EFI University

giskard · Joined: 24 Oct 2005 Posts: 251

Being an electronics nerd I decided to play with methods of improving boost control. I built an analog circuit that implements a fast feedback loop for the wastegate can pressure. It has a pressure sensor to sense the can pressure, tee'd off in the line from the boost solenoid to the can. The main feedback loop in the ECU then tells my circuit what can pressure it wants.

The composite feedback loop is called a "nested" feedback loop which is a good solution for some control problems. In the case of boost control, the filling and emptying of the wastegate can, (and thus moving the actuator / flapper) is relatively slow, slowing down the whole system. The fast inner loop speeds that up.

The end result is very fast, stable boost response.

Car is a 1.8L miata with a GT2560 with target max boost of 15 psi, using a can with a 5 psi cracking pressure. Internal wastegate, and for now, the (fast) Greddy solenoid. I plan to re-test with a cheapie GM solenoid.

http://www.miataturbo.net/showthread.php?t=64996

Let me know if you guys can't see the graphs.

Aurélien · Joined: 30 Jun 2009 Posts: 137

Hello !

Interesting job.

Do you have any comparaison/graph versus good PID ?

Roberto Arano · Joined: 22 Nov 2005 Posts: 361 Location: colorado

nice work, way to think more outside the box than most !

I forget what I was testing for , but at one point I remember making a "string-pot" to log actuator position. maybe that would also be useful in your experimenting, but I suppose "can pressure" does largely the same thing for your useage.

it's starting to come back to me, i think i was looking at what frequency (while maintaining steady boost) and how fast the actuator would move under conventional control methods. I might have been thinking about making an electric actuator at the time.
_________________
www.circuitse7en.net
http://stores.ebay.com/Circuit-Se7en

HOODEY · Joined: 20 Jan 2006 Posts: 169

A picture is worth a thousand words..I got lost in all of those words.

A PID boost controller could be implemented quite easily using microcontroler.

I just have not got around to it as yet.

giskard · Joined: 24 Oct 2005 Posts: 251

Aurelien,

I've always had (standard) PID control using my AEM, (using a special circuit to add the 'D') with lookup tables. The PID output is of course, the duty cycle for the solenoid. The AEM has lots of feedforward lookup tables which helped a lot.
Feed-forward:
http://en.wikipedia.org/wiki/PID_controller#Feed-forward

However it was difficult to get it stable with accurate, fast settling over different operating conditions.

For example, if you hold part throttle at 5500 RPM, then slam the throttle open, there is a strong tendency to overshoot. If you go WOT at 3000 RPM, it has a tendency to oscillate when it gets to 4000 RPM. The system wants to have different values for P and D under different conditions, which is (a) not doable in the AEM and (b) time consuming to tune.

To get it stable in all conditions the loop gain has to be low, and because the can is slow to fill, the D term has to be high. This means the boost rise (spoolup) slows down a lot before reaching its final target.

What I did was to add a 2nd, fast local feedback loop for the can pressure. It accepts a command input for can pressure. It controls the solenoid in order to hit the target can pressure (which of course it monitors with a pressure sensor)

The main PID loop output is no longer the solenoid duty cycle. It requests a can pressure from the fast local loop.

The main effect I saw was that it was now easier to tune the main PID loop for fast, stable response.

Hoodey: PID in code is easy. The hard part is adaptive control a la the OEMs, which means it self tunes its PID parameters and lookup tables as the system ages or is changed. i.e. you put a bigger exhaust, it initially overshoots, then it learns and re-tunes to get rid of it:

http://en.wikipedia.org/wiki/Adaptive_control

MrDomino · Joined: 01 Oct 2009 Posts: 188 Location: Indianapolis, IN

What is "can" pressure?
_________________
I am a Mechanical Engineer.

Roberto Arano · Joined: 22 Nov 2005 Posts: 361 Location: colorado

Giskard is refering to the wastegate actuator diaphragm part as a "can" 'cos it looks like one . Wink

_________________
www.circuitse7en.net
http://stores.ebay.com/Circuit-Se7en

sohc_mshue · Joined: 04 Mar 2006 Posts: 366

This is similar to the boost controls implemented by NLR and some other boost control manufacturers. The "can" pressure is the setpoint rather than boost pressure.
_________________
www.shuetuning.net
efi101/102 graduate

giskard · Joined: 24 Oct 2005 Posts: 251

I verified that the system works well with a cheapie GM boost solenoid. Wastegate can cracks open at 5 psi, boost target is 15 psi. Without the local feedback loop, I found it difficult to stabilize the main PID feedback loop using the GM sol'd. (Was easier with the big fast greddy solenoid). Even better, the PID settings that worked with the Greddy work with the GM. So the local loop (mostly) removes the dynamics (and slowness) of the solenoid/can system from the main PID loop.

Pantera EFI · Joined: 12 Feb 2005 Posts: 1268 Location: So. California

MY interest : Desert races 250-1000 miles long, some at night.

My platform : Our ECU-882 with control of boost (x/y using RPM/MAT) and "boost by gear" guess using Wheel Speed to select gear. (seven choices)
The internal boost PID control is very smooth.

My goal is to allow for long life from the Fords provided Echo-Boost engines and to win the race.
There will be a "power button".

What are some ideas for an "advanced" boost control (power/life) strategy?

Lance