrying about real world conditions. Sure, 500mV active range at high impedance is plenty in an electrically noisy environment, right?

Any "tribal lore" on how to handle this? Thanks!

[bdunham7]

You can change the current and 'switch ranges' to keep the sensor voltages within reason.  If your maximum power input is indeed 0.5mW, for example (that sounds low to me, even for the small fast type of IAT) you can feed it 100uA from 5 volts to 1 volt, which is 50k to 10k, then switch to 200uA--which at 10K will be 2 volts and 0.4mW--and even less below that.  Systems that do this build in some hysteresis, which is not an issue because the ECM always knows what range it is in.   Typical systems with ranges like this in actual use don't use constant-current AFAIK, the simply use two pullup resistors in parallel with a transistor in series with one that the ECM can command on or off.  I think if you pick the right resistors the math works out even better than constant-current as far as keeping the voltage sensible over a wide range.

[ace1903]

Isn't IAT placed in air stream at place with maximum air velocity?
This 100uA max is usually for still air measurement. For an engine I think measurement is important when engine is running and there is considerable air flow.
I think that air flow negates effects of self heating and allows larger measurement current.

[langwadt]

Isn't IAT placed in air stream at place with maximum air velocity?
This 100uA max is usually for still air measurement. For an engine I think measurement is important when engine is running and there is considerable air flow.
I think that air flow negates effects of self heating and allows larger measurement current.

if it is in the airstream and self heating you get a flow dependent result, that's how a mass air flow sensor works

[Renate]

I'm not sure what you're looking for.
Are you a car manufacturer and trying to come up with a design?
Is there some problem that you're trying to solve? Are you seeing this on the OBD?
If you're just concerned, use a digital sensor on a µP and do a curve fit with what the OBD reports.
You can do all that with the engine off.

[IDEngineer]

A customer already uses IAT's and wishes to use a second identical device to sense temperature in the overall engine space. So I'm investigating out how to use an IAT as a temperature sensor potentially many wire feet away as described.

[eugene]

You said that you can't run it deferentially, but I don't see why not if you're adding a new sensor with new wiring and new signal conditioning. You are not obligated to ground one end of the thermistor near the sensor just because that's the way some auto manufacturers do it. You can do whatever you want!

Alternatively, since it's a slow moving signal, maybe you can get it clean enough by averaging the crap out of it over as long a period of time as allowed.

[bdunham7]

A customer already uses IAT's and wishes to use a second identical device to sense temperature in the overall engine space. So I'm investigating out how to use an IAT as a temperature sensor potentially many wire feet away as described.

Not to go all Captain Obvious on you, but why wouldn't you just use the exact same method that they use now with the identical sensor?  Noise can be integrated out and I think the main problem with an engine compartment temp sensor is going to be figuring out where to put it to get whatever results they want.  Most such sensors that I know of are built into some component or another (sometimes even the ECM) and aren't known for precise or stable results.

[IDEngineer]

You said that you can't run it deferentially, but I don't see why not if you're adding a new sensor with new wiring and new signal conditioning. You are not obligated to ground one end of the thermistor near the sensor just because that's the way some auto manufacturers do it. You can do whatever you want!

As I mentioned above, some IAT's ground one side through their mounting bracket. We could require an insulating mount, maybe....

But yes, I'm starting to think that we force some changes back upstream. Possibilities include:

1) Require both sides of the sensor to be electrically isolated so we can do what we want (no manufacturer-grounded sensors)

2) Just package our own sensor in our device, and require that they mount our device at the desired sensing location. This keeps all our stuff tight and compact. All we need is +12V, Ground, CAN-H, and CAN-L which are far more noise tolerant than a thermistor at the end of a long harness run.

I'm leaning toward the latter because it puts us in control of everything. Standalone thermistors are ~$0.40/100's so this might actually SAVE them money in the deal as compared to packaged IAT's which appear to be several dollars and up, plus requiring a mating connector.

[langwadt]

You said that you can't run it deferentially, but I don't see why not if you're adding a new sensor with new wiring and new signal conditioning. You are not obligated to ground one end of the thermistor near the sensor just because that's the way some auto manufacturers do it. You can do whatever you want!

As I mentioned above, some IAT's ground one side through their mounting bracket. We could require an insulating mount, maybe....

But yes, I'm starting to think that we force some changes back upstream. Possibilities include:

1) Require both sides of the sensor to be electrically isolated so we can do what we want (no manufacturer-grounded sensors)

2) Just package our own sensor in our device, and require that they mount our device at the desired sensing location. This keeps all our stuff tight and compact. All we need is +12V, Ground, CAN-H, and CAN-L which are far more noise tolerant than a thermistor at the end of a long harness run.

I'm leaning toward the latter because it puts us in control of everything. Standalone thermistors are ~$0.40/100's so this might actually SAVE them money in the deal as compared to packaged IAT's which appear to be several dollars and up, plus requiring a mating connector.

LM35 et.al.

[viperidae]

From what I can tell, modern IAT sensors are two wire devices, designed for use with a pull-up to 5V of 1k to 3k.
When wiring it, you would supply it with a signal ground.
Some cars will use a map sensor with integrate IAT sensor.
I got my information from Bosch, GE and Delphi data sheets

[IDEngineer]

From what I can tell, modern IAT sensors are two wire devices, designed for use with a pull-up to 5V of 1k to 3k.
When wiring it, you would supply it with a signal ground.
Some cars will use a map sensor with integrate IAT sensor.
I got my information from Bosch, GE and Delphi data sheets

Right. I see the same info, plus quite a few IAT's appear to "presume you want a ground" and helpfully    tie one side to ground through their mounting bracket.

1K seems awfully low for a thermistor. That's 5mA at 5V, whereas most of them speak in terms of tens or hundreds of microamps. Who knows, though... one of the Ford IAT spec sheets recommends a pullup of 3R. That is not my typo (though it's probably theirs!)... a THREE OHM pullup to 5V. And the example schematic shows 5V, to 3R, to the IAT, to ground. No active device switching, nothing but a sensing connection from the junction between the 3R and the termistor to the ECU.

Meanwhile, that same spec sheet says max permissible current is 244uA. I guess Ohm's Law works different in Detroit.

[bdunham7]

Right. I see the same info, plus quite a few IAT's appear to "presume you want a ground" and helpfully    tie one side to ground through their mounting bracket.

Really?  Example?  That would be odd because most modern (meaning less than 30 years old!) systems would either have a separate, dedicated spot for signal grounding on the engine block and run all signal grounds back to that point, or go one step further and go all the way back to the ECM.  I've never seen a temp sensor, IAT or otherwise, that had a two-pin connector but grounded one of them.  Perhaps if you stick with plastic-bodied sensors you won't have that issue!

Meanwhile, that same spec sheet says max permissible current is 244uA. I guess Ohm's Law works different in Detroit.

Without looking at the same data sheet as you, I can't comment on that number or what it means (surely it isn't an absolute max without damage number?) but I can tell you that a fairly typical old-school IAT/CT/CHT sensor application might use a nominal 10K NTC, a 5V nominal voltage range and a working range of 1.0-4.0V, with fault codes set for voltages <0.5V and >4.5V.  It makes little sense to have the same maximum current spec over that whole range.  As a first-order estimate, if you assume/approximate an NTC that is 10K at room temperature, 1k at 100C and 50k at -20C, then a 5k pullup resistor more or less works to keep things almost in the working range over those temps.  You'd need to tweak the NTC and pullup to make it actually stay in the 1-4V range.   But that max current spec would only seem to work at the high end of the resistance range and I really don't see why exceeding it would matter at lower voltage.  The two-range approach I mentioned earlier is even better at doing all of this. 

[ifrythings]

@IDEngineer, I have reverse engineered a few different Ford computers and they haven’t changed sensors from at least the early 90’s to 2010 as far as I have tested. The below circuit is what Ford used and the graph is also taken out of the computer, not sure if this helps or not.

[robzy]

I'd recommend checking out how the Speeduino project handles it: https://speeduino.com/home/

They have both open source software and hardware.

[emece67]

.