EAE with or without prediction? (Success with no prediction)

All questions about MS2/Microsquirt/Microsquirt module. See also MS2/Extra manuals

Moderators: jsmcortina, muythaibxr

Post Reply
KaPower
Master MS/Extra'er
Posts: 501
Joined: Wed Dec 22, 2004 9:43 pm
Location: Bonney Lake, Wa

EAE with or without prediction? (Success with no prediction)

Post by KaPower »

It's time to get this ball rolling.

note... I'am one of those guy's that's always had problems with Accel/tip in, until I switched to 4squirts alternating.

With limited testing and tuning in the driveway (my nieghbors hate me!), I have had some success EAE no prediction. Mainly with a mapdot below 200 kpa/sec, above that I start getting a lean spot.

I think it is becuase of the lag time between mapdot and inj pw output from MS. A slow increase in map allows MS time to respond.

With prediction turned on, I can limit the lean spot to no more then about 1 afr. Things get really rich after the accel right now (limited tuning & testing).

I tried setting the prediction to (1 at 500&600 kpa/s & also 1 at 600 kpa/s)

I think some prediction at lower kpa/s might be needed.

With prediction turned on, I get an EAE prediction event, up to 10.8ms on my datalogs. This is way to much fuel with my giagantic 740cc injectors, I should only have half that much fuel.

Usually I get the those huge numbers during my highest map dot events, which are up to 1400kpa/s.

More testing tomorrow... Hopefully this thread will eventually get normal Megasquirters with single TB's, an idea of were to start.

Here is a small datalog.
Last edited by KaPower on Sat Apr 07, 2007 4:51 pm, edited 2 times in total.
muythaibxr
Site Admin
Posts: 8228
Joined: Thu Oct 14, 2004 12:48 pm

Post by muythaibxr »

The large pulse-widths are the result of where the prediction is going in your table... If it's overshooting into a boost area, or just overshooting too much at all, you'll get some errors.

I wrote that prediction because I noticed the same behavior though... the MAP seems to lag a bit, and the pulsewidth change lags a bit behind that... I did it to make the MAP peak sooner so the pulse-width would follow.

With the ITBs, another way I was considering fixing this was to use alpha-n in one table, and speed density in another, and combining them in such a way that I get good tunability where speed density doesn't make sense, and get better response even in the area where speed density does make sense.

Ken
KaPower
Master MS/Extra'er
Posts: 501
Joined: Wed Dec 22, 2004 9:43 pm
Location: Bonney Lake, Wa

Post by KaPower »

Ok, let me get this straight. The prediction code looks at the ve map, and not the EAE map?

If I lower my max load kpa number will that prevent so much overshooting? I have it set to 255, normally my max boost is 170-180kpa. I would try setting it to 200 kpa, but I don't like that, becuase if my engine overboosts for some reason, as you know, it will go really really lean...

Prediction looks very, very promising, but I need to limit this overshooting. I need the normal amount of AE, just sooner then standard AE. I don't need more then a 2.5-3ms pw increase for AE. Right now I'am getting 5-8ms AE with the spike.

Can you put a user definable cap on the amount of prediction without affecting transitions into boost?

JM
KaPower
Master MS/Extra'er
Posts: 501
Joined: Wed Dec 22, 2004 9:43 pm
Location: Bonney Lake, Wa

Post by KaPower »

I don't think prediction will work on a boosted engine. It's not safe to set the upper limit to a low value, and *I think* thats the only way to limit overshooting PW's.

JM
KaPower
Master MS/Extra'er
Posts: 501
Joined: Wed Dec 22, 2004 9:43 pm
Location: Bonney Lake, Wa

Post by KaPower »

Success!

No prediction is needed for me.

Ken, you should look at this datalog, It's short but sweet!

When first trying to tune EAE, it felt like I just started Megasquirting all over again... (Still does actaully..:)

My tune is still ruff, & fast tip in during idle in the driveway needs some work, but under load, while driving things are looking close to where I want them, and feels great. No stumble's at all.

The thing that helped me is running higher Adhere to Wall numbers then I thought was necassary.

20% at idle, and up to 60% at 100kpa. (65% @ 120kpa not sure what to do in boost right now)

I jump up to 35% off idle, and then increase 3-5% every 5kpa.

-My map lag factor is 80... That may need to be calmed down some. My sucked from wall numbers are still very raw.

-4 squirt alt injection strategy

For the first time ever, I have no lean spikes while driving (small spike during tip in, @ idle w/ no load), and I still Can't get 2 squirts alternating to work for me. It still gives me a stumble no matter what AE strategy I use.

I can see EAE making adjustments while in boost, but my AFR was still Flat.


I played around with Launch control once or twice, so don't worry about that in the datalog.


Jason M
Last edited by KaPower on Mon Apr 16, 2007 8:13 pm, edited 1 time in total.
muythaibxr
Site Admin
Posts: 8228
Joined: Thu Oct 14, 2004 12:48 pm

Post by muythaibxr »

I think when you let of the throttle it's going a bit rich... that can be fixed by increasing the sucked off the walls amount at the lowest-load part of the graph. If you increase the amount in the graph for the sucked off the walls value, the MS2 will assume that more fuel is getting pulled off the walls, and will inject less fuel under those conditions to compensate.

The MAPdot prediction may need some more thought to make it work for boosted engines.

Once the engine acts right at mid revs, you should adjust for idle and high rpms using the RPM EAE curves.
KaPower
Master MS/Extra'er
Posts: 501
Joined: Wed Dec 22, 2004 9:43 pm
Location: Bonney Lake, Wa

Post by KaPower »

When using the prediction code, does a 0 mean no prediction?

Or is it still predicting at all times?

JM
muythaibxr
Site Admin
Posts: 8228
Joined: Thu Oct 14, 2004 12:48 pm

Post by muythaibxr »

0 means no prediction. 1 means just normal prediction, above that, it's "amplifying" the prediction... or predicting that many ms*10 ahead of actual conditions... so the more you increase this, the more likely you are to overshoot.

Ken
KaPower
Master MS/Extra'er
Posts: 501
Joined: Wed Dec 22, 2004 9:43 pm
Location: Bonney Lake, Wa

Post by KaPower »

I set all of the rpm correction tables to 100%. I'am going to get my added to walls & sucked from walls curve's adjusted first. (I'll try 3000rpm area while cruising/loaded in 3rd gear) Then I will adjust the RPM correction bins to suit high, and low rpm's, & of course tip in at idle w/no load.

Although, I first got my added to wall curve ruffed in, in the driveway with tip in off idle...

I also set up a 10kpa VE column on my VE map. I think that should help with the rich decel (I tried maxing out sucked from walls (25.5%) and it's still rich after lifting)

Is it normal to have EAE drop to zero, and bring the injector PW to zero as well?
KaPower
Master MS/Extra'er
Posts: 501
Joined: Wed Dec 22, 2004 9:43 pm
Location: Bonney Lake, Wa

Post by KaPower »

Idle pw drop out/EAE spike to 0%?
muythaibxr
Site Admin
Posts: 8228
Joined: Thu Oct 14, 2004 12:48 pm

Post by muythaibxr »

That can happen if the numbers are too far off, I saw some odd behavior once or twice when tinkering around on the bench with numbers that didn't make sense. It looks like there were some other odd things here and there while you were idling, but only the one time caused a complete drop in pulse-width.

When you're testing EAE, you should also log "wallfuel" as that can help me tell what's going on.

It looks to me like the combination of settings too far from correct, combined with the MAP jitter may have caused that problem.

Your 10kPa idea might be good, as well as using fuel-cut. It looks like running rich on boost causes a lot of fuel to get stuck on the port walls... Thinking about it, you might just need to have the EAE tables cover boosted areas, and make the curves continue along the same curve you've got tuned to 100kPa now.

Ken
Post Reply