[jackbauer]

So I have started a project to design an open source logic board for the Tesla Model S large drive unit. Basic idea is to use the Heubner inverter system as the heart of a new logic board that will be a drop in replacement for the tesla original. Already proved the motor will run on the bench in slip control and just today worked out the connections for the igbt driver boards in the inverter.

I'll use this thread and github to keep track of progress and discoveries.

Things learned so far :

Motor is 4 pole

Encoder: 36ppr , 5v supply and has internal pullups to 5v

All internal board connections made by JST connectors (pdf attached). Readily available from mouser.

IGBT driver boards have 8 connections. GND , +12v , High drive , Low drive , High fault , Low fault and 2 unknown. Suspect they could be a temp sensor , fault reset or maybe both. Pinouts to follow.

Videos:
https://www.youtube.com/watch?v=BASio7Tfcnk
https://www.youtube.com/watch?v=CU1G4is9moY

 

[Tomdb]

This thing should be as easy to get working as with the Chevy Volt unit I reverse engineered.

If you need any help doing some board layout or something let me know.

Ofcourse thank you for posting this in the public domain.

 

[jhuebner]

This will be a major improvement on Tesla drive train recycling, thumbs up

I guess the drop in board won't be 2 stories high but the stm32 will be directly on it?
Make sure to use the rev2 main board schematic and maybe double check the IGBT fault detect pins as they are new. BTW some STM32 pins are 5V tolerant and all others can take 5V with a simple 10k series resistor. More in the data sheet.
Once you figure out the temp sensors I can integrate them in the software.

 

[jackbauer]

Yeh it will be a single sided smt board. I've already ordered the connectors from mouser. Good to know about the STM pins. It seems the driver will accept 3.3v or 5v signals although the fault pins are 5v.

The temp sensors should not be too much trouble. May have to pump warm water through the unit to get a second setpoint.

Given the problems we have seen with usb comms and the fact this thing will be pumping 1200A+ into the motor i'm thinking of using a MAX232 or similar for serial. What I would really like is to be able to communicate over CAN

Thanks for the offer Tom

 

[jhuebner]

CAN communication is in the making. All parameters will be settable via CAN.

I recommend you use RS485 over twisted pair, it is by far more fault tolerant and you can get cheap Chinese USB adapters, too.

 

[riba2233]

Wow, nice work and thank you for doing this!

 

[Arlo]

Great work. Its good to see someone working on this. There is a couple different drive units. I believe the "P" models have the more powerful version then the rest of the models in the rear all the fronts I think are the same.

 

[jhuebner]

Oh, forget what I said about using RS485. It needs a transmit enable pin and there isn't one... Perhaps RS422 is an option as it's also differential but bidirectional.

 

[jackbauer]

Oh, forget what I said about using RS485. It needs a transmit enable pin and there isn't one... Perhaps RS422 is an option as it's also differential but bidirectional.

So something like the SN65HVD379 on the logic board?
http://www.ti.com/lit/ds/symlink/sn65hvd379.pdf

and a usb adapter on the pc :
http://www.ebay.co.uk/itm/USB-to-RS...991839&hash=item361a7deffe:g:6kQAAOSw8GtZQezG

Looking forward to the CAN configuration. Will need to look at some sort of a pc front end for that.

 

[jackbauer]

Great work. Its good to see someone working on this. There is a couple different drive units. I believe the "P" models have the more powerful version then the rest of the models in the rear all the fronts I think are the same.

Thanks Arlo. Do you know if the P model drive units are any different in hardware or is it just firmware limited?

 

[jackbauer]

So the current sensors are plain vanilla 5v supply and 2.5v out at zero current. Positive current decreases the voltage and negative current increases.

Pinout :

Black = GND
Red = +5V
Grey = screen GND
White = 2.5v referrence
Green = current output. 2.5v at 0A. Scale TBD.

This will interface perfectly with the inverter current sensing circuit.

 

[jhuebner]

Lovely, none of that old-school +-15V stuff. I really like their straight forward design. I reckon their BOM cost is lower then that of far less powerful inverters out there (including my own in the Polo)

 

[jackbauer]

More news : Worked out the DC bus voltage sensing , DC bus discharge circuit and temperature sensors. The two case mounted temp sensors read around 15k at ambient and decrease resistance when warmed. Resistance is the same in both polarities so i'm thinking plain jane NTC. The sensors buried in the motor read 146k at ambient in both directions. No way to warm them .....yet!

The DC bus voltage sensor outputs 30mV for 10V on the bus , 60mV for 20V etc in a linear range.

Video and pinouts to follow.

 

[Arlo]

Thanks Arlo. Do you know if the P model drive units are any different in hardware or is it just firmware limited?

The hardware is different this is why you can't pay for a over the air update to get a P added to the car at a later point.

I believe all the front units are the same just the rear is different.

 

[Arlo]

So the current sensors are plain vanilla 5v supply and 2.5v out at zero current. Positive current decreases the voltage and negative current increases.

What part number?

I have found it is common to get a bit of residual magnetism on the current sensors as well as temp plays a small roll in the calibration of 0 amps.
I am using 2 types 1 is the melexis type for a buss bar and the other is a Honeywell csla2en for the inverter phase outputs them selves and both types have a bit of a problem with showing a true 0.

 

[jackbauer]

They are actually shunts to the best of my knowledge. If you want to eliminate drift when reading hall type sensors you need to take a differential reading between current output and refference output. I have used this on the LEM HTFS800 series to good effect. You would need at least 16 bit accuracy though.

 

[jhuebner]

The hardware is different this is why you can't pay for a over the air update to get a P added to the car at a later point.

I believe all the front units are the same just the rear is different.

Different as in weaker/less IGBTs? I read about 14 in parallel, maybe they populate a few less for non-P versions?

 

[Arlo]

They are actually shunts to the best of my knowledge. If you want to eliminate drift when reading hall type sensors you need to take a differential reading between current output and refference output. I have used this on the LEM HTFS800 series to good effect. You would need at least 16 bit accuracy though.

I didn't realize you were using the OEM sensors yes they are shunts.

Thanks for the pointers on the hall type.

 

[jackbauer]

Working out some more parts of the inverter :

https://www.youtube.com/watch?v=PzazVv5q20o

 

[jackbauer]

Arlo , you might find this interesting. 0.114v Vce drop on the Tesla igbts using your diode test method.