Tuning Boost procedure

[AVP]

Can someone please verify if this is a correct procedure to tune boost?

1) Set PID : ZERO

2) Set target boost: ZERO

3) Configure RefDG boost: use gradual rising numbers to see OPEN LOOP boost reaction. Configure to desired boost levels on all gears. (boost may be slightly different on one gear than other on a given DC.

4) Enable target boost: by dialling in the desired boost figures per rpm

5) Tune PID: start with P, keeping I and D ZERO
   -P: start with a high number (ie 150-200). Highest the number, the slower the spool and the easiest way to avoid overshoot. Do runs on a rev range that the turbo is within the spool range and go WOT. Watch the reaction of boost. If there is a spike, then increase. If it doesnt reach the desired boost, then decrease. Leave it once overshoot is at it's minimum

   -I: Once P is configured, start giving small rising numbers to I. This will enable the boost vs rpm curve not to stray away from the desired target.
   
   -D: Start with small numbers. D will help the P in overshooting, BUT too high numbers on D will cause oscillations.


I have a question to whoever has used mac valves, or in general.

What kind of Hz do you use for the boost valve control? And what does the hz mean? The higher they are the quickest the response?

thanks
vasilis

[GintsK]

Mainly it is just one way. Number 5 is wrong regarding P: more P means more overshoot. And D is like damper - it counteracts to fast changes.

My way, especially for external w-gates is different: start with PID like 150, 50, 70 but with exact target - may be lower than final, and low or even zeroed ref table. Then review resulting DC. Then fill ref table with little lower values but same shape as PID keeps. Lower values prevent overshoot. If overshoot is acceptable, then ref DC table can be same or even higher than actually maintained by PID.

Gints

[AVP]

in megasquirt, higher P meant lower overshoot as P is reacting more to an error.

ill have to do some trial and error on this.

[AVP]

as i fixed my wastegate, i ended up with 120-8-12 as my previous were 130-5-14.

i did something in between the 2 above methods.

i tuned the open loop on the DC that is closer to my target. Then dialed the target where i wanted, and finally used my previous PID setting to see if i get an overshoot. Actually if rpm vs time is correct on my logs, then it was a bit late, so i reduced the P and D a bit and i think it is a bit better.

I dont know what vems uses to read the rpm, but in action i think it is 'seeing' different things than what it is logging.

on my last run today, the high boost was coming in my rpm dash at 3500 hitting hard, but ok i wont trust that, but the digital vems was also before 4000rpm. My laptop decided to die off battery at the time, so no log, but i bet on the log, it will be coming later than 4000rpm.

I wont bother with it for a while. Car drives nice! Im happy!

[gunni]

VEMS is where any RPM is coming from.

It uses your crank sensors to figure out where your crank is, rpm is also a result of that.

All other RPM functions in your car will be based on what VEMS is saying.

[AVP]

so? i still dont understand why the flat spot on the rpm is there.

if the crank sensor is having a problem OR due to the 7A flywheel, the crank sensor is having trouble reading right the rpm, then it may be that i am getting a poor rpm reading? hence the flatspots??

its known with the 7A flywheels on the audis, that vems has trouble seeing the actual pin, and it has to be enlarged so the VR sensor can read it.

can the problems occur from that?

if you look at my project thread there is a link with 2 logs which one has a very peculiar rpm logging

[andreNL]

I tuned my boost pid like this,
Set boost safety pressure to 102 kpa
Set all PID settings 0
Turn up P till valve opens
Set P one lower
Turn D up til valve opens and closes frequency like set one lower.
Dial in I till valve ticking stops and or valve opens.
Check again worked fine for me almost flat line with little overshoot at Opening.

[AVP]

what do you mean valve opens? how do you check that?

[andreNL]

Just listen to the clicking of the boost soilenoid in the garage with the engine off,
Or fit a test light to the boost soilenoid wire

[AVP]

So you put the boost safety to 102 so that it marches the map sensor reading with the engine being off?

I'll try that to see if it is any better actually. Thanks

[andreNL]

Yes you see the kpa reading hopping between 101 and 102 use that to trigger the valve going open, the I is zero so it imidetly closes then you set the d to open it again, now it opens and closes rapidly generating a clicking sound. Now set the i, when the valve stops clicking it starts hovering. The ecu turns it on and off so fast the valve cant close or open fully.
Goodluck

[GintsK]

PID is for tuning whole system not just solenoid valve. System is greatly affected from w-gate chamber volume and solenoid valve flow, membrane/w-gate valve ratio, spring, exhaust gas back-pressure... Could be method is usable for initial settings. But especially for built-in W-gates tuning procedure require more.

[andreNL]

Ziegler–Nichols method to calculate the PID settings.
Pu is oscillation period maybe you can use 21hz the value if the soilenoid is running on. so 255/21= 12.

PID   0.60Ku   2Kp / Pu   KpPu / 8
P => 0.6xoscillation  value
I => 2xP / Pu
D => (Px12)/8


I hope it helps, merry Christmus 


Update:
Pretty complicated, mostly trial and error to find right values
here a link interesting reading http://www.advice1.com/reference/pidalgorithmref.html

I first used 18-60-120 with integral at 30. Changing integral limit to 255 made the controller more responsive. Fiddling with numbers and formulas gave 119-40-102 wich pretty nice results.