MAP sensor sampling

[Laaf]

I want to try this

[prj]

Accel. multiplication, going to see if it's going to be off of TPS or MAP delta.
I call it multiplication, because this is for calculating air charge. Motronic has a wall film model to calculate fuel enrichment on top of that, but no accel. multiplication, as that's not needed with a MAF sensor due to the much quicker response of the MAF signal.

I have both calculations working, going to stab the throttle and compare.
The MAP/LOAD delta calculation is running crank-synchronous, the TPS delta calculation after some testing I put into a 10ms loop.
Both output fairly similar values, the question is only whether the MAP/LOAD delta is quick enough or if I need the TPS.

[quattro84]

Now also wrote the algorithms - MAP * VE(MAP, RPM) * IAT...
Normally ECU stock acceleration/deceleration enrichment is based on load gradient.
Maybe I will need to change acceleration enrichment to use TPS gradient - will need to test.

But sampling every X teeth, then summing it provides a good and stable signal regardless of RPM.

MAP/alpha is quick enough. Of course MAP is slower compared to MAF because you measure the same air at a different phase in time.
Ignition retard on transient is good, actual retard needs to be mapped rpm/alpha for optimum results, giving better for response on turbo cars and saves the engine from hick-ups under rapid throttling, not sure about the amount of fueling enrichment to use, would map that as well if possible.

I wouldn't use TPS as a direct input for acceleration enrichment because it is far more dependant of other criteria and i think this can give unwanted "multiplications".
Next to that it will be more robust on MAP/alpha for using it under different ambient conditions and of course systems a.k.a. cars/customers.
In the most ideal situation you would like to map it for different conditions and interpolate directly in between those maps. (i think)
These engines don't rev extremely fast so i think the actual delay between actual/measured mass flow is acceptable.

But, if understanding your thoughts right, this will need to be dialed in for every different configuration in order to work? Or do you want to scale the axis for MAP? Could be done "easily" once you would have a MAP reference map based on rpm/alpha.

Next to that with running MAF vs. Speed/density you will be able to map the actual VE (perhaps not actual VE but at least the trend) Why not try to model something like this?
This is how the OEM ECU's of this era work, use reference mapped MAF, calculate transient norms in order to shift setpoint for fueling/timing/boost/etc. using mapped VE and using estimators with offset errors based on inlet temp/humidity/pressure/...
I know this is far too much to get crammed into Motronic, and nowhere needed since you want to find optimum power & driveability without overfueling.

HTH.

[prj]

MAP/alpha is quick enough. Of course MAP is slower compared to MAF because you measure the same air at a different phase in time.
Ignition retard on transient is good, actual retard needs to be mapped rpm/alpha for optimum results, giving better for response on turbo cars and saves the engine from hick-ups under rapid throttling, not sure about the amount of fueling enrichment to use, would map that as well if possible.

I wouldn't use TPS as a direct input for acceleration enrichment because it is far more dependant of other criteria and i think this can give unwanted "multiplications".
Next to that it will be more robust on MAP/alpha for using it under different ambient conditions and of course systems a.k.a. cars/customers.
In the most ideal situation you would like to map it for different conditions and interpolate directly in between those maps. (i think)
These engines don't rev extremely fast so i think the actual delay between actual/measured mass flow is acceptable.

But, if understanding your thoughts right, this will need to be dialed in for every different configuration in order to work? Or do you want to scale the axis for MAP? Could be done "easily" once you would have a MAP reference map based on rpm/alpha.

Next to that with running MAF vs. Speed/density you will be able to map the actual VE (perhaps not actual VE but at least the trend) Why not try to model something like this?
This is how the OEM ECU's of this era work, use reference mapped MAF, calculate transient norms in order to shift setpoint for fueling/timing/boost/etc. using mapped VE and using estimators with offset errors based on inlet temp/humidity/pressure/...
I know this is far too much to get crammed into Motronic, and nowhere needed since you want to find optimum power & driveability without overfueling.

HTH.

I have all that and more in Motronic, don't worry.
Also you are using mass flow and speed density interchangeably a lot.
Anyway, will test how delta behaves today and will just use the quicker signal.

The accel enrichment I do from a base map that is RPM x Cylinder filling and multiplied by a factor that is mappable based on current delta. The delta can be the delta between calculated cylinder filling or it can be the TPS delta (rate of change). I will use whichever responds better.

The VE is modeled as a RPM x MAP table just like every other ECU, no problem at all and additionally has IAT correction.
I am not worried about speed, I really don't lose any, because what I lose I gain by throwing out the MAF algorithms out of the ECU completely.

Fueling enrichment as such is done based on a wall wetting model in this ECU based on delta load (cylinder charge).

The retard is done based on knock sensors. If there is knock in any of the cylinders during a rapid TPS gradient it retards all cylinders by a set amount. Works very well.
It then learns this value, and next time there is a gradient this value is instantly substituted. If no knock occurs, it's stepped back bit by bit, remembered, then stepped back fully once rapid operation of the throttle ceases. So next throttle bashing will remember the learned value...

Basically - I am discussing only the capturing of cylinder charge. You are getting a lot into the fueling side of things as well - but the fueling side of things does not get affected whether you have mass flow or SD, as long as the cylinder charge is calculated correctly.

[prj]

I just re-read, and I am giving too little info to start with, so you didn't have a lot to go on.

Basically, stock ECU:
* Capture LOAD directly from MAFLIN(MAFVoltage), divide by RPM, no correction done to LOAD from transients, as the MAF is fast enough.
* Do accel enrichment based on a wall wetting model.

"My" ECU:
* Capture LOAD from MAP, correct by VE(MAP,RPM) and IAT(Degrees).
a) Calculate delta from captured LOAD between cylinders, multiply LOAD further based on Base(RPM, LOAD), Multiplier(Delta LOAD)
or
b) Calculate delta from captured Throttle position in 10ms loop, multiply LOAD further based on Base(RPM, LOAD), Multiplier(Delta Throttle position).
* Do accel enrichment based on a wall wetting model.

With the MAP, the problem is that LOAD does not keep up on transients, so LOAD has to be corrected, so that ignition is correct too.
The wall wetting is a separate problem and it's exactly the same for SD and Mass Flow.

Once I get this working, it's time to write this for ME7

[quattro84]

Hmm okay i get what you are after.
But in the end it is still leaning on the stock Liquid fuel effect model, or do you know which inputs & booleans to tweak in order to update it to current situation?
Since emissions weren't that strict at the time i am not sure they divided the model up for droplets & wall film.

So you at least need stock corrections for IAT/charge temp @ injector which will be there i guess, but needs lots more such as fuel pressure, injector spray pattern/efficieny if changed compared to OEM calibration.

How do you want to tune this model anyway? On a testbed i would go for HC emissions, combustion delay, dQ/dt

Good luck.

P.s: What error percentage do you see now? (or expect) How great are the gains or what are your final achievements?

Route A sounds like a more robust plan.

The way you describe stock operation makes me think that it is not a great deal, not correcting sensor reading to actual cylinder filling, so a charge error is not noticable during transient.
That actual gap is what i called a Transient norm. Bigger air mass gap = bigger throttle actuation => asks for more reaction in order to overcome response and emissions. Still think that a corrected speed/density load is close enough.

I have to think how i would like to have this working for port injected corrections. (have not looked into it because it never was an issue, models most likely build by the same people that did the bits you are looking at right now )

[prj]

So you at least need stock corrections for IAT/charge temp @ injector which will be there i guess, but needs lots more such as fuel pressure, injector spray pattern/efficieny if changed compared to OEM calibration.

I am not touching wall film model. I will leave it like it was mass flow, because I am not tuning injectors! I am tuning the load charge calculation. We are actually using stock injectors right now.

The way you describe stock operation makes me think that it is not a great deal, not correcting sensor reading to actual cylinder filling, so a charge error is not noticable during transient.

I think it's a big deal because right now it's way off. And it's easier to correct air charge when I have a good wall film calibration.

I am hoping to compare delta calculation within the hour, but right now it seems like TPS is faster than MAP. Might turn up the data logging sample rate up a little, right now it's 20hz, maybe 60hz will show better.

[prj]

P.s: What error percentage do you see now? (or expect) How great are the gains or what are your final achievements?

Not achieved much yet - but right now if you go anywhere near the throttle the car bogs down, unless you abuse the wall film compensation really hard
The MAP signal is not even close to being as sensitive to rapid changes as the MAF signal.

Also if you compare pushing throttle MAP vs MAF, then with MAF LOAD ramps MUCH faster. With MAP it takes a while to go up, and that's why there is a transient lean condition right away.
That's also why I am correcting LOAD rather than messing with fueling. I'll post some logs afterwards...

[prj]

Tests show the signals very similar, but TPS is ahead as expected during hard gradients.
Need to tune the VE map steady state first, then transients.

[quattro84]

Starting to understand what you are after.
I did some similar calli work for reducing PM emissions using throttle actuation as a main input as well.
But the on diesels you tend to lean more against a fuel limiter based on PM emissions, tuning it for max response up to an allowable PM production.

MAF readings are always faster, but if the actual reading is of no use then responding to Alpha would be a better because it lacks the system inertia.
Your load corrections will work once mapped against rpm,alpha so low throttling actuations can be ruled out for fuel wetting corrections, not sure if you can run high MAP i.c.w. low load (antilag on EGAS?) but that needs some extra inspection.

The way you approach this way of development would ask for some more insight in combustion engine modelling, info which can be found if one looks a bit further...
They come with similar results, MAP being slow, using time for reaching it's asymptote

http://e-collection.library.ethz.ch/eserv/eth:27550/eth-27550-02.pdf

http://www.macchine.unisa.it/MECA01/Papers/MECA01_paper05.pdf

Can you make me a high freq log with a load step which shows MAP vs MAF vs AF?

[prj]

Well, I can't do MAP vs MAF, since to read the MAP I have to remove the MAF code.

I altered my code somewhat to sample differently, and I think I have a bit better response now than before, so not sure the TPS will be needed at all.

Right now I am just enriching off of wall film like with the MAF and with a hot engine this seems perfectly fine.
Have to run some more tests when cold.

Interesting literature, thank you.

[quattro84]

Ah, i figured you had found a way to run MAP & MAF simultaniously with an enable/disable map for real time switching

[prj]

Ah, i figured you had found a way to run MAP & MAF simultaniously with an enable/disable map for real time switching

No, I don't want to load the ECU too much.
I actually like the fact that the MAP code is more compact and takes less time to execute
It can be swapped back and forth, but it's either one or the other, as I overwrite a lot of the factory code.

I am actually getting better results sampling once at 10 bit just before the air calculation for the cylinder. Seems more responsive than sampling 3 or 6 times per 54 teeth and then using the summed value, yet still very stable - no issues with the MAP signal being jumpy at all.

[Witrotseun]

This is abit above my understanding, and might seem abit like a stupid question but how would altitude differences play with your algorithms, for example:

Here in SA we have very high altitudes 1800 odd meters above sea level right down to sea level. As all we have known is aftermarket ecu's for the last couple of years and software changes only coming into fasion recently. Over here anyway. Where a car had needed remaping where ever it goes. A car would start behaving strange 3 or 4 towns down the road.

[prj]

This is abit above my understanding, and might seem abit like a stupid question but how would altitude differences play with your algorithms

Altitude changes do not make any difference whatsoever to this, because it's based on absolute pressure, not on relative.