Reasonable to force users to run both MAT AND IAT always?

[Fred]

I've been thinking about heat soak and airflow and having an accurate idea of what temperature air is entering the engine.

By forcing the use of two air temperature sensors, one immediately after the intercooler in a location that does not see heat soak, and one in the manifold that does, we can best calculate the temperature of the air entering the engine at all times (i.e. immediately after start and while running consistently, or have some table of interpolation based on time/rpm/airflow/etc.

Obviously this has some significant advantages (hence OEMs use it) but it also eats an extra ADC channel that could be used for other things. Given we have 16 of them total there will be a few spare regardless, but is it reasonable to always expect people to install two of them?

I definitely DO NOT want to get into the situation where you have a choice about running two with X algorithm or two with Y or one with Z etc. We want the best reasonable approach for most vehicles that can be used by all. If you only wanted to run one sensor, you could either tie the input to both appropriate ADCs or better, link both channels to that pin in software defeating the purpose and reverting to the basic one sensor method.

With one sensor I always strongly recommend running it as close to the front of the car as possible and as far from the engine as possible as startup conditions don't matter anywhere near as much as running conditions that can be easily corrupted by having the sensor in the manifold etc.

Thoughts?

I'll put a poll for people to vote in.

[GartnerProspect]

Well, are we really at risk of running out of ADCs? I'm guessing we have a ton of general purpose I/O availible though, right? If ADCs are the concern, what about using a single ADC and a normal I/O to trigger a solid state relay (unsure of switching time or switching cycle lifetime so take this with a grain) to switch between them? As a non-expert on the subject, I don't think they BOTH need to be sampled 100% real time, do they?

I don't think it should be forced but I like it a lot and think it should be the default configuration IMO.

Generally Ambient air doesn't change drastically or frequently And really it should only be slightly higher than MAT except in boosted application.

So sample the IAT occasionally to identify heatsoak. This would also keep the algorithms easier too, no? One flag to completely eliminate the whole circuit.

This is all coming from someone that doesn't know what he's talking about, FYI!

Also on a side note: Do the ADCs sample continously in the background without a cycle penalty until they are polled or do they only sample once polled? And what is the penalty to poll, if any? Maybe I don't even understand anything about it...

[thebigmacd]

Also on a side note: Do the ADCs sample continously in the background without a cycle penalty until they are polled or do they only sample once polled? And what is the penalty to poll, if any? Maybe I don't even understand anything about it...

I'm pretty sure they only sample once polled, because they use successive-approximation, and you need to synchronize when the result is going to occur. If they sampled continuously, you wouldn't know the exact time when the result can be read.

[GartnerProspect]

Thanks. More things to google when I should be sleeping.

[shameem]

Could a better temp sensor (e.g. - http://www.bmotorsports.com/shop/produc ... cts_id/400) be the way to go instead of increasing code/hardware complexity to compensate for old sensors/design.....

This one is a combined MAP/MAT - http://delphi.com/pdf/ppd/sensors/et_mapmat.pdf

[Fred]

That's a good idea Gartner, the mat sensor is only really of use at low loads and at start up time. the rest of the time air flow is sufficient to keep things close to ambient and therefore consistent. Thus you are right, only IAT needs continuous sampling. The digital pin to swap inputs is a nice idea too.

About ADC, there are multiple modes of sampling that you can setup. I have it setup to sample continuously circularly through all of them at the highest rate (least efficient on power). Or at least, I think I do. I just got it working at all and moved on. There is no penalty to poll as you just read the register that contains the value and you have it. Easy as that.

The trouble is that the MAT sensor is representative of the temperature of the air at startup, but quickly becomes wrong as fresh air flows in at a rate higher than the sensors response (even if it is open element). I use an open element hyundai sensor on my ute that is immediately after the intercooler right at the front and has no chance of heat soak. In consequence, the air that is in the plenum and heated when I hit the key is less dense than that where the sensor is measuring, but only during start up and for a decreasing amount a short time later. It wouldn't be a matter of reading MAT and using it, but rather calculating or tuning a curve based on time and or running conditions and the relationship between the two sensors.

Shameem, that sensor you linked is much like what I use, the toyota ones are closed and next to useless IMO. Note the time constant for it :

Thermal Time Constant: **< 15 sec

That's a long time for us, and the reason you want your primary reference well away from the engine in the air that it is actually seeing most of the time.

I guess the most important part is that all future users are made aware of the importance of
A sensor type : open and fast
B sensor location : forward, outward, and least susceptible to heat soak.

If that is the case, one can be used with minimal negative impact, and that impact only for a brief moment at start up when hot.

Admin.

[shameem]

Instead of using an RTD or a thermocouple - Is it possible to use something out of the automotive "box" - like this one - http://www.national.com/ds/LM/LM94021.pdf
or this one - http://www.analog.com/UploadedFiles/Dat ... D22100.pdf

The main issue would be mounting i guess - but it could be epoxied on to a connector (with all the conductive paths covered ofcourse).... The main concern i could think of at the moment would be the effect of fuel vapors and other corrosive elements on the chip itself....

They have possibly better sensing elements than whats being currently used (which was probably designed in the 70s) - possibly better reaction time and the signal processing is built in as well......

[Fred]

I think those are designed for sensing the temperature of things like hard disks and heat sinks etc. Things with large thermal masses that move slowly. The open element sensors are pretty hard to beat for response. The one you linked is likely to be about as good as it gets. The one I used registered my breath on it fairly quickly, however breath is high humidity.

I don't think we can say to a user "you must use this temperature sensor". I think we need to advise them to use open element sensors mounted intelligently (even if the oem setup requires a retrofit for it, this is DIY efi after all.) and let them know that if they don't use a good sensor, the performance of the system will be less than it could be.

Admin.

[AbeFM]

I'm a little confused here, so I'll ask some questions. Basically, everything till the last couple paragraphs has a big question mark after it.

MAT - Manifold Air Temperature
IAT - Intake Air Temperature

The MAT is the temp measured inside the manifold. This should be susceptible to heak soak, as the hot manifold will heat the sensor.
The IAT is the temp in the IC piping, but after the intercooler. This is the temp of the air as it inters the manifold.

The MAT, depending on where it's mounted, is really just there to get a feeling for the drift in the temperature away from what the IAT was reading. This will including heating due to pumping effects through the throttle plates, and heating due to heat absorbtion from the manifold. If your MAT is mounted right at the entrance to the manifold, and you have long or variable/complex runners, you would expect it to read low anyway (though this overlaps with the errors inherent in measuring MAP)

So, since MAT is being used with MAP to get the actual count of air molecules in the manifold, that is what you should be using most of the time, and the IAT is just there to show you how wrong you were due to heat soak.

Certainly, to my mind, multiplexing a few inputs is the way to go. This is not super-time-critical information, as it's a second out of date at all times anyway. So, this would be a great case for something like an external chip/circuit that either switched between them or whatever.

My best suggestion would be to use a thermocouple in the manifold. They are much much much faster. People avoid them since the readout circuitry is more complex. But heat soak isn't an issue, it would equilibrate in a very short time, radiative thermal coupling to the manifold notwithstanding.

Really, how long does it take to empty the manifold? Just fudge the numbers for the first few engine cycles (or, minutes?) to be a bit leaner, based on coolant temps, and be done with it. If the coolant temps are low, there's no heat soak, you can trust the sensor. If not, there's a fudge factor you can let for for a while.

The option for more sensors is nice, but I don't think it should be forced, I'm not sure I see the gains.

[Fred]

The MAT, depending on where it's mounted, is really just there to get a feeling for the drift in the temperature away from what the IAT was reading.

Agree.

So, since MAT is being used with MAP to get the actual count of air molecules in the manifold, that is what you should be using most of the time, and the IAT is just there to show you how wrong you were due to heat soak.

Disagree.

My best suggestion would be to use a thermocouple in the manifold.

They also don't come stock on your 1995 rb25det etc etc. We want the core feature set to be applicable to anyone/everyone.

Really, how long does it take to empty the manifold? Just fudge the numbers for the first few engine cycles (or, minutes?) to be a bit leaner, based on coolant temps, and be done with it. If the coolant temps are low, there's no heat soak, you can trust the sensor. If not, there's a fudge factor you can let for for a while.

Excellent idea. Coolant and manifold are going to be fairly closely related for certain. Additionally, if such a correction was used, it would be in a user definable table of sorts. A super strong recommendation to put your IAT in an intelligent place could be plastered in bold flashing neon on the website front page and people that adhere to it should have minimal issues.

The option for more sensors is nice, but I don't think it should be forced, I'm not sure I see the gains.

Agreed, and esp in light of you stating the obvious (which I was blind to) about coolant temp for heat soak of air.

Thanks for the post, good work.

Admin.