Hello Off Road World!

[defender15]

Hi all!

Project is:

buggyPhoto.jpg

Built from scratch by my friend in his garage...
This is a Mercedes OM613 engine, which is a straight-six, common rail, solenoid injector diesel: https://en.wikipedia.org/wiki/Mercedes-Benz_OM613_engine

It is currently running of the original ECU (Bosch EDC15) with a few nasty tricks (Arduino based) to avoid panic from not having the other usually fitted units (like ABS, stability control, auto-gearbox...). This is a temporary fix to get it going and enable testing of the remaining parts, but we have been looking to move to a more tunable system that is open enough so we can, worst case, even program any new functionality ourselves!
I don't have any experience with "standalone" systems, and certainly not with petrol which seem to be the main motivation for the Rusefi, but I was really happy to find there is some effort already in developing a diesel system as well!

Constraints are:
- I am currently living in a different country, so no everyday access to the actual engine
- My current "home setting" is not really compatible with having diesel spills or any elaborate test rigs around...
- Time/money as I bet for everyone else in this forum

Plan is:
- Spend the next few days reading more about Rusefi
- Start experimenting with the stimulator and TunerStudio
- Follow this thread very closely for rail control logic: https://rusefi.com/forum/viewtopic.php?f=2&t=1393
- Try to find out as much as possible about the injectors as it seems the MC33816 might not be able to generate sufficiently high voltage? - TBC
- See what I can do to contribute to the diesel side of Rusefi!

Is there a thread where the driver for the diesel solenoids is being discussed?
Also, as a first suggestion, would it make sense to have a separate index for diesel on this forum? Just an idea...

[Abricos]

Welcome to the candy shop ...
https://rusefi.com/forum/viewtopic.php?f=2&t=1393&hilit=Noxz&start=180

[AndreyB]

Probably too early for diesel section since so little diesel engines running so far

Where are you on the globe? How much software development skills do you have?

A lot about common rail is discussed on slack at https://rusefi.slack.com/messages/CFP3QAJCW/ see https://rusefi.com/forum/viewtopic.php?f=13&t=1198

[mck1117]

- Try to find out as much as possible about the injectors as it seems the MC33816 might not be able to generate sufficiently high voltage? - TBC

If the injectors are solenoid (not piezo), the mc33816 can probably drive the injectors. Noxz will eventually be testing with modern Denso diesel injectors, and maybe some Bosch GDI injectors. If the injectors are piezo, then the mc33816 isn't the right sort of driver to handle them. Piezo injectors are much higher voltage.

[kb1gtt]

Welcome along, and good to hear from you. Can you tell us more? What voltage do you expect to need for the injectors?

[defender15]

Hi again and thanks all for the replies!
Here:

Welcome to the candy shop ...
viewtopic.php?f=2&t=1393&hilit=Noxz&start=180

Yes that's the thread I mean to follow. It seems to be the one who's getting injectors clicking sooner!

Probably too early for diesel section since so little diesel engines running so far

Where are you on the globe? How much software development skills do you have?

A lot about common rail is discussed on slack at https://rusefi.slack.com/messages/CFP3QAJCW/ see viewtopic.php?f=13&t=1198

Thanks. I registered to Slack a few days ago, will join the discussions when I actually have something to add!
I am currently in Solihull, UK but the buggy is in a garage back in Portugal... Not sure yet to which extent, but I think we can manage to simulate/prototype/test things from here and when I have confidence on the system I'll buy my plain ticket to go back there fit it and tune it!
I am an ok programmer (definitely not great at it)... Background is Control Systems Engineering, actually my MSc project was done using an STM32F427ZI in a DISC1 dev board (no RTOS though...). My DIY "career" has seen a few Arduino projects and custom boards which I usually simulate using SPICE and design in EAGLE. Code-wise I have been trying to move towards a "model-based" approach: use Matlab to layout the boards' functionality and then generate the code automatically (basically it just means I got even lazier to write code than before xD )

If the injectors are solenoid (not piezo), the mc33816 can probably drive the injectors. Noxz will eventually be testing with modern Denso diesel injectors, and maybe some Bosch GDI injectors. If the injectors are piezo, then the mc33816 isn't the right sort of driver to handle them. Piezo injectors are much higher voltage.

Yes I'm sure they are solenoid. But I seem to remember seeing somewhere that they need pulses up to 100V or something. I think the limit on the mc33816 is around 72V?
Please let me get back with some certain info, I used to have the car manual some time ago when we got the engine, to help with the wiring and stuff but that pc died so I need to get it again... It's a pain to try work with OEM parts! Worst case I can find a nice scope and get my friend probe the injectors and record some data I guess.

So as said, time to get me some real info on the setup since I can't even remember what valves it has to control the rail, etc.
Something I already know is that it has a VGT (turbo) controlled by the ECU using an electro-valve, so I think this might be something new to Rusefi (if not please point me to the discussion!). One of my projects was a VGT controller for my own off-roader (Defender Td5), so I have some experience on this which might be useful!

Again many thanks to all and keep up the nice work, I'll be spying

[defender15]

So here is the progress regarding the injectors:

- Mercedes PN: A 613 070 05 87: http://mbepc.net/1/m/om613/65c/613960/0/07/105
- Bosch PN: 0 986 435 063
- OEM driver chip is Bosch 30421, or CY220: http://www1.futureelectronics.com/doc/BOSCH/0272230421.pdf

I am struggling to find a voltage rating...
This seems to suggest 80V peak pulse: https://markstammersdiagnostics.com/reference-waveforms/diesel-injectors

My understanding of the CY220 datasheet is the VBOOST can be whatever the "user" finds suitable by adapting the voltage divider to keep it within input range, and it seems this voltage is only required if there is not enough voltage build up from the solenoids already stored in the capacitors for the next pulse? Not suggesting to use this driver, just trying to understand whether the mc33816 will handle the job! Any help appreciated...

I've also got myself the car manual including repair/description/schematics etc so hopefully most stuff is described there (going through it)
I've also got the OEM ECU "SW function description" (not really the same cause info is not readily available like this, but close enough - EDC15C, 6 cyl BMW engine) but it's in german which is a pain most people who have ever tried to study these ECUs probably know... except german guys of course! :p

Meanwhile been thinking of what sort of test setup would be needed as a first approach. I'd like to keep things simple and clean, not really looking to get a wet rig at least for now. So... What about:
1) Have the simplest possible model (Simulink/Simscape) of:
- Engine Block (air cycle)
- Fuel System
2) Deploy model to some target (been able to use an Arduino Due for simple models using Simscape before for instance)
3) Develop simple "breadboard style" signal processing to be able to emulate the sensors and actuators in order to interface with the OEM ECU (I got one around) - speed pulses, rail pressure, rail valve, pump pressure/flow, need to put more thought into this...

I understand I will probably have to emulate the actuator loads so that the ECU doesn't start detecting a bunch of nuisance faults.
It might not be feasible to simulate some things like for instance the injectors.
Might as well buy a set of used injectors or even only the solenoids which I could find on Aliexpress for around 50 GBP/set of 4.

Does anyone know if it would be possible to create typical running conditions for the injectors without fuel pressure involved (ECU-wise...)?

Same goes for other actuators.
But would be good to have the simulation read the actuation from the ECU and calculate the "plant" response so the sensors can be emulated back to the ECU where possible...
Not looking for crazy precision here, if OEM ECU can be made happy, we'd have some confidence this can be used to develop the Rusefi based ECU, test the driver and different controllers,etc. Always in comparison to the OEM ECU??

I realise this might not be the most "scientifically correct" method, but as a first idea it seems it might be able to get that engine running from 2000km away! :p

[kb1gtt]

I understand diesel rail pressure is commonly around 20k PSI, while gasoline direct injection is around 2k PSI. I understand the GDI commonly uses a boost voltage of around 72V, while diesel commonly uses a higher voltage like 180V to 300V. Of course these terms are vague, so it's entirely possible your injector is around 80V.

I also understand that a chip which is desired for around 80V can drive higher by changing the resistor divider. But there are still other limits. For example, you can't drive 1M volts, as the volts would arc over various gaps, and such. However I think a chip which could drive 80V can drive say 300V if you of course change the resistor divider, but also you'll need to add an external driver chip(s) which can uphold the higher voltages. As well the design will need to consider creepage and clearance differences, etc. However I believe it is all completely do-able.

We currently have our version of the MC33816 board, as well as the PT2000.
https://rusefi.com/forum/viewtopic.php?f=4&t=1337

Beware the PT2000 board below currently have some known issues, I need to get around to completing them.
https://rusefi.com/forum/viewtopic.php?f=4&t=1511&p=33043

Also please note that most of these injectors will "operate" at 12V if you have a low fuel rail pressure. AKA you can have low cost cleaning equipment which flushes detergents, etc and it only needs a 12V battery. However if you have your rail pressures, then you need the higher voltages to open the injector.

I would suggest making a flow bench. Do you think you can create the high pressure fuel rail some how? If you can make a physical flow bench, then you could see if one of the above mentioned boards can drive the injectors and produce the proper flow rates. AKA you might be able to open your injectors with 80V, but you might also discover the flow rates are not as expected. A flow bench would help identify those kinds of issues.

Many high pressure pumps are cam driven. Perhaps you can find a head with pump and cam, flip it and fill it with oil, such that the cam bearings are oiled. Then spin it with a drill or similar motor. This could perhaps make your rail pressures. I've also seen some daring people use a CO2 canister with some compressed gas like N2 to pressurize a canister. Then this small canister was connected to fuel rail. This is very daring, but is one potential method to get the high pressure fuel rail.

[mck1117]

chip which is desired for around 80V can drive higher by changing the resistor divider. But there are still other limits.

And those other limitations are significant. To be able to turn on the high side nfets, you have to get the gate above the supply voltage, which means the gate driver needs to be able to level shift up to the high side voltage. The mc33816/pt2000 can't drive high side transistors over 72v.

[defender15]

Right, from what I've read so far, we are not expecting voltages as high as 300V for a solenoid type injector (piezo injector is more likely). At least I hope not!
I finally had some time to read the OEM driver datasheet as well as the PT2000 datasheet.
It seems the OEM setup would be something similar to this:

OEM_Circuit.pdf

2 banks with 3 injectors each.

We have in the OEM ECU from what I could identify (subject to my lack of electronics skills but...):
- 6 x BUK9640 (2 x 3 Low Side FET, YLSx in the schematic above) -> max Vds=100V
- 2 x BUK9640 (2 x Booster FET, YBOOSx in schematic above) -> max Vds=100V
- 2 x BUZ100SL (2 x High-Side FET or battery voltage FET, YHSx in schematic above) -> max Vds=55V
- 2 x big blue capacitors (VBOOST) with markings:

42160 ts10
15u K 100 L9

I thought the V rating of the capacitor would provide a hint on the max voltage of the injectors, but I can't seem to find these caps anywhere...
Could it be 15uF, 100V?
Would it be safe to assume my VBOOST shouldn't go higher than 100V minus some design margin (so say the 80V I've seen elsewhere)?

So about a fuel rig, I think the only chance of that happening is to have the smallest possible thing since I don't have a garage at home.. looking to get some parts of a Smart W450 800 CDI which has similar ECU (should be at least). If I can't get a setup compatible to my place I'll have to do an electronics only rig and simulate the rest...

Would it not be possible to add a second gate driver on the PT2000 schematic to enable switching the higher voltage on the high side fet?

Thanks a lot!

[kb1gtt]

Yes the PT2000 can be modified to drive higher voltages. I'm not exactly sure what is required, but I know I could make it happen. However we should first make the board we have operate, then we can modify it once we know it works for the normal design intent.

[defender15]

Right, I think I see my chance to actually contribute with something for the project!
What if I get 6x this (I've seen this cheaper somewhere else...):

http://www.starsdiesel.net/common-rail-injector-solenoid-valve-assembly-f00vc30318_p12899.html

and your PT2000 "dev" board? With this be a good setup to check the board is working ok? By probing the relevant voltages, etc?

[AndreyB]

I believe one injector is enough to start. Please note that part of the challenge is understating the pretty advanced control of these NXP chips and developing their firmware.

I would suggest we focus on the same MC33xxx chip just to have more people focused on the same area, unless there are some fundamental differences. I know mc33xxx board was produced. I know issues were revealed but I am not 100% if all the issues were properly logged A way to help could be to read https://app.slack.com/client/T4NJHQ8QZ/CFP3QAJCW/details/members and make sure that all the issues were recorded and tracked property. While we have some mc33xxx firmware, it's not ready to click injectors. A way to contribute could be to get in touch with @Noxz over phone/skype/etc and write down his feedback.

The hardware is only a part of this. All the code around the firmware is also tricky.

It could be best to get NXP dev board and work on the firmware even for just GDI not diesel injection. By contributing to 4 cylinder GDI firmware with dev board, you are making steps towards 6 cylinder diesel later. Please use your judgement considering what resource are available to you. If you are interested to get me a dev board it could also be a way to contribute

Please consider joining slack so that we can chat. I believe that forum is not the most efficient way to coordinate people.

[kb1gtt]

33816 is more developed than our PT2000 board. As well you can purchase the 33816 from NXP already assembled. I have made some changes to 33816 per first build feedback, if so desired I could flip out the gerbers for another spin. I was waiting for additional feedback before doing the next spin, but if you think you'd make it, I could make more gerbers. Or if schedule is more of a concern, then it would be faster to get the NXP boards which are probably sitting on the shelf some where. It would likely take month to get PCB's manufactured, BOM's purchased, and then assembled.

Hmmm, looks like @russian needs to merge my copy of 33816 with official repo. Take note that there is a file knownissues.txt, which details the changes that I've made. It should contain a list of changes. I'm sure @russian will merge these when he gets a minute. For now, take note that before manufacturing the board, it should be R0.2, as the R0.1 has known issues which have are probably resolved.

[AndreyB]

Hmmm, looks like @russian needs to merge my copy of 33816 with official repo. Take note that there is a file knownissues.txt, which details the changes that I've made. It should contain a list of changes. I'm sure @russian will merge these when he gets a minute. For now, take note that before manufacturing the board, it should be R0.2, as the R0.1 has known issues which have are probably resolved.

My copy is in sync with you, I just forgot that so many changes were made to MC33816. I am sorry for causing confusion.

https://github.com/rusefi/rusefi/blob/master/hardware/Common_Rail_MC33816/known_issues.txt

Looks like MC33816 is ready for another fabrication attempt if you make fresh gerbers and if Andre @defender15 is interested. I would probably go with NXP board to focus on microcode, I could be wrong about microcode as well

[defender15]

I'm sorry work has been crazy as we have an important build to deliver, I haven't had much time for anything else!
Yes I would be interested in building the dev board which I can hopefully do it at home using my "electronics oven". I've built a few boards before and they turned out ok so I will give it a go.
Problem is I am still tempted to have a better look at the OEM circuit for the injectors (. Pdf in a previous post) cause it seems simpler.
I've been thinking if it would be possible to achieve the same kind of logic of Bosch CY220 chip but using the M33816 microcode. This way I think I could have both the "reliability" of a proven circuit and the flexibility of the MC33816 + avoid the (lack) of availability of the bosch chip...

[mck1117]

I had a chance to look at the OEM circuit, and it looks like the CY220 uses a very similar architecture as the MC33816 or PT2000. There's really only one architecture out there for running DI injectors, and they both implement it. Because peak-and-hold injector drivers like these run closed-loop control on the injector current, the injector voltage isn't nearly as important as the current. Running at 65v vs 100v probably won't make a huge difference, other than it may add a little bit of time to how long it takes for the injector to open.

More valuable will be determining how much current and duration are needed by the peak phase of injector operation, and the current for the hold phase. Do you have an oscilloscope that you could use to probe around inside the ECU while the engine is running? Probing the current sense resistor (labeled R_sh in the diagram you have) will let us see how much current the controller is delivering during each phase of injector operation.

[kb1gtt]

It sounds like NOXZ is not going to get a chance to look at the 33816 for a couple more days / weeks. When do you think you would try to make the proto board? Also do you plan for the 33816 or are you going to try the PT2000? I have more work which needs to be done to the PT2000, and the MC33816 is much closer to a successful build-able design. Should I make the gerbers for the 33816 or should I get the PT2000 to the point where gerbers could be made.

[defender15]

So, I'd be looking to order the board any time. The thing is I'd like to take the best possible first attempt with regards to my project, so, I'm thinking a 100V peak compatible board (not sure what is the rating of the parts on your board at the moment...) plus the 2 * 3 inj/bank thing...

Bear with me a little while I figure this stuff (still trying to get KC to work... well really I tried it once last week and didn't have time to try again with the libraries from Github yet...)

Also this https://markstammersdiagnostics.com/reference-waveforms/diesel-injectors/ seems to suggest 20A peak.
This is the low side (6x) and the Boost (2x) FET in the OEM ECU:
https://assets.nexperia.com/documents/data-sheet/BUK9640-100A.pdf

At the same time I am negotiating some parts to see if I can actually build a rig or if I need to keep everything in simulation + physical loads.

[kb1gtt]

Did you get your copy of the repo with git command line utility or some git GUI utility? Or did you try it as the zip file?

(edited by r)

[defender15]

Right! After some attempts I think I got it working ok (by ok I mean no symbols show the "?" and no errors come up when saving or opening at least the MC33816 schematic and PCB):
- Downloaded .zip from https://github.com/rusefi/kicad-libraries into user/Documents/KiCAD
- Follow this tutorial https://forum.kicad.info/t/library-management-in-kicad-version-5/14636 from which I created a path variable: KICAD_RUSEFI_LIBRARY (similar to Digikey library in the tutorial)
- Added symbols (all .lib files) and footprints(.pretty folder + all .mod files) to global library/footprint and not project specific
- When opening the schematic, it prompted to remap some symbols due to new version (?), accepted, KiCAD created project specific "-rescue.lib"

I'm zero familiar with KiCAD so I don't know if any of this is already a red flag for something not being quite right, but it seems to work both schematic/PCB at least for read purposes which is kind of what I need now...

[kb1gtt]

That sounds like the mess with KC4 files opened with KC5. That should work just fine for viewing the files. It would likely be an issue with new development efforts. We typically like things to be more clean, so we avoid the entire rescue thing. If you had KC4, the entire rescue thing goes away. It appears that KC6 is just around the corner, and will likely have this same kind of mess to deal with. It is very time consuming to port over the KC4 stuff to KC5, so I'm holding off on doing this port for now.

Any how, glad to hear you got a copy of it to open and to function. I believe what you did, should be functional. A bit messy on the back end, but functional. Let me know if there is anything you need.

Also I took about 10 minutes pondering how to get more voltages out of it. It would be significantly difficult to get the high side drivers to switch properly. As @mck1117 has mentioned, the gate drive for how it works now, needs to be above 72V. The 33816 and PT2000 have a small charge pump circuit which generates this small increased voltage. Those will likely not work above the 72V they are rated for. I think a modified version of the below picture could be used to drive a higher voltage.


This was from here https://electronics.stackexchange.com/questions/31594/mosfet-when-can-we-not-assume-that-the-gate-current-is-0

In the above, I think we can change Q2 for the now N-MOSFET, then change the R14 and Q15 GND, to the 72V rail. The load resistor would be connected to the injector. I think this could work. We should probably model that up in LTSpice or QUCS or something similar.

[mck1117]

I think trying to run those extra 30 volts is probably unnecessary. I bet it will work just fine at 72v, though the injectors may open a tiny bit slower.

From the link with scope traces, the part circled is the only part that would change:



The slope would be around 25% less steep, but that's probably perfectly acceptable for our needs.

[kb1gtt]

I agree that the 72V vs 100V is not a big deal. I would expect the potential issue is that your fuel delivery at idle is a bit less accurate. There is a small chance this less accurate could cause some oscillation, but that is not likely and for a race vehicle, who cares if your idle is nice and silky smooth.

I'm not sure I agree with @mck1117 about the 25% slope. I think the difference is less than that. I think the normal time for the injector to open will be around 0.10mS. With the 72V rail I think it would be more like 0.11mS until fuel is flowing. I then think your typical idle injector pulse will be around 1mS open. I just don't see this slight delay as a real issue. I understand the concern and desire to stay with in OEM specs, but I'm a skeptic it's of any real world issue. I also think that it's a smaller variation than your gut might be telling you.

[defender15]

Hm. Ok. We are going with 72V then.
So only thing to sort is I need 6 injectors.
I need to check again when I get home, but I think I saw one injector channel is currently a spare in the original board (Gate 5? Don't remember now).
And I think I can drop the pump control circuit since I only need a rail pressure regulator which I'm hoping I can hook to any PWM capable output, so no need to have a dedicated current control for that (?) .
How hard would it be to modify the current MC33816 to add these extra 2 injectors?
Would it work?
Thanks a lot for your help guys!

[AndreyB]

I believe 6 injectors means two MC33816?

I believe we are kind of far from 6 injectors?

[defender15]

Hi! Replied in Slack to continue discussion...