Recently I've exposed the actual air density calculation as a tuneable curve: http://www.msextra.com/forums/viewtopic.php?f=125&t=46816
I still think that the MAT/CLT concept has to have some merit, even though it seems nobody has yet had much improvement with it !
One factor that might be throwing off testing is that the VE table and warmup will almost certainly need re-tuning once MAT/CLT is enabled. We all seem to agree that the air in the cylinder just before the intake valve closes has to be a different temperature from the measurement on the MAT sensor. With regular non-alcohol fuels, this actual temperature is going to be hotter than we measured.
1. On a cold day with a cold engine at startup.
Say 0C outside temp, and 0C actual temp.
2. On a cold day with a fully warmed up engine and a remote air temperature sensor, there will be a large difference at low rpms.
Say 0C outside temp, and perhaps 100+C actual temp.
At at theoretical zero rpm, the ingested temperature will surely be equal to the engine temperature. Heywood cites an example of modelling an intake manifold as a 1.4kW heater.
3. Same deal at high rpms
Say 0C outside temp, and perhaps 30C actual temp ?? (guess)
The difference between 1 & 2 is presently handled somewhat by warmup enrichment. We have to apply a lot of WUE when cold in part due to this airdensity difference. EDIT: Not sure WUE will be any different.
The difference between 2 & 3 is presently handled through the VE table.
2 & 3 in particular highlight what I think I'm getting at. Our air temp sensor still says 0C, but the actual temperature ingested is wildly different. (24% absolute.)
Using MAT/CLT would likely alleviate a lot of the low rpm or idling "heatsoak" issues because the air temperature (it seems) is already dominated by the engine temperature. If we've already tuned for that reading and set the MAT/CLT curve to take little account of outside MAT at low revs, then the impact should be greatly reduce.
Perhaps the curve should be MAT/CLT% vs. air flow.
air flow = MAFMAP * RPM or MAP * VE * RPM depending on the algorithm in use.
However... all of this needs testing. Testers would need to start from a guessed MAT/CLT curve. 100% CLT at 0rpm to 100% MAT at infinite RPM. What are the numbers in between?? Then re-tune WUE and VE. Then see how it behaves in hot traffic and hot vs. cold days.