AFR not targeting?

Jon k · Post by **Jon k** » Fri Sep 14, 2007 11:41 am

Hey guys, tuning an MS2/E Beta 9 car and we have an AFR table setup but for some reason it seems that during autotune and EGO correction both the AFR is 1 - 2 points richer than target, ie. 13.2 instead of 14.2. I've looking during autotune and noticed it pulling from 85% (limit) EGO correction to 100, but the afr is still richer than the target table.

Wideband is an LC1 0-5v 10-20 afr, setup in MegaConfig and MT.

Thanks!

Post by **muythaibxr** » Fri Sep 14, 2007 11:48 am

You might have your PID settings wrong. The EGO code has not been changed since we took it from ms2-standard.

Ken

diegom6 · Post by **diegom6** » Fri Sep 14, 2007 1:48 pm

Pardon my ignorance, what is PID settings?

Post by **muythaibxr** » Fri Sep 14, 2007 1:50 pm

The PID/Smith Predictor closed-loop settings. They determine how quickly and how close the actual AFR can get to the target.

Ken

rb26dett · Post by **rb26dett** » Fri Sep 14, 2007 1:57 pm

here ya go :

http://en.wikipedia.org/wiki/PID_controller

fred.

Jon k · Post by **Jon k** » Fri Sep 14, 2007 4:21 pm

Hm thanks guys - so are the defaults not good enough

Right now it goes:

10ms
4000 rpm
100%
20%
0%

Can anyone lend a hand. I am taking the car 70 miles tomorrow to do some road tuning on the way to a location.

UnaClocker · Post by **UnaClocker** » Fri Sep 14, 2007 5:59 pm

I wish we could get rid of the d*** PID O2 correction, O2 correction was flawless in MS1. Seems hardly anyone can get it to work right in MS2-E. Sometimes reinventing the wheel is not the right way to go.

Jon k · Post by **Jon k** » Fri Sep 14, 2007 9:35 pm

UnaClocker wrote:I wish we could get rid of the d*** PID O2 correction, O2 correction was flawless in MS1. Seems hardly anyone can get it to work right in MS2-E. Sometimes reinventing the wheel is not the right way to go.

I semi agree though I don't remember having issues back when I was running 1.0.2 or what not - but, I forget what my settings were!

Keithg · Post by **Keithg** » Sat Sep 15, 2007 8:32 am

I got tons of overshoot with the default settings. I had to back off P to 50 to get o2 correction to be more smooth. I do not know what my I and D are. Just play with it until the GEGO tracks reasonably. Until this is done, EAE tuning is a waste of time. Once I did this, I noticed that PID works *much* better than the old MS1 version and that I could actually tune EAE .

KeithG

Jon k wrote:Hm thanks guys - so are the defaults not good enough

Right now it goes:

10ms
4000 rpm
100%
20%
0%

Can anyone lend a hand. I am taking the car 70 miles tomorrow to do some road tuning on the way to a location.

Jon k · Post by **Jon k** » Sat Sep 15, 2007 8:38 am

Thanks Kieth - the P is what I have @ 100%?

What does the RPM value due?

And, by overshoot, do you mean it was going rich? Or lean? Or off in genaral?

Trying to figure some s*** out!

Keithg wrote:I got tons of overshoot with the default settings. I had to back off P to 50 to get o2 correction to be more smooth. I do not know what my I and D are. Just play with it until the GEGO tracks reasonably. Until this is done, EAE tuning is a waste of time. Once I did this, I noticed that PID works *much* better than the old MS1 version and that I could actually tune EAE .

KeithG

Jon k wrote:Hm thanks guys - so are the defaults not good enough

Right now it goes:

10ms
4000 rpm
100%
20%
0%

Can anyone lend a hand. I am taking the car 70 miles tomorrow to do some road tuning on the way to a location.

Keithg · Post by **Keithg** » Sat Sep 15, 2007 9:12 am

P at 100 is 100%, Back this off.

Overshoot in both directions. Too much increase and too much decrease of PW to follow o2. The transport delay values are for determining how much time shife is between the engine and the 02 sensor. You need both values correct. Look at a log and try to determine how long it takes from a PW dropoff to when the o2 sensor sees any change.

This is tough to do. If I use the fuel cut and look at a log, I look at the engine speed and how long it takes in s for the sensor to sense a change (when it starts to rise in AFR) and it still may not be time coherent with the engine control and you may have to tweak these values a bit.

KeithG

Jon k wrote:Thanks Keith - the P is what I have @ 100%?

What does the RPM value due?

And, by overshoot, do you mean it was going rich? Or lean? Or off in genaral?

Trying to figure some s*** out!

Post by **muythaibxr** » Sun Sep 16, 2007 5:44 am

The old ms1 way of doing things resulted in a nasty oscillation that wouldn't allow you to say... run closed-loop while free-revving.

The new PID/Smith predictor works much better when tuned right.

I generally start by setting P to 50, and I to 10 right off the bat. Then I work on the predictor numbers... set the time to the measured response time that a change of fuel takes to go from you changing it, to the AFR reacting... set the RPM to the revs where you saw that delay.

Then I fine-tune the P,I, and D settings by driving around, and getting it so that the AFR quickly gets to the target with minimal overshoot.

Once I get this close, I tune EAE, which usually gets me to the point where the AFR stays flat except for hard stomps...

The hard-stomp behavior means it revs, but there's a small lean-spike. Right now it's possible to get close to getting rid of this, but it can't be removed all the way. I'm going to be working on code for 2.1 that'll get rid of what remains of that spike.

Ken

redmist1 · Post by **redmist1** » Mon Sep 17, 2007 10:54 am

I gave up on 02 correction. I just do lots of logs with all enrichment off and let Megalog viewer do the correction of the VE tables and tweak the idle settings.

Post by **muythaibxr** » Mon Sep 17, 2007 12:00 pm

redmist1 wrote:I gave up on 02 correction. I just do lots of logs with all enrichment off and let Megalog viewer do the correction of the VE tables and tweak the idle settings.

Shouldn't be necessary if you tune everything correctly.

The biggest problem people have is when using normal AE with PID closed loop. If you don't set a delay after AE to enable closed loop, PID often overreacts. You can fix this with PID settings, but most people would rather just set the delay.

Ken