1st time enabling closed loop boost. Sanity check please?

[hjtrbo]

Hey team, just wondering if I could get a sanity check on the included log? The log is from the file "91_In Car_Closed Loop Boost 2.tec"

This is my first time switching on closed loop boost control. I followed the guides etc. The control looks okish overall. There are 2 spots that I have since adjusted. 

The first spot is the initial transition from underboost to closed loop. I have softened up the target boost and dropped the duty cycle a bit to try and get a ramp in as opposed to the light switch like action that I have in that log. Hoping that will avoid the initial boost spike I get.

The 2nd spot is the large correction around 5000rpm. I added a bit more to the duty cycle in that area. Started raining so will check again next time I go out for a run. The file labelled "92_Ready to load.tec" is the adjustments I made to address the 2 spots noted here.

Interesting to get 100% duty cycle up top. I did not expect that. Have bought some plumbing fittings to do a boost leak test on the pipe work. Turbo is also starting to blow a bit of smoke so time for a new cartridge. 

98 fuel, wideband matching nicely, FG Mk1 XR6T, F6 turbo. Turbosmart actuator with 12psi spring. Boost target is 17.5psi. Butt dyno got 370rwkw (no correction) not bad considering the air wasn't the greatest today. Still need to get those knock ears on....

92_Ready to Load.tec 91_In Car_Closed Loop Boost 2.tec 91b.teclog Master List.tlo

Edited May 14, 2022 by hjtrbo
added screen shot to save opening log

[Puffwagon]

In my experience virtual dyno lines up with a real dyno with 6 smoothing.

[hjtrbo]

I'll check it out. Thanks. 

[hjtrbo]

Hmmm, more work is needed.

3rd gear roll on WOT up to redline, all is well.

Do it again but shift to 4th. No good. Boost oscillates under and overboost.

5th gear roll on, oscillates under and overboost. 

Time to have a play....

[Roland@pcmtec]

What gate are you using? It might just be a mechanical limitation due to high backpressure and a generally non linear system.

The gt3582 will overboost if you hold it steady state 18psi at 4k on a dyno with the internal gate, has it been ported at least?

[hjtrbo]

It's a turbosmart IWG75. Yes flapper mod has been carried out. 

I have no problem in open loop. Based on that I point to the closed loop control being my issue. Most of it is still set at default. 

I noticed it's quite fast to react and the oscillations are the same period. That suggests an over damped control loop. 

I'll cut the gains and soften the clips to try and make the response slow and doughy. I can tighten things up later once I have control.

[Roland@pcmtec]

8 minutes ago, hjtrbo said:

I have no problem in open loop. Based on that I point to the closed loop control being my issue. Most of it is still set at default.

Good news

If you can get to a dyno and do some steady state tuning this will get much easier. Then you can do repeatable tests with the same system gain (eg rpm and engine load) per cell.

Doing it on the road gets tricky as you've found.

My advice is as follows (sounds like you understand PID control theory anyway).

Short oscillations is too much P gain (can also be too much I as well, but most likely P)

Undershoot is not enough I gain

Boost creep it taller gears with longer runs is not enough I gain.

Also you can widen the closed loop min/max hysteresis in case its kicking in and out of closed loop/open loop. I've seen this be a problem when the open loop table is massively out, it will be closed loop, reach the max error, go open loop, then it will kick it wildly off course and it all goes pear shaped. Not suggesting this is the case but it can help with honing the closed loop control to widen the range temporarily.

[hjtrbo]

Thanks Rolls. Leave it all with me. Plenty to play with  

[Roland@pcmtec]

If you are bored download one of these tools https://www.pid-tuner.com/software/ https://pidtuner.com/

Then you can tune a pressure control system with a large variable system gain and see where it starts to fall over. 

We used to use PAS TuneWizard when they had a free version in Industrial Control and I found the simulators were great for reminding my brain which direction to move each gains to affect different issues.

Some will even auto import data and do a auto tune. But this relies on good steady state data and a known PID algorithm which you really need a dyno for.

edit: Even found some screenshots of the old tool from a post I made!

http://www.plctalk.net/qanda/showthread.php?t=93068

[hjtrbo]

Plctalk!, crickey haven't been on that site for years since I switched careers. 

Thinking out loud, the fact I already have my oscillations I was just going to Ziegler Nichols it and see what happens....

Thanks for the link to the software. I'll definately play with it. 

[Roland@pcmtec]

If the period is repeatable then that might be an easy win. 

[hjtrbo]

damn, that was boringly easy to fix. Still not perfect but good enough to keep moving forward and get some nice tight control.

Cut the gains in half, zero'd out the adaptive feed forward, halved the underboost duty cycle.

3rd to 4th Before

After - Still some tweaking to do but heaps better and very driveable

5th gear pull before, fucken disaster

After - Close to perfect

Edited May 16, 2022 by hjtrbo
Adjust headings and sort screen shot order

[Puffwagon]

51 minutes ago, hjtrbo said:

halved the underboost duty cycle.

That will make it spool up slowly and isn't how you should fix it. Yeah it works but there are better ways to do it that don't involve touching the underboost value or the pid control.

Try other stuff, you're very close and I'm sure you will get it working properly eventually.

[hjtrbo]

6 minutes ago, Puffwagon said:

That will make it spool up slowly and isn't how you should fix it

Yeah understand. This is just the start of the journey. Just needed to get it under control to start with.... Give me 2 weeks

You'll hate me for this: Against your better judgment the PI gains adjustment seems logical. A (slightly) larger turbo, aftermarket wastegate and port mod have changed the overall control system characteristics. In the Industrial Automation world, a change like this would trigger a re-calibration of the control loop so I am just calling on that experience as I have nothing else to go on...

[Puffwagon]

Fair enough, I see the logic and playing with the pid controls will probably help fine tune it after the wgdc table and other setting are correct.

Leave the underboost value over 80% to keep it spooling as quick as it used to. The overboost value can be doubled (still a negative number obviously) or slightly more as it's got a very slim margin to start with. This may be the majority of your issue.

Get the wgdc table slightly under the boost level you want to run, it works better that way. I've done all those mods you've done and have never had to back off the underboost duty.

In any case there's nothing wrong with playing with stuff to work it out yourself, it's encouraged providing you kinda know what you're doing, which you do.

[hjtrbo]

3 minutes ago, Puffwagon said:

Leave the underboost value over 80% to keep it spooling as quick as it used to.

In the next tune for tomorrows run to work. 

Thanks man, I always appreciate your help! 🍻

Edited May 16, 2022 by hjtrbo
spelling

[Puffwagon]

Eh I wasn't going to say anything but you guys were overthinking it by a nautical mile. It's a very simple thing to do once you know how.

[hjtrbo]

Bloody engineers

[hjtrbo]

As per @Puffwagon have the underboost DC raised and working at 0.8 now. Next run I'll set it back to 1.0 and see if I can get it to work. Few more adjustments to the gains and started modifying the underboost table. 

Its actually quite a challenge to tune. Getting it right for a idle 3rd to WOT rip up through the gears is one thing, and I think I got it. Then when I'm doing 100km/h and pull the shifter back to 3rd and to simulate passing a car its not quite right. So I dumb it down a bit for that scenario then it requires a redo of the WOT run. Anyway, lots of back and forward and getting closer. 

Fully understand now why people leave it full open loop.

Now, if anyone out there would be willing to share a nice cheery picked screen shot of a closed loop WOT log run through a couple of gears I'd sure appreciate it. Need to know what I'm aiming for. 

 

[Puffwagon]

Did you change auF0296 to -5psi?

[hjtrbo]

Negative. Still at default -2.2psi. I'm trying to leave that scaler alone as I have already been bitten before when I had a wastegate signal line leak and I was running max boost for a little while before I knew better.

Extending the underboost range has helped alot. It gets it onto the PID controller earlier into the boost ramp.

I've nearly got it not hitting auF0296. If anything it barely just clips it. I think once I implement the change below I'll be golden. What it should do is still let boost come on quick when ramping up from low rpm but when I am already at mid to high rpm (I.e. the getting ready to pass a car in 3rd gear scenario ) it should ever so slightly taper in the commanded boost which will reduce the error term when it clicks over to the PID loop. You're going to tell me I'm over thinking it again.....

 

 

Edited May 17, 2022 by hjtrbo
screen shot edit

[Puffwagon]

Nah mate you do you

[Roland@pcmtec]

Over thinking is fun. Without over thinking we wouldn't have PCMTEC!!!

[hjtrbo]

Calling it done. Is it perfect no. It's good enough. As per @Puffwagon advice dropping the wastegate down a tad worked well. My target was 18psi so I went back to open loop and set the WG up to give me 16psi. Then turned on closed loop. Pretty happy with the results. Left overboost at stock -2.2psi. Underboost was significantly altered to get the PID to work earlier. Also managed to get underboost WG back to 1.00 for maximum spool up response. P & I have been modified a fair bit. All other hysteresis stuff etc I left stock.

Before and after tune for comparison if anyone finds them useful. Start tune is full open loop. Final tune is full closed loop. Wastegate is IWG75 with flapper port onto an GT3582 (F6) turbo. Final log also included so you can see whats what. Can also compare to stock so nothing to hide. Mk1 FG XR6T.

I wish they had they had derivative gain in the tune. I found with just having P & I there is compromise between staying on the target and rejecting disturbances. Until someone shows me something better so I know what I need to aim for I'm sticking with what I've got.

Can't wait to get the E85 back in. Got 17deg of timing to try out at 18psi. Already starting to droll at the thought of it. 

5th gear WOT starting at 1500rpm with intermittent throttle lifts to disturb the boost controller. A little overdamped but gets it on the 2nd swing

3rd gear WOT pull starting at 1700rpm. WG at 1.0 to spool then catches it. WG at 1.0 at the end of the pull so no more boost

3rd 4th 5th rip starting at 2500rpm. Stayed within the bounds until the last shift which had a bit of spike. More work to do on torque reduction strategy

97d_Final.teclog 90_In Car_Open Loop.tec 97_In Car_Closed Loop Boost 7d.tec

Edited May 18, 2022 by hjtrbo

[Roland@pcmtec]

"I wish they had they had derivative gain in the tune. I found with just having P & I there is compromise between staying on the target and rejecting disturbances. Until someone shows me something better so I know what I need to aim for I'm sticking with what I've got."

There is also camshaft intake and exhaust adders based on closed loop error. These are usually zero'd by most tuners but if you put them back you may find it assists with the control.

[hjtrbo]

Righto, more to learn!  

[hjtrbo]

Spending more time on it. A bit ashamed, those charts above are disgusting. Playing for a few more weeks. No excuse for your control. Learning.... 

[hjtrbo]

Now I'm getting somewhere. The log is looking much tidier than my first attempt I showed earlier on in this thread. Quite a bit involved. Lots of parameters to play with. Also fitted a new 4" dump pipe and got rid of the stock FG one. Surprised what it did, I was able to take heaps out the OL WGDC to get it to make commanded boost. Pretty cool to see it first hand. 

Probably the biggest thing I'm finding so far is getting the open loop wastegate DC right. From 2000rpm upwards there are 2 situations that I have found beneficial to run. 3rd gear WOT and 5th gear WOT. First I tuned in the WGDC for 5th gear until it was making full boost, then I switched over to 3rd gear to fill the rest of the OL WGDC table out. Reason is, if you just get the 3rd gear OL WGDC table spot on, then do the same again in 5th gear, you'll over boost it pretty early on in the boost ramp. 

Then onto some tweaking of the gains, under boost table, & integral clip. Over boost table is still stock for safety. New tune attached for reference. Had to end the testing early as I stopped making boost until like 4500rpm. Thinking the wastegate is stuck open or the blow off valve is shagged. Back onto the hoist tomorrow...  

 

99_In Car_Closed Loop Boost 9a_New Dump.tec

[Roland@pcmtec]

You could cheat and use boost by gear to get the same boost in each gear but with different OL tables.

[hjtrbo]

1 minute ago, Roland@pcmtec said:

You could cheat and use boost by gear to get the same boost in each gear but with different OL tables.

I wanted to ask you about that. I was running boost by gear but then reverted back as I has some strange behaviour in 6th gear. I just wanted to get a bit further with the closed loop waste gate tuning before I re-enabled boost by gear and tested again.

What was happening is that I was commanding 12psi in 6th gear (WG spring) and had the OL WGDC set to 0% for 6th. Then I went for a run and noticed in the log in 6th gear at WOT the boost target went up to like 17psi? Thought it would have stayed at 12psi like I had in the table. Am I on crack?

[Roland@pcmtec]

Did you also set the Desired Boost to 0psi for 6th gear? Also check it actually thought it was in 6th gear. I've seen issues where the detected gear drops out or picks a different gear. Usually in manuals, but I've seen comms loss also drop it out in a ZF

[hjtrbo]

I had it at 12psi (spring pressure)

 

[Roland@pcmtec]

Would need to see a datalog with detected gear in the log to comment further.

[hjtrbo]

Naturally. Lets leave it for now as I could have been on crack. Once I'm done with this part of the tuning I'll re-enable it and see what happens. If it turns to shit, I'll raise a ticket through support. 

[hjtrbo]

Who would of thought. Wastegate actuator was binding up. The plastic guide in the picture has chunks missing out of it. Some of the plastic fragments were floating around in the lower housing of the actuator. Sometimes a piece of fragment would wedge itself in between the guide and shaft. Proved it on the bench by shaking it around and moving it with my mityvac. Turbosmart were really cool about it and are sending out a new one. 

 

[hjtrbo]

Want to bring this thread to a close.

Probably the easiest way to sum it up is if you have a F6 turbo or GTX3582 on a XR6T and you're wanting to run closed loop boost control the best thing I ever did was yesterday I copied over the F6 settings for PI gains and clips, cam adders, overboost, and underboost. If you're still running the GT3576 then you'd leave all that alone. Don't disable the cam adders which some tuners will do with an open loop tune. They gave me the butt dyno feeling of more low down torque and the data showed I got into boost quicker. 

Set up base tune as per turbo (stock or F6 or use F6 for a pulsar copy as a base)

Force open loop as per the guide on this forum and enable boost by gear

Spend stupid amounts of time dialling in wastegate. It doesn't have to hit target exactly but you want the boost error to be consistent. 

Do hits from low rpm in all gears (TCC locked), then set up your desired boost table ramp ins to follow the actual open loop ramps. Adjust the x axis if required to get more resolution on the ramp. Then fill out rest of desired boost with what you want. If you get this right generally you'll avoid integrator wind up which can cause a minor overboost scenario (more prevalent in the lower gears due to ramp speed) or if you've gone the other way it will lag behind the target for a bit till it catches up. 

Enable closed loop and tweak wastegate and desired boost table to get the smoothest operation. Log P & I, desired boost, MAP, WGDC, underboost and overboost flags.

This screen shot shows desired boost and underboost settings not right. The first part is the desired boost is set to ramp in to fast and the turbo can't physically do it. This is kinda ok so long as our underboost tables are set right. Its a balancing act. In the case below both desired boost and underboost are not optimal. As a result the integrator starts to wind up to early. When the boost target is finally reached we have this 'huge' integral adder term acting on top of the open loop WGDC settings that cause overshoot. It takes time to bleed the integrator adder away so the P cracks the shits and drops heaps out of the WGDC and then we get a little unstable. In the worst case, overboost kicks in and opens the wastegate fully and then we are in full catch up mode, graphs look like shit and we wonder why we don't just leave things in open loop and walk away.

Here is the opposite. Desired boost set to ramp in to slow. You can see the wastegate spike to 0 during overboost which further slows down the boost ramp. From there the integrator winds down massively and it takes ages to undo it and get back to close to normal.

And the money shot which includes a desired boost step change to check wastegate DC settings. Tight underboost settings cause the wastegate to kick closed a few times as we ramp up. Not a bad thing as we all love the boost ramping in hard. P and I having to do very little work is exactly what you want to see. This result is open loop WGDC, desired boost and underboost tables all working hand in hand.

If you do it right it boost should ramp in just as hard as it can on open loop.

The interesting part will be to see what happens when the weather changes. Current air charge correction was 1.06 so would like to see how the boost control goes when it is hits like 0.98 or something. Have to wait till summer...

Further edit: Also the wastegate overhaul worked wonders. I think for longevity reasons Turbosmart need to consider changing the plastic guide out to something like an aluminium 6061 piece. And also they need to ditch the swivel boost reference fitting to a basic straight or angled barb fitting. I've had issues with both of those 'wear' items that are solely due to age. There is no reason for this to be the case if they made minor specification changes to the design. I mean who wants to be buying a new wastegate every 25 - 50,000km in a stock + bolt on application?

 

109_In Car_Added a few F6 changes_This is your new good fall back.tec 109_4th Boost Response Test.teclog

Edited June 8, 2022 by hjtrbo
Extra info. Hope someone finds this useful.

[abs351]

Hjtrbo, well done mate, interesting reading and findings. I will be going down this path for closed loop, you have saved me heaps of time.

Appreciate your time.

[hjtrbo]

Another thing that is important. The overboost scalar, timer and DTC's. Don't be a dick and max it out like I've seen on some 'professional' tunes. Leave it on as it'll protect your motor in the case of a stuck wastegate. You know you've tripped the protection when you see the below response. I'm not logging pedal position but be assured that it is flat to the floor. You can see the ETC dropped down heaps then gradually continuing to close in as the rpm rises. Additionally the desired boost is dropping off and the wastegate is tending to command full open. It will keep doing this behaviour until you restart the car. 

Don't get frustrated by it, consider it a fuck up tax and a sign that your tune is not there yet.  

For interests sake the ETC will follow this table in the event overboost protection is tripped.