Preferred Method of EBC Tuning MS3 1.4.1
 

Finally going to start tuning EBC. I read through this ADJ's Quest but am at a bit of a loss on which FW version all this pertains to:

1) Is there a difference in methods or results between 1.4.1 and 1.5.x?
2) Do I start with Open Loop and then transfer, or does the Closed Loop Set-up mode make that approach obsolete?
3) Is there a thread that I have missed?
4) I presume that Jason's approach to throttle vs boost target is still valid.

Now I found this, which, though it may say what the manual says, makes a bit more sense to me: ADJ's Answer

@aidandj, when you say getting earlier spool, do you mean:

A) At 3500 RPM, punch throttle and see how many mS it takes to get to, say 180kPa?
B) At 2000 RPM, punch throttle and see where the boost threshold (crosses above 100kPa) occurs?
C) Something else that I'm still not getting?

Thanks.

1.4.1+ are all the same.

Setup mode negates the need for open loop tuning

#2. Start at 1500 go full throttle, record when you hit your boost target (careful, you will easily hit overboost this way). This allows you to get a dynamic idea of what the lowest duty cycle that will affect your spool.

Think about it this way. The code will hold the duty cycle at your max value until you reach the lower limit Delta, then the control loop kicks in. The smaller duty cycle range the control loop has to work with the better off it will be. If I asked you to guess my number between 1-100 and 30-80, which would you be more likely to get right?

Well, I already know that mine will hit 180kPa very quickly at 40, so the 8-30 range seems likely.

Thank you for the clarification.

80*. Missed a 0

This is where I am so far on 1.4.1 FW:
1) Set up Min Max, as you prescribed, Min = 10, Max = 40 (I recently reset down to 38, but that is hair-splitting).
2) Generally have the bias table and target table set up.
3) Originally set max boost to 180kPa in the targets with Boost limit at 185. I could not leave it there because I hit the overboost so much even in closed loop.
4) To try to be able to set "Basic" CL control, as seen in the image, I set top rows to 160. This allows me to not have so much power that the controller reaction can be seen. But what I'm getting, is that with slider at 350, seemingly no controller reaction. The Boost Duty just sits at the bias settings.
5) Unless coached to do otherwise, I will try slider at 500 (max), and targets to 170 and see what happens.
6) In ADJ's quest, there was a time that he reported that slider to "0" was too much, so the FW guys toned it down.
Questions:
A) Did they go too far reducing Slider Sensitivity, or is there a bug or bad setting?
B) Should I just go ahead and go to Advanced?
C) If I go to Advanced, is the slider still going to have no effect?


https://cimg7.ibsrv.net/gimg/www.mia...106b7dc175.png

I believe the slider still has an affect. It's essentially an overall multiplier for the reactions.

May I ask why your boost target is higher at 80% throttle than 100%?

Probably for testing.

Yes, for testing. When goal was correctly set higher, at 100%, two things: First, I kept hitting over-boost. I have not dyno'd at such a high boost, so I don't know where I am yet; Second, time at high boost is very short, so no time to determine when I get to oscillation (which, as I stated, I have not gotten to yet).

Here is a snapshot of the non-control in "Basic" mode with slider at 350. Same as the MSI posted above:


https://cimg4.ibsrv.net/gimg/www.mia...a3d25f58e7.png

This was very helpful, thanks for sharing guys.

DNMakinson, from what I've learned your issues with holding target boost and overboost are probably related to your Boost Control Lower Limit Delta setting of 15kPa. From the docs:

• Boost Control Lower Limit Delta - Boost pressure must be within this many kPa of the target boost before closed loop control will activate. Outside of this range the valve is held wide open (keeping the wastegate shut) for fastest spooling.
  

Agreed.

I am now at 30, and it is much more tamed.

Another thing I found. Do not use Boost Tolerance. Lifting throttle brings on overboots cut. There may be a use for it, but not that I can find.

EDIT: Also, I found that, in the end, I needed 65% for my maximum setting to truly keep the wastegate closed. And 10% minimum, as sometimes 13% is needed for actual control.

Here is today's picture of boost following target with throttle:

https://cimg6.ibsrv.net/gimg/www.mia...dd6767ad09.png

I've been working on this also, getting good results...

Just curious what is the idea behind different boost targets at different TPS values? Being as i'm still tuning, I just have the whole table set to one value (150 right now).

So you have have linear throttle based boost. Half throttle, half boost, 3/4 throttle 3/4 boost.

Right. Ends up being roughly throttle position = torque.

One other reasons n: it reduces the amount of boost that is built, and then throttled. I other words, if at 1/2 throttle, you ask for your full boost, and get it in the manifold while dropping some pressure across the throttle, that is inefficient and hard on things. But if you don't ask for full body until you fully open the throttle, that is more better.

This is of none effect after 70%, because a throttle valve has little effect on flow between 70 & 100%.

Good info, thanks! Guess ill start working on that.

I found this over in the Megasquirt Support Forum...The last few posts on page one are especially interesting. PID values can go up to 200. I was just assuming the max was 100, but really that is the mid-point.
Basically they are saying "Using a smaller delta and higher PID numbers will help it reach and stay on target."

I was having the same issue as this guy, where it appeared to work well in the lower gears and not as well in the higher gears...

Megasquirt Support Forum (MSEXTRA) ? Closed loop Boost Controller not working (View topic)

Thank you for sharing this, I thought I was going crazy with PID #'s in the 150/100/100 range and still seeing lazy duty reactions. I'll try ramping it up and see if I can finally get rid of the boost drop off I've been fighting.

I've been fighting with this as well...
My situation is aggravated by a 2000ft elevation change and usually a 15degF drop in temp during my drive home.
Everything is within ~5kpa of target at 70degF, but as I get closer to home it shoots into overboost.

I believe the manual still states the P is back asswards with 0 being the highest coefficient and apparently 200 the smallest?
Good to know that 200 is the lower (or upper?) limit. I'll start venturing into that realm.

Say what?!? 200 is LESS responsive than 0? Can anyone confirm this?

According to Ken its now a standard PID control. So more P is more coefficient.

Basic mode is
P: 100
I: 100
D: 100

Lol, I know its freeware so I can't complain too much... but I like the "don't RTFM, sift through my rambling forum posts"
Does Ken ever explicitly state if they're still arbitrarily inverting things or not?

From the manual:
2. Set Proportional gain to 100% and slowly lower - While tuning Proportional gain, higher numbers
mean slower boost climb and lower final boost. For safety, start with a very high gain (100% should
be sufficient). Find the RPM that typically spools quickly, and fully and quickly depress the
accelerator. Note how much boost is reached. If boost overshoots the target dramatically, increase
the Proportional gain. Otherwise, reduce the Proportional gain and try again. Do this until boost
reaches the target with a small amount of overshoot.

I believe that is incorrect in the latest firmware. But I could not tell you. It should be an easy test though

I had a multi paragraph rant typed up on how tuning is made to be artificially hard......but I'll just thank those that are willing to share and start working on a better sticky post with links and explanations on the gotchas that aren't obvious to the beginner.

What you need to realize is msextra and all the firmware is done by 2 guys...In their spare time...After their day jobs. They have families, lives, etc. And I find it amazing that we have the documentation we do. I sure as hell couldn't produce that. Megasquirt was not meant to be perfect and super easy​ to use. It's a diy solution that has been commercialized. There are other platforms that promise that.

Absolutely, I get that, but the folks that are making darn good money off of their efforts should be helping the community........all I see is them making money off of others hard work.

Here are my present settings with results:

Slightly more over-boost on stab at 50 MAT than 70 MAT, but pretty good. I have not played a lot with D as usually PID need very little D, unless this is somewhat scaled. I have tried up to D=2 with no noticeable change. D is to get large Control Output very quickly when Variable is far from SP.

https://cimg4.ibsrv.net/gimg/www.mia...fbb9109095.png

Note Control Lower Limit is set so high as to be non-existent, but if you look at log, the Control Output goes to max due to high P and I.

https://cimg2.ibsrv.net/gimg/www.mia...5ee57053e9.png

210 would use full SP duty, but I have not called for that much kPa yet.

https://cimg3.ibsrv.net/gimg/www.mia...3d5a63882c.png

Here are my actual targets. I drop at 7K to maybe help save my stock oil pump. Conjecture on my part as to why Mazda put 6500 rev limit on MSM.

https://cimg1.ibsrv.net/gimg/www.mia...d52361388c.png

See Control Output in green on 2nd chart maintaining the PV at SP. Look a few posts up for another log of Boost following throttle.

Note: There is not "standard" PID. Some controllers use P = Proportional Band, and bigger is smaller, some use Gain, and bigger is more. This is all Gains. Some use I is actually P*I. I think this is separate I in the equation.

Set bias table at typical values, like an Open Loop. In leboeur's case, he would want to set his bias table in the middle of his low land needs and high land needs. Then the PID corrects for the error between those bias outputs and the true needs of Boost-Target on a real time basis.

So DNMakinson, it looks like you are getting acceptable results with larger PID values, but your Delta is fairly high at 90. Do you recall how it behaved at lower Delta and the higher PID values? Did you try that combo?
Cats for sharing, btw.

It behaved similarly. I made it really high as to be ignored to help the overshoot. But, as I mentioned, DC still goes to max at punch, which is the purpose of that setting. So, seems to not matter with the rest of my set-up.

Not having used 1.4.1 yet, here's my :2cents:

There is a small icon to the left of each parameter describing it. I think for each firmware, if you are in doubt, you should refer to that description in your project with your firmware.

Just to be a punk, I am re-tuning my EBC currently with a stiffer spring to hold more boost up top. My PID's are currently p34 i9 d136, MS2PNP 3.3.1 IIRC. I am not quite there yet and it took probably 30 or 40 good logs to get where I am.

Dear Son... What is the title of this thread?

You do have me inquisitive about high "D", though. I may try some and see results.

Uh, is dis da title ;) I said to be a punk

FYI
1.5.1 beta 3 was updated to include SD card logging for the clbc PID values (closed loop idle too). My tuner was able to use this info to finally sort out the non-responsive closed loop boost control.

Just so I fully understand. Are you saying that it will actually plot the instantaneous values of the contribution to the output from the bias plus each of the three gain equations?

Could you post a log?

I still have not gotten the car back, but yes that is my understanding.

My tuner had been in contact with Matt from DIY, who helped make the logging of these values possible, so Props and Cats to Matt.

To be honest, my understanding of the PID formula inner workings is sketchy at best. From what he said, once my tuner had the values real time it only took four runs for him to get the CLBC to respond as you'd expect. I had personally seen many threads where folks had a hard time getting the CLBC to engage consistently, and those who had were having to deal with slow spool in some areas to get the overshoot under control in higher gears, or very rapid escalation of boost.

So, I apologize for the lack of detail, but I thought it would be good to let folks know this feature was available/coming. I will post more info once I have the car back.

OK, thanks. My settings are 145, 190, 10 @ 469 slider. I agree that these are not sensible to me, but are giving decent results over time and temp.

I used DNMakinson setting and it completely smoothed out my boost control (exception i have the slider setting down to 350). I am using the DIYautotune boost control valve set at 12lbs 185KPA with boost cut at 205KPA. Initially i was in open loop boost control which worked well, but kept hitting boost cut which i still cannot solve. I was initially hitting boost cut in closed loop with the slider at 470, but as i reduce the slider the tendency of hitting boost cut lessened. The boost cut still activates when i drop a gear on the freeway and stab the throttle WOT. I just not sure how to fix that problem yet. Thank you for this great information guys.

Finally got around to posting the PID logging function available in 1.5.3. I'm having to redo the clbc as I replaced my boost actuator. Just trying to handle it myself rather than bugging my tuner for it. My maps have solid fuel and timing higher than I am seeing on boost overshoot, and I have boost cut enabled, so I don't expect I'll break anything. It is still somewhat rough though. There are a couple more pulls past 118 sec. First time snipping one of these files.

You need to enable INTERNAL_LOG_FIELDS in your project properties, then you can add fields like these to your SDcard datalog. These are not available in regular datalogs.
pid_log_boost_p
pid_log_boost_i
pid_log_boost_d
pid_log_boost_count
Settings:
P=35
I=1
D=195
Slider=490
Lower Delta=30

Per the developers...
Note that these are in internal units which won't necessarily correspond to the externally visible units e.g. for idle the 0-100% scale is handled 0-255 internally. Viewing the "idle steps" field gives you 0-255. The "count" field increments each time the calculation is run, so log this as well.

What is in your bias table at 190 kPa and 5000-6000 RPM?

This is absolutely awful, by the way:


https://cimg9.ibsrv.net/gimg/www.mia...c938debf4c.png


20 kPa oscillation with huge swings in duty cycle



https://cimg1.ibsrv.net/gimg/www.mia...4345959cb4.png

Insufficient Integral to pull PV (boost) back to SP (Target)

Yep, it is awful. I haven't set up the bias table for 90 & 100% yet.
There is an Auburn football game today. Yesterday when I went out to work on these sections there was a State trooper literally in the turnaround I use on one end of the stretch of hwy I use. Didn't get anything done at all.
What I try to do is to have the bias set where the initial set once I'm in range be a value the is very close to that needed to maintain that target. I need to do some runs to establish those values.

I thought it would be good to finally show where you can watch the PIDs.

And we appreciate the data. My point is that, I don't think your tuner utilized that data to obtain a good end. I had questioned the PID settings when you mentioned them before. Seeing the log only confirmed my suspicions that they are way off.

EDIT: go https://www.miataturbo.net/megasquir...3/#post1437218 to see the results of my very different settings. Not saying that my aggressive setttings would be appropriate for a larger (slower responding) Turbo, but they make more sense to me in general. P & I driving things with a sprinkling of D.