NefMoto

Background...  I've noticed a temporary (about a second or so) lean out condition immediately after camshaft advance (about 30 degrees).  I have 2 settings on my camshaft advance 0 degrees and 30 degrees.   The spike in lean out varies from 20% leaner when cold to about 10% leaner when warm.  Generally, I don't notice this when driving with a warm engine.

Question...  Is there a good reason for this, is it by design (e.g. cutting spark during the changeover)?  If not, have any of you corrected the lean out? 

I've tried to search for some answers on this, but have found anything.  There wasn't anything obvious to me in the FR.  So, I'm asking for a little help.

Can you feel it while driving?

Lol, yeah, that's what happens when you change your cam timing.... It shouldn't come as a surprise.

Cam timing makes a massive difference in how your engine flows air.

Most tuners who deal with forced induction don't really know what to do with cam timing. However, having tuned 2 N/A 4.2 V8 S4's I've learned that cam timing can make some HUGE differences to the torque curve when done correctly.

Mister T, I experience the lean out described above at STOCK cam timing -- I didn't make any changes.

Porsche 996 Turbo Variocam Plus doesn't have an EGR.  They use cam overlap at low RPM (cam advances no later than 1,500 rpm, even when cold) as a de facto EGR.

My thinking continues to evolve on this lean out, but ultimately I think the lean out is a direct result of the major difference in KFURL at low RPMs between the 2 cam settings 0 degrees and 30 degree advance.  The root cause of the problem is aggravated by insufficient enrichment on cam change (KFNWSL/H), likely caused by my upgrade from the OEM Siemens injectors to EV14s.  Since I don't have an FR specific to my car and the available FRs don't have any info on the KFNWSL/H tables and have been flying blind trying to adjust these tables.  The values vary from 0.92 to 1.00, but I don't know if these are lambda targets, multipliers or divisors of a target lambda. 

Just today, I tried a different approach...  I pushed the cam change above normal "street driving" range.  The ME7.8.4 seems to be a little more sophisticated regarding cam change than the Audi ME7.  I can specify the changeover point by 1) rpm, 2) load and, get this... 3) gear.  OEM settings are the same across gears, btw.  I basically logged to determine my street load profile by rpm and gear, then updated the table to "push" the cam change much higher, from 22% load (essentially idle) to 75% load.  Just think about what the OEM did for the sake of emissions... They switched from "torque" setting to "power" setting immediately off idle.  What a terrible emissions vs performance trade off!

Since the retarded position (0 degrees) is the "torque" setting, my part throttle, off boost behavior improved significantly, as you might expect...  Of course, this benefit comes at a cost of worse emissions due to the loss of the EGR effect (not that I care about the environment -- already stripped out my secondary air).  Still have some lean out, likely related to the the injector conversion and imperfect WFRL/BAKL/VAKL values, but it is much lower than when combined with the cam change (1.05 lambda peak vs 1.20 with a 1,500 rpm, cold engine cam changeover).

Now, I'm wondering if I should take it higher still... I've read many have pushed their cam change all the way up to peak boost or further (for faster boost spool up and improved boost PID behavior), which would probably be around 3,000 to 3,500 rpm and 220% load for my billet K16s.  That seems like a long way from the stock settings.  At my current settings, 75% at 3,000 RPM, the cam change happens before I build any boost.  So, I'll probably try higher load values for changeover, but I'll do it in multiple steps.

Care to offer any advice?  This is the first car I've tuned with variable cam/valve timing... For that matter, I'd be interested in how you optimize the variable valve lift too.  I can't find shit on variable valve lift regarding the ME7 control tables/scheme.

interesting topic you are in:  I just looked into KFNWNGNG , and I see where you changed the load values to go to "torque" profile.  I suppose at idle rpm you do keep the 105% load value.

for the Valve lift control I see in VS:  MIVSNG  . It seems that valve lift is controlled by load and above certain load values valve lift increases.

As also cam time is load dipendent  both are influencing each other. 

In 1 month I can try this on a car with logging, time permitting I will report back.

Thank you for your findings
Andreas

Andreas,

Thanks for the response. I'm glad to have someone else reviewing this.

I did some research on the Variocam Plus system and learn some interesting things that are especially relevant. First, the system varies the intake cam timing and intake valve lift in a binary fashion. Intake cam has either 0 degree overlap, used only for idle (really), and a 30 degree overlap, used everywhere above about 1,200 rpm.  The intake lift also has 2 settings. It is either 3mm (idle and low rpm) or 10mm (high rpm).

This information really changes the cam discussion. It isn't worthwhile to maintain zero overlap WITHOUT also altering the valve height to 10mm at low rpm.  Here's why... The overlap improves intake flow at low valve height because air can be pulled from both the exhaust side (recirc) and the intake side, minimizing pumping losses.

Delaying the cam changeover while still in low valve lift actually REDUCES your ability to produce "torque" and impedes boost spool due to reduced mass flow (from pumping losses arising from high cylinder vacuum).

I verified this myself in my logs.  At higher rpm, my boost immediately increases upon cam changeover WHEN in low valve lift.

So, altering cam changeover on the Variocam Plus system isn't as straightforward as what has been done by others on their Audis with different VVT strategies.

At this point, I'm not sure how to move forward. Because I'm not sure of my valve clearance, I don't know if I can safely run a retarded cam (zero overlap) while also running the valve in high lift. If the Porsche engineers really milked the design (resulting in limited valve clearance), you can't do it without valve-piston interference.

So, for now, I'm left with a choice between lean out on cam change (stock settings) or lower power from a higher cam changeover at low valve lift. I guess I'm back to looking for a way to address the lean out with fueling.  I'll post some logs of the issue so you can see the undesirable lambda coupled with a drop in load and requested torque, which has to be the reason for the fueling shortage.

Any thoughts?

I'll tell you how to move forward.
Don't do anything at all. Literally.

If you have serious backpressure issues on top you can try to move it, but the way the engine is designed I think it will hurt performance everywhere.

The reason the cam even moves is to make idle acceptable and not sound like a bag of sh...
As for leanout, well, the lambda will show a bit leaner during spool due to scavenge effect.
It doesn't mean the mixture is that lean, it's kinda tricky there.

And without proper logging it's difficult to tell. I could check my logs from when I was tuning to see?
I had some 800+ cc injectors... and full ram logging.

PRJ,

Thanks for joining in...  I can really benefit from your experience.  I'll take any advice you have to offer.  Hell, I'm so frustrated with this niggling issue that I'd pay for help.

I've attached a short log that shows the lean out following the cam change on a cold engine (45C).  This is a good example of my issue after adaptation has had a chance to "work" (the lambda spikes above 1.2 immediately after a fresh flash).  Unlike the stock 996 Turbo, I've enabled transient adaptation (CUKA) and this reflects the best it can do.

Please know that I'm logging with Durametric and they don't use any of the Bosch terms (like PSSOL for Boost Setpoint).  So, I'm left to my own devices to figure out what variable they are really logging.  I have no idea what "calculated mass flow" represents, as an example.  BTW, I called Durametric and they wouldn't tell me...  as they're clearly part of the glass bead society.  They said, "Call your tuner".

As you review, note that my requested torque and ignition angle drop on cam changeover, which probably has something to do with the lean out.  My problem...  I don't know how to appropriately address it.  I found cam changeover enrichment tables (KFNWSL/H), but don't know what to do with them without an FR that explains how they work.  I'm not even sure they are enabled at this point (since I've tried various values and seen no detectable change).

BTW, this lean out isn't a scavenging event.  RPM is too low for that to happen.  I doubt backpressure is a concern since I have high flow headers and cats and an exhaust much less restrictive than stock.  The K16 turbine is perfectly sized for my application, but the compressor side got a larger, billet wheel upgrade.  Anyway, this lean out results in stuttering and poor throttle response, albeit only briefly and noticeable only when cold, but I still would like to address it.

Any and all advice is welcomed.

if lean out is only a concern at warm up,  in the 997S tables there is KFBAKL, enrichment based on engine temp. Should exist also on 996tt.

p.s. wenn du Deutscher bist, schick mir doch dein mobile per PM.

By
Andreas

KFBAKL is mostly throttle movements IIRC, if your throttle is static, it's not gonna do that much.
Durametric sucks for logging :( And they won't know which value it's really logging because it's logging MVB's not RAM.

Also from that log it doesn't look like cam overlap itself is causing the spike, but rather load increase. Try tweaking the wall wilm to be a bit bigger.... and possibly KFBAKL.

Alright...   I've taken another run at my WFRL.  This time I've examined the derivatives of WFRL because the FR says the curve should be smooth.

I tried this out this morning and it helped some, but the throttle response was definitely smoother.  I'll need to take it further...  (removing the dip described below).

Anyway, just wanted to ask if anyone else has reviewed the derivatives of their OEM WFRL curves.  My problem ocurrs between 30% and 60% load, which is also where the WFRL derivative has a "dip" in it.  Do the AUDI OEM WFRL curves have the same dip?

My thinking is that the 1st derivative slope should be steep to start and then gradually flatten out.... clearly, my OEM curves don't do that.  And the second derivative is really telling... Note the OEM curve is very inconsistent from point to point.

So, why did the OEM have these weird undulations in the 1st derivative?  Does the cam overlap and valve lift affect wall film?  I'm not sure...

Looks like you're talking about getting a tiny lean spike when you lift off the throttle on relatively light loads correct?

Like prj said, leave it alone. A slight lean spike is to be expected on throttle lift as the fuel film on the manifold walls is momentarily depleted.

I can promise you that it's not causing your engine any harm. Hell, you should see how lean my S4 with headers goes when I lift, it's literally off the scale lol.

The one suggestion I would make is to park the exhaust cam at full lobe separation unless you go over about 60% load. The EGR function is only for fuel economy and emissions.

I found that keeping the lobe angles separate at part throttle made for much better transient response as it always takes the cams time to go from overlap (EGR) to separation (torque). It may only be a tenth of a second, but it was noticeable enough for me to do it on my 3.0 AVK Audi. 

YO SUPERMEGANATUNER, nothing what you've said is relevant whatsoever to this topic. In this topic we're talking about cars with 2 point NVS on intake cam + variable lift. I understand you've never even seen one, but at least pay attention to what's written in the thread.

@jpurban
I attached a dyno pull from a 996TT I was tuning.
I think I went kinda nuts on KFBAKL and wallfilm to sort the entire issue with the Dekas that were on it. So much so that I have a rich spot where you have a lean spot.
But I don't let it bother me too much because actual tracks very well and elsewhere mashing the throttle and lifting off the throttle my lambda correction was reasonable.

You need WinlogView to view the file, that you can download from the 14point7 website (or use whatever else you want).

ohhhh, SUPERMEGATUNER, I like the sounds of that lol. Actually, my post was right on point if you cared to look at the title of the thread "leaning out on cam advance" AND my point about parking the exhaust cam for transient response is also an excellent point that no one else has ever raised to my knowledge.

You'll also note that I also mentioned the wall wetting as being the root cause of the lean spike (and gave you credit for being the first one to point it out (although I did know it in advance of reading your post).

Further, parking the exhaust cam on an FSI engine also has the advantage of eliminating EGR which is the cause of most, if not all of the carbon buildup on the intake valves that plagues that engine series. Again, a point that I have yet to see anyone raise around here.

PRJ and Mister T, thanks for the responses.  I'd like to run some more thoughts by you...

Mister T...  How would you recommend altering KFNW?  I've attached a screenshot of my OEM values.  You'll see I quickly drop to the low 20 RL as a changeover point.  My cam changeover seems to aggravate lean out by 10% or more when it occurs at low rpm in the range of 30 to 50 RL.  So, 60% seems like a good call on your part.  I get the feeling you know the crap I'm encountering.  Not undriveable, but not ideal either.

Also, have you altered NNWS or DRLNNWSG?  I ask because raising the changeover point from near idle to the middle of the cruising range could aggravate oscilliations in cam position -- Think uphill vs downhill in cruise control.  You'd cross the 60% RL threshold a lot, right?  May not be a concern, but you'd probably increase the number of cam changeover events, which could cause less-than-smooth transients in cruise as well as increase wear on changeover mechanisms.  So, it seems to make sense to increase protection against unnecessary cam changes...  just thinking out loud, if you will.

PRJ...  I completely agree with you -- Appropriate transient compensation is the ideal solution.  Altering cam changeover point isn't ideal with this engine design -- reducing overlap with the low valve height ultimately reduces the engine's ability to breathe (as rpm rises and scavenging benefits grow).  That's never good.  BUT...  I have tried to alter my ESUK tables so many times and only seem to make matters worse.  Frustrated is an understatement.

I'm willing to keep working on ESUK adjustments, but could benefit from your experience.  First question... Do you have to have a dyno to effectively update ESUK for new injectors?  Second...  Do you disable KFBAKL/KFVAKL (Set to 0) when trying to tune WFRL, as the FR instructs?  I guess simply want to know your process for ESUK tuning... Any key steps?

I've attached a screenshot of my relevant ESUK tables.  I mentioned the lack of continuity in my WFRL derivative.  Is this normal?  I see similar weirdness in my KFABAK table, where the short term portion drops to nothing a 60C.  You can definitely feel the "numbness" in throttle response, but I'm sure the OEM had good reason.  Only reason I can find is that LRA is enabled at 60C.  Since the OEM has ESUKA disabled in my file, could it be that the OEM expected LRA to address the transients above 60C?!?  Do you see the same weirdness in your Audi files?

Revisiting a solution to my cold start lean out that I first read in the EV14 thread...  Adjusting ESWL warm up tables.  I did just this, but found limited benefit.  Think I now know why... My warm up is super short -- My ATISLATM is only 250 injections at a 40C TMST.  That's about 30 seconds, barely enough to get out of the driveway.  My warm up ends and I end up with a lambda target of 1.00 while my motor is still quite cold.  Are your Audis similar?  (Edit:  Ithink ATISLATM is the threshold for switching from lamns to lamwl, but the Audi FR shows a less-than symbol to set lambwon.  A little confused, obviously.)

Perhaps this isn't a problem if you directly from short warm up to cat heating, which has it's own rich lambda requirements.  However, I've removed my secondary air, disabling KH as a result.  That may be the root cause of my lean out when cold...  That begs the question... How do we replace the KH rich lambda targets?  I'm guessing the answer is setting CWWL bit 0 or bit 1, which enables LASWLTM (inactive LR) or KFLASWLR (active LR), respectively.  My OEM CWWL = 0, and I have neither enabled.  Am I n the right path?  What settings do you use in the cars you tune that lack KH?  Please forgive me if I'm asking a silly question.  This area (ESWL) is especially new to me.

Absolutely, especially when cold.  It isn't violent or anything.  The throttle just goes limp/numb and the car doesn't respond for a split second or so.  You can feel the stumble in your butt dyno.  ;-)

*** Update -- Pushed the WFRL changes too far and made matters worse.  Slight surging under constant acceleration as I transitioned from one load steo to another.  Logs show the lambda bouncing around more, confirming that my changes weren't the right ones.  At this point, I'm going to revert to stock values and start again. ***

Dude, this car is not FSI nor does it even have adjustment on the exhaust cam. No one cares.


Yes, see below :)

I kinda mash stuff around until it's right. I don't do it as the FR instructs.
WFRL makes global changes, KFBAKL/KFVAKL are more precise. Try KFBAKL.... I don't remember anymore what I changed in the end.
And yes, it is normal that it drops to almost nothing, because the stock injectors are perfect for the application, whatever you are using now isn't though (as it was in my case).
Try setting the 90 degree to about the same as the 45 degree and test what happens.

Warmup stuff - KFWWL? I don't have a great deal of Porsche cal experience, some maps are different / called different in ME7.8.

Btw for KFBAKL to have an effect there has to be some sort of load gradient or throttle gradient during the actual change. If load is fairly static, then it's not gonna do much.

This suprises me, same method I use but don't usually get quite right. I don't know what you are talking but what about the turbo enrichment which I don't have because originally NA and due that slight lean when boost rises?

Such thing like "turbo enrichment" does not exist. You have problem with hardware, car or incorrect mapping. Probably all of it at the same time.
Please keep this on topic guys, we're talking about Porsche 996TT here. Thank you.

An update...

I reset the ESUK transient comp tables to stock values and logged today on the way to work.  There was an accident and traffic was horrible, clutching stop and go for several miles.  That sucked, but it gave me lots of low speed, low load logging, which is exactly what I needed.

Here's what I noticed...  Most of my lean outs occurred when slowly accelerating AFTER a slow rolling deceleration (no throttle) with NO gear change/clutching -- even with a fully warm engine.  When decelerating in gear, the RL falls well below idle RL (even with DFCO disabled) -- down to a range of 15% to 18% RL.   Once I accelerate slightly -- no cam change -- my lambda spikes to 1.1 or higher and the car stumbles/hesitates (but does not stall - so the issue isn't bad).

Important learning #1...  Since I don't have much problem accelerating from a stop and my RL doesn't dip below 18% (usually) when doing so (a different load step range), I have to suspect my WFRL value between 12 and 18.  Question...  If I wanted to apply a fix to this range only, what is the best way to make that change?  Lower the value in WFRL at 12 and below?  Or, raise points 18 and above?  If raising 18 and above, how do you keep from disturbing the other ranges?  This may sound like a dumb question, but I'm really asking if it is more important to maintain the percentage difference or the absolute value difference of the steps you don't want to change.

Important learning #2...  Decelerating in gear (no throttle) drives the RL down to the values in KFRLMN/KFRLMNSA.  Knowing this -- and the issues I have below 18 RL -- one solution could be to avoid the range below 18 by simply raising KFRLMN to some value above 18, but still below idle RL. 

Since KFRLMN is a pretty easy solution, I'm thinking of changing it and then logging again -- ignoring the WFRL for now.  If the majority of the lean out goes away with this change, I may not need to mess with WFRL since the rest of the load range (OEM values) felt pretty good today.

PRJ, looks like you have a nice dyno set up.  That's an AWD dyno, right?  What I'm really wondering though...  How the hell do you keep it so clean? 

Any thoughts?  Anyone else increased KFRLMN in their tunes?

Holy shit!!! we agree on something!!!

Quick, go look out your window and tell me if you see four horsemen in the sky lol ;)

Seriously though, cam adjustment of ANY type on a system as dynamic as a 4 stroke engine with constant torque demands (with or without a turbocharger) is a black art.

Although PRJ is correct in stating that your question doesn't really involve cam phasing, I'm gonna add my 2 bits since I'm on my soapbox.

Speaking from my experience, I've literally spent hundreds, if not thousands of hours looking through my datalogs to see how mass airflow and measured load respond to cam phasing.

The best way I can articulate how cam adjustment works in practice is thus; Any engine will have natural resonance peaks in accordance with its' hardware (exhaust, intake, throttle, etc...).

Where cam adjustment shines is that it allows you 1) maximize the torque from those natural resonance spots, or to 2) eliminate the dips in the torque curve. Basically just move the cams around the spots that you want magnified or filled and eventually you'll come up with a superior torque curve than you would with a fixed cam.

With NATURALLY ASPIRATED engines, the way I view it is that the airflow is dynamic, To fill in a hole, I dial in advance (overlap) in a smooth, radiused drop-off. However, you eventually  'bottom out'. To 'recharge' the airflow, I quickly pull the cam timing back up towards lobe separation over 200 RPM or so. Then, I start dialing advance back in. Rinse lather repeat. Most (if not all) of my cam timing sweeps at WOT have a smooth, saw tooth look to them. I'll try and post some pics up to show what I mean.

While there is some debate as to whether or not changing the OEM calibrations on a stock vehicle will do you any good, I am of the opinion that you can improve on them performance wise. Reason being, is that you're not concerned with emissions and trying to make everything as smooth as possible for the 'average' driver.

Then, once you start changing the intake/exhaust and alter the VE of the engine, getting the cam adjustment dialed in is a MUST!!!!

I cam promise you that 95% of tuners don't mess with the cams on naturally aspirated Audi engines. Want proof, just look at their dyno plots, if the SHAPE of the tirque curve remains the same, they only added timing and touched up the fueling.

Anyway, I know this is off topic, but I've just spent the last 4 hours reading SAE and Academic papers on the subject and thought I'd take the time to drop some nollij on 'yall ;)

Here's a pic of one of my latest logs.

Notice how the airflow picks up along with how the cams move after 4000 RPM and stays just about dead straight to 7000 RPM?

Same with measured load (pink line). Measured load (at least with N/A engines) is a carbon copy of the shape of your torque curve. Here, it only varies by about 5% from 2000-7000 RPM!!!  ;D

Finally, the other light line is _ps (manifold pressure). It drops a bit past ambient (898 mbar) then after 4000, it goes past 1000 mbar for the rest of the pull. Anyone in the know care to say why that is?

I should also note that this S4 is modified to the hilt. Headers, dual X pipe, 3.5 silicone intake, re-engineered intake manifold (w 'mid length' runners), shaved half shaft throttle body, all on my own tuning file...

Driving wise, the car almost feels slow because the torque curve is so flat and the acceleration is so linear through 3rd, 4th etc...

(https://farm2.staticflickr.com/1824/29060829848_da9072803e_h.jpg) (https://flic.kr/p/Lh1mfu)560A S4 dialed in (https://flic.kr/p/Lh1mfu) by zimbu themonkey (https://www.flickr.com/photos/156667009@N03/), on Flickr

KFRLMN might have some other implications... Check your TEMIN also, that it's not pegging the injectors.
It was obviously cleaned for the photos :D And yes, AWD with adjustable wheelbase and toggleable front-to-rear lock (required for the 996TT).
But we try to clean it every now and then and keep it looking fresh.

Does the lean point happen always a certain MAF flow values?
As you changed intake geometry,e.g. different compressor wheel,maybe KFHFM or the relative compensation tables are out offerte tune.

Andreas

Andreas,

I think you're essentially asking if my steady state fuel trims are good.  If they weren't, then I'd need to focus on TVUB/KRKTE to dial in RKAT/FRA in rough terms... and then KFHFM for any lean/rich spots across the MAF spectrum...  then FKKVS/KFKHFM across both engine speed and TE/MAF.  Working from broad to narrow, so to speak.  Right?

I've spent hours and hours learning that process, logging, trial and error, etc.  All the stuff I'm sure many here have done as they climbed the learning curve.  I think my steady state fueling is as good as I can expect.  I've attached a screenshot of my steady state lambda error, showing that I'm +/-3% across the "part throttle" street driving spectrum.  The error shown INCLUDES LRA adjustments (as part of the error) so you have an idea of the quality of my steady state tune -- I'm not relying on LRA to save my ass. 

I have to pause to give some props...  As for the "dynamic" lambda control at higher loads/airflow near WOT, Bosch ME7 is f'ing awesome.  I've had my car for 5 years and I'm still amazed at the accuracy/control of the widespan O2 closed loop control at WOT.  The ME7 ain't your daddy's Holley 4 barrel carb on his '69 Chevelle.  ;-)

All this said, I hope I've convinced you that I've done the necessary homework to be on the ESUK tuning step.  I'm at the point of turning to ESUK and NWS to find a solution to my transient lean/rich conditions.  I really didn't want to tackle this additional step, but I don't feel I have a choice since some of my transient lambda errors swing from +25% rich (decel) to -25% lean (accel) in the span of a few seconds.  So, I'm going through the pain, failures, frustration of dealing with ESUK as a newbie. 

John

*** Update - Another day of logging ***

Tested changes today...
1.  Smoothed out WFRL from OEM.  Overall WFRL has the same span as OEM, but I smoothed out the slope changes across load steps.  First step to refining WFRL.
2.  Increase KFNW slightly at low engine speed to raise cam changeover to improve driveability with imperfect transient adjustments
3.  Raised KFRLMN to 19.5% minimum (just below warm idle, no a/c load) to provide additional margin to transient errors

Good news:  I really like the improvement in throttle response when underway, which I attribute to the WFRL update, especially the increase around 30 to 40%.  Also, higher cam change point kept stuttering to a barely noticeable level, but it is clearly still there as the logs show.  Cam changeover shift confirms lean out is poor transient response - not cam changeover.  See the log to watch the rich/lean decel event play out while the cam stays in overlap position.

Bad news:  Raising KFRLMN out of the 12 to 18 load step range of WFRL did NOT have the desired effect.  The big issue of rich decel to rolling idle leading to idle/accel lean out is still present.

Analysis at this point...  Throttle off decel at mid-level, cruise load drops to RLMN.  During decel, mixture goes extreme rich (sensor limit, about 0.75) initially, then leans out.  This suggests one of the following corrective actions:
1) Increase WFRL over load range to increase enleanment on decel.  This doesn't seem ideal because the accel transient seems pretty good when not following a decel.
2) Increase short term portion of decel enleanment (KFAVAK) to shift enleanment to address initial richness and simultaneously reduce late decel lean.  This seems to make the most sense as a next step.  Reducing the long term enleanment, that can effect accel that immediately follows, may be more of my issue than pure accel enrichment.  I've included a log sample that shows a decel to rolling idle that illustrates the point.
3) I don't think it makes sense to raise total decel enleanment (KFVAKL) until the short/long split is corrected.

Sounds like a plan, but...  A review of KFAVAKL/KFAVAK reveals that the short term factor is already pretty high at 0.30+, above the FR suggest value of 0.25.  And...  The KFVAKL values are low relative to the FR suggested values.  Perhaps that doesn't matter, but the FR says if you stray too far from the suggested values, something is probably wrong with WFRL/ZUKK/DUKK/etc.  So, I'm a little lost at the moment regarding the next step.

I need to stew on this a little more.

You will need to bump KFVAKL a lot higher than what is said in the FR, maybe 5 times as high, when you have injectors not made for the application...

Also, did you make sure TEMIN is low enough?

I came across this information in a Porsche training document. It gives some general information about cam adjustment in the 996 turbo, 997 turbo and GT3 models. Hope it helps :)

Fuel trim is always a very slow adaption of the mixture (in the range of 0.75 to 1.25 where 1.0 is no correction).  So at a steady engine speed after let s say 0.5 seconds the fuel trim will adjust to correct lambda.  Long term fuel trim is even slower.
So it is too slow for your transient lean condition.

Basically to measure HFM  calibration the whole lambda - autoadjustment of ME7 is a problem.
I usually set  all KFLF to 0.9  and also KFLBTS and LAMFA to 0.9.
I found that in this way the lambda  regulation is turned off (at least for several seconds at constant rpm and throttle) as the target is too far away from 1, see FR somewhere.

Then I do a simple log drive with steady slow acelerations at different throttle plate angles.
From the log I then  fill an excel table where I prepared the x and y points to correspond to the KFHFM table. Y - yalues are measured lambda (I use my own lambda sensor).  It is a wise idea to write yourself a little program that does parse the log data .

Ideally all y - values should be lambda 0.9  (they are not in my screenshot as it is taken from a different setup , no time to search through all my measurements.)

Where you see differences, correct the KFHFM table and try again.




Good luck
Andreas

Andreas,

Do you have to follow this process even if you disable LRA (NOLRA = 8, I think)?

If my rkat and fra are small across most loading, engine speeds, etc., I'm good, right?  Other than low load, low speed transient situations (1 second or so), my car behaves well.  Idle is perfectly smooth.  After the initial shift, lambda settles pretty quickly. 

I guess what I'm really asking...  Are there other lambda regulations than those shown in LRA (fra and rkat)?

John

PRJ,

VAKL increase is the next step...  I'll keep you posted.  My VAKL is about 2/3rds of my BAKL.  Do you find this relationship to be typical?

My TEMIN is set at 0.  Daz helped me with that a few years ago -- My A2L file isn't a perfect match to my version and we had to search for it.

John

*** Update ***

Feel like I'm making progress.  Current theory...  Porsche tuners optimized VAKL for DFCO, which explained why it had some very low levels - it wasn't actually used very often (DFCO forces error vales to be used when B_SA is set).  So, I set the VAKL and KFAVAK to the recommended values in the FR. Magitude looks good, but the short/long splits need adjustment.  Higher revs require greater short term.  In the process of logging and tweaking.

Still hung up on the cam change lean out...  See the attached log.  Could the 10% RL drop at cam change (which is confirmed by a nearly 10% drop in MAF) be triggering ESUK compensation that leans me out 10%?  If so, then maybe I need to drop short term decel percentage around 1,200 to 1,500 rpm...  Sounds like something to test in my next attempt.

Can be.

Well, shit.  I thought I had this whipped (as outlined below), but I spoke too soon.    I'm still having lean out at some of the cam changeovers, about 10%.  So, I'm back to thinking I don't know wth the KFNWSL/H tables do.  Does anyone have an FR for the 996TT/ME7.8 or know where one can be had?  I'm not sure I'll figure it out without one.

I guess I'll have to look for alternative solutions...  One thought...  I have 2 tables, FKKVS and FKKVSNWS.  Maybe I could bump fueling at the boundary cells in each table?  Anyone else try this approach?  BTW, the Porsche factory tables don't appear to take this approach.  So, there's no evidence it'll work (if it did, then why do the KFNWSL/H tables exist?).

Honestly, the whole cam changeover could be an ESUK tuning issue.  As mentioned earlier, my RL and MAF drops during the change about 10%, which could be driving an poorly tuned ESUK response.

This is just baffling.  I'm kinda at a loss.  The factory ESUK settings are so far from the FR default values that I can't figure their strategy.  Example...  DUKK is 0.94 and ZUKK is 0.3.  When you couple that with a KFABAK short term portion of 0.2 (short term tip-in enrichment is only getting 2% of the total ESUK enrichment above TMOT of 60F), it sure looks like the Porsche calibrator nerfed the short term portion of ESUK, especially for acceleration.  Could it be that the OEM Deka injectors were that non-linear?  Or was Porsche just nerfing short term acceleration to smooth out the transition from decel fuel cut off to accel fuel cut-in?

Besides the cam change lean out issue that started this thread, I have one other lean out out issue that seems to be related to the weird ESUK settings, but this is just a theory.  DUKK and ZUKK effect both accel and decel. 

So, picture this...  You're in 2nd gear and accelerating moderately (60+ RL) -- quickly, but not enough to be into boost and your lamsoll is still 1.00.  Very typical street driving.  Then, you prepare to shift, lifting off throttle and your RL drops from 60+ to 18.  Since you've disabled decel fuel cut off, that load reduction sends lambda from something near 1.00 to the sensor rich limit (0.73) in about 0.4 seconds (because ESUK isn't doing shit due to the high DUKK value).  You press the clutch and shift into 3rd.  While you were doing that, ESUK and LR kicked in to correct the massive rich condition and put you in a slightly lean condition.  Then, you let the clutch out, applying moderate throttle (not enough for boost or a rich lambda target).  As you do that, you're starting from a lean lambda and ESUK isn't doing shit to enrich because the high DUKK is keeping it from doing so.  That causes your lambda to spike higher and you feel the sluggishness that lasts for a brief moment before LR and ESUK kick in to help get you back to a lambda of 1.00.  Sure, the car is driveable, but you know something doesn't feel right.

Assuming my interpretation of events is right, I'm left with 2 choices...  1) Enable DFCO at lower RPM, going back to factory NWEM settings, hoping the factory FUKE and ZUKE settings work for EV14s or 2) be prepared to change everything in ESUK from factory, which appears to be optimized for DFCO to low rpm settings.   Since I want to make #2 work, I'm going to set DUKK, ZUKK, KFBAKL/KFVAKL, KFABAK/KFAVAK to the FR values and test using the factory WFRL.  If that isn't a total disaster, then we'll see if WFRL tuning gets any easier.  Note that all my previous attempts at ESUK tuning only made things worse -- I'm hoping that was because of the high DUKK value, which basically undermined my attempts to address short term lean out in accel and vice versa in decel.


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

***  Success ***

Hell yeah, muthafucka!  Finally, a cam change with no (meaningful) lean out! See attached log for proof.

Solution:  In ME7.8, KFNWSL governs mixture correction factor during cam change from late (retarded intake, zero overlap) to early (advanced intake, 30 degree overlap), BUT it isn't intuitive: Rich is numbers less than one, suggesting it might be a lambda-like approach or lambda-modifying value.  Since no available FR covers this table, I'm not sure exactly how it works, but I've confirmed that it does for my issue.  Word of warning: Do NOT set any values in KFNWSL above 1 or you will lean out (1.3 or more), usually in acceleration, and that is a recipe for disaster.

In my log, you'll see the cam changeover has been raised to 40+ RL from stock 22-23 RL level, based on advice from Mister T.  This actually makes the cam changeover lean out worse because the KFNWSL table values rise from 0.92 or so at the expected OEM point to 1.00 (no enrichment) at the higher RL.  So, contrary to intuition, raising the changeover RL makes things worse, if you don't also adjust the KFNWSL table.  In my test (shown in the log below), I reduced my 10% to 15% lean out (KFNESL value of 1.00) to 5% by reducing KFNWSL to 0.91.  So, the table seems to have a linear effect on cam changeover lambda.

I'll fine tune this with more testing and let you know if anything changes in the perception expressed above.

fantastic, thanks for your tenacity

Well done :)

I remembered somewhere in the back of my head there being a cam change map, but didn't remember the name, and didn't have really time to look.

Update...

Step 1...  With FR defaults in all of the ESUK defined tables/scalars except for the factory WFRL, I unsurprisingly ran lean on increasing load and rich on decreasing load.

Step 2...  Again with default values as above, I scaled up WFRL by 5% across the spectrum and returned to the factory value for ZUKK, increasing from 25% to 30%.  Regarding ZUKK, I think I see what the factory was doing -- This is kind of a global increase in short term portion of ESUK, which you can use to balance short vs. long term adjustments.  If you get ZUKK, right, then the factory values for KFAVAK/BAK only require small deviations from the FR default values.  Results were great in mid-load situations (judging by the response of the car -- logs still show some leaning), but more WFRL/ZUKK is required for low load situations, which are still well lean/rich on accel/decel. Since the problem is symmetrical, that points to WFRL increase and/or ZUKK increase.  With some more analysis, I'll figure out my best guess for each to test in Step 3.  After I get the car performing reasonably well with the global settings, I'll shift to the accel/decel specific tables.

Step 3...  I increased ZUKK to 0.35, effectively adding 5% more to short term.  That helped some, but still have more to go to balance short vs. long, especially on decel.  Throttle lift to idle while rolling still goes 15% rich, followed by 15% lean, which I'm taking as a good indication that a higher ZUKK is required.  Since I still feel I'm overall lean (short + long) on accel, WFRL is getting close, but needs a little more in the low to mid range (18 to 60 or so).

Step 4...   Next step is a ZUKK of 0.40 and higher WFRL down low.  I'm going to smooth out the factory WFRL curve in this step.  More to come once I get to test.  

As a general note...  In retrospect, I shouldn't have been scared to start from scratch on ESUK using the FR recommended values -- Nothing bad happened.  Hell, starting from scratch is actually easier (when large adjustments are needed) than trying to alter the factory BAK tables (only, as some have recommended).  Even though I'm only tuning to the warm motor results, the car is still very driveable at a cold summer start (40C) using the FR defaults that adjust for motor temp (BAKL/VAKL).  Since cold performance is improving with each iteration, I'm thinking my cam change lean out may become a non-issue if I get ESUK good enough.  I guess all the FR instructions prompted me to think you had to have a dyno and the ability to do controlled load steps to get ESUK "good enough".  Couple that thinking with some failed attempts at altering just BAK and you might understand my trepidation.  Now that I'm making good progress -- car is coming alive, throttle response much improved -- I'm feeling less intimidated, but I also realize I'll probably be fine tuning for some time to come, especially as we get into the cold winter starts (0C or less).  My goal, at this point, is to keep transient (0.2 second increments) lambda excursions within +/-0.05 when under normal driving (lamsol = 1.00).  I think that'll be doable without a dyno/load cell based on what has been achieved thus far using only the "global" adjustments.

Comments, criticism welcomed...  I'm still learning.  Feel free to tell me what I'm doing wrong.
    
Edit:  Added a graphic that illustrates the "spread" between max rich and max lean for accel and decel segmented by max/min loads/rpm of each load change event.  Goal is to get them as close together as possible.  Since the spreads are pretty uniform across the spectrum, I'm still focused on the global settings as the next steps.  Note how rich the the decel events can get, which becomes a problem when you disable DFCO on a tune that is optimized for DFCO (my current working theory).  Believe it or not, this is actually an improvement over my starting point.

*** Update ***

New information...  
I did some very slow speed driving around the neighborhood in 1st and 2nd gear with a hot engine.  I found that I could recreate the cold lean out (issue that kicked off this thread) even when hot.  

What I noticed...  
1.  The lean out occurred without a cam change in many instances --> Cam change may aggravate the condition, but it isn't the root cause.  
2.  The lean out occurred under near constant load -->  Confirms transient enrichment/enleanment (ESUK) is not the root cause.
3.  The lean out always follows a sizeable ignition retard (e.g. 30 advance to 0 advance), even under constant throttle/pedal angle.
4.  The lean out only occurs in the region LLR is active, which is NSTAT (760 RPM Hot Idle) plus DNLLR (1,200 rpm), which means anything below 1,960 rpm when warm.  Higher when cold.
5.  The lean out occurs when operating near the idle RL (40% cold, 24% hot) in gear, rolling.  The higher value at cold explains why I encounter the problem more often when cold, but hot testing reveals the problem exists there too.    

So...  These new facts point to a non-optimal LLR.  

Thinking out loud...  I'll need to do more research, but my first thought is that my fully revised KFMIRL/KFMIOP (linearly scaled, like OEM) isn't playing nice with the stock torque reserve settings.  Since the torque vs relative load relationship is changed, the torque reserve tables probably need the axes scaled appropriately to mimic the stock behavior (which was fine).   If this doesn't solve the issue, then I'll raise the minimum ignition advance, which should be okay since my cat heating is disabled and there's no need to allow values like 6 degrees ATDC (as required by KH).  Side benefit to increasing ignition advance is less burbling exhaust on decel (which some might like).

Anyone else run into this issue?  Do most of you leave the low portions of the KFMI tables stock?  My first purchased tune for the car only altered the last 3 (high load) lines, perhaps to avoid the need to scale other items?  Maybe that is a better approach, but it must have it's compromises too (non-linear load torque relationship requiring KFPED custom tuning to avoid large load changes -- which my purchased tune did not have, explains why the pedal didn't feel "right").

Don't touch IOP/IRL at lower load unless you want a ton of trouble.

Now, you tell me?  jk...  I've updated my IOP/IRL to be stock at 100 RL and below.  Definitely an improvement.

Do you typically leave stock up to 100 RL?  Or, do you start the scaling at a lower value?

The reason I ask...  Starting above 100 RL creates a very non-linear relationship between pedal angle and net requested load.  I'd really prefer a more linear pedal, but the limits of KFMIZUFIL won't allow for much correction in KFPED.  You're left with a pedal that is very sensitive at low load (not necessarily bad), but quite insensitive at high load (something I don't like).

How do you overcome this?  Or, do you live with a non-linear pedal?

Uhmm, KFPED you can set to nearly whatever, monitoring should not do anything.
Well unless you go inverted. But I've never had issues with KFPED.

PRJ,

I thought (perhaps mistakenly) that you should make sure KFPED requested torque stayed below the values in KFMIFIZUL to avoid torque intervention.  If you don't, then you end up with mifa (request) being set to the ceiling of miszul, right?  If so, you end up with something worse than an continuous, non-linear (pedal vs torque) curve -- a discontinuous, non-linear curve -- because actual torque can never become requested torque if that means actual would exceed miszul, right?

I guess if actual torque is lagging requested (like during accel), there could be no problem...?  Or, no problem if the engine can't achieve the requested torque (in a particular load cell), but then the KFPED design would be poor.  It doesn't make sense to me to request a torque that isn't possible...

Am I missing something?  Thanks in advance for educating me.

John 

***  Final Update ***

Car is running fantastic.  Smooth idle and silky part throttle response, that rivals OEM.  I need colder weather to test cold/warm-up performance, but the improvement is such that I don't foresee issues.  Here's a summary that documents what finally fixed this issue for me.  Maybe it'll help someone...

Issue:  Lean out/stuttering most pronounced with cold engine at cam change under accel following rolling idle.

Root cause:  Poor steady state fuel tune aggravated by inappropriate IOP/IRL tuning.

Contributing factors: 
0) Novice tuner:  Lack of knowledge and context regarding appropriate FKKVS and KFMIOP/KFMIRL best tuning practices
1) Hardware change:  New, larger EV14 injectors and increase to non-standard fuel pressure supporting larger turbo upgrade
2) Logging software:  Durametric limited logging variables, their weird nomenclature and slow sample rate (about 5/s)

Solution (that produced good results fo me):
1)  Ensure KFMIOP/KFMIRL are stock values for RL below 100 to avoid unintended consequences, like weird ignition angles due to untuned torque reserve effects.  I'm sure there are a myriad of other reasons too.
2)  Ensure KRKTE, FRLFSDP and TVUB values set appropriately.  Be sure to account for any changes in fuel pressure too.  Note that KRKTE has a resolution of about 3%, but you can use FRLFSDP to set it more accurately if you have a non-returnless fuel system.  My stock Porsche file did exactly this (all stock points were 1.016, adding 1.6% flow to the poor resolution of KRKTE).  So, I followed their lead, which ensures the best results in the next step.
3)  Use stock FKKVS table values with a scaled injector axis as a starting point.  I found scaling the injector axis (ms) by ratio of new injector flow rate versus old injector flow rate to be effective. Scale TEMIN by the same factor.  Scaling in this manner makes sure your FKKVS goes low enough and has sufficient part throttle resolution, maintains a stock response. 
4)  Test (drive, log, analyze) for reasonableness and adjust KRKTE/FRLFSDP/TVUB as needed.  Don't move forward unless long term trims (only fueling reported by durametric) are within a few percent of 0% RKAT and 1.00 for FRA.  This minimizes risk in the following step.
5)  If using Duramteric, you have little choice -- you must disable lambda regulation to effectively tune FKKVS.  Set RLLRUN to 100 for all values to disable below 100 RL.  No need to go any higher for part throttle tuning.
6) Test part throttle performance progressively.  Only test after idling to full warm first (84C coolant temp on my Porsche) -- Expect cold idle lambda to be about 5% to 10% rich w/o LR.  Plan to do multiple tests, gradually expanding the envelope of each test.  Start with a low load, first gear only test (neighborhood driving) -- stay below 25 mph, for example.  Analyze it, make appropriate adjustments to FKKVS and test again expanding to 2nd gear.  Rinse and repeat until you've covered your part throttle driving range and your no-RL lambda error is small across the spectrum.  As you test, keep your load low until you're confident lambda stays below 1.05 or so.  With lean mixtures and higher loads, you can burn exhaust valves.  You'll know you are done with this step when the car drives like stock in part throttle without the help of LR.
7)  Reset RLLRUN to default values and perform a final test.  There should be little noticeable difference (from previous step final adjustment) AFTER driving for a few minutes.

Special note regarding FKKVS tuning/log analysis...  Filter the data to screen out transient (accel and decel) events (load changes). I found that calculating the median RL value for each FKKVS cell and screening data for a +/- 5% RL variance was an effective way to filter out transient noise.  Also, make sure you have sufficient test sample size for each FKKVS cell before making changes.  Be sure to smooth out FKKVS surface.

BTW...  My FKKVS increases (versus stock) around the cam changeover point.  This appears to be an effective way to address cam change related lean out issues.

Yeah I solved the cam leanout mostly through FKKVS as well on the dyno.

In retrospect, I think that is why Porsche has FKKVS and FKKVSNWS tables, unlike many other cars.  The Porsche tuners eventually realized -- even with well characterized, linear injectors -- that 2 tables were really needed to properly address the large differences caused by a 30 degree change in overlap.

I'm envious of the dyno.  It would be nice to be able to dyno and more easily/quickly cover a broader range, but I think I got a pretty good tune with street driving.  Trick was realizing RL had to be disabled, multiple tests were required and the data had to be carefully filtered/screened to minimize noise from street data.

Another key learning for me...  Because the EV14s are so linear (compared to the OEM Siemens DEKAs), I assumed I needed to set FKKVS to 1 and start over.  As part of this process, I realized a non-linear FKKVS doesn't mean the injectors are non-linear.  A lot, if not most, of the non-linearity in those tables are simply engine characteristics (varying efficiency, etc.).  So, starting with the stock values, scaled as appropriate, is a huge step forward versus starting from scratch.

Similarly, I currently use stock KFKHFM values even though I've changed the MAF to the RS6 from OEM.  I figure KFKHFM is really MAF housing characteristics -- not MAF.  So no need to alter it, if using the same housing.

This may not sound like a big deal, but my stock FKKVS and KFKHFM have variances across the table greater than 10%.  That is huge.  Not using stock as a starting point makes tuning much more difficult than it needs to be. 

I flattened the entire fkkvs and tuned it from scratch, but I could do it steady state...
Sorry for the one-liners, but I am quite busy and you are very verbose :D