@LamboLover would you be able to modify the code to use the P/D factor as well when using commanded wg for us? That would be a big step. I just bought a dynojet so I can do whatever testing you need.
DIY tuning
- Thread starter Sbrach
- Start date
Do we think we really need commanded? F series doesn’t have it anyway right so it would be E series only. I agree it was the easiest option for me to start tuning my car but I have started wrapping my head around the turbine power model and will be working on a PID tune when I get back in town. Spool and post shift boost control are much better on PID tunes and boost/load by gear is much easier as well.
You have looked at it much more recently than I, but do you think that there is a useful feedback variable to log with which to tune it? If it was wastegate duty base/final and target/actual boost, you can log both and tidy it up. The disconnect is probably the mindfuck not that you tune it by power and target/actual boost?
We have WGDC base, WGDC after PID, and WGDC final already available to log via MHD as well as boost actual. Boost target has two channels, labeled ‘Boost Target’ and ‘Boost Target Ex’ that behave oddly but Boost setpoint, although scaled incorrectly, I think should be all we need.
For me it is in fact the mindfuck of the compressor characteristic map to get base WGDC. The turbine pressure model is a neural network so trying to actually understand how it all works isn’t likely for me. If I can get the desired results out of trial and error with the table though that should be sufficient and a compressor table for my turbo should work for anyone with the same turbo and similar exhaust. Just need the time to experiment.
For me it is in fact the mindfuck of the compressor characteristic map to get base WGDC. The turbine pressure model is a neural network so trying to actually understand how it all works isn’t likely for me. If I can get the desired results out of trial and error with the table though that should be sufficient and a compressor table for my turbo should work for anyone with the same turbo and similar exhaust. Just need the time to experiment.
@LamboLover Commanded Wg is extremely useful and speeds up large single turbo tuning tremendously. All the of the stock frame N55 cars I've tuned use the stock boost control. However, when tuning the 6266's and up, it saves a ton of time in logging and on the dyno.
The reason people don't use commanded on the F series is because its simply not available.
The reason people don't use commanded on the F series is because its simply not available.
As we still are developing the N55, it would be a good idea in my opinion to use what we know works. Those of us that tune as a profession really need to save time. That's why im super stoked about tune on the fly! Every hour we spend fighting the stock boost control is time wasted. Just my.02.
Lambo, whatever you need from me and if I can be of any help in testing please let me know.
Lambo, whatever you need from me and if I can be of any help in testing please let me know.
I guess I don’t understand how adding PID to commanded is different than using the stock control other than how you get your base WGDC value. Commanded is just setpoint x rpm whereas stock is compressor and turbine model.
I would think it would be better to spend the time developing solid base compressor characteristic tables for different turbos so that work could then be reused on all N55 cars and likely on s55 and other MEVD engines.
The problem I have with commanded is that 22psi at 4K during spool, 22psi at 4K in 2nd after short shifting from first at 5k, and 22psi at 4K in 6th take drastically different WGDC to achieve. The commanded table is really only good for a single gear pull and if you adjust it to have decent spool you will overboost like crazy shifting below redline.
So if we get PID added to commanded you are still going to have to fight it because trims are going to be huge in those scenarios.
I would think it would be better to spend the time developing solid base compressor characteristic tables for different turbos so that work could then be reused on all N55 cars and likely on s55 and other MEVD engines.
The problem I have with commanded is that 22psi at 4K during spool, 22psi at 4K in 2nd after short shifting from first at 5k, and 22psi at 4K in 6th take drastically different WGDC to achieve. The commanded table is really only good for a single gear pull and if you adjust it to have decent spool you will overboost like crazy shifting below redline.
So if we get PID added to commanded you are still going to have to fight it because trims are going to be huge in those scenarios.
If the models and neural network are as good as they should be given the expertise, time and cost they took to develop, it should allow the drop in of the correct tables.
On the GT-R I simplified the boost control to WGDC = lookup(MAP, RPM) * lookup(gear, TPS). It didn't even have a boost target, but it was implied from the slope of the first table as MAP was below and above target, so proportional gain if you like, but no I or D. It seems to work even at quadruple stock power without spikes on all sorts of cars in all sorts of conditions. Trouble is on BMW, the nannies and inter-dependencies would ravage that idea. Otherwise I'm all for intelligent simplification.
I could take over any table lookup and replace it with a string of custom lookups. The problem is defining exactly what is needed and avoiding the consequences of it not fitting other models that depend on it.
On the GT-R I simplified the boost control to WGDC = lookup(MAP, RPM) * lookup(gear, TPS). It didn't even have a boost target, but it was implied from the slope of the first table as MAP was below and above target, so proportional gain if you like, but no I or D. It seems to work even at quadruple stock power without spikes on all sorts of cars in all sorts of conditions. Trouble is on BMW, the nannies and inter-dependencies would ravage that idea. Otherwise I'm all for intelligent simplification.
I could take over any table lookup and replace it with a string of custom lookups. The problem is defining exactly what is needed and avoiding the consequences of it not fitting other models that depend on it.
Unfortunately, this is my experience with the 450 as well. Anything above 18 psi and E 30 my high-pressure fuel pump tanks. I guess I need to look into getting a DW 400.
Stock boost control uses several other tables in the process in order to output a duty cycle. The cobb diagram from the N55 guide comes to mind. From my experience, it makes it difficult to get a nice smooth boost curve without at least 3 to 5 revisions... if not more. Stock control is just time consuming. Can you use it and get great results, eventually you can. But every setup is different and every car doesnt react the same way. With the complexity of the stock system, I just see having pre-set base maps, a very time consuming thing. There's definitely more than 1 way to skin a cat in this case for sure though.
As for the issue youre having with commanded at 4k, a lot of the issue you're describing is typically fixed by the p factors instant correction to boost error if i understand what youre saying correctly. What you can do if you have the p factor adjustment is set the duty cycle in the table to whatever it should be post spool. So say your wgdc limit is 70%. You know that after youre at target boost, you end up around 22-23 on your wgdc value from your logs. Youd then set the entire table to 22. If you need additional duty cycle, PID will add it up to 70% (if setup properly during spool).
heres an example.
Attachments
I really don’t get how stock boost control would be worse than commanded with PID.
Tune compressor characteristic and turbine pressure to give you a base close to what you want and PID handles small errors due to environmental and dynamic factors.
Vs. tune commanded table for one specific ideal scenario and then rely on large PID corrections and the undershoots/overshoots/oscillations inherent in a wound up PID loop to handle all the non-ideal / dynamic cases.
Tune compressor characteristic and turbine pressure to give you a base close to what you want and PID handles small errors due to environmental and dynamic factors.
Vs. tune commanded table for one specific ideal scenario and then rely on large PID corrections and the undershoots/overshoots/oscillations inherent in a wound up PID loop to handle all the non-ideal / dynamic cases.
Presently all the code mods are done in one automated build for all N55/S55 and with more automation could potentially handle 500 BMW model variants. If I did something it would probably be something aimed at all and based on a more generic automation method to add new tables and logic across platforms. Patching only 98G0B to make it rough like a piggyback would certainly be something to avoid.
I completely understand. I wouldnt want to spend time on 98g0B only wouldnt make a lot of sense. But keep in mind, powerful stand alone's use similar strategies. its not simply a piggy back thing.
AEM for example:
I really think we should peruse a less complex boost control rewrite especially if we're going to spend time making creating tune on the fly. The closer we get to the speed of a stand alone is a plus. Simple, fast, with proper control should be the goal imo.
Anyways, I think SBrach is in favor of the stock system, I would prefer to at least have proper commanded with full PID control on a toggle. If youd like to help @LamboLover , id appreciate it. If not we'll just keep doing what we're doing. Im just trying to save us time and energy. Especially for the big single kits being released by Motiv and a few others in the coming months.
AEM for example:
I really think we should peruse a less complex boost control rewrite especially if we're going to spend time making creating tune on the fly. The closer we get to the speed of a stand alone is a plus. Simple, fast, with proper control should be the goal imo.
Anyways, I think SBrach is in favor of the stock system, I would prefer to at least have proper commanded with full PID control on a toggle. If youd like to help @LamboLover , id appreciate it. If not we'll just keep doing what we're doing. Im just trying to save us time and energy. Especially for the big single kits being released by Motiv and a few others in the coming months.
Last edited:
It is primitive compared to what is in the DME already and jumps from boost target to WGDC which makes a dependency on the hardware (for example it does not linearise wastegate duty to wastegate position and loses a lot of other models).
That might not matter if it actually gives better results in difficult setups with no downsides as appeared the case with the remarkably similar strategy I used on the GT-R, except there it allowed 100% WGDC so the cars immediately gained power as that engine wanted 100% at the top and was efficient doing so, and also removed an overboost on DCT gear changes. No one bothered much that it did not have I and D feedback, but with BMW commanded wastegate, when I noticed it does not have PD, people asked for it, which suggests it might not be ideal without it?
So having not used it but noticed why N55 moved away from "commanded wastegate" that looks quite like the N54 model, is commanded wastegate in the realm of significant improvement on N55 is my main question, and in 2 years, majority feedback is that it is not. I can be convinced but have to weigh it against flashing a new car, or putting real flex fuel on an RS3 etc.
That might not matter if it actually gives better results in difficult setups with no downsides as appeared the case with the remarkably similar strategy I used on the GT-R, except there it allowed 100% WGDC so the cars immediately gained power as that engine wanted 100% at the top and was efficient doing so, and also removed an overboost on DCT gear changes. No one bothered much that it did not have I and D feedback, but with BMW commanded wastegate, when I noticed it does not have PD, people asked for it, which suggests it might not be ideal without it?
So having not used it but noticed why N55 moved away from "commanded wastegate" that looks quite like the N54 model, is commanded wastegate in the realm of significant improvement on N55 is my main question, and in 2 years, majority feedback is that it is not. I can be convinced but have to weigh it against flashing a new car, or putting real flex fuel on an RS3 etc.
The commanded is definitely not ideal until PD are added. Once we add PD all the problems that Sbrach mentioned above would be an easy fix. Tuning everything from a pure stage 2 to a big single 6466 would be a breeze. In my opinion, it is a very significant improvement over what we have now in terms of simplicity. BMW designed the boost control system to work with the stock turbo. With the complexity of the current system combined with the introduction of a completely different turbo charger, manifold, etc it makes things get extremely time consuming. SBrach has probably over 20 to 30 revisions on his personal stock turbo car alone.
I feel that the reason for the 2 year delay is because for the N54 community, Cobb did most of the dirty work for us. They developed the XML and even gave us base maps to work with along with an ATR guide. When they released the AP it had about 85% of the tables we needed. With the N55 cobb cancelled on us so that same level of development never got done. Now we're doing it ourselves and its a much slower process. I think we have to consider things from a tuning perspective and its just not practical to fight a neural network for something as simple as boost control.
Speaking of the GTR. Tim Bailey @ cobb will even tell you that the GTR is one of the most simple cars he's worked with and i completely agree. I could tune a GTR in my sleep. Thats why its so easy to get nice smooth power. Boost, timing, fueling... all easy. We should look to mimic that on our platform.
One last thing id like to note, theres going to be more 335i's on the road than RS3's. I think it would be more beneficial to a larger group to work on us first
.It is exactly that it needs PD to be ideal that concerns me, it should work well without it if it is a good strategy I feel. I have not had the time or inclination to tune MEVD whereas in GT-R development I started by trying to tune it and only when it could not be tuned did I design and code a replacement wastegate control. For 2 years I have been hearing opposite opinions about whether the MEVD factory system is feasible for wild setups, whereas with the GT-R it was universally acknowledged as untunable on standard logic. The difference with GT-R is that you could ignore boost target with no complaints. Also on MEVD the present PD is feedback to compressor power, not just a switch to turn them back on. It can be done but I am a long way short of having the info to help decide what/how to do it rather than how to code it.
Last edited: