I was amazed when I originally asked multiple people local to me who had quick turbo cars, including tuners, how they set up their PID settings. The most common answer was, "I don't know, I just throw the same numbers XXXX has in their car and it works." This thread was an eye opener and helped a lot from where I started. I will use the same methods now with matching gate springs!

 

tuning PID's is easy. just need to swing a black cat by the tail at midnight on a full moon and then read the assortment of voodoo chicken bones you've scattered out on the pentagram you've carved in the dirt

 

I deal with PID control in my work and often have to explain its function to folks much more educated than I am.
My way of learning is very visual, mind you, so this may not make sense to some. I also tend to oversimplify complex things, so please don't take this as me being patronizing.

Think of a range, lets say 1 to 5. That's the control range, or overall "gain".
Now make a set point, say 3, that you would like to maintain.
The P or proportional part of the control model considers that my desired value is in the middle of the gain range, with equal values on each side of the target (set point).
Let's say that upon approaching the setpoint (3), the control overshoots to 4. This is largely controlled by the I or integral portion. I think of this as how large of a change based on time and distance my control model makes (from my set point).
Lastly, the D or Derivative- this determines the rate of change or slope in reaching my setpoint. How fast in time it gets there (which can also cause "overshoot" or hysteresis).

I know there are many that can fully explain it mathematically much better, but that's my .02.

 

I followed your I and D, could you clarify your P? I was under the impression that the P is the first initial rapid change based on the delta error of measured vs. desired.

 

Probably not, but I'll try-

P = how far the result is from your setpoint any instant. How far is key- that's why when my simple mind is setting up PID controls I think in terms of "how large is my range or "window" that I'm trying to control?".

Waaaay back when my instructor taught this using calculus, but my brain has seen better days, lol.

 

so if the target is 3 and the measured is 4, it's the Ps job to make an instant adjustment of 1 to bring it back to target?

 

Not to sound like a wise acre, but...Yes, as long as your I and D (both can cause overshoot) are correct. It's a very much a relational thing- I tend to look at I as "how many steps and how fast" and D as "the response / acceleration rate to my target". If I'm controlling something unfamiliar, I usually stay on very conservative end of integral.

I hope this helps and doesn't actually confuse anyone.

 

Are there really no advanced “TEST” options for the Holley ECU for Boost control?

I’ve not worked with the Holley much at all, but was attempting to help a friend last week at the track. Using dome control and playing with PID as mentioned here.

All of my setups have been Mega Squirt based. In megasquirt software I can go into “test” and set each boost solenoid to operate on/off or at any freq/ I desire to test everything out. Is there nothing like this in the Holley software? Or did I simply miss it? To “test” I’d need to have IGN on and supply variable shop air to the map sensor?

Our issue was the setup would make 14lbs from 0-25 on the dome. Then anything over that it would jump to 30lbs. (S498 89/105 5.3 6.0 setup). Typical twin 4 port MAC setup w onboard air.

 

i havent really messed with PIDs at work, but i have at home on several quite varied applications, and ive found its basically backward terms on 'slow' vs 'fast' reacting systems. another thing ive played with that i never really see talked about much is the frequency. on one of my systems (dbw throttle motor control) i turned the frequency up and it completely changed how the entire system reacted. ive also done nested pid loops and oh wow does that get tricky.
pid stuff is fascinating but my last attempt for work i spent a couple days beating my head on the wall and ended up replacing the whole code block with an array of if elses and it works 10x better and more reliable ¯\_(ツ)_/¯
one thing i found interesting lately while poking around at a gm e39 is it has a 3d table for each of the pid terms in its throttle control. i have done real basic 1x3 kinda tables in the past but never anything that complex. i may try implementing a strategy like that some day. i should see if my haltech can do that....
edit: looks like the haltech can do it, pretty neat...

 

as far as i know, mostly yes. if you have a pro dash you can use virtual switches to activate and deactivate, or use tps to activate it, then sit with the car on engine off pushing on the gas pedal to cycle it.

this sounds like it is less bc pid and more how the engine is reacting and not possible to test off the track.

 

The static pressure tests are laughable and don't count for much
​​​​​​
looks all perty and nice then you make a pull and it's ****..... Welcome to the WG feedback making large swings in dome volume
You want to know the easiest thing you can do to perty much make the tuning idiot proof? Add volume
I can run p from nothing to 99 and still hold stability

 

Where is it best to add volume? between the dome and the MAC valve? How much volume we talk'n? -thks!

 

Pop can size is good, say like a old AC dryer works well.

dome sensor installed in tank

.Solenoids feed the tank.

Line from tank to dome.

Say with the gate shut you have 100 units volume
With the gate wide open 50 units.
You can see as the wastegate moves this greatly changes the volume taxes the loop.

Typically seen when you let Boost slam in you see dome overshoot Target even if you have the PID very aggressive and the vent noid wide open can't control it.

Now let's say you add a 300 unit tank
With wastegate closed you now have 400 units... With the wastegate wide open you have 350 units...
.

You can see how this greatly reduces the impact of the gate feedback.
Solenoid feedback and everything



​​​​​​

 

Don't let anyone tell you it will be slow either
here is a static test on my car, all this is located inside my car with a lot of volume and a single line that's about 8 ft long that goes up to my dome
This test really doesn't show you important stuff like I said, it just shows you speed still

 

And here is good one of actual function.
.
Note the dome stability.....there is zero filtering on that.
.
It's a good one to show you the potential one of the system, where I had to pedal and how it all responds
.
.

​​​​​​

This is out on the street where i really need to manipulate power accurately.
.
Also note that I am running in map mode... My target boost is actual Target boost (which is silly everybody doesn't in my opinion, even if you zero out and make the second controller inactive)

.
.

​​​​​

 

I'll throw a small post in here. I don't have really any experience to speak of tuning P.I.D., but having went through Holleys Tier 3 training semi recently, they do teach that and provide a hands on working setup you have to use in front of the class while they manipulate it and make you figure it out. With that said, this is the basis on how they teach P.I.D.:

They compare each term, to a suspension setup/part so you can kind of visualize how each works. I can not copy, upload, or reproduce their manual, but my notes are my own.

If you think of the P term as the spring rate
And the I term as the load leveler/overload spring
Finally the D term as the shock absorber

If your P term is too low, you probably won't reach full boost The I term is going to add or vent to get you on target, a small value will be much slower to react and a large term will operate more quickly.
The D term reacts to how quickly the feedback is changing, and tries to put the brakes on to slow things down a bit. You can move a shock in and out slowly by hand but if you hit it hard, it will take much more effort.

The TRL is the Target Rate Limiter, if it is too low, the P term will be lazy and slow to react.

Basically, if you have your safeties setup to protect yourself, you should be able to really make big swings on these, to get you where you need to be. Once you have a general idea, and again, as long as your have yourself covered, you should be able to swing these 50% and see if it's better or worse, and then change the value again by 50% and within a few swings, get yourself headed in the right direction.

Here is a hand drawn picture, showing the desired target, and how each term kind of reacts. I hope this helps someone.

 

thats good info, thanks.
another thing of note, is it seems like behind the curtains these pid algorithms are all implemented a little differently. holley has their nested loop algorithm for the dominator which is what gave me the idea to implement a similar strategy on a personal project. its a clever idea but ive never tuned on a holley before.