Megasquirt Support Forum (MSEXTRA)

milesinfront wrote:To me Motronic 1.0 and 1.1 feel crude compared to this code base

Yeah EAE had it's moments of greatness for me, but over-all it was a dog. I'm hoping for better results from round 2!

milesinfront wrote:I may check with Gtech this weekend anyway.

Datalogs can help here! I hit the space bar as soon as I launch then use the speed in gear feature of MegaLogViewer to determine the stop point. I do have a G-Tech as well though...

Thanks, I'll try the spacebar trick

Here's the datalog and msq from last night

Looks pretty good around the 2,000 rpm mark. I love how flat your AFR gets at times.

Some of the time your AFR's move in steps. Do you have a sampling delay set in the LC1 controller?

Just like my car, your AFR's get pretty 'whack' once the PW's go down to around the 1.0ms range.

By using the 'changing the squirts' method I determined my opening time to be 0.76ms. I'm constantly questioning myself on this one, but no matter how many time I rerun the test, 0.76 comes out as the winner. (i.e. @ 0.76 I can alter the number of squirts from 1-3 without any major change in AFR.)

I think that making a VE table with a constant AFR Table would make EAE tweaking a little easier. ANY deviation from a constant AFR of say 13.5:1 would indicate if EAE changes were necessary. Of course this means retuning an entire VE table...

I keep forgetting to ask! Does anyone know how MegaLogViewer deals with EAE? Can you use MegaLogViewer with EAE enabled? (VE analysing...)

I'm going to check 80-100kpa at 3000 to make sure it's ok
Also 4000 and 5000.
Why? Because I'm autocrossing tomorrow (last minute impulse signup)
Will take some in-car video of so-so tune and crappy driving.

Tonight I did my 1st couple of EAE runs with 3 squirts sim. The original settings were much closer to the mark this time round and the car is mostly drivable. Still bucks a little during shifts, but not as bad a before. I've datalogged and analyzed twice this evening and things 'seem' to be heading in the right direction (~2500rpm) but I'm still not so sure that I'm in total control. I still have trouble deciding what to alter when the map is dead flat and the mixture is lean due to EAE% slowly ramping back to 100% from 85-90%. If the map is flat, why is EAE still 'pulling' fuel? I'm sorta ignoring those conditions for now and altering the easy ones. Adjusting A2W when map is increasing and adjusting SFW when MAP is decreasing. I'm not convinced that this is entirely the right way to do this, but it's all I know so far....

Going back to 100% slowly is an indication that Added to walls and sucked from walls are too far apart.

Ken

muythaibxr wrote:Going back to 100% slowly is an indication that Added to walls and sucked from walls are too far apart.

Cheers! I'll use that!

bouncing this thread...

I had EAE mostly playing nice. The major problem I had was that dropped throttle would go full lean then back to stoich. The problem is that this lean spike would mess with O2 correction and it would over correct. It was most prominent on gear change.

Ken says 'slow return to 100% means that the values are too far apart on AWC and SWC'. Hmmm. The scales differ by a factor of 10. WHat, practically, does this mean. My assessment was that my dropped throttle was 'too hard' and too much fuel was pulled. Another comment I have heard is: 'the size of the puddle drives things'. I thought about this a bit and seem to agree. My reasoning is that if dropped throttle on gear change causes a huge lean spike, the puddle size is too small. Also, I noted that the EAE % went pretty full range and swung very quickly.

My curves were very flat and SWC was way up in the 20s max to min was 24 to 20, maybe. AWC was very low as well and very flat 2 to 18, maybe. SO, I started over. I flashed a new 2.0.1 and saved the default MSQ. Woah! curves are very vertical. I put these in my MSQ manually. Tried to drive it... Woah, again! EAE% needle swings very slowly in comparison. Much bigger lean outs on gear change and lots of pig rich accels. I have gone and flattened both curves a bit. It'd be nice to know how to tune it to:
1) make EAE swing faster vs slower
2) make the size of the puddle shrink or grow

so far, not a lot of improvement.

This is on a turbo 2.0L 16V engine. You'd think that with 16V, low RPM would have lower velocity and less fuel would be sucked from walls and less AWC would be needed to feed the puddle, but I have not figured out how to get this to play nice.

I'll look at Peter's MSQ and see what his curves look like. Anyone have a turbo set of curves they'd like to post (for EAE that is tuned pretty close)? Is anyone successfully running this on a similar engine to mine: medium disp 4 cyl turbo 16V (or 20V)?

I know this can be tuned, but I just cannot seem to eradicate the dropped throttle leanout. 40-120kpa accels were fine at 2krpm 80 to 30 kpa decels also looked fine. It was the gearchange that messed things up. If O2 correction could be turned off <30kpa, I am sure mine would act much better, at least until I get this tuned. Grrr!

KeithG

Note I concentrated on tuning one RPM bin. I left EAE all the time and didn't use any O2 correction. I also left the RPM correction curves at their default shape around the 3000 rpm bin I chose and then flattened them out above and below that plane.

I turned EAE off last night to do some datalogging and hated the feel of normal AE.

I found my LM1 was a little out of cal (about .75 AFR lean, so tune was too rich), so I'm retuning the VE map with my new LC1 and will resume testing next week. Last night's datalog has the VE reasonable close.

But even this morning the car feels better with EAE than with normal AE.
I'm hooked.

Pick one RPM bin and get it working there.
When you make changes, make them equal geometrically, not arithmetically. (not same count larger or smaller in adhered vs. sucked from)

Keithg wrote: 1) make EAE swing faster vs slower

Increase the number of squirts is the first thing that comes to mind. How fast or slow it reacts depends on this, because it only adds and subtracts from the walls once per squirt.

2) make the size of the puddle shrink or grow

Bench testing would help you a lot here. Basically, increasing added to walls, or decreasing sucked from walls will make the puddle grow, and make the response slower. Any time you make the curves further apart, you get this effect. The curves aren't on the same scale because the amount you're squirting in ms is much lower than the total number of ms worth of fuel that is on the wall... a low percent of a high value matches a high percent of a low value. Added to walls is a percentage of a single squirt, but sucked from walls is a percentage of wallfuel.

This is on a turbo 2.0L 16V engine. You'd think that with 16V, low RPM would have lower velocity and less fuel would be sucked from walls and less AWC would be needed to feed the puddle, but I have not figured out how to get this to play nice.

Lower velocity means less sucked from walls, and possibly more added.

I'll look at Peter's MSQ and see what his curves look like. Anyone have a turbo set of curves they'd like to post (for EAE that is tuned pretty close)? Is anyone successfully running this on a similar engine to mine: medium disp 4 cyl turbo 16V (or 20V)?

I know this can be tuned, but I just cannot seem to eradicate the dropped throttle leanout. 40-120kpa accels were fine at 2krpm 80 to 30 kpa decels also looked fine. It was the gearchange that messed things up. If O2 correction could be turned off <30kpa, I am sure mine would act much better, at least until I get this tuned. Grrr!

KeithG

Dropped throttle leanout means there is way less fuel coming out of the walls than you have tuned. Lowering sucked from walls would probably fix this in those areas.

Also, I'm assuming that you're talking about when you're on boost... do your EAE curves go into boost? If not, then there is something going on in the puddle that you're not representing in your curve, which could have an effect too.

Ken

Peter Florance wrote: Pick one RPM bin and get it working there.
When you make changes, make them equal geometrically, not arithmetically. (not same count larger or smaller in adhered vs. sucked from)

This is a good way of describing it.

Ken

muythaibxr wrote:

Peter Florance wrote: Pick one RPM bin and get it working there.
When you make changes, make them equal geometrically, not arithmetically. (not same count larger or smaller in adhered vs. sucked from)

This is a good way of describing it.

Ken

Thanks Ken.

"percentage of percentage" didn't sound right to me, but that was my first answer.

muythaibxr wrote:

Keithg wrote: 1) make EAE swing faster vs slower

Increase the number of squirts is the first thing that comes to mind. How fast or slow it reacts depends on this, because it only adds and subtracts from the walls once per squirt.

I mean the gage needle in EAE%, not amounts of fuel. With the default curves, it is much slower in moving. With mine, it appears as a quick stab.

KeithG

Yeah, I knew what you meant. My comment still stands. The WF value is only updated once per squirt (although now that I'm thinking about it, I want to go back and verify that), so if you decrease squirts, you decrease the responsiveness of the algorithm.

I'm going to have a whole lot of fun trying to get this to work with sequential injection. Can we say 8 wallfuel variables?

Ken

I'm not saying anyone is not thinking this way, but I'm careful not to think of these as commands or settings but rather letting the software know what the model is doing at different operating points of the engine. My gut feeling is model-based software is part of our future, so we need to be able to think this way.

It's validated by the fact that this is the best tune I've ever had, even with the half-assed EAE tune I have so far.

It did however, give me a headache for a while trying to understand it.

Again Peter, you're right. I was already thinking about it the way you describe, but of course it doesn't occur to me to say it out loud. (Or in this case type it).

When you make changes, make them equal geometrically, not arithmetically.

Do you mean that the ATW & SFW curves should look similar? So if it goes rich or lean at a given point to adjust both tables?

So, at the risk of appearing an idiot, does anyone feel like looking at these and giving me any hints... and no, I cannot increase # of squirts. I do have lag comp turned on, or have had. This is not a stomp, but a lift and accel.

KeithG

milesinfront wrote:

When you make changes, make them equal geometrically, not arithmetically.

Do you mean that the ATW & SFW curves should look similar? So if it goes rich or lean at a given point to adjust both tables?

The curves should have roughly the same shape... at least similar.

Ken

Keithg wrote:So, at the risk of appearing an idiot, does anyone feel like looking at these and giving me any hints... and no, I cannot increase # of squirts. I do have lag comp turned on, or have had. This is not a stomp, but a lift and accel.

KeithG

I'll take a look this weekend.

Ken