Holset VGT Electronic Controller Project

[mulcibre]

Hey guys, first subaru owner checking in :P

I've been mucking with this turbo for a while, I'm about to mount it on my impreza GC, with custom flanging for the turbine. But we're not here for information about mounting are we?

This turbo is brilliant. The first time I saw a dodge super duty tow 15,000+ pounds up a 30 degree dirt driveway, I knew I had to have it. The low-end torque is unbelievable, and the response and longevity are unparalleled.

I've been poking around and this seems to be the one thread where people have their heads on straight.

First: the way to do this is with the 1939 CAN control. PWM would be nice, but doesn't look like it's happening.

I have access to a dodge super duty currently running this turbo, and have one I got off ebay like everyone else. The truck is local, I use it frequently for trailering my car (it always breaks)

My dad is mega-CS guy, with a foundation in EE (PHD in information theory) he is helping me as far as getting the signal interpretation going, and when we've got that, the control logic for the VGT.

The logical start is to use a CAN capture program on the signal lines to the holset. Here's some data on that:

http://www.kvaser.com/en/about-can/high ... ls/36.html
http://www.j1939forum.com/viewtopic.php ... 5db67f8237

an alternative would be using a Digital Storage Oscilloscope, like signal capture. This is less ideal, your "brute force" method.

I've been trying to gain an understanding of how CAN-bus works

If I am interpreting the information presented in this thread, and other info online, I need to connect to the hi/lo CAN signal wires to the turbo on my uncles trucks turbo? That will at least get us the signal we need to feed the actuator?

There are a few guys out there that have gotten some of this working, but to my knowledge nobody has gotten a stand-alone controller working

Here's a guy who got CAN analysis working on a prius:
http://www.vassfamily.net/ToyotaPrius/CAN/cindex.html

I've been trying to get this figured out for a while, I feel that the vacuum controlled VGT rigs are a good start, but this technology is so powerful, the vac-actuated holsets are like taking a loaded gun and beating somebody to death with it. If we can get CAN control of this thing, the sky's the limit.

[WhatsADSM]

Hey guys, first subaru owner checking in :P

I've been mucking with this turbo for a while, I'm about to mount it on my impreza GC, with custom flanging for the turbine. But we're not here for information about mounting are we?

This turbo is brilliant. The first time I saw a dodge super duty tow 15,000+ pounds up a 30 degree dirt driveway, I knew I had to have it. The low-end torque is unbelievable, and the response and longevity are unparalleled.

I've been poking around and this seems to be the one thread where people have their heads on straight.

First: the way to do this is with the 1939 CAN control. PWM would be nice, but doesn't look like it's happening.

I have access to a dodge super duty currently running this turbo, and have one I got off ebay like everyone else. The truck is local, I use it frequently for trailering my car (it always breaks)

My dad is mega-CS guy, with a foundation in EE (PHD in information theory) he is helping me as far as getting the signal interpretation going, and when we've got that, the control logic for the VGT.

The logical start is to use a CAN capture program on the signal lines to the holset. Here's some data on that:

http://www.kvaser.com/en/about-can/high ... ls/36.html
http://www.j1939forum.com/viewtopic.php ... 5db67f8237

an alternative would be using a Digital Storage Oscilloscope, like signal capture. This is less ideal, your "brute force" method.

I've been trying to gain an understanding of how CAN-bus works

If I am interpreting the information presented in this thread, and other info online, I need to connect to the hi/lo CAN signal wires to the turbo on my uncles trucks turbo? That will at least get us the signal we need to feed the actuator?

There are a few guys out there that have gotten some of this working, but to my knowledge nobody has gotten a stand-alone controller working

Here's a guy who got CAN analysis working on a prius:
http://www.vassfamily.net/ToyotaPrius/CAN/cindex.html

I've been trying to get this figured out for a while, I feel that the vacuum controlled VGT rigs are a good start, but this technology is so powerful, the vac-actuated holsets are like taking a loaded gun and beating somebody to death with it. If we can get CAN control of this thing, the sky's the limit.

Yup CAN control is the "correct" way to do it. I actually had CAN logging including dissecting working a while back... But simply couldn't find a truck local to me so I could get some logs on it therefore I had to give up on it.

Take some logs, disect and try to reverse engineer the proprietary PGNs. Once you understand the format then you are done (with the hard stuff at least)

[Mistaken]

So i am going to resurrect this post from the dead with an update...

http://www.youtube.com/watch?v=D-6kNMzEJf4


Thought you guys might enjoy seeing a little video of my recent progress. As of last friday i got my own stepper motor setup moving under its own power, but didnt have a lot of time to play around with it more since i went out to the bar. I am continuing work on my code today and will be creating a more assembled setup for testing. As it seems right now i am going to need a bigger stepper motor if i want to be able to cycle the vanes faster, that one in the video is fairly maxed out. If not i might try to gut my spare actuator and reuse the factory BLDC motor like others have done.

More to come.

[240z4u]

Great to see you haven't given up on it!

Keep up the good work man, looking forward to seeing how it works out.

[RustspecS13]

Talk about timing, I was just going to go look for this thread.

I'm considering using one on my VQ30 swap. Goal is a very broad responsive power band in the 300-400whp range. I really wont need that much for drifting for a while, but I hate peaky 4cyl turbos and I don't want a V8.

I'm really excited about using a stock style actuator on a VGT. A WG actuator is ok but it has so much more potential.

~Alex

[StricNyne]

i am sooo glad you made this, i am a programmer, but i dont have the experience you have !! i am seriously looking foward to this i want to run it on my setup !!

[Mistaken]

Stricnyne, its funny that you say that because I am actually a 21 year old mechanical engineering student as opposed to a EE/CS/ etc. I am just kind of teaching myself as i go along. Although i do have quite a bit of experience with automation work.

I think if i ever get to the point of selling this it is going to be as a kit, with either a modified upper actuator housing or a custom cnc'ed top for it. I am hoping to get the first setup on a vehicle in the next few months, but progress is kinda slow and its going to be winter soon.

[StricNyne]

lmk i wish i could do something to help ya, i have minor automation work with plc but not much

[Mistaken]

Quick random fact.... i got the factory brushless dc motor to move the actuator/vanes. I will try to make a quick video tonight or tomorrow, but it is WAYYYY faster then my original stepper. The vanes can basically snap open and closed.

[RustspecS13]

That's completely awesome. But is that just to open close it? Or did you get it to actually control it like oem.

~Alex

[diesel2fast4u]

Hi, can you control the original mounted motor from the holset?

I would be very interested in knowing what hardware you are using, I need to build te map controller for my engine myself as it is for a mercedes diesel where the EMP is very important.

Do you have an email adress? I send some pics of the project (diesel raceboat ...) thanks

Raf ([email protected])

[Mistaken]

I am changing my focus to use the factory brushless d/c motor instead of a stepper, since through the massive gear ratio it takes a huge amount of steps to go from open to close. I currently have the motor just going full open and full closed, but im working on the code to have it move in coarse steps so that i can control the actual position. I am using the factory hall effect sensor ring to hopefully achieve a 120 degree step, which in the end is a fairly fine step after going through the gears.

I am using the same components as someone that was attempting this project, found at the link posted below. He has the code posted and everything for MAP sensor position control. So basically anyone could copy what he did, do the final trouble shooting and put it on a car.

I personally wanted to be able to have full control over where the turbo would be at any point in time. This is mostly because i figured it would be a tuning nightmare if the turbo would be changing the a/r unpredictably, also his program stops increasing the a/r after it reaches maximum boost. I used his wiring diagram as reference in order to properly wire my motor, but from there all similarity ends.

He used a picaxe controller, im using an arduino so his code will not work with my controlling solution.

http://www.homemadeturbo.com/f7/holset- ... er-117261/


Also if any of you guys ever want to reach me just email me at [email protected] . I get email on my phone so i check it constantly. Feel free to email me anything you guys find or questions/comments. This is a turbo community project more so then my own personal project so everyone plays an important role.

[Mistaken]

One more thing i thought of. In order to use the factory motor you have to de-solder all of the motor attachment pins, which is some what time consuming, but generally fairly simple. Guys over on homemadeturbo tried sending the pwm pins they found on the board signals and wound up just torching a few microcontrollers.

[l0nestar]

*re-subscribed*

[BigADaan]

G'day Guys,

First time visit here, but not new to the VGT project. I have been messing around with the CAN control idea also.
I'm trying to get this thing going for my engine (sorry no Nissan/Infinity) but a '96 5.9 Cummins.

Followed some of the other (impressive) on-line threads, but I like the direction this one is going with the J1939 control.
Have a actuator wired up like the guy has with the PIC controller and can make it move, but the CAN idea is so much cooler.

Please understand I'm no CAN expert nor electronics expert, just have a beyond moderate interest in this stuff. I still have no idea if there's a low cost programable CAN controller or something that can do the job but this is what I gathered so far;
Out of the SAE J1939 papers two PGN's the actuator should be using. One is for actual position of the actuator (which it gets from the hall sensors). The other one is desired actuator position. Both in 0..100%.
Then on the actuator I have(when it was still intact ), I sniffed with PCANview what the CAN node had to say when it was powered up. It detected some messages and data but since I'm still waaay in over my head with these 29-bit ID's and stuff I still can't make anything usefull out of them.
If anyone is interested in this data let me know.

I don't have access to a 6.7L Cummins (hard to find in Holland) so I can't get any "live" CAN logs.

But with the right PGN's, anyone has any ideas on hardware??

[Mistaken]

I think most of us have kind of given on the j1939 route for the time being. If I could get my hands on a cummins for a few days im sure i could data log the necessary command codes, but without access im pretty sure CAN control is fairly impossible.

I did some more work with my stepper retrofit actuator today, in that i moved it to a smaller gear ratio. From what ive tested so far, even running the stepper at full blast its still wait to slow to be effective. Also using the lower gear ratio i can very easily stop the actuator. As for my brushless DC controller i lost my sensor ring that i desoldered from the board, so im making myself a new one. I started soldering all of the connections but its a bit of work. I should hopefully make some progress this week as i have quite a few days off this week from work.

[karlo]

Out of the SAE J1939 papers two PGN's the actuator should be using. One is for actual position of the actuator (which it gets from the hall sensors). The other one is desired actuator position. Both in 0..100%.
Then on the actuator I have(when it was still intact ), I sniffed with PCANview what the CAN node had to say when it was powered up. It detected some messages and data but since I'm still waaay in over my head with these 29-bit ID's and stuff I still can't make anything usefull out of them.
If anyone is interested in this data let me know.

Post it, someone might be able to decipher the data. I'll certainly have a go.

As Mistaken points out, the ideal situation would of course be if someone could pull a log from a Cummins. That would most likely provide us with enough information to piece together the puzzle.

[BigADaan]

OK, I realize I'm no futher then where WhatADSM was a year ago.
Figured out that the messages the actuator is spitting out are proprietary messages and thus mfg unique.
Guess I'll have to join you in the search for a 6.7 Cummins.

[Mistaken]

Hey everyone, just wanted to let you all know i havent fallen off the face of the earth. This project is still on going, but i recently began working on replacing my s13 coupe with a 240z. The Z is getting a full cage and thrown on a rotisserie before the motor goes into it, so progress on the VGT setup has been at a stand still.

On a side note a friend recently bought a 6.7l cummins pickup, so if anyone knows of a piece of hardware i can use to datalog CAN commands easily we are in business on the CAN control.

An engineer at work suggested something like this.

http://www.usbee.com/sx.html

Anyone have experience with these or any other CAN logger?

[Fuse]

Hi,

Want help?

[Fuse]

Hi,

I post this because someone may find this information usefull in th future.

I planed to use one of those actuators as an electric servo motor but it turned out that it initially spins a whole turn to find its "calibration point" (there is a magnet on the final gear and a hall-element on the PCB). This makes it very difficult to use as a servo though so I have to find something else for my application. However I got it to respond to CAN traffic and the protocol is quite straight forward. It is sensitive to source address so you will have to get the whole ID-field right which is extended, and I can confirm that these actuators communicate using proprietary frames (but its very easy to hack if you can access the bus and take a snapshot of the traffic). Depending on the brand you take your turbo from, there is most likely to be a difference in the ID but I have found some indications that the rest of the data in the frame is the same or very similar regardless of brand.

I would have posted more information about the protocol and the length and startbits of the signals i hacked, but that info lays on another computer... I guess you could mail me if you are really keen of knowing. But ID is brand dependent, so I don't think I can help you with that unless you use a Volvo VGT. So depending on your brand; Cummin, Nissan or whatever.. You will need to read the whole CAN ID (guessing will take you forever) but once you know that, the protocol is straight forward without any inititializing or time-critical stuff (atleast mine had noting like that).

Good luck!

[Fuse]

Forgot to mention, it's using J1939 and that means 250kbps...

[Fuse]

G'day Guys,

I sniffed with PCANview what the CAN node had to say when it was powered up. It detected some messages and data but since I'm still waaay in over my head with these 29-bit ID's and stuff I still can't make anything usefull out of them.
If anyone is interested in this data let me know.

I don't have access to a 6.7L Cummins (hard to find in Holland) so I can't get any "live" CAN logs.

But with the right PGN's, anyone has any ideas on hardware??

Please post the data if you still have it, we can compare the differences between my findings then check if the two brands are the same... If we are balls-lucky its the same as mine and I can help you get it going.


Kr
// The Fuse

[BigADaan]

@ Fuse,

Here ya go;

message length data period count

18EEFF02h 8 CD 0B A2 12 00 85 00 00 84875 4
18FF0A02h 8 17 0B FF FF FF FF FF FF 100 2174
18FFC502h 8 08 FF FF 3F 17 FF FF FF 25 8701

One of these should be vane position obviously. I haven't touched the project for a while now, but I think the 25 ms message was vane position since the data changed when moving the vanes.

Cheers!

[zengineer]

Any progress on this? I am very intrigued.

[Mistaken]

Hey guys, just thought ide check back in. Ive been super busy as of lately with way to many projects. Since the last time i posted ive parted out my 240sx and have been working on my 240z build instead. I also have been building an 3.3l flat six swap impreza to have as my daily driver/daily rally car, once i finish up this little project ill be swinging my focus back towards continuing my Z build and the VGT project.

Ive got on more semester left for my undergraduate and then i should finally have a normal schedule back to do some work. Awesome to see the collaboration building up on this project, and cool to see that someone has retrieved the CAN command. I can show those to my professor next term when im at school, hes a CAN wizard when it comes to automotive applications.

[karlo]

I took at look at the data log that BigADaan did. The messages are in order:

PGN EEFF - Address Claimed
PGN F0A0 - PDU2 format broadcast (don't know what this message is, though)
PGN FFC5 - proprietary PDU2 format broadcast

As far as I understand, this is what the holset spits out. If so, there's no guarantee that it will respond to those messages.

[FarmKTM]

I know this is a couple years old, but it has had the most information that I have found. I have a 6.7 that I have done a turbo swap on, and kept the turbo to put on my Corvette. Before I pulled the Holset VGT I connected to the CAN bus and logged all the messages. I did a coast to full throttle to engine brake run, cruising over the hills, and a couple others. I am hoping that I get some time this afternoon to sift through all the messages and test the commands on the turbo and get it running. Hope that this will answer questions that everyone including me have been searching for.

[FarmKTM]

I have the motor moving, but not exactly how I want it to. I started a new thread of my findings. The link is below. If you have any questions feel free to ask.

[FarmKTM]

Ok. I got it all working after about 20 minutes of playing with it here. It was as simple as taking out the delay between CAN messages to make it work the way I want it to. The electric motor in this will snap the veins to the position that is wanted. Is there a way to post a 10 second video of the motor movement?

Also I have not figured out quite how to engage the exhaust brake yet, but if there is interest I will get that done also.

As for normal operation of the veins it is really simple...

Message ID in hex: 0cffc600

For Data the first byte is a range of 0 - 255 for how far open the veins should be. 255 is all the way open.

Data Structure is 8 bytes, example is for veins half way open in decimal and hex.
125, 2, 1, 255, 255, 255, 255, 255
7d, 2, 1, ff, ff, ff, ff, ff