I have been trying to tune the VE table and yesterday I noticed my AFR went in the 17's after sitting still with the engine running because the IAT was 111°.
Last night driving home when the IAT was 64°, the AFR was in the low14's to high 13's.
Shouldn't it be the other way around? Seems like a colder denser air would make the car lean, and hotter air would make it rich!
Is there a table for adjusting this?
And how do you guys running SD full time get around this?

 

There is the "charge bias temperature" deal which
tries to do a realistic model of the real air temp at
manifold pressure, which is a key piece of the speed
density air mass calculation.

There is IAT-only or complex model option, and for
the complex option you have the blending table for
IAT and ECT relative weighting.

Then there is the question, of how the IAT reading
relates to the true air temperature in the pipe vs the
MAF wall temperature or the air lid temperature. The
air really probably did not change much but the heat
soak of IAT sensor did. People sometimes seem to
do better relocating the IAT to sense outside air
somewhere the engine compartment heat doesn't
interfere.

 

jimmyblue hit the nail on the head...I used to have this problem (especially on a hot restart in the summer). I adjusted the ECT/IAT bias towards the ECT side of things (since it is a bit more stable than the IAT readings in stop and go traffic), and I also located my IAT sensor outside of the intake completely. It now sits in a little radio shack black box just behind my driver's side headlight. I let the car cook today while eating lunch, came back, fired it up, and the AFR was right where I wanted it to be.

My methods are a bit controversial, though...after all, wouldn't you want to know the IAT in the intake tract itself? Apparently not in my application...that's why I don't go around advertising half the **** I've done to my car since it isn't along the mainstream set of beliefs about how a car should be tuned.

 

As long as the car is N/A, it seems like there should be no controversy surrounding your method, sounds like a good idea.

How is it that the EFIL users say this isn't an issue, what is different in their COS's that handle IAT heat soak in OL?

 

It is definitely N/A, and probably will be for the rest of its natural life.

And that's a good question about EFILive...I'm running a 1-Bar SD Enhanced OS from HPT, and this has always been a problem (until I went in and fixed it).

 

I wish the charge bias table was active in the '98s *grumble*

 

It's one of our Aussie members who touts OLSD and EFIL regularly... He says that IAT heatsoaking isn't an issue for him but has never gone into details over it. He goes as far as calling it a selling point for EFIL over HPT. I didn't think their 1-bar SD OS was that much different from HPT's (well, other than they don't get real time tuning along with their COS ).

 

Will the Cyl Charge Temp Bias setting not have any affect on this on a 98 tune? I know it doesn't have the "Complex" model capability but thought you could still work the bias table with some results?

 

No, I thought it didn't have any effect on 98-00 cars; thus why they say to upgrade to an 02 file if you can.

 

it's still not an easy table to work with... and you chase your tail a bit back forth from it to to VE trying to find a balance.

 

My approach was to nail down a VE at a consistent weather condition, then chase my tail for months screwing with the charge bias tables. I figure if the VE and IFR tables are right, the IAT charge bias must be the problem. However, making sure the VE table is only messed with in comparable weather conditions can be difficult by itself.

 

it's not that easy, because both temp and VE determine airmass, so you gotta figure out which one of these is the real cause of the airmass change. until you get that, you're solving x*y=1 sort of equations, so you got an infinite number of answers, but you dont know which one of them is the correct one. you gotta introduce more equations, basically constraining the set of answers. i couldn't figure out how to do it with histograms, i had to do a lot more complex modeling.

 

That is the beauty and EXTREME downside to multi variable equations. Some can be fudged to make up for others, but you never really know if you are right without vast (beyond my ability) data collection. My approach is to quantify specific variables under specific conditions, then use modifiers for other conditions. This is almost certainly NOT correct to math and physics standards, but I'm not running NASA in my garage either. BTW, glad to see you are still posting here. Your input is ALWAYS welcome.

 

Even then, you still need to know what the REAL IAT is. In an FI setup that needs the IAT sensor in the charged pipe before the TB, it's going to get heat soaked and not represent the reality of the air's temperature in the pipe. I can imagine that if it worked then the math could tune it, but how can it compensate for that massive error?

 

So is it the Cylinder Charge Temperature Bias vs Airflow table that can me used to adjust the differential? I don't see anything that relates to temperature.
I have to do something, tonight I was running over 16:1 AFR on the highway when I got into traffic and the IAT jumped to 136°. I thought I was going to have to pull off the highway and reflash it to richen it up. But I babied the car and it made it until I got up to speed and the IAT cooled down. It seems to really go lean over 100 °.
I have been saving my scans with the IAT as the file name so I can make a library as a reference of before I make any changes. I just need some help in knowing what to change, and which way to change it.

 

Charge bias is the relative weighting factor for the two
temperatures you have available as data (ECT and IAT).
ECT is pretty believable. IAT in my experience is not. If
you had a cleaner setup where the thermistor had only
air to warm it, no conductive path from the tube-wall to
the element, it would be more realistic. If the IAT were
inside the manifold then the speed-density equation
would be self-consistent (pressure and temperature of
the exact same air mass, not one before the throttle
and one after, etc.). There's too much fudging goes on
in the factory system. I think it would be cool to put an
IAT sensor suspended inside the intake next to the MAP
(why they did not integrate MAP and IAT, I don't know)
and try running with straight-IAT input. But do not know
if GM does other fudgery like calculating the "effective"
manifold air mass temp based on MAP/BARO expansion
factors and the like. Probably three people in the free
world (so called) know and they ain't talking.

Anyway, the purpose of the charge bias table is to
model how the real air temp changes as airflow does;
at low flow you're sitting in the runner and warming
to ECT; at high flow you're in and out before air can
change temp so it looks like IAT (here is where the
question, "which IAT?" comes in).

You can play with the IAT sensor itself, you can get
bare thermistors of proper calibration that respond
more / faster to air temp and have not much thermal
conduction path from the leads to the element
relative to the factory piece. But the factory piece
makes damn sure that the head does not go sailing
down the intake tract, which is nice.

If you are comfortable with assuming that the air
temp does not change much from outside to TB,
then relocating the IAT sensor is worth a shot.
At high flows this ought to be true. Relocate it to
somewhere in the inlet air path and away from the
radiator and motor. Or wrap the area around it in
insulation maybe. Anything to make it see more
inlet tract air and less wall temperature.

 

Frost: yes, you still need to know what the real aircharge temp is, however the only way you're gonna get it is through modeling from its components: IAT, ECT, Bias, and Lag Filter. This is why by regulating Bias and Lag Filter you effectively control the aircharge temp, but it is still only an estimator, not the real, actual aircharge temp. Because that temp is only one of the two unknown variables in a single equation, you still need more constraints to more finely define your optimum solution. The key though, is to derive a proper metric to assess how well your model works in reality, otherwise all your modeling and changes cannot be quantified, thus allowing comparing all of the different scenarios you come up with.

Once you have all these things, THEN you can actually attempt trying to figure out the whole bias/lag filter. however, because you're not really solving for a single variable, your result is going to include a new ve table as well.

The biggest obstacle so far is the injector data: offset, short pulse, and the actual flow rate. without these being dead on, your airmass and everything that gets derived from the airmass is going to be imprecise to the point of being useless.

 

EFILive has a "fudge" kind of table allowing you to mutiply VE by a factor based on IAT. I think getting the Bias table right is the "proper" way to do it. But a nice quick and easy option is the A0014 table in efi as discussed. This table is also used by FI guys who might need it BTW.

 

How can a fudge table make up for the fact that the sensor gets heat soaked?