S54 airbox +288/280 schrick on CSL software tune

[BMWfanz]

I use a wideband to log my drives. Once you tip in the throttle pedal, the DME will use the calculated value (+trims) to control fuel injection, but after a few seconds? It starts going closed loop with the O2 sensor.

If you check some of my logs, where pedal position is constant, you can see the drift here: https://datazap.me/u/bry5on/brysontu...g=0&data=4-5-6

Which logger are you using? Testo?

[Bry5on]

Which logger are you using? Testo?

Yessir, Testo

[BMWfanz]

Yessir, Testo

You using mss52/mss54/mss54hp ecu in Testo? Quiet annoying to select parameters there

[Bry5on]

You using mss52/mss54/mss54hp ecu in Testo? Quiet annoying to select parameters there

Yeah it’s a bit of a process every time. I’m only logging a few parameters as you get data faster with fewer selected, but it’s still an effort. I’m grateful to have the tool though, most of all.

[BMWfanz]

Yeah it’s a bit of a process every time. I’m only logging a few parameters as you get data faster with fewer selected, but it’s still an effort. I’m grateful to have the tool though, most of all.

Currently fighting with my notebook- on old one was windows 7 and my AEM wideband O2 worked properly- now on new laptop on windows 10 can't add external device (

[Pavlo]

I’ve been interested in tuning my part throttle CSL maps as well. How are you disabling fuel trimming to stoichiometric while driving, and what AFR are you targeting when it’s disabled? Between 2-3k there seem to be a bunch of resonance modes that cause AFR to swing wildly. I’ve tuned it out of WOT by changing target trims and RPM in my WOT map but haven’t figured out how to start with the part throttle map.

You can use the fuel trim values to tell you what to change and where. Lambda integrator scales into the VE, you can multiply VE by this value and it won't overshoot because of how the math works out. I semi automated it with Excel and Matlab and takes about 4-5 good drives until the entire table is done. My entire usable RPM range is near perfect, regardless of the conditions.

[Bry5on]

You can use the fuel trim values to tell you what to change and where. Lambda integrator scales into the VE, you can multiply VE by this value and it won't overshoot because of how the math works out. I semi automated it with Excel and Matlab and takes about 4-5 good drives until the entire table is done. My entire usable RPM range is near perfect, regardless of the conditions.

Maybe I’ve been getting this wrong the whole time, but I thought fuel trim was just a single number for short term and a single number for long term. How are you applying this to your whole map?

Also, I think I used the wrong term above when I said trim. What I meant was disabling closed loop control. It sounds like your method above actually works *with* closed loop control. Any chance you can point me to the table you’re using to adjust trims? I must have just overlooked it somehow.

thanks!

[Pavlo]

Maybe I’ve been getting this wrong the whole time, but I thought fuel trim was just a single number for short term and a single number for long term. How are you applying this to your whole map?

Also, I think I used the wrong term above when I said trim. What I meant was disabling closed loop control. It sounds like your method above actually works *with* closed loop control. Any chance you can point me to the table you’re using to adjust trims? I must have just overlooked it somehow.

thanks!

Ohh, I see what you are saying. I read "fuel trim" as instantaneous correction, not the adapted value (I do turn adaptations off when doing this, usually by raising the minimum temp to something high).

And yes, you are right that I use the closed loop to help me tune. The process is: do large data logs of Lambda Integrator values (just need one bank, but I sometimes average the two) >> create a histogram of this data in a table that matches your DME VE table >> Fill in all the unused cells with number 1 (means empty cells wont make any changes) >> Do a "dot" product of the Lambda Integrator table into your VE table (this is matrix multiplication step, so tables have to be identical in size). I can make a video of this in the next few weeks. Excel is all you need, but Matlab makes it quicker for me. I've played around with making a formula that uses wideband Lambda values, but Lambda Integrator works too well and also defaults to 1 when computer isnt using it to trim. I still do use my wideband to tune all the high load sections, but those are really easy compared to the rest of the drivability sections.

It was taking me waaaaay too long to make it perfect before figuring this out.

Edit: Tools I used to log is Testo, Megalog Viewer pro to create the histograms, Excel for math and to fill in unused values with "1" and Matlab to do the dot products.

[Bry5on]

Ohh, I see what you are saying. I read "fuel trim" as instantaneous correction, not the adapted value (I do turn adaptations off when doing this, usually by raising the minimum temp to something high).

And yes, you are right that I use the closed loop to help me tune. The process is: do large data logs of Lambda Integrator values (just need one bank, but I sometimes average the two) >> create a histogram of this data in a table that matches your DME VE table >> Fill in all the unused cells with number 1 (means empty cells wont make any changes) >> Do a "dot" product of the Lambda Integrator table into your VE table (this is matrix multiplication step, so tables have to be identical in size). I can make a video of this in the next few weeks. Excel is all you need, but Matlab makes it quicker for me. I've played around with making a formula that uses wideband Lambda values, but Lambda Integrator works too well and also defaults to 1 when computer isnt using it to trim. I still do use my wideband to tune all the high load sections, but those are really easy compared to the rest of the drivability sections.

It was taking me waaaaay too long to make it perfect before figuring this out.

Edit: Tools I used to log is Testo, Megalog Viewer pro to create the histograms, Excel for math and to fill in unused values with "1" and Matlab to do the dot products.

I would certainly appreciate this! I have some part throttle odd behavior between 2-3k that I’d really love to work on. My long term fuel trims are 1.1-1.2 range, and my hypothesis is that there’s a lean section in the tube somewhere that’s forcing 2-3k to go too rich after applying long term adaptations.

Dot products/matrix math are no problem here for me, I’m mostly interested in exactly how you’re logging and pulling out the data. Thanks again!!

edit: and also where the adaptation min temp is stored. I’m not in front of my windows machine to look it up, but this field isn’t ringing a bell as I’ve not played with it before.

[magaspif]

Pavlo, did you ever get a chance to write all of this up and maybe supply your tools?

[Bry5on]

Pavlo would you mind writing this one up? I’ve found the correctly labeled XDF and got the maps identified and am planning to get into it this weekend. I’d appreciate something that’s copyable if you’re willing to write it up before I recreate the wheel!

XDF with alpha n maps identified: https://github.com/Bry5on/MSS54-XDFs...mplete.xdf.txt

[Bry5on]

Back on this - is everyone using throttle pedal voltage (%) for y axis, or should we be using throttle plate opening (%), or something else? I understand Pavlo’s process here, other than exactly how to disable adaptations. Thanks in advance!

Disable adaptations in tune (how?)

For logging:
1. Lambda integrator (either bank)
2. RPM
3. Throttle voltage? Plate opening? Other?
4. AFR just for fun?

Then matrix math existing tune map with integrator values, or the number 1.

Then modify tune and repeat until integrator values are close to 1

[S54B32]

Back on this - is everyone using throttle pedal voltage (%) for y axis, or should we be using throttle plate opening (%), or something else? I understand Pavlo’s process here, other than exactly how to disable adaptations. Thanks in advance!

Disable adaptations in tune (how?)

For logging:
1. Lambda integrator (either bank)
2. RPM
3. Throttle voltage? Plate opening? Other?
4. AFR just for fun?

Then matrix math existing tune map with integrator values, or the number 1.

Then modify tune and repeat until integrator values are close to 1

Sorry, but imo this will never work. If you don’t have the (base)knowledge how to tune, I highly suggest to go to a professional (local or on board here). Or if you really want to learn and spend countless hours on this, I would start in the megasquirt community, how they tune VE table. Maybe watch some HPA videos (some nice input/knowledge there!)

[Bry5on]

Sorry, but imo this will never work. If you don’t have the (base)knowledge how to tune, I highly suggest to go to a professional (local or on board here). Or if you really want to learn and spend countless hours on this, I would start in the megasquirt community, how they tune VE table. Maybe watch some HPA videos (some nice input/knowledge there!)

I appreciate the sentiment, especially given the difficulty of establishing credibility on the internet. I’m just hoping to skip the work of going through all the decompiling work done 15 years ago many of us participated in (my memory isn’t THAT great from that long ago) by leaning on the community to grab the axes and apply them to what’s available to log in testo. Stoichiometry is not particularly difficult engineering/science! Megasquirt is simpler for me than the mss54 as the routines and maps are clearly labeled in my native language, and they’re generally running much simpler algorithms with closed loop feedback on wideband oxygen sensors with faster calculations and more granular tables.

I agree with the caution for readers that if you don’t know how VE tables and combustion efficiency work, don’t play around yourself in your tunes! And even if you do, you can make big errors that can grenade your hardware. Thank you for the heads up, I hope this doesn’t discourage folks from opening up in conversation here.

edit: For what it’s worth, my original plan was to figure out how to disable the lambda integrator and adaptations and tune directly via AFR, but if lambda integrator is dynamic enough in the logs, this idea seems that it may have merit. It’s the transients while the integrator is hunting for stoich that I’ll have to log to understand.

edit2: The tuners on the board here don’t seem to tune partial throttle fuel maps, and the answers I’ve gotten from those experts are that adaptation values of ~1.2 are normal. I disagree with this, as the whole table doesn’t shift when you change parts that affect VE. I have tunes from the most popular tuners here and they all use default CSL part throttle fueling.

[ppm008]

I've decided to tune my swapped E30 S54 by myself, currently already tuned idle and low speed jerks, now I'm tuning maps for partial load on the road and have some issues. I'm using XDF from GitHub for 0401 software. Now I'm stucked on fuel tuning at low rpms- issue is next: lean AFR (15-17 on AEM wideband logs) between 1500-3500 on partial load. I'm adjusting Fuel injection duration load base (KF_TI_N_RF 18x12 map) and problem is that injection time is not rising as it in my map and on logs I see lean AFR. Seems like on this duty injectors are maxed out and injection time is miximized. So I'm going to rise injection end angle map (KF_TIENDE_N_RF 8x6 map). But now I have some doubts- can my stock injectors survive on WOT tests?) Another question is Alpha N map- KF_RF_N_AQ 18x12 map) - is CSL software using this map?

So how did You tune idle and low speed jerks