Karter16's Silbergrau E46 M3 Journal

[wahsm]

Hey yes I will! I'm away for a few days but will be putting this up next week when I'm back in front of my PC :-) thank you!


Sent from my iPhone using Tapatalk

That's awesome, thank you. Such a clean solution.

[karter16]

Yep, you've got it. If you're logging via D-bus and not CAN, it's really hard to trust the data at 100rpm fidelity. There's a lot of latency on the D-bus for request/response (testo logging is alternating requests) and the polling frequency is super low. For the higher fidelity logging TPS, RPM are both on CAN and a wideband should be plenty responsive.

While you're in there disassembling, if you'd like to continue your current approach, you should really see if there's a way to broadcast more useful values over CAN.

Yeah fair point re resolution/sample rate. I'll have a look into the CAN functions when I get home next week - that would be pretty cool!


Sent from my iPhone using Tapatalk

[karter16]

You ever end up sharing the file for the adapter? Have the CSL rail and same sensor, your adapter is very clean, would love to print it. Your build journal is dangerous to follow lol.

Hey yes I will! I'm away for a few days but will be putting this up next week when I'm back in front of my PC :-) thank you!


Sent from my iPhone using Tapatalk

[wahsm]

You ever end up sharing the file for the adapter? Have the CSL rail and same sensor, your adapter is very clean, would love to print it. Your build journal is dangerous to follow lol.

[Bry5on]

Yes I've found the same with the first round of VE tuning - threw one cell to the lean side. The other thing I've done is built out a histogram of RO/RPM/Lambda at 100rpm intervals to better see where there are peaks and troughs between the RPM interpolation points in the CSL VE table which is interesting, as I fine tune I may well end up adjusting the interpolation points to smooth things out more.

Yes am keeping the VL table in mind, but wary of doing too much there until I get a Wideband and get it in play. I've found logging STAT_VL_EIN useful for understanding in the logs exactly when the car is in the VL operating mode.

If I can get the VE map tuned well enough I theoretically should be able to then use the lambda measurements to help evaluate tweaks to camshaft timing to identify where VE is improved/worsened (will have to experiment to see if this works in practice).

Yep, you’ve got it. If you’re logging via D-bus and not CAN, it’s really hard to trust the data at 100rpm fidelity. There’s a lot of latency on the D-bus for request/response (testo logging is alternating requests) and the polling frequency is super low. For the higher fidelity logging TPS, RPM are both on CAN and a wideband should be plenty responsive.

While you’re in there disassembling, if you’d like to continue your current approach, you should really see if there’s a way to broadcast more useful values over CAN.

[karter16]

Nice. Some areas of the map are sensitive to fueling and can ping if they’re not right on. Others are sensitive to throttle and cam timing. I found the same, translating 100% didn’t really work but it gets you real close. My advice is always sanity check your VE tuning with logic and ideally the wideband. I’ve found it’s thrown a few cells lean, but mostly works. Keep in mind how the full load table is incorporated while you’re doing this.

Yes I've found the same with the first round of VE tuning - threw one cell to the lean side. The other thing I've done is built out a histogram of RO/RPM/Lambda at 100rpm intervals to better see where there are peaks and troughs between the RPM interpolation points in the CSL VE table which is interesting, as I fine tune I may well end up adjusting the interpolation points to smooth things out more.

Yes am keeping the VL table in mind, but wary of doing too much there until I get a Wideband and get it in play. I've found logging STAT_VL_EIN useful for understanding in the logs exactly when the car is in the VL operating mode.

If I can get the VE map tuned well enough I theoretically should be able to then use the lambda measurements to help evaluate tweaks to camshaft timing to identify where VE is improved/worsened (will have to experiment to see if this works in practice).

[Bry5on]

After a few days of Mr Three being in hospital we got him home yesterday morning. Yesterday evening and today I tested v3,4 and 5 of the tune. 3 and 4 were duds (in comparison to v2) in various ways (pushing the boundaries of the theory covered above too far, but on the plus side helping me get a feel for the various adjustments) but I'm very pleased with how v5 has turned out and am happy that I have a good base now to build on. To that end I ran the first round of logging for VE tuning this evening as the tune is currently rich across the board. After a few rounds of VE tuning I'll then start making fine adjustments to the tune as needed once the larger adjustments are out of the way. I managed to force some pinging under heavy load from about 1100rpm, which is interesting as I'm entirely Euro tune in that range so would be interested to know if that happened on my original tune as well - I don't recall having come across it before. But aside from that it's feeling good.


Sent from my iPhone using Tapatalk

Nice. Some areas of the map are sensitive to fueling and can ping if they’re not right on. Others are sensitive to throttle and cam timing. I found the same, translating 100% didn’t really work but it gets you real close. My advice is always sanity check your VE tuning with logic and ideally the wideband. I’ve found it’s thrown a few cells lean, but mostly works. Keep in mind how the full load table is incorporated while you’re doing this.

[karter16]

After a few days of Mr Three being in hospital we got him home yesterday morning. Yesterday evening and today I tested v3,4 and 5 of the tune. 3 and 4 were duds (in comparison to v2) in various ways (pushing the boundaries of the theory covered above too far, but on the plus side helping me get a feel for the various adjustments) but I'm very pleased with how v5 has turned out and am happy that I have a good base now to build on. To that end I ran the first round of logging for VE tuning this evening as the tune is currently rich across the board. After a few rounds of VE tuning I'll then start making fine adjustments to the tune as needed once the larger adjustments are out of the way. I managed to force some pinging under heavy load from about 1100rpm, which is interesting as I'm entirely Euro tune in that range so would be interested to know if that happened on my original tune as well - I don't recall having come across it before. But aside from that it's feeling good.


Sent from my iPhone using Tapatalk

[Bry5on]

That last bit is the ticket. Now compare the tune numbers from CSL vs euro and you’ll find that the euro intake cam timing should actually be making more power (and run smoother) above 3k rpm than the CSL values in most cases

[karter16]

Ok I think this approach is helping my brain. Here's another interesting view - this is the difference between the Euro M3 tune and CSL as it pertains to the crank degrees at which the intake valve is completely closed.



And correspondingly for when the intake valve opens:



Arguably the point at which the intake valve completely closes is the most important of the 4 (intake close, intake open, exhaust close, exhaust open), so I wanted to isolate that and examine the difference between the two maps. The intake open difference is directly related, plus an offset to account for the extra duration of the CSL cam.

As could be expected we can see that at higher RPM the CSL is configured to to close the intake valve later (by between 5-10 crank degrees) than the Euro M3 tune which will drive more top end power. we can see at low load between 4300 and 6300 RPM that the CSL closes the intake valve earlier than the Euro M3 tune. I presume this is to keep drivability decent with the bigger cams, as presumably is the other island at low RPM thru mid and high load.

[heinzboehmer]

I've peeked at similar spreadsheets on Bryson's laptop, but the visualizations towards the bottom of the post are awesome. Thanks for sharing.

[karter16]

To Bry5on's point I need to make sure I understand the theory of cam timing well. I understand the individual concepts, but where I've been struggling is putting it all together and visualizing what it means in different scenarios. To this end I decided that some spreadsheets would help (spreadsheets always help).

First up I took the interpolated AVAN and EVAN tables for both the Euro M3 and CSL tunes and calculated the cam overlap for each for each point on the tables:





As you'd expect we have low overlap at idle, high overlap in the midrange, sloping down to high RPM.

So how does that work out as a difference between the two?



The CSL generally runs more overlap everywhere apart from high rpm low load, and particularly in a few different islands in the low and mid ranges.

However - we need to remember that the CSL cams have somewhat greater duration than the standard M3 cams. For the purposes of comparison then we should account for this.




That's cool and all, but what it still doesn't tell us is whether the Euro M3 tune and the CSL tune differ in where the centre of the overlap occurs (e.g. before or after TDC) which is also important information in understanding the differences between the two and why.

So I calculated the centre of the overlap for each of the tunes





And from there we can compare the difference in overlap centre between the two:



Interesting stuff to look through anyway - I feel now with this view I can better visualize and understand the impact of the various tweaks etc. that I'll make. I've only just put these together and haven't worked through them in detail to look at various areas of interest but will update here as I do.

[Bry5on]

Haha - I would love to compare with your work but at this point agree - sometimes it's good to have separate efforts and then we can compare later. In terms of my process, I'm still figuring that out haha. At the moment all I really have is a tune that transitions from the Euro M3 tune to the CSL tune, with injection timing erring on the side of the CSL tune (which is generally richer) to be safe. I intend to try out a couple of different things from here to see how it goes. I ran a logging session this afternoon with LTFT off and adaptations reset to confirm my assumption that the midrange would be running rich - which it is. Plan from here is to play round with exactly where in the various maps I bring the CSL values in as I suspect I could bring them in at a lower RPM than I currently am in higher RF conditions (we'll see if that's a correct assumption or not ). Really what I'm battling at the moment is my own level of understanding.

One thing I'd love to know (it doesn't relate to any particular issue I'm having, I just was reading the other day and saw it and am super curious). In one of your first posts on the Mullet Tune you said you had to adjust the exhaust cam advance at mid throttle in the high 3k rpm range. How did you work out that was what was needed? From my (admittedly limited) understanding it seems like that would be one of the less consequent things to change and I'm curious as to how you zeroed in on that being what needed to be adjusted?

Good question, honestly hard to think back on that specific moment now that I'm through it. If I can venture to guess, there was probably a lump in the power delivery as you rolled into the throttle. This is one of the tricky things about the tuning in all honesty, throttle application rate changes fueling and can lead you to chase your tail a bit (think, "that just worked last time, what the hell?!").

In the end, doing some homework on cam timing, first intake and then exhaust, would be a smart move if you're still at the basics. You might find that you'll prefer more euro values than CSL in many cases

[karter16]

I’m enjoying watching your approach as well and resisting the temptation of sending you the files to compare. I don’t want to interrupt your process and would really like to see if we converge on the same answers, particularly since I’ve been doing some maps with the wideband O2 sensor in the loop.

Most of my imperfect running is the first 60 seconds from full cold starts in ambient temps below 60F/16C. Lots of variables to chase in the cold start tuning, and to tune via AFR you’ve got to disable secondary air, etc etc. Plus you only get 60 seconds per day to tune while you’re taking meetings for your day job.. Fun.

As you get some seat time, I expect you’ll find idiosyncrasies to chase - looking forward to you getting more time in it, and happy to share notes if you find some odd stumps here and there. It was helpful for me to have a few friends cars here (thanks Heinz) to load tunes on to hone in on what worked best, as each car responded slightly differently to some changes, but all run well on a single converged file.

Haha - I would love to compare with your work but at this point agree - sometimes it's good to have separate efforts and then we can compare later. In terms of my process, I'm still figuring that out haha. At the moment all I really have is a tune that transitions from the Euro M3 tune to the CSL tune, with injection timing erring on the side of the CSL tune (which is generally richer) to be safe. I intend to try out a couple of different things from here to see how it goes. I ran a logging session this afternoon with LTFT off and adaptations reset to confirm my assumption that the midrange would be running rich - which it is. Plan from here is to play round with exactly where in the various maps I bring the CSL values in as I suspect I could bring them in at a lower RPM than I currently am in higher RF conditions (we'll see if that's a correct assumption or not ). Really what I'm battling at the moment is my own level of understanding.

One thing I'd love to know (it doesn't relate to any particular issue I'm having, I just was reading the other day and saw it and am super curious). In one of your first posts on the Mullet Tune you said you had to adjust the exhaust cam advance at mid throttle in the high 3k rpm range. How did you work out that was what was needed? From my (admittedly limited) understanding it seems like that would be one of the less consequent things to change and I'm curious as to how you zeroed in on that being what needed to be adjusted?

[Bry5on]

I’m enjoying watching your approach as well and resisting the temptation of sending you the files to compare. I don’t want to interrupt your process and would really like to see if we converge on the same answers, particularly since I’ve been doing some maps with the wideband O2 sensor in the loop.

Most of my imperfect running is the first 60 seconds from full cold starts in ambient temps below 60F/16C. Lots of variables to chase in the cold start tuning, and to tune via AFR you’ve got to disable secondary air, etc etc. Plus you only get 60 seconds per day to tune while you’re taking meetings for your day job.. Fun.

As you get some seat time, I expect you’ll find idiosyncrasies to chase - looking forward to you getting more time in it, and happy to share notes if you find some odd stumps here and there. It was helpful for me to have a few friends cars here (thanks Heinz) to load tunes on to hone in on what worked best, as each car responded slightly differently to some changes, but all run well on a single converged file.