Boost Control - Closed Loop Setup Mode

[Dennis930]

Nigel,

I had the exact same problem you have. A pull from 2000 to 6500 RPM was fine. But, when I shifted into third (RPM's drop to 4200) and go WOT I would get an overboost spike that hit the boost cutout (set to 30 KPA above max boost 217 KPA). It did it in both setup and basic modes. I could not change the DC in the 4200 - 4500 RPM range because it would drop the boost in that area when doing a 2000 to 6500 RPM pull. I went to the advance mode and adjusted the PID parameters, but I still could not get it react fast enough. I concluded the setup could not catch the sudden increase in boost. My max open DC is set to 65. My Porsche wastegate has a rather large diaphragm chamber (around 4" dia.), so I thought maybe it could not vent off pressure from the top fast enough. I ended up installing a high frequency MAC valve that runs at 39Hz and I left the muffler screen off the vent port. Since then, my system has worked fine in the basic mode (slider set at about 200). I previously had a 4 - port MAC valve that I plumbed as a 3 - port, so that could have been the problem to begin with. I should drop my overboost limit from 30 t0 20 KPA over max boost now.

[nathaninwa]

I understand the scenario. My LS drag car requires nest 75% for second gear 220kpa, but only needs 65% in high gear for the same 220kpa. I was actually asked to quiet my car down at the track after overboosting and tripping the decibel meter!

My solution was to switch to boost by gear, target 220 in second and 221 in high gear with the bias table having its own 220/221 rows. Worked flawless so far

[NigeT]

Thanks Dennis and Nathan,

Since I've generally only been considering 3rd gear boost comparisons I don't think this problem is gear related. Although that could be a problem to come of course.

I've been looking at dMAP/dt. On the long pulls it is well below 200, but on the mid-RPM (well into the meat of available boost) switch to WOT (from say 15-20% throttle) I'm seeing close to and even quite well above 200. It looks like when boost is climbing rapidly either the boost control, or the mechanical elements of the system, are not able to respond quickly enough to close down the increase in boost.

I'm using Garrett GT2560R with just the standard wastegate - is that likely to be restricting the ability to control the boost? The control valve is a Pierburg. This was fitted by the tuner that did the tuning of the car - he replaced the valve that was supplied by Rev with my Basic MS3.

[pit_celica]

My take about it is that a turbo is a system that has an inertia.

At low RPM, the system's inertia isn't rolling. The bboost control loop needs to fullfill this inertia before the boost can come-up to the target. You tune the PID closed-loop to take this inertia into account and this gives you good resuls ina 2500-6500 RPM run.

Then, you do an upshift at WOT and the starting RPM is 4500. The inertia is still in the system (as the turbine/compressor is still turning at high RPM) and you shouldn't need to compensate for it. But, you tuned the PID closed-loop parameters to compensate for it and this is what it does. The result is that the loop is reacting too slowly to the high delta-MAP and this results in overboost.

I think that the boost control should have a kind of "Inertia" parameter that gets full when in full boost, but decay slowly as RPM decrease (not instantly). The loop would use this inertia parameter as an input and would modify his responsiveness based on it.

Maybe a good point to discuss for further boost control code enhancement.

As nathaninwa, my solution was different bias table for different gears. In my application, for a given tuned boost control loop, everything was perfect in 3rd gead, 2000-7000 RPM, but the target was slow to be reached in 1st and 2nd and I could have overboost in 4th and 5th gear. No PID tuning could solve this.

Sam

[NigeT]

I think it is sort of like inertia, but more to do with the volume of gas spinning the turbine. At 2000 RPM a 1.8 MX-5 has insufficient flow to get a 2560R to do anything much. By 4500 RPM it is fully on song. So you go from low throttle cruise at 4500 RPM to WOT, shifting up a few boost targets and the boost control switches on, goes max duty to get things moving quickly towards the target but then, as you say Sam, it can't reign it back in quickly enough to not overshoot target by over 20kPa in my case.

We shouldn't be tuning the duty bias table to cater for spool up, at least not for speed. I think that the closed loop boost control algorithm caters for that by going max duty till it gets within the lower limit delta. It should just be the duty necessary to maintain the target at that throttle / RPM coordinate. Of course we have to make some compromises here as we can't really test and measure this very precisely because under the test conditions the coordinates change rapidly.

I still don't understand what are the set of triggers required to activate boost control. On my standard boost target table most of the lower throttle openings have a target of 140kPa, and the higher throttle openings all have 160kPa. If it was actually starting at 140kPa it would be within the lower limit delta and shouldn't go max duty at all.

I do realize that I may yet hit problems when changing gear but I don't have access to a track to test WOT in higher gears at RPMS that will achieve target boost, so I cannot test those scenarios. I think I need to get control of boost for throttle transients when the turbo has enough energy to deliver max boost first and worry about gear differences later.

Does the differential term off the PID algorithm act to slow the approach to target? If it does, how high can it reasonably be set? If the algorithm can't control the effect of sudden changes in throttle opening then what does it give over open loop? Of course my mechanical system may be too slow to react, but I need to rule out tune issues before changing hardware.

Should I be setting targets for lower RPMs (below where full boost is achievable), to levels that are mechanically achievable, our is it ok to continue what my tuner did and just have the same target for all RPMs for a given throttle opening?

Thanks.


Sent from my iPad using Tapatalk

[NigeT]

I now have tried differential gain values of 100, 150 and 200 (P and I @ 100) with really no discernable change in reaction to high differentials. I thought there would be a detectable reduction in overshoot, or at least an increased reduction in boost duty to try to to reduce the overshoot but I'm just not seeing it in the logs.
As others have said, it does not seem to be possible to tune the PID algorithm to react quickly enough to high differential in MAP. I guess I have to increase the lower limit delta to give more cycles for the algorithm to react. This presumably will slow spool up which would be a shame.


Sent from my iPad using Tapatalk

[muythaibxr]

[NigeT]

Thanks but which one? Current has D=200, I=75 and boost tolerance increased to 25kPa, but the reduction in I gain does not seem to have reduced overshoot but has increased oscillations around target. I suspect I=100, D=(anything) and tolerance = 25kPa would work better, but 25kPa is more overshoot than I'd like.
I'll attach the msq tonight when I have access to it.

Thanks.

[NigeT]

Here's the last MSQ and log. The log may seem confusing - IIRC I started with PID 100200, stopped in a lay-by and changed to 75200 and then changed the tolerance up to 25kPa.
I'm going to revert the PID settings to 100100 (essentially Basic Mode I think) and see if the increased (now 25kPa) tolerance stops over-boost on 4500rpm+ cruise to WOT changes.

Thanks.

Can't attach the log - it is too big even when zipped.

[muythaibxr]

If you could do 1 run and 1 log with 1 set of settings, then make a change to settings and do it again that would help.

So several small logs with only the behavior for those settings.

[NigeT]

I don't disagree, I just had (and have) very limited time and very limited battery in the old laptop.
Can you suggest a sequence of runs to try? I'm really quite happy with the 2k to 7k pulls, it's just going from low throttle to WOT at RPMs that can give the required boost level almost immediately that lead to > 23 kPa overshoot. I'm pretty convinced that PID settings of 100150 and 100200 had no significant effect compared to Basic Mode.
Thanks.


Sent from my iPad using Tapatalk

[NigeT]

I managed to get some data logged today - it was dry enough to use the car for my commute. I reverted to P100:I100:D100, but left the lower limit delta and tolerance as they were. MSQ attached. I have cut down the log to a section where the car was well warmed up and it just covers (I believe - this was 15+ minutes into the journey) a long boost pull from low revs followed by a number of mid range cruise to WOT transitions starting from different RPMs showing which cause overboost and which do not.

On the return journey I changed just the PID D term - to 200. I'll try to post another log subset over the weekend for this setup as the overboosting actually seemed more prolific.

[NigeT]

Here's the return journey MSQ and a couple of log subsets.

[muythaibxr]

Are you using the slider in advanced mode? If not, if it isn't responding fast enough, turn up the slider.

[NigeT]

I believe the slider is set to something like 311, which seemed optimal in Basic Mode particularly for the long boost pulls.
Thanks.


Sent from my iPad using Tapatalk

[NigeT]

Just double-checked and the sensitivity slider is set to 311.4. I found higher settings unsettled the control for long boost pulls, say 2500 - 7000 RPM.

I'm pretty convinced that the problem I have is related too high MAP differentials when boost gets "switched on" when the mechanical system is capable of making three target boost almost immediately. On the long pulls we start below where any real boost can be generated and move up the rev range with a mechanical limit on the differentials.

I haven't found anything to help me understand why boost duty isn't always above 0 given my lowest target is 140 and MAP is frequently below the target.


Sent from my iPad using Tapatalk

[CRSTune]

Increase "Boost Control Lower Limit Delta" (you normally start at 100) and decrease your initial duty values. This will give the system more time to react before hitting the target and start it at values lower than the known good duty cycle. You want the system to bring itself up to the target instead of overshooting.

Then go back to basics. You're too hung up on the D value compared to P and I. Set your D and I values to 0 and P to 100. Adjust P until you're happy with the oscillation (it will a little) in the system that creates low overshoot (this value reacts to target difference). Then tune I to bring down the size of the peaks and reduce oscillation (this is the value that gets you to the target). At this point, you can adjust P a little more to increase response. Lastly, tune D.

Section 7.6.1.2 in the manual goes over this subject.

Check out these videos done by Motec that go into fantastic detail regarding CL boost control as well as general PID theory:

PID Theory: https://www.youtube.com/watch?v=16Clfh5eBzg
CL Boost Control: https://www.youtube.com/watch?v=A_EP5AdvuPk

Keep in mind that Motec uses a few settings not present in Tunerstudio but knowing what they do may help give you an understanding on how to avoid the problems they're meant to prevent.

Lastly, check out the Wikipedia article on PID controllers:
https://en.wikipedia.org/wiki/PID_controller

If you take a look at the various graphs they have on that page, it will give you a very clear view of what to expect from each of the parameters. What you're trying to get is an over-damped system (one that does not oscillate nor overshoot).

[NigeT]

Thanks CRSTune, loads of information to review. I thought I was using the recommended approach starting with Basic Mode and tuning the full rev range pulls. Although not quite critically damped, I thought these are working pretty well.
Should I be tuning the mid-high RPM cruise to WOT transient primarily instead?
Thanks again.


Sent from my iPad using Tapatalk

[CRSTune]

Since that is the area with fastest spool, likely yes.

Tune it so it responds exactly as you want in those RPM sections especially considering those sections have the fastest transient response with the least amount of time for the system to react. Once it works well, you can go back and set boost targets in the low RPM range that ramp up to the desired target. This way you can avoid issues where the ECU is trying to hit a target it will never reach.

[NigeT]

Thanks, that makes a lot of sense. At the low RPM end I imagine the boost available is well below the standard Garrett actuator pressure. Lots of reading... to do this week.


Sent from my iPad using Tapatalk