Mercedes Benz 48V MHEV BSG - Help needed

[LeoDJ]

Hi there, (I hope I'm posting this in the right topic)

I recently acquired a 48V belted starter generator (BSG) out of a mild hybrid Mercedes for cheap.
The main goal I have right now is driving it as a motor (ideally with recuperation).
I've already made some progress, but need help now, as I'm stumped by the behavior of the BSG.

On paper it has amazing specs: 12.5kW and 165Nm peak, integrated motor driver, manufactured by Bosch SEG.
Sadly I only got it as a spare part and don't own or have access to a vehicle that has such a BSG inside (Mercedes calls it "EQ Boost" when there's a BSG/ISG inside), so I don't have easy access to CAN logs. It's also a fairly new technology, built into cars from 2018/2019 onwards.

I got the pinout from the factory service manual and the resistance/capacitance values of the pins are also plausible, so I'm reasonably sure that's correct:

1. GND
2. NC
3. Terminal 30t
4. CAN-L¹ H
5. CAN-L¹ L
6. Terminal 30c (crash signal)

I've also acquired a promising looking DBC file that defines some CAN message IDs the BSG expects (not including actual nominal message contents tho, just some definitions).
¹ From the FSMs, CAN-L refers to the "Hybrid CAN Bus" whose only participants are the BSG and the CPC (Central Powertrain Computer). But according to the DBC file, the CPC forwards quite a few messages from other components to the BSG.

Now I wanted to try sending some of them to the BSG.
So I applied GND to pin 1, 12V to pin 3 and 6, and connected a PCAN-USB adapter to CAN H/L. It draws 12mA from the 12V supply.
Then I try to send any CAN message with SavvyCAN or PCAN-View.
But no matter what baud rate I try, the bus goes into the "Busheavy" error state, because it accumulates acknowledge errors. (The CAN transceiver in the BSG doesn't appear to even receive the message?)
I also tried it with and without 48V on the big screw terminals and swapped CAN-H/L for good measure.

I double-checked with an oscilloscope and as far as I can see the CAN message sent by the PCAN looks good, but the ACK bit slot remains recessive.

This is where I've hit a bit of a roadblock. Does anyone know if I'm doing something wrong or have misunderstood something fundamental?
Could it require a special wake-up message to wake up the actual transceiver itself? (If so, I'm not sure how to fuzz this easily, since e.g. SavvyCAN doesn't recover from a CAN bus error itself?)


PS: I also have a Mastodon thread with more pictures etc, if you're interested.

[LeoDJ]

Alternatively, I'd be super grateful for a CAN dump of the CAN-L bus communication between the CPC and BSG. If somebody happens to have (access) to one of the following cars and could capture a dump, that'd be really great. (I can also provide some more detailed information once required)

Plan B would be to get rid of the integrated motor driver and build it myself. But that would involve quite the big effort, as this BSG is a "Synchronous three-phase machine with 5-phase alternating voltage" with excitation coil. (It actually appears to have 5 phases labelled UVWXY in addition to the excitation coil). I don't think there's anything off-the-shelf or easily modifiable for something like that ^^

[tom91]

Can you please share the DBC. It is most likely that you are not waking it properly.

Due to it not having a ignition line it needs waking over CAN. 12mA on 12V does sound a bit high for a single component in a modern vehicle, does this change when not sending it any CAN?

[LeoDJ]

I'm not sure if I can send it, but I can send you a PM, I guess.
I didn't see any obvious wakeup message in there, but I don't have have an eye for such things yet ^^"
I did try a few message IDs listed in there tho, that could have something to do with "ignition" etc, but not on all baud rates.

If I remember my research correctly, "Terminal 30t" should be switched 12V, so it would only draw that current when the ignition is on, I guess.
I don't think it changes when sending CAN messages. But I can test that in the coming days.

Thank you for your help.

[LeoDJ]

I have found my mistake!
I'm glad it actually was something stupid ^^
I did test it with and without 48V connected in the past, but apparently I neglected to connect the two GNDs (power and signal GND) together.
I did now and the behaviour changed completely.
It now draws 250mA instead of 11, goes into sleep (0mA) after 5s of no CAN activity and most importantly, spews out a bunch of messages after detecting any CAN activity, wohoo \o/

Another minor stumbling block was that CAN H/L were swapped, compared to what I had gathered from the official(?) wiring schematics. (But everything else was correct).
Idk why, but I triple checked it again and they're definitely the wrong way around in the documentation. Interesting

Now I can finally get to the fun part of actually trying to send the BSG all the many many messages it wants to see and figuring out what values it needs, to do anything at all ^^

A CAN dump would still help a lot, but I fear I have to just poke around in the dark for now ^^

(Pictures: https://chaos.social/@LeoDJ/111543158801662532 )

[Scrappyjoe]

Some googling tells me the OEM part numbers for these are -

MERCEDES-BENZ (2649001100)
MERCEDES-BENZ (A2649001000)
MERCEDES-BENZ (A2649001100)
MERCEDES-BENZ (A2649062300)

In case this ever gets made into a wiki...

[Bienne24]

Hello

Me and a friend are currently studying for our bachelor's degree and are preparing our bachelor's thesis. As part of our project we have to source a 48V generator, which can come from any brand. It's tricky to find one for which there exists a Dbc from the manufacturer...
The BSG from Bosch is interesting as it fits well into our budget.

We want to ask if you would share your achievements with us and give us the dbc file ?
This would make our lives a lot easier.

Greetings from switzerland.

[threeReefs]

Hi Leo -

I am working on - I believe - exactly the same unit as you - although trying to get it working as a 48v generator rather than a motor. It seems to be part of a standard package which Lucas sells to vehicle manufacturers wanting to do "light hybrid" versions of their cars.

I am documenting my progress (roughly) at https://microcogen.info/index.php?topic=4052.0

I took a different approach from you, reckoning that as I simply wanted the generator function I did not need the controller module, so I ripped it off and opened it. I skinned my knuckles enough times doing this that I got a bit rough and it ended up being quite a destructive process (see the photo in that thread) but it's a complex picture inside and you might find it useful. My plan was to wire the five stator outputs into a bridge rectifier directly, and take manual control of the field current, and that would give me a 48v alternator.

What is really doing my head in is this : I cannot get the base alternator unit to generate more that about 20V without using crazy high rpm (I have put some graphs of performance into that thread but basically I think I would need about 11,000 rpm to get 48V+). This cannot be how it works in practice, because it's running at about 2:! off the crankshaft and the engine simply won't be turning that fast much enough time to be useful. I am also struggling to understand the "why" of the 5-phase setup. On the face of it, the alternator/motor is wound with five phases in a "pentagon" configuration (kind of like a 3-phase done in delta). Why would you do this instead of 3-phase like everyone else? It provides more smoothness in motor mode, I imagine, but am I missing something more? The documentation talks about both three-phase and five-phase operation, so maybe the controller does something more than just bridge-rectify the five-phase output. Maybe it's using the phases to do some sort of voltage-doubling, of the controller module could contain something like a boost converter to push 12-24 volts up to 48. This seems a bit clunky and I can't see the components in it that would be needed to do this. In particular I was expecting to see some large FETs in there, and there's no sign of them.

I figure if I can work out the Canbus commands I might get it into generator mode that way and see what happens. It's a nicely-made device, SKF / German-made bearings so I think it's worth persevering a bit !

Richard

[bigmotherwhale]

That sounds exactly right, the rectified output must always be below battery voltage to avoid charging while the battery is full, otherwise when you were revving the engine the diode bridge in the mosfets would be rectifying it into dc back into the battery whether it needed charge or not.
The motor winding itself works as a boost converter via regenerative braking to increase the voltage to provide the current for charging the battery, see: https://ww1.microchip.com/downloads/en/ ... motors.pdf

[threeReefs]

Oh - that's really interesting, and the principle makes sense. Although I had reckoned that when you didn't want the battery to charge, the controller would simply lower the field / armature current (or lower the duty cycle, if it's using PWM), and the generated output voltage would drop.

But I get that using a boost converter gives you a lot of control. The PDF you linked to is very helpful and I'm with it up to where it talks about using a simple / classic boost converter (LC combination) on page 11.

Then it switches to "Boost Converter Based on a 3-phase MOSFET Bridge" (page 13) and my brain stalls. Which is annoying because there are two FETs on the input of each phase in this alternator (I found them after realising that they are not conventionally encapsulated, but just sitting "open" on the circuit panel) so I suspect that is what's going on in this device. There are no inductances in this diagram on page 13. Are you saying that the motor windings act as the inductance ? That would be neat. Do you have any pointers to more information on how that works ?

In the diagram on page 13 there is no load, so no current flowing, and I think inductances need current flowing through them to provide the boost effect ? Where am I going wrong here ? I'm intrigued because half the FETs in his picture have their gates connected to pulse driver, and the other half have open/disconnected gates.

Sorry for so many questions - this really is an interesting device.

[bigmotherwhale]

The current comes from shorting the winding which will then rise in voltage on release.
To do what you want you could pick up a Ebike controller or similar with regen and use that to charge the battery, you would have to make your own controller for the armature winding using a pwm of some kind and an Arduino etc..

[threeReefs]

Actually this explains a lot about the design I'm looking at - in particular each of the five phases from the windings comes into a separate pad with components on it, and one of these is a 2mOhm precision shunt https://www.isabellenhuette.de/fileadmi ... er/BVB.PDF one leg of which is grounded. The other leg goes to a large FET so that's going to be how you short / disconnect the winding.
Thinking about this, presumably because you're shorting to ground, there's current flowing but very little voltage so very little power dissipated as heat? All the energy goes into building up stored field in the inductor (field winding) so you're not wasting it, you get it back during the next stage. At first I thought it would be a very inefficient way to boost voltage but I'm starting to see that was wrong.

[threeReefs]

I still wish I understood what they meant by "Synchronous three-phase machine with five-phase alternating voltage".