Somebody PLEASE do a direct PWM turbo controller already!!!

ryanp

Vendor
Joined
Jun 22, 2008
Location
Barnsley, South Yorkshire, UK
TDI
Arosa CR - 550hp - 9.7 @ 150mph 1/4 Mile, Citigo 4x4 CR TDi - 340hp, Caddy 2.0 CR 4x4 TDI - 300+hp, Golf Mk2 Van 1.9 TDI - was 290hp, Mk5 Ibiza 2.0 FR TDi - 270hp, BMW 135d - 360hp, BMW 330d - 335hp, BMW 335d - 380hp + a few more ........
I do agree for a control perspective that the E-actuator makes a vacuum actuator look old tech but rarely would it be possible to install the turbo in the correct orientation so alterations to the bracket at least would still be required.

The actuators fail for fun in some OE situations, Ford being the most common i know of!!
 

majesty78

Veteran Member
Joined
Dec 7, 2007
Location
Austria/ Europe
TDI
Skoda Superb 3T5 CFFB
That's correct, the biggest advantage in using the modified vacuum actuator is the free clocking orientation of CHRA to suit an non OEM application....

In my case situation is a little different as I can keep stock orientation of turbo and actuator....

Regards, Alex
 

TDIMeister

Phd of TDIClub Enthusiast, Moderator at Large
Joined
May 1, 1999
Location
Canada
TDI
TDI
Ryan, what prevents an e-actuated turbo from being oriented properly compared to a vacuum-actuated one? Or did I misunderstand?
 

majesty78

Veteran Member
Joined
Dec 7, 2007
Location
Austria/ Europe
TDI
Skoda Superb 3T5 CFFB
Electronic Actuator has it's fixed mounting position, mostly on compressor housing.
Sure it can be changed by making a custom mounting bracket for a different location, but then you also have to fiddle around with correct basic setting which is not as easy like with a vacuum actuator and stop screw :-(
 

TDIMeister

Phd of TDIClub Enthusiast, Moderator at Large
Joined
May 1, 1999
Location
Canada
TDI
TDI
Now I get it. :) Clocking relative to the turbine/CHRA and compressor is the problem.
 

dieseleux

Théoricien -TDIClub Contributor
Joined
Nov 14, 2006
Location
Pas assez loin pour vider ma tank!
TDI
Jetta TDI 02
Last edited:

Dakta

Veteran Member
Joined
Oct 22, 2008
Location
UK
TDI
Non Vag :(
dunno how relevent it is as my vnt is vacuum but I tune rover diesels and there is no vnt solution so i've been having to build my own.

Unfortunately I'm not a fan of arduino, nor did I even know it existed at the time, so I had to go the long route of learning electronics, coding and so forth. Even had to design my own pcb's :( result is it looks a bit rough -







haven't finished the software but this was an artists impression:



Eventually after a hefty amount of time i've started to see results:



Green line is requested (i intend this to be mappable to either quantity adjuster feedback line or tps so you can base target boost on load) though on that test it was set static. Blue line is actual boost, red line is an unused channel (I sometimes use it to report duty cycle back to laptop so i can check its working).

It uses a very simple mainly proportional control loop though im still working on the control engine. The units are in byte value (i.e what the controller computer sees) rather than real world values i'm afraid so it might not make much sense to you.


here's another log with duty logging enabled:




Not sure if this is of any interest to anyone, but this vnt control has been my 'baby' for about a year now, it's been slow going and i've had more failure than success, but it's getting there. All configs are held in eeprom so technically once my protocols finished it's 'mappable'.

My main aim now is to try and dampen the duty cycle response.

I have wondered how difficult it would be to make it work with an electronic actuator - pwm wouldn't be too bad i dont think, CAN would be harder but some microprocessors i use do have can support so you couldn't rule it out i guess, though it'd need new schematic.
 
Last edited:

Dakta

Veteran Member
Joined
Oct 22, 2008
Location
UK
TDI
Non Vag :(
cheers, these days there's quite a few vnt control projects about, but back when I started, there was none so you pretty much 'on your own' lol.

Bit of a scary experience - that 'lda' project linked above is more advanced I think, but i'm not a control engineer at the end of the day, and there was nothing else available for control so I couldn't just sit there, Ihad to do something about it.:p

PID really blew my head when I first started researching it, I know it's quite simple in theory but turning it all into practice *puts gun to head*

I can't do math :(
 

TDIMeister

Phd of TDIClub Enthusiast, Moderator at Large
Joined
May 1, 1999
Location
Canada
TDI
TDI
Green line is requested (i intend this to be mappable to either quantity adjuster feedback line or tps so you can base target boost on load) though on that test it was set static. Blue line is actual boost, red line is an unused channel (I sometimes use it to report duty cycle back to laptop so i can check its working).

It uses a very simple mainly proportional control loop though im still working on the control engine. The units are in byte value (i.e what the controller computer sees) rather than real world values i'm afraid so it might not make much sense to you.


here's another log with duty logging enabled:




Not sure if this is of any interest to anyone, but this vnt control has been my 'baby' for about a year now, it's been slow going and i've had more failure than success, but it's getting there. All configs are held in eeprom so technically once my protocols finished it's 'mappable'.

My main aim now is to try and dampen the duty cycle response.
Veeerry interesting! Reminds me of work I did in my Master thesis developing a boost controller scheme for 2-stage turbos... :)



Edit: The answer to your response damping is to have an integrating component of your control system in addition to just a proportional scheme - a "PI" and not just "P" controller. :)
 
Last edited:

dieseleux

Théoricien -TDIClub Contributor
Joined
Nov 14, 2006
Location
Pas assez loin pour vider ma tank!
TDI
Jetta TDI 02
Some D need at spool time, make spool a little bit slower but add stability at target boost point, after reach this point, D is turn off or reduce.
You need state switch in programming code to change and adapt your algorithm to regulation system. (state switch like: Far lower, Near lower, on regulation, Near upper, Far Upper)



Dieseleux
 

Dakta

Veteran Member
Joined
Oct 22, 2008
Location
UK
TDI
Non Vag :(
Thanks....that's interesting.

D is quite easy to implement, integral I have a bit of difficulty with because quite often (when driver is off the throttle etc) the right level of boost can't be attained exactly leading to windup, which could cause instability whilst it winds back down i've been thinking of ideas to avoid this, perhaps having a regulation mode like you say where boost is near setpoint so things like the 'I term' come into effect etc.
 

TDIMeister

Phd of TDIClub Enthusiast, Moderator at Large
Joined
May 1, 1999
Location
Canada
TDI
TDI
One of the things I did was to implement a form of a D-type predictive feedforward that's only active when the actual boost pressure comes within a certain range below- or above the requested value. The idea is that when the actual boost is approaching the requested value, the controller measures the trajectory or slope of the transient boost. If the slope is high approaching the requested value from below, it's a good probability that an overshoot is going to occur. The same is true that if the slope is a large negative value coming from above the requested, an undershoot will likely occur. Below are pictures of the overall schematic of the controller (this is for a WG setup, but a VNT would not be much different). the PI controller in this model has built-in anti-windup. @Dieseleux: As you can see, state-switching was implemented. :)



Schematic of the feedforward alone:


Signal response:


As you can see, the feedforward reacts faster to open the WG between about t=1.7-1.8 seconds and begin to close it faster after the peak than with simple a PI controller alone. Beyond t=2.6 seconds the feedforward doesn't play a role. Simulations showed large improvements in stability and convergence time of the boost control with properly chosen gains. Edit: obviously you can't have a negative WG open area; the actuator signal was tapped before a limiter, so any negative signal just means zero. For context, t=0 represents a WOT load-step from second gear starting from 20 km/h.

To be clear, a proper controller, especially in the context of a two-stage turbo system, is not as simple as just having a PID controller. Here's a schematic of how BMW controls its 2-stage turbos:
 
Last edited:

Dakta

Veteran Member
Joined
Oct 22, 2008
Location
UK
TDI
Non Vag :(
wow that's some pretty good info, I assume that's a matlab affair?
(my controller is nowhere near that complex lol). That said it's easy math to calculate a slope and proximity to setpoint so i'm sure dampening could be improved somewhat easily :)

What i'd probably do is have a subroutine that activates after a certain interval (after a new boost sample has taken place) and does a subtraction of the last measured boost sample from the current one, and that's your slope over the period (I know technically it's linear and boost could theoretically dip in that period, however with a high sample rate that becomes less important). You can then use that slope as a factor in how much you want the actuator to react against the pi terms etc can't ya. Typical D.
 
Last edited:

TDIMeister

Phd of TDIClub Enthusiast, Moderator at Large
Joined
May 1, 1999
Location
Canada
TDI
TDI
wow that's some pretty good info, I assume that's a matlab affair?
It was done in GT-POWER but similar things could be done in Matlab as well. My model had both detailed engine- and vehicle physics so I can do simulations of the engine going through transient load- and speed changes as well as the accelerating vehicle accounting for driveline friction, rolling resistance, inertia, drag, road gradients (if desired), etc. :)

What i'd probably do is have a subroutine that activates after a certain interval (after a new boost sample has taken place) and does a subtraction of the last measured boost sample from the current one, and that's your slope over the period (I know technically it's linear and boost could theoretically dip in that period, however with a high sample rate that becomes less important). You can then use that slope as a factor in how much you want the actuator to react against the pi terms etc can't ya. Typical D.
Exactly what I did! :) Except I'm using engine load/speed as the input signal for the PI and boost pressure for the D placed on the feedforward.
 
Last edited:

Dakta

Veteran Member
Joined
Oct 22, 2008
Location
UK
TDI
Non Vag :(
fair does, i don't use matlab but i did once have a bit of a sniff at it, thought it might come in handy though i'm anything but a scientist lol

so whats with these electronic pwm actuators, can't they be driven with the same signal as goes to the n75 valve, or do they need a different frequency/voltage? surely we're talking some simple conversion here.

I'd be tempted to forego the controller and build an adapter board that can listen to the n75 signal perhaps nd replicate it in a format the electronic actuator can use.
 

TDIMeister

Phd of TDIClub Enthusiast, Moderator at Large
Joined
May 1, 1999
Location
Canada
TDI
TDI
so whats with these electronic pwm actuators, can't they be driven with the same signal as goes to the n75 valve, or do they need a different frequency/voltage? surely we're talking some simple conversion here.
Yes, the above was a bit of a tangent considering you were talking about your trials and tribulations of building your own controller. :eek: Apparently from posts above, depending on the actuator used you can directly employ the same PWM signal used to drive the N75, so long as you have a PWM and not CAN actuator. Some actuators even use the same 140 Hz base frequency, so it should be amazingly simple :confused: Some have feedback of vane position but the early ECUs have no use for this information.

I'd be tempted to forego the controller and build an adapter board that can listen to the n75 signal perhaps nd replicate it in a format the electronic actuator can use.
Do it and I'll be a customer! :)
 

Dakta

Veteran Member
Joined
Oct 22, 2008
Location
UK
TDI
Non Vag :(
yes, the above was a bit of a tangent considering you were talking about your trials and tribulations of building your own controller.
Don;t get me wrong, i love discussing it - there's not many people to talk to about it on the rover forums, so it's all been solo so far (which, as a non-electrical engineer or engineer in any capacity has led me to crash frequently lol)

It's nice to see it getting there though :D

Do it and I'll be a customer!
if there's enough people wanting to fit these vnt's to justify the costs of producing a design then i'd give some serious consideration - i've re-read the thread, it seems theres several pwm actuators so i'd have to plan on how to make it all interoperable (i think thats the word). Probably can't do much about the canbus right now, though some of the microcontrollers ive used in the past do have CAN facility, i'd have to do my research in that area to be able to use it though.
 

ryanp

Vendor
Joined
Jun 22, 2008
Location
Barnsley, South Yorkshire, UK
TDI
Arosa CR - 550hp - 9.7 @ 150mph 1/4 Mile, Citigo 4x4 CR TDi - 340hp, Caddy 2.0 CR 4x4 TDI - 300+hp, Golf Mk2 Van 1.9 TDI - was 290hp, Mk5 Ibiza 2.0 FR TDi - 270hp, BMW 135d - 360hp, BMW 330d - 335hp, BMW 335d - 380hp + a few more ........
As usual Kris, i can donate an E-actuator to help things along, i have quite a few including early types.
 

Dakta

Veteran Member
Joined
Oct 22, 2008
Location
UK
TDI
Non Vag :(
i'll shoot you an email Ryan, i've got to write you a response anywho :p

I am not 'waist deep' in the vag scene, so would something like this be realistically useful?
 

snakesausage

Veteran Member
Joined
Jul 6, 2006
Location
Golden, Texas (East Texas)
TDI
2015 Audi A3 TDI 2002 Jetta TDI Wagon 5spd 300k miles 2000 Jetta Sedan 5spd 500k miles
As usual Kris, i can donate an E-actuator to help things along, i have quite a few including early types.
So ryan, what type e-actuator are on the GTB turbos that you source for your kits, the type 1 (I am assuming the early type) or the type 3 CAN bus?

I do agree for a control perspective that the E-actuator makes a vacuum actuator look old tech but rarely would it be possible to install the turbo in the correct orientation so alterations to the bracket at least would still be required.
Why would it be not possible to install the turbo in correct orientation to use the e-actuator? Is it interference with the exhaust header or some other part of the car or some other reason?

Edit: I see above that it is clocking of the turbine/CHRA but since you are making the header cannot the GTB factory clocking be preserved?
 
Last edited:

Fix_Until_Broke

Top Post Dawg
Joined
Aug 8, 2004
Location
Menomonee Falls, Wisconsin, USA
TDI
03 Jetta, 03 TT TDI
Someone should start a requirements/specifications list of what we want this thing to do.

1) Use existing EDC15 PWM as input
1a) Details needed about what this is

2) Drive which actuator?
2a) PWM Input
2b) CAN Input
2c) Stepper Motor

3) Power source?

4) Load requirements?

5) Actuator feedback?
 

Dakta

Veteran Member
Joined
Oct 22, 2008
Location
UK
TDI
Non Vag :(
That'd be a good idea.
My initial idea would be to use a small low spec microcontroller to analyse the input signal and reconstruct a usable signal at high frequencies (talking about being able to recalculate signals within 1-2 cycles, including the sample time)
I have done something very similar once when trying to find a tacho signal for my rover, it didn't have a W terminal on the alternator, so i used the dash signal, that turned out to be a square signal with the tacho readout represented by it's frequency so I converted that to a more conventional signal where the frequency was fixed and the duty was modulated to represent the value.
I think I need to read up on how these actuators actually work. Direct PWM is easy enough but i think one or two types are a bit more complicated.
A full spec would be nice tho. something does however bother me -
if its such a straight forward problem why has it not been done?

either it's not a very big problem, or it's a BIG problem :D
 
Last edited:

Fix_Until_Broke

Top Post Dawg
Joined
Aug 8, 2004
Location
Menomonee Falls, Wisconsin, USA
TDI
03 Jetta, 03 TT TDI
At the moment it seems like a moving target - if we're serious about building something, we should first figure out what it is that we want to do.

Pick an ECU
Pick an Actuator
Analyze both
Design interpreter between the two

I personally like CAN based controls because there is so much you can do with them - but you have to know what you're doing (which I really don't).
 

Dakta

Veteran Member
Joined
Oct 22, 2008
Location
UK
TDI
Non Vag :(
That's why i'd do that last. I've always wanted to get into the CANBUS scheme because I've always thought it would be nice if I could modulate the fuel quantites on vp44/vp30 systems 'tuning box style' (but perhaps in a more controlled and adjustable manner).

CAN would be a vastly different circuit than another controller, and if it was all included in one design you'd probably have something that cost a fair bit, so I'm in agreement of sorts that you want to pick one route through the problem, solve it, and if needed crack on with the next.

The CAN system i think would be the hardest, because there will need to be some protocol analysing and disassembly required too no doubt. Have to work out what 'ID' the actuator is listening to, whether it responds to the ecu to confirm 'orders' and all that.

I can give an indication of what my design can do, but i'm not very well informed on actuator types and signals, so i'd have to do some research to decide anything atm :(
 
Last edited:

keaton

Veteran Member
Joined
Apr 21, 2009
Location
Mesa AZ, 85202
TDI
TDI-less ATM
I have done some CAN Bus work, PID controller (mainly just PD), lots of motor controller and control systems work as well i hacked a BMW 402/403 & 413/504 ECU... so this is right up my allie

got a BSE in Electrical & Robotics

one of my CAN videos rev1
http://www.youtube.com/watch?v=2QuRBQzGs-c

My current CAN Library covers all 11-bit PIDs, working on adding all 29-bit PIDs

rev2 is much snappier, cleaner video + GPS and other goodies... soon to be HD & for sale :-D

let me know how i can help

i just haz no diesel to hook it up to for testing :-(
 

Fix_Until_Broke

Top Post Dawg
Joined
Aug 8, 2004
Location
Menomonee Falls, Wisconsin, USA
TDI
03 Jetta, 03 TT TDI
keaton - thats pretty cool.

If we know what frequency the bus runs at and what node addresses the turbo actuator and ECU are, the rest is easy(ish). Record all can messages on the bus and pick out the messages to/from the actuator over a wide range of commands to get a feel for the data range and resolution of the messages and if there are any enable messages or checkup bits that need to be set.. The controller (yet to be designed) then just needs to take a physical input (from N75 command) and spit out CAN messages (which is the easy part).
 

ryanp

Vendor
Joined
Jun 22, 2008
Location
Barnsley, South Yorkshire, UK
TDI
Arosa CR - 550hp - 9.7 @ 150mph 1/4 Mile, Citigo 4x4 CR TDi - 340hp, Caddy 2.0 CR 4x4 TDI - 300+hp, Golf Mk2 Van 1.9 TDI - was 290hp, Mk5 Ibiza 2.0 FR TDi - 270hp, BMW 135d - 360hp, BMW 330d - 335hp, BMW 335d - 380hp + a few more ........
So ryan, what type e-actuator are on the GTB turbos that you source for your kits, the type 1 (I am assuming the early type) or the type 3 CAN bus?



Why would it be not possible to install the turbo in correct orientation to use the e-actuator? Is it interference with the exhaust header or some other part of the car or some other reason?

Edit: I see above that it is clocking of the turbine/CHRA but since you are making the header cannot the GTB factory clocking be preserved?
The only situation where the factory clocking would nearly work is with a top mounted turbo and all custom pipework and oil lines, might need a funky layout to work well in a packed bay too.

VAG GTB's have the CAN-BUS actuator AFAIK
 
Top