NefMoto

Hey guys,

So i've been lurking around here for a while and I'm trying to learn as much as can, but I could use a hand here.

So the problem I'm having is mild overboost (~150 mbar) issues.

I've spent a lot of time working on the "base" tune that was posted by Gonzo some time ago: http://nefariousmotorsports.com/forum/index.php?topic=1272.0 (http://nefariousmotorsports.com/forum/index.php?topic=1272.0)

When I first started messing around with gonzo's tune the boost pressure was super unstable at WOT and would surge and be generally crappy. So I changed the PID using the information posted in this thread: http://nefariousmotorsports.com/forum/index.php?topic=1449.0 (http://nefariousmotorsports.com/forum/index.php?topic=1449.0) 

NOW, the boost is super solid and the car pulls great BUT I get the overboost issue.

I'll attach, my current tune version, the tune I used as a base,the stock tune for my car, and some logs.

I'm still tweaking the PID to prevent the overboost but I can't seem to figure it out. If anyone could help me/point me in the right direction it would be most appreciated. Is there anything I should change since I'm running a K03 rather than a K04?

If you are running K03s, all of your PID maps should be closer to stock.

In particular, DRL

Also, when working with the PID, you should be logging:

ldimx_w, ldimxak_w, lditv_w, ldptv, ldrdtv, ldtv, ldtvm, ldtrv_w

Ok, I can do that on my way home and post them.

I set KFLDRL back to stock, I'll see if that gets me any closer to where I'm trying to be.

Thanks

Keep us updated! I will try to help as best as I can.

Ok so I flashed an updated tune before I left work (Same as AWEdit2.5 except KFLDRL was reset to stock values.)

Changing KFLDRL back to stock made partial throttle pretty "chunky" and the WOT overboost issue was worse if any different. A lot of the PID tables are still totally different than stock, mind you. I could get some logs with the altered KFLDRL tomorrow if that would help.

It was hard to get a good pull without getting lots of wheel slip or hydroplaning and dying since it rained a bunch this afternoon.

Anyways, here are what logs with the PID. (both are 3rd gear pulls fwiw)

Ugh. Put them all back to stock and start from scratch :/

How do I use the PID results to determine what I should change for future reference?   

Good question :)

There is no easy answer, tuning a PID is VERY tricky business... but if your turbo and wastegate are stock, you should only have to make minor changes.

I'll try to help as best as I can.

If only someone wanted to clarify the PID section of the wiki but lacked the appropriate help to do so!

Dude. Tuning a PID is its own art... I wouldn't even know where to start considering what I do to the PID is generally hack it to shit until it does more or less what I want it to.

There is literally decades of literature on the topic, most of which is beyond my understanding.

lol

I know it. That's the problem I'd like to rectify.

So in KFLDRQ0,1,2

- one axis is RPM (or so it seems)

what is the other a measurement of? Mine goes 100, 200, 400, 700

for S4 Mbox, (actual pressure - "ground" pressure), or pressure over ambient in hPa (mBar)

Not sure of your file, but i assume your "ground" pressure is also ambient.

You'll probably have to find an ME and a math PhD to advise :P

So is it deviation from the requested load? e.g. Motor running at 4k, requested load is 1500 actual is 1000, apply value at (500,4000)? Or am I totally off ?  Are your values also under 1000 like mine?

whoa whoa... no

pressure and load are two completely different things

it is actual - ambient.

ambient is (probably) near 1000hPa ... :P

do you mean mBar?


So if the turbo or whathaveyou is pushing 1 bar then the x-axis of the table will be 1000(but actually 700 cause that's as high as it goes)?

The input is lde.  lde = measured - setpoint which is basically the error part of the PID controller.

Checkout the LDRPID section of the FR, the diagrams might make some sense.

I'm assuming the 700 you speak of is the lde axis, which means that if lde>699 then it will be in that section of the map.

The other input is nmot (rpm)  The output of he maps, say KFLDRQ2, form part of the PID math which becomes the duty cycle of the controller (afaik).

I'm researching this at present and am looking through the code etc so someone might correct me.

Forgive me for being a nob, but what is the FR?

I had figured the axis was something along the lines of error/deviation, so being LDE makes sense -- I vaguely remember reading that at one point. I'll start logging LDE so I can try to piece this shit together.

I flashed back to AWEdit2.5 this morning. here are some logs. The load curve looks better than last evening (KFLDRL changed).

To me, the curve looks good, or at least it looks like a typical pid result curve. So in theory, wouldn't I be able to tweak the P or something to control the overshoot without making my rise time go to bot? At least I think it's the P that I should tweak (LDRQ0S?)

http://nefariousmotorsports.com/forum/index.php?topic=400.0title=


The axis is not lde



No, you should see a pronounced dip in wgdc right before peak boost


No. P will not help overshoot. You will need to increase D.

Also, please consider using ECUxPlot rather than doing all that graphing by hand..

Sorry I meant the input, the axis is HPA?

http://nefariousmotorsports.com/forum/index.php?topic=5673.msg53380#msg53380

Right, thanks :)

IIRC, lde is used for the control term axis'.

Base boost pressure, plsolr, is used for I limitation.

Oops. You are correct

yay! :P

Ok, I'll mess with D, that makes sense. Thanks for posting that example. That helps a lot

also, i'll start using ECUx. I really have no idea why i wasn't using it. It's nice as shit.

BTW that one is VERY extreme because that is a tight wastegate system. You will not need as much of a dip for k03s :) A very slight dip is probably more than sufficient.

Noted. I dont really have a dip at this moment, is more of a taper. So i'll see what i can get it to do and post logs later.

This answers the axis values question:

http://nefariousmotorsports.com/forum/index.php?topic=315.msg22201#msg22201

Sorry for going dark for a while. I've been making some progress though not much.


what is "i" in these algorithms?
 

"i" is "this sample"

so lde(i-1) is "the last sample of lde"

numerical integration and differentiation... might want to brush up on your math, since you are delving into the PID.

out of curiosity, what kind of Dr. are you? :)

Thanks. I have a pretty fair understanding of PIDs and while i'm familiar with calculus, I'm going to take a 90's kid approach to this (at first) and just poke things until it works better. I'll keep looking at my logs and adjusting accordingly 


It is merely my online alias. My name is actually Alan :P

Numerically integrating means simply summing samples over a period of time, i.e.

I(i-1) = f(i-n) + .... + f(i-2) + f(i-1)
so I(i) = I(i-1) + f(i) is the easiest way to get a running I(i)

Differentiation is dx/dt. Numerical differentiation is simply (f(i)-f(i-1))/(i-(i-1)), or f(i)-f(i-1)


In this case, f(i) is lde(i).

You aren't the verbal judo guy???

Unfortunately no :C I have failed the nefmoto community.

How does KFLDRQ0 apply? Or does it? i cant seem to find anything about it.

?

Q0 is the proportional component.

From your OWN post:

Right, But that uses LDRQ0S. Does it use KFLDRQ0 when the system is not static?

Depends on the version; there are many variants of PIDs. If the FR does not mention it, you'll have to disassemble the specific image you are looking at to figure out how it is used.

guuuhhh. Well, the wastegate is acting better now. I'm making progress.

You gotta lower that KFLDIMX, way more :)

I'm glad you said that cause that's what I'm currently tuning :)

Is there a particular fashion in which i should tune it? Where should i start? I've started on the last column. Is that proper?

Disagree. Needs more Q2.

I think it needs both, personally. But what the hell do i know?

I've decided to give up on tuning and pursue my dreams of being a modern artist. I will be selling prints of this in fine painted frames for $150. I accept paypal and travelers checks.

I have ruled out the DIMX, So i think you were right.

Here is my Q2 table, it is currently set to the stock values. If i want to control initial overshoot, Which part should Increase? By how much should i increase it? Should i increase each column by the same value or adjust each one individually?

It seems pretty stable most of the time after the initial overshoot, so i feel like i need to adjust the 700 lde column up. Is that the right idea?

 

So I have been tuning KFLDRQ2 and i'm getting closer to where i want to be. Here is a 3rd gear pull.

I like the progress I'm making. But i have a question about requested Boost now.

My ldrxn is flat until i pass 3500. Why is there a dip in req boost with a trough at 3500 instead of a proper taper like i imagine there being? More specifically, are there a few tables i should look at to remedy this?

Take a look at your IAT's and see if KFTARX is causing it. If not, KFLDHBN.

Or this

http://s4wiki.com/wiki/Tuning#Cam_changeover_effect_on_requested_boost

I'll log Intake Temp and check it out.


I don't even know where to start with this. What should i log?


P.s. Thank you guys a bunch. You've both been a lot of help :)

Log cam position. I am betting you will see a correlation.

Will do. After looking at other logs from the same flash, it doesn't seem to occur regularly. Either way I'll see if I can replicate it.

If that is the case, DD probably has it right, and it is IAT.

It doesn't look like intake temp to me since it's flat until the dip is over.

I attached a log with the camshaft angle logged as well. I apologize for it being mildly crappy.

here ya go

something screwy though, not sure that is the right var for cam advance

Do you know the variable off the top of your head?

It's wnwi_w, and that looks right to me. Nye, gotta look at the scale. All that movement is still only 1 degree.

Ok, good because That was the only cam variable i could find other than intake cam. So what should i do to control the requested boost? Should i decrease KFPBRK? Or KFURL or....

I cant seem to figure this notch out. I've tried making the last column of KFPBRK and KFPBRKNW the same (i saw it as a solution on another thread) but that didn't really help. Does anyone have a suggestion?

KFTARX-did you change it?

Ah, it appears that has been changed in the gonzo tune i was using as a base. I'll try changing it back to stock.

It's a weighting factor for requested load. Stock is not sufficient.

it will show up in req load though...

Only Some parts of it were set lower than stock. Should I adjust the affected rpm range (3-4k)?

KFTARX-Set to 1.0
KFLDHBN-Set sufficiently high to prevent hitting the max allowable pressure ratio for the altitude you're at.

That should get requested load a bit closer to where you want it. After that, it's time to rescale KFMIOP as it's generally not sufficient in the 1.8t to run much load over stock.

I see your AFR is like 14ish just after 3000 rpm during the 19 PSI boost spike. Be at like 12.5 from 2500 rpm up and taper AFR down in the rpm range where EGTs go up too high. The 1.8t engines power band is between 2800 and 5500 rpm with stock cams.

if Gonzo changed KFTARX then it's to your benefit over stock values as he most likely allowed for more load at higher IATs and rpm.

I did this. Generally, the requested boost was lowered. The notch was still as present as it's ever been. I can post a log when I get to my laptop. I'm pretty content with how my actual boost performs before I changed KFTARX to 1.0. It doesnt seem to be affected heavily by the notch unless the actual is crossing set point in that range. I was thinking that it might be KFMIOP but I am fairly unfamiliar with that table. I'll have to research.


Thanks for that. Thats an old log, I don't have that much overshoot anymore but I'll be sure to keep an eye on that. I haven't messed with fueling yet as I'm trying to get boost under control first.

Could be the stock timing that is giving you the "notch"? your boost stays pretty consistent, fueling is ok.. load, maybe you'll have to add some timing in the rpm range between 3000 and 4500 rpm? I would get your fueling on point before anything else though. maybe someone else will look into your log/s.
If you have a new one post it up.

He're's a recent log. I kinda cheated and adjusted ldrxn to remove the notch (i know it's probably not the right way to do it).

Fueling I want to fix, but I'm not sure where to start (lamfa?). I don't have a wideband o2 sensor, so i don't know if my lambda readings are accurate. It seems like it's running WAY too rich most of the time (high egt protection?). I checked block 032 on vcds and the ltft was only at 1.6% enrichment after some fairly heavy driving, which seems good to me. What variable should I log for EGT?

At this point I have a pretty good understanding about tuning the boost but with fueling i'm a little lost. I need a gentle nudge in the right direction. 

I've attached my latest tune revision

Full stop. Fix this first. Go no further.

This was my thought. That's why I haven't been touching any fueling stuff, i've read plenty of stuff on the forum about how accurate narrowband is. I'll pick up a wideband some time soon. I mean it drives really well otherwise, at least on my car.

Yes you do have a WB 02 sensor, in the B1S1 position if it is an AWP motor with stock ECU.

Don't worry about block 32 at this point, you are running stock fueling so only use it at this point to look for boost and vacuum leaks. As long as those numbers are less than 5%+- you are in the good. closer to 0 is ideal but both or 1 being 0 is a problem and never should be like that. There will always be some sort of correction. the 1.6 you have is about perfect for both numbers in block 32. AFAIK Block 33, 1st number is what you will want to be watching because it is like real time correction via the o2 sensor.

as far as the variable to log for EGT, I'm not sure what to log in VCDS, I know there is a calculated catalytic converter temp in one of the blocks but I am forgetting right now which. I imagine this will be what you want to log as one of the primary goals of the ECU is to keep the catalytic converter "happy". (unless that has been coded out along with the 2nd O2 sensor)

As far as the gentle nudge, hit up the S4 wiki on fueling, really on everything, it is very accurate as Nyet spent much time preparing it.

Then how come my actual actual only reads down to .75? I'll look into it.

As far as logging I almost always use me7. I was just using VCDS to check trims. I logged egt model pre cat and it never got high enough to activate the part protection,  I'm pretty sure at least. I need to sit down and go over it all again.

Are you talking about actual lambda? What do you mean by it only reads down to .75?

 

Lambda actual doesn't read below .75 but requested Lambda typically falls to around .65 which is considerably richer than I want; check the latest log I posted.

AWP is correct, it's in the title of the thread. I'm running an HS tune that is mislabeled as an HN, I just never got around to fixing it. Sorry for not stating that at some point.

Anyways, the requested and actual not lining up is what bothers me about tuning the fueling.

yea saw that after I posted about it. I was never very good at finding Waldo.. :-[

I even edited out the comment but you got here first.

Add " uulsuv_w " in your ECU file to log. that is " O2 sensor voltage wideband" convert to lambda or afr and I think voila


ADD: uncheck the EGT sensor variable, there isn't an EGT sensor and log other parts of the EGT model, even the whole thing if you can.

haha no worries man.

I'll add that var.

I have since removed EGT sensor; the flat value it showed didn't provide me much information haha. I'll add the other parts of the EGT model

Now i'm seeing I-reg adaptation. I can fix that easily though.

How does one do this? I've looked a few places and cant seem to find a solution that gives me reasonable results.

I've attached a log

I did a little research and I am still looking for what an (ADC-wert) is in the FR. Apparently the output of that variable is the ADC-value of the precat 02 sensor voltage. the actual translation for that var is: "Probe voltage before Kat a broadband lambda probe (ADC value)" So when i find out what an ADC value is, we should be able to use that factor against the data and convert it back over to 0-5v like a normal wideband. I'm wondering if this ADC value converts the wideband signal voltage to short band sensor voltage and that's what is showing on the gragh. I'm not sure yet. your log doesn't have actual AFR only desired, so I can't really compare. I just started looking into this though .

I'm seeing I adaptation after about a week or so of running a flash. The wiki says:

You can numb positive I-Regulation adaptation via increasing LDEIAP and/or shifting TLDIAPN upwards RPM wise (which you might want to do regardless, if you are running turbos that spool slower than K03s).


When it says increase LDEIAP, does that mean move increase it in a positive manner or increase the magnitude of the number? I.e. it is at -40 stock. Is -50 or -30 an increase?

Also shifting TLDIAPN, does that mean moving all values to the right by a column or two?

I'd be wary of following the wiki to the letter. Best consult the FR and come to your own conclusions...

So the FR reads:

" If the actual limiting value is too small order to correct fully, i.e. (a) deviation > LDEIAP (approx. -20 mbar) ... the actual
limiting value ldimxr is corrected to larger values until the current demand for integration is just met, and the
prescribed safety margin to the integrator limiting value is maintained. "

So, after reading this, I think I should make LDEIAP like -50 or negatively greater. My goal here is to set the value high enough so the deviation won't be greater than LDEIAP, and thus the I-limit will not change over time. Is this the correct assumption?

Lets put it this way: i read the FR and I'm still a little baffled why they are comparing negative values and whether greater than means greater than, or greater than the absolute value of.... I apologize for the vague wording in the wiki but I really didn't want to lead people astray with my bad assumptions.

Honestly, it would be best if you could convince somebody like phila or prj to review the actual assembly code :(


IIRC my LDEIAP is now actually positive; i couldn't get it to do what I wanted to by making it more negative...

No worries about the wiki. I've set it to -60 and i'll run it for a week or so. But I think i'll have to follow what you did and make it positive. Do you know off the top of your head roughly what you set it to? Is it a bad idea to completely disable the I-reg adaptation?

+50 maybe?

And no, i don't think disabling entirely is a bad idea, particularly if you are near the MAP limit.

Certainly letting it adapt *up* is a bad idea.

Alright.

Right after flash, it peaks and holds about 16.5 psi, which is dandy. But it gradually climbs and today it peaks around 20 and falls to where it should be after a second or so. Upon examining logs, I could see the WGDC was following the raised I value. I'll keep poking around with it and see if i can get it under control.

 Once i get that under control i think i'll start a new thread about tuning timing and fueling of this bin.

Wait, hang on, i keep thinking you are running a stage3 s4

if you are well below the MAP limit, i'd let the I adaptation do its thing... but probably disable it while you are finding the right I-limit.

Here is my latest log.

A week or so ago I found a pin hole in my intercooler. I have since replaced the intercooler with a better one. My EGTs seem to be at a much more reasonable level now, as well as my fueling.

As of late i have been tweaking small areas of the pid to fine tune the wastegate action.