Leaning out under boost

[Kimmo]

Been a while since I gave the car any love, and it started making some new noises, so I paid Benny a visit for some clues... We took it for a spin and he gave me a bunch of advice, and also said to show him a log and LV.

Yeah, something's up...

BtSsm_LV_20230915_1558.png (53.4 KiB) Viewed 3941 times

Just trying to figure out how to get a log up at Datazap... linkage pending

[Kimmo]

Okay, went for a spin and generated another couple of logs and LV.

BtSsm_LV_20230917_1215.png (52.62 KiB) Viewed 3931 times

Not sure what's up with the logs finishing at 200 seconds though, I didn't notice it did that until I got home and looked at them. The first one is using the last of a bit of cheap ethanol blend (fuel light on) and the next one is with a full tank of BP Ultimate.

https://datazap.me/u/kimmo/cheap-ethanol-blend
https://datazap.me/u/kimmo/bp-ultimate

Bloody pain in the arse how the logs got cut off though; I went for a spin up and down the freeway for nothing, it cut off before that >:(

[Kimmo]

Worked out the 200s cut-off was because I was using the rear demister switch, and it runs on a timer.

Went for another run, but like before, I forgot to put it on S# until just before the last two WOT runs onto the freeway... at least it actually got recorded this time.

https://datazap.me/u/kimmo/bp-ultimate-s

[bigBADbenny]

Thanks for posting here, as I mentioned, I feel there’s no tangible benefit to doing diagnostics via PM.
Info should be shared for education and pm’s… well they’re ultimately personal.
I help out worldwide so the last thing anyone wants is a notification (what notifications lol) in the middle of the night, usually from me.

IMG_0286.jpeg (211.35 KiB) Viewed 3924 times

Ok so what is learning view?
It’s a graphical representation of the ecu learned values from when the LV (learning view) was made.
It’s a useful addition to a RomRaider install and is included in the BTssm app.

Let’s assume you’ve read the fine print in the infographic above, or read up on the RomRaider forum and we’re all on the same page.

[bigBADbenny]

Your Learning View:

IMG_0287.png (52.62 KiB) Viewed 3924 times

Top line: IAM is still at 1.00, the ecu isn’t yet pulling timing globally, eg due to knock events etc.

Next line: fuel learning over maf airflow ranges, at idle your long term fuel trim is within spec, but as airflow increases, it’s dropping off sharply.
That -10%… that’s the ecu pulling 1/10 of the normally required fuel, to keep the afr at 14.7.
How would you feel if we pulled your fuel? A little lean?

Indeed the FSM workshop manual (red link in my signature) specifies +/-10% fuel learning as the sane limit for safety, rich lean DTC’s trigger at +/-15%. Tuners and the aftermarket recommend +/-5% as the take action to diagnose limit. Either way you’re headed in the wrong direction.

The ecu is judging the engine is too rich, but why?

If the airspeed past the mass airflow sensor was slowed down with everything else being equal or as per usual, we’d have too much fuel to go with less air?

Well that’s the effect of having an inlet tract leak before the throttlebody.
That it’s less at idle, more-so under load suggests it’s maybe intercooler or inlet pipe related as opposed to say the pcv system. It matters not since an inlet pressure test will get to all leaks.

We don’t need to guess, because we know how to test.

Ok now we’re looking at the third section: timing correction.
This is a graphical representation of Fine Learning Knock Correction…
That’s learned knock events, these tend to be more tune related than mechanical knock, but like the second row, we want to see less of non zero records here.

Overall this LV looks like the engine is headed in the wrong direction with little chance of self improvement, yet driving it, you’d never know. We made some logs, reasonably gentle driving and the issue didn’t look so bad, but that’s the kicker: stay out of boost where practicable until the issue is diagnosed.

Another important clue is your car has some mods including airflow mods, yet is on what looks to be a stock rom!
We can see this in both the fuel learning airflow grams per second ranges, and in the rpm ranges in the timing correction table: eg 10,000rpm is one signifier.

Anyway, I wanna say, any engine mods and no tune, that’s nuts.
As in looney. Let’s fix the issues and get it tuned.

Next we’ll use Datazap and MLVHD apps to plot the datalogs.

[bigBADbenny]

WWW Datazap dot me, what a convenient way to share and visualise data logs.

It works great on phones and computers, the less reflective can share logs without needing to get their heads around how the cloud works or usually doesn’t when they fail to grasp the nuances of actually sharing a link to their files, it’s simply too abstract for the peripherally interested layperson. So datazap, it’s self evident really.

IMG_0290.png (74.42 KiB) Viewed 3922 times

It’s a boon for sharing then but for datalog analysis, having only one graph window for many parameters, and only two data scales, you’re best off comparing up to say two to four parameters at one time before the view becomes overcrowded.

Here’s rpm vs the FLKC & FBKC, it’s intermittent, on load…

IMG_0291.png (50.65 KiB) Viewed 3922 times

Somehow this screen shot is upside down, but it’s a good point, versatile data visualisation software enables full manipulation of said data to suit an end. That’s where MLVHD provides solutions for taking a consistent diagnostic approach.

[bigBADbenny]

So nextup, go to the linked DataZap page, let the page load, find the download button, put the .csv file in the Kimmo folder in my BTssm logs folder, open said log in MegaLog Viewer HD (paid version) app on my MacBook....

Screen Shot 2023-09-18 at 08.58.01.jpg (171.8 KiB) Viewed 3914 times

So here's the initial view of the supplied log. All that data is a bit overwhelming I agree, however I want to see all the most relevant data in context to get an overview and to spot any outlier anomalous events or trends.

This is a rescaled full screenshot so the quality is a bit iffy, but I wanted to show the whole interface.

[bigBADbenny]

Log Viewer graph 01.png (134.69 KiB) Viewed 3913 times

Here's the same view, but with MLVHD generated graph: more detail but a smaller file...
I've parked the cursor on the FKC event at the beginning of the log, we can see the parameter min/max values and the parameter scaling at the left margin, and on the right, the cursor values.

So I want to mention that FKC (FBKC) is an immediate record of a knock event, whereas FLKC is a learned response to knock.
Whilst we can see the engine isn't in immediate danger we want to get to the issues maike your LV's look concerning.

In the log we can see the FKC event takes place during an accelerative gear change, and this is inline with a potential issue with your engine mounts, and your PW tmic probably hitting the Ultrex strut brace during engine twisting moments.

But again from the LV we see your fuel trims are out of range, and we can see this in better detail in MLVHD, which will inform the diagnostic approach to resolving that issue. Usually beginning with an inlet pressure test.

Lets break down the graphs in the screenshot:

The first graph is about what the car is doing, speed, rpm, tps and boost: primary context essentially.
This is super important, logs without vehicle speed drive me nuts, theoretically that can be extrapolated from gear position and rpm, but you've included it so great.
I'm always peeping idle vacuum, max boost as these have known specs and can indicate boost vac leaks or issues.
TPS is very useful and in conjunction with accelerator pedal angle we could estimate what S-drive mode the car was in (throttle map)...

Then AFR vs knock vs whatever you feel is relevant, I put closed loop switch in there, but often its IAM or knock sum etc.
Here, AFR is scaled so stoich 14.7 is on the centreline.

Then there's fuel trims: long and short term (learning and correction) and total fuel trim which is LTFT+STFT, a custom parameter that allows fuel trim to be seen in context.
Here, the parameters are scaled to +/-10%: the sane range as specified acceptable in the FSM, again this means that these parameters have zero on the centreline. But I have other custom parameters that use tighter, looser or no scaling. It's all about how one wants to visualise the data for clarity.

Then load and spark timing afain for context. Note the somewhat inverse relationship between them.

Next is MAF in volts and grams per second, the latter being an ecu generated parameter. For sanity checking an clarity these are duplicated with custom scaled parameters that place zero values on the centreline: if there's maf signal dropouts, we need a way to easily spot the fault.

Then there's VVT angles and time parameters, Time Accumulate being a custom parameter that normalises the time scale across multiple log files opened in the same MLVHD window.

So thats an explainer of my first look diagnostic MLVHD Log Viewer screen.

[bigBADbenny]

Log Viewer graph 02.png (107.3 KiB) Viewed 3910 times

Ok, so we look at that overview and get a general idea of how the car was driven, what happened (especially anything learning related) and if there's anything odd or anomalous happening.

Overall it looks great apart from the little bit of FKC and the fuel trims.

So let's zoom in on that FKC event, see what gives. It initiates around a second after reapplying the throttle after a gear change, the engine is still in vacuum, closed loop fuelling.

Remember that the more parameters logged, the more parameter load on the ecu, the less log data point per second and less data resolution. However there’s limits to the data rate depending on the ecu & connection type. Here the data rate is around 10 samples per second per parameter. Canbus parameters can have a faster acquisition rate and better resolution, but for applicable ecu’s you’ll need the can patch for RomRaider or Btssm and experiment with parameter load.
Suffice to say the ecu is not an oscilloscope, but we can make up for data rate by capturing as much data as is practical: mlvhd can handle massive datalogs.



You're still on the stock tune so the conservative open/closed loop delay has not been adressed and we see this in your total fuel trim dropping to -9% with the clutch in, AFR briefly leaning, then TFT (total fuel trim) going to +17%, and as the knock event initiates, it's back to -4 then -12%.
Now all this is fairly normal ecu emissions activity, no need to panic, but you can see how LTFT holding negative will influence everything else.
Plus that FKC might influence subsequent FLKC which is triggered later in the log at the same load & rpm conditions.

Remember this is a fault somewhere between the air filter, and the intake valves or potentially even the exhaust, but we won't assume anything, even that our tests are working: we'll use logging to see if the fixes are having a positive effect.


To me it's not particularly clear if the FKC is AFR or mechanical knock related, but this doesn't matter as we use the log data to help us verify that the real world fixes are making progress in the right direction.

At this point thats just:
Verifying the engine, pitch stop and gearbox mounts are not separated or failed: recommend new OEM engine mounts and STi groupN pitch stop and gearbox mounts for decent NVH: engine shakes and rocks during cranking and engine brake testing.

Isolate the PW tmic from the strut brace.

Inlet pressure test.

If you're considering eg a silicone inlet pipe (currently the last remaining stock piece in the inlet tract apart from airbox and maf) I recommend PSR V2 type to suit the OEM PCV piping returns, you'll require a tune, to get tuned you'll need a clean bill of health/pass basic health checks regardless.

Next up we'll use the MLVHD scatter graph feature to make very pretty histograms of various engine parameter combinations, then drill into the graphs using a variety of filters, for clarity.

[Kimmo]

Wow, man. That's about 500% more info than it takes to make my head spin; I'll have to reread that a few times to get it all. I think I'm getting the gist though, and first priority seems to be the engine mounts to eliminate the possibility of the inlet tract leaking because it's smacking into stuff. Then leak test, yeah?

Gonna owe you a few favours after all this...

[bigBADbenny]

Two separate issues, the grinding noise you hear on throttle application might cause FKC, the oil you spotted in your tmic piping or elsewhere, whilst normal in moderation, can rot the stock inlet pipe at the turbo end.

Pulling fuel under load, that might cause knock, so the circle completes.

That’s to say FKC, I don’t think it can affect fuel trims, perhaps maybe via FLKC affecting timing?

But we can approach the issues separately, regardless since it’s two different systems ultimately, mechanical vs emissions.

It sure is a lot of words, but the process is relatively quick since I have years in the preparation and development of the MLVHD app settings.