VW Mild Hybrid Starter Alternator

[Kleband]

Hi bigmotherwhale.
What do you mean by Audi test platform?
Is it a log from Audi Belt Starter Generator test platform?

[bigmotherwhale]

Yes, like the one on AliExpress that linked to.

There is also a chance that some messages may be for DC DC and battery, although they were not connected at the time.

[Kleband]

Thanks bigmotherwhale!
Quick dokument view reveals that the BSG is manipulated with implemented engine ECU signals and not with diagnostic tool function like signals. Engine rpm in CAN message with id A8 is likely calculated from alternator speed. Don't know if engine CAN messages checksum like byte 0 and byte 1 are calculated or just copied. Looks like buying the Chinese tool makes no sense, still don't know how to calculate numbers for CAN messages to be sent. Some experiments to send CAN messages can be done though.

[bigmotherwhale]

You need to use the magic byte cracker at the start of this thread and work out with the log i sent which ones need to be modified for your unit and then played back.
Looking on savvycan there is an an area in one of the messages that looks like a torque request to me, I will let you know if i have any luck when i have a unit here and has been repaired.

[bigmotherwhale]

I ran the script to compare this log to the previous captures, and the numbers seem to match so does that mean that you can just playback the same to any alternator?
I don't really understand how these numbers are used in the message structure as a check, can someone explain this in a simple manner?

[royhen99]

I think the crc is performed on bytes 1 to 7 and the magic byte, and the crc result is placed in byte 0 ( this may vary for different messages).
Damien got the Audi/VW alternator/BSG working today by replaying the log above.

[bigmotherwhale]

Nice! there should be a negative torque portion of the log as well, which you should see if you have a battery and spin the alternator from an external source, I wonder if it is voltage controlled in some way with a limit set in CAN or if that is done entirely by the ECU based on SOC.

I wonder if you can ask if all messages seem to be needed as, it would be good to get the DC-DC and battery up and running as well and if they are present then that would mean the possibility of a whole system making retrofits possible with a suitably programmed CAN controller.

[Swing]

The Audi 3.0 TDI mild hybrid is not the same as the VW 1.5 TSIe mild hybrid.

They use the same DC-DC converter I think, and might be using the same battery.
The battery is 12s, and since there are many 12s modules in the market I don't think it is worth it to do a different voltage, unless of course there is an existing system.

The battery is only 600Wh, very bulky, and the VW mild hybrid only uses 250Wh of that battery.
Knowing that the VW BSG can do 9kW power output, it means that it was only intended to run at full power for 100 seconds, and generate at 12kW so only 75 seconds of regen at full power.

Of course this could be different for the Audi BSG, since it is a different unit with different power levels. But it gives you an idea.

The battery reports SOC and temperature of I tember correctly. It has current sensing on the BMS.
Since the battery reports SOC through CAN and it is quite a lot of power into a small battery, I don't think the charging level is regulated by voltage.

[bigmotherwhale]

The alternator is voltage regulated in a sense, its rated voltage is 51.5v in generator mode and 40v in motor mode. this would fit in pretty close to a 12s pack maximum and minimum voltages for operation, 4.29v and 3.3v, so all that would be needed is to make sure the current falls within the right range for acceleration and regen and some sort of de-rate for both time duration and temperature.
see https://github.com/damienmaguire/VW-Mil ... ergen2.jpg

[bigmotherwhale]

Got my Alternator in the post today and connected it up, managed to get it to attempt to spin using the logs first try, I say attempt as it has a broken power stage like most of these so it just buzzes, I will try again after i have bodged some mosfets together for testing.

[bigmotherwhale]

Anyone see an issue with using an aluminium PCB to replace the blown fets?

Planning on using 3 x FDBL86366-F085 per phase as that is all I can fit in the footprint.

[jrbe]

Anyone see an issue with using an aluminium PCB to replace the blown fets?

I think an aluminum pcb makes a lot of sense. The only downside I see is when soldering the bus bars & jumpers they will need a lot of heat.

Doing a quick design review, maybe prematurely..
Do the corners of the pcb need to be clipped to fit into the housing? Or are those used to hold them down with the black plastic?

Are you planning on running the gate traces like the original, using short jumpers?
The bus bar pads look different too, new jumpers for those?

What's the 5th jumper wire from the capacitor area, is this a temp sensor?

Also, the original design seems to be split in the middle, for thermal expansion? I'm not sure that's worth copying when using an aluminum pcb or not. The aluminum used in pcb's is very soft, it might bow from heating up. I think I'd try 1 piece first with thermal paste to the housing. Maybe a few feet on the cover to push the mosfets / pcb against the motor / generator housing.

[bigmotherwhale]

Yes i was worried about that too, soldering on these Al boards is difficult at the best of times.

I have taken out the old mosfet boards and measured them to get the size of the PCB they are square edge.

I dont think they are split for any reason other than ease of manufacture.

I was planning on using a thermal epoxy but a hardware mount might also be a good call.

Im only using a single layer board as it will cost too much for double, thats the reason im using a wire per gate due to routing, i also considered using external resistors to maximise copper area on the source drain connections.

The original bus bars are cut leaving as much copper as i could and im soldering them onto the board, at least thats the plan, i will be using lead solder and a large iron, i hope its enough.

There is no temp sensor, its +/ - /phase/ low side /high side

Im only using half the number of power connections to the capacitor as the factory as i have separated the phases left and right rather high low repeating like you see here. I hope this will be okay.

I have no data for the original fets so its just a guess as to what capacity the gate driver can handle.

There are some better fets available from Infineon but they are very expensive, i will see how i get on with this.

[jrbe]

Thanks for adding that picture, it's very helpful.

If you're going with individual gate wires it might make sense to add a short trace off the gate resistor and a pad for each wire (if you didn't already.) Should help reduce soldering effort some.
The pads for the bus bar connections might be on the small side. It might help to have exposed pad on a side you can fit the iron to get the solder puddle started.

I think he capacitor / battery + / - should be ok as long as the capacitors are paralleled well between all the legs. If they're individual you may end up losing capacitance from missing half the bus bar / direct path.

I think it looks like a good plan overall.

[bigmotherwhale]

pads on the gate resistors are already there, and the pads have already been enlarged on the power and phase, this was a print out i used for getting the right alignment, as you can see they are not yet perfect on the right.

The capacitor is one big one with two plates +/- insulated with plastic that spread out from under it and to all the phases so i think im ok there.

Im wondering if i should also make an exposed copper area to fill with solder between all the phase and power area, with just 1oz copper board and a full fill and im still wondering whether it will be enough to handle the current, the motor is a real work of art with the thickness of the stator wiring.

This is my first PCB ive made so im still learning how do do things, its taken quite along time even tho its a very simple design, mostly learning the software.

[bigmotherwhale]

I have added some exposed copper areas around the pcb for filling with solder to help with the current distribution between the fets.
do you think this is needed on a 1oz board or is this overkill?

[jrbe]

pads on the gate resistors are already there, and the pads have already been enlarged on the power and phase

Adding a short trace between the resistor pad and adding a dedicated wire pad means less heat input required to solder on a wire. The resistors, copper planes, and aluminum pcb base all wick away heat when you try to solder. Dropping down to a small trace helps narrow down the escape path for the heat from the soldering iron. You will often see similar things where ground pours connect to a smd pad to help as heat breaks. But you don't have much room left in your design for this.

this was a print out i used for getting the right alignment, as you can see they are not yet perfect on the right.

Understood, and it's a great way to test before ordering boards. Making the pad larger means the solder has a wider base to connect the bus bar. If it's too small of a pad the solder won't shape like a Hershey kiss around the bus bar to help spread the connection both electrically and mechanically.

The capacitor is one big one with two plates +/- insulated with plastic that spread out from under it and to all the phases so i think im ok there.

Yup, that's good.

Im wondering if i should also make an exposed copper area to fill with solder between all the phase and power area, with just 1oz copper board and a full fill and im still wondering whether it will be enough to handle the current, the motor is a real work of art with the thickness of the stator wiring.

Solder is not as conductive as copper, it doesn't help as much as typically expected. You could solder a copper bar or copper bend to the pcb then attach the cut bus bars to them. Lots of heat required to solder. Some people leave a missing # pattern in the solder paste layer to allow flux / gasses to escape the large soldered area. Doesn't matter if you're hand soldering though. It could mean you could solder it in an oven /hot plate then just worry about connections in the motor.
You could leave the extended sections as exposed copper and decide not use it too.

This is my first PCB ive made so im still learning how do do things, its taken quite along time even tho its a very simple design, mostly learning the software.

There are techniques to bring down the power and switching loops. It might not be as critical here, as long add they didn't use GANfets originally (faster switching requires low inductance designs.) That would mean mosfets wouldn't work. I think it's unlikely, GAN fets are pricey still.

[bigmotherwhale]

PCBs arrived from JLC today and seem to be a good fit, waiting on the PID controller and FETS so I can populate.

[bigmotherwhale]

Did the soldering today on a pid controlled clothes iron, came out quite nice, Im waiting on some new tips for a Weller 200w iron to arrive before i can attempt install.

[bigmotherwhale]

Bodge completed however I want to have a good look over the gate driver again to see if all the transistors are OK before its powered up.

[bigmotherwhale]

Hmm it seems I must have damaged something or it didn't like being powered up without the motor phases being connected, it seems to have stopped transmitting can messages over the bus.... I was very careful to go thought all the driving transistors on the broken phase to check they were working. not really sure what went wrong

[jrbe]

Hmm it seems I must have damaged something or it didn't like being powered up without the motor phases being connected, it seems to have stopped transmitting can messages over the bus.... I was very careful to go thought all the driving transistors on the broken phase to check they were working. not really sure what went wrong

That sucks, I was rooting for you.
Time for a micro open inverter board for these?

[bigmotherwhale]

Yes its a shame, I shouldn't have powered it up with it apart, i don't think there is anything wrong with it physically, I think its gone into some error mode. I suppose you could make it work with something like that, but i got a Kia HSG and got it working with a Lebowski board and gen1 Honda IMA inverter in a few hours, its not as compact and will require a separate inverter its likely to be more reliable, its also certainly easier than obtaining CAN logs, which seems to always be the issue when i want to use parts.

[ama]

Sorry if I resurrected this. I have been following this for a while.

I just bought a brand new unit from Audi, with the new part number 4N1903028.
Supposedly, it would not fry itself anymore with some updated hardware and software.

I am starting to build a bench and then I will start with replaying the log posted earlier to get it spinning.

[Doig5710]

It could possibly be going into component protection mode after a certain time, vag stuff sometimes has a time delay built in so that it has some functionally until the time limit is reached then it shuts down until coding has been done to link it to the car.