Tesla Small Drive Unit Support Thread

[bulletbug]

I'm using SDU from a 2015 Model S. It has the odd connectors from HVJB to the inverter. I don't have the correct wires. I'd like to make up a solution where the inverter can be taken all the way out of the housing. Then remove the receptacles for these cables so that 4/0 wire with a lug on it will come through each wire opening in the case.

To make it waterproof, I'll design and print 3d parts that will replace the bolt down points for each cable as well as integrated o-rings.

I've attempted to remove the power stage from the case and run into challenges.

Does anyone have experience with removing the power stage from the case of an SDU? Can you share the steps of the process?

Alternatively, does anyone have a blown SDU that they would like to sell cheaply so I can experiment on a non-working part?

Thanks,

[jason_arnold]

I have something of a procedural question: in Damien's "SDU howto" video from Jan 2018, the SDU board with current sensors and metal shield have been removed as a single unit. I'm trying to tear down my 2017 Model X front SDU to the same stage to simplify the R&R of the current sensor leads, but can't get the plastic shroud that holds the phase cables in place to budge - any advice from the veterans?

[jason_arnold]

Well how do you like that, not too often you get to answer your own question, but I got it sorted. Figured I'd share here just for posterity: my mistake was trying to pull the plastic shroud UP and push the phase cables DOWN through it. Turns out the opposite is what's needed: start by bending the phase cables over so that the shroud is parallel to the PCB, then push the shroud away from the lugs on the phase cables; once the square part of the lug is poking through the top of the shroud, give it a twist and the insulation will squeak through the slot nicely (the middle terminal in the pic below shows that part pretty well, I think). I went from right to left as I found the slack in the cables was best this way.

[jason_arnold]

Just a little note to let people know we have finally been able to source the 2 pin JST connector for the HV interlock and coolant temp sensor so all future opensource boards will be supplied with these connectors fitted. At least until Elon buys Mouser out of stock again

So it seems Elon did go on a buying spree as my V6 SDU board doesn't have these connectors (CONN3 & CONN4) fitted. Does anyone know what the part family/number is so I can try to source some?

[Jack Bauer]

[jason_arnold]

Thanks Damien, believe me I have watched this video and know I can just swap the parts over, but figured since the parts number for the Melexis current sensors as well as the 2 pin JST connectors are now known, I'd rather populate the SDU board without savaging the OE board

EDIT: found the CONN3/4 p/n in the BOM for the V4 SDU: S02BA-AIT2-1AK from JST. Mouser has heaps in stock.

[Roger]

I found this after market drive axles for Model S 2012-2018
https://aftermarketintel.com/trakmotive ... a-model-s/
https://www.rockauto.com/en/catalog/tes ... ic,3310381

[georgS]

Anybody knows where I could get a pre-crimped 20pin connector for SDU?

[georgS]

Anybody knows where I could get a pre-crimped 20pin connector for SDU?

I will answer my own question: Ford 7U2Z14S411GA / Motorcraft WPT889. Unfortunately they seem to be only available from US.

[cloudy]

Having some trouble sourcing CONN1 (IGBT DRIVE CONNECTOR)

3M 929975-01-36 828-2058 3M 929 Series 929975 Series Number 2.54mm Pitch 72 Way 2 Row Straight PCB Header

I've bought one of these as per the BOM - but the connector is 8mm high (correct as per the datasheet) The tesla part is lower profile - about 5mm deep. Is the difference OK?

[cloudy]

Looks like mouser have the low profile equivalents for CONN1 now (maybe they didn't originally) Will try ordering one
CES-112-01-L-D
https://www.mouser.co.uk/ProductDetail/200-CES11201LD

[cloudy]

The part above arrived and is exact match for the tesla part - fitted and mates well. @Damien if not using already may be worth looking at?

I got my first turns of my SDU tonight - very exciting! I have some questions if anyone is able to help?
My test power supply is 50V/10A - not sure if this is enough for basic function testing.

- I can get the motor to spin in manual mode at fslipspnt 5 and ampnom about 40%, increasing the values starts to sag the input voltage and the motor starts to judder. When using normal run, I need to up the boost to around 5000 to get the motor to start at all (I guess down to the low voltage), and it starts to judder at anything more than a crawl. Do these sound like symptoms of the weak power supply, or encoder issues?

- Secondly, I see the A,B,C hs temps and the fluid,stator,case temps are multiplexed - how are these presented in open inverter? are they averaged out, cycled or available individually?

EDIT: Managed to dig this up around the LDU, but I'm guessing it's the same:
There are four motor temp sensors. Two are implemented as I was given the temp/resistance info. Two were not as up to now no one had bothered to find out or share. The mux runs through all 3 heatsink and all 4 motor temps and displays the hottest.

I'm getting some odd values on both tmphs and tmpm, if I sample at 1:1 tmphs oscillates between values between 22.50 and 100 and tmpm between 0, 49.21 and 190. Very odd - guess it's an issue with the way I built the board...

- The params_SDU.json from Damien wont import fully there are some mismatched field names - is that just due to running the latest firmware? Do I need to change the switching freq/overcurrent mode to avoid the inverter blowup issues?

[cloudy]

I am getting very odd temp readings, and have been tracing the source code for reading the multiplex vs the hardware layout on the actual board (SDUv4).
There seems to be some strangeness in the mapping at least in 4.76 against the SDUv4 board... Am I interpreting this wrong? If even the logic is inverted, things don't seem to line up....

Logic Table for 74HCT4052 multiplex on SDUv4



And here is the source with **** comments (I'm assuming .Set is a pulldown and .Clear is a float high)

[cloudy]

I think this mapping is actually correct for the LDU. The issue is the mux inputs are different for the SDU in both pinout and sensor type, but I can't see a hardware switch in code for running on the small drive unit...

[Jack Bauer]

Moved to support thread.

[cloudy]

Just to update if anyone is interested. I found that S1/BMS_IN is not switching. You would expect to see it cycle logic states as the stm polls the multiplex. S0/ERR_OUT IS toggling. This means I guess some of the analog value is being read from an inactive input.
I flashed a simple test program to output pulses on both S0 and S1 and that works fine, so it's not the hardware...

Wondering as BMS_IN is used as an input on some hardware platforms, whether this is being initialised as an output correctly?

[cloudy]

I get the feeling I am going to be talking to myself here, but spent some more time on this tonight.

Compiled a new version of openinverter with the temp values set statically - motor now works normally without any temp derate. I can only assume the SDU sensors are simply not handled differently from the LDU in software and therefore have no chance of being correct. Still a little confused as I can't be the first person to come across this... A forum search didn't turn up much.
I will try and fix the mux select logic issue and then map the correct sensor inputs. Anyone know the best way to contact Johannes to see if we can put in a fix for all?

I'm still new to this, so if I'm barking up the wrong tree here - please let me know!

[Jack Bauer]

I get the feeling I am going to be talking to myself here, but spent some more time on this tonight.

Compiled a new version of openinverter with the temp values set statically - motor now works normally without any temp derate. I can only assume the SDU sensors are simply not handled differently from the LDU in software and therefore have no chance of being correct. Still a little confused as I can't be the first person to come across this... A forum search didn't turn up much.
I will try and fix the mux select logic issue and then map the correct sensor inputs. Anyone know the best way to contact Johannes to see if we can put in a fix for all?

I'm still new to this, so if I'm barking up the wrong tree here - please let me know!

welcome to my world:)

Sounds like you have uncovered a major bug so first off thanks for sharing. I too doubt you're the first to find this. Most likely discovered and patched but not shared Maybe if we rename to closedinverter.com and charge everyone $10k to join...

I'll ping Johannes and we'll get it sorted.

[Kevin Sharpe]

I get the feeling I am going to be talking to myself here, but spent some more time on this tonight.

Thanks for your debug efforts and sharing the results

[cloudy]

No problem, glad it's useful!

I'm able to get sensible real values at least from TEMP-A & FLUID now by setting fixed low mux states. Next step - figure out why temp1_out.Set() isn't actually sending the pin high.

JCurve isn't quite right for the SDU fluid temp sensor, (reading 20c at actual 11c). Will try and get some measurements to build a curve for it. Unless anyone knows what sensor it's using?

[Kevin Sharpe]

welcome to my world:)

Sounds like you have uncovered a major bug so first off thanks for sharing. I too doubt you're the first to find this. Most likely discovered and patched but not shared Maybe if we rename to closedinverter.com and charge everyone $10k to join...

I'll ping Johannes and we'll get it sorted.

Can you remind us how many Tesla inverter controller boards you've sold?

[Jack Bauer]

About 280

[cloudy]

I'm really confused then! I can't believe there are 280 out there and nobody has picked up on tmphs derate, it would be pretty obvious... Perhaps the temp1 switching is a unique issue to my setup and people just didn't notice/care about mis-mapped sensor readings? Or maybe this is only manifested in newer firmware?

Either way, If anyone is able to power up an SDU board and just scope S1/Pin 9 on the temp sensor mux IC that would answer that...

[Jack Bauer]

The heatsink temp sensors are the same in both large and small du. Also the mux pattern should he the same at least for the heatsink side of things.

[cloudy]

Yes, good point - all 52K type on 0,1,2 - but port 3 grounded out HS side on LDU (2Y3) and both sides on SDU (1Y3/2Y3)

The issue I think on the HS side is S0+S1 both high would pick the 0V state on input 2Y3 & this is polled for in case3. This would read as temp 100+ If I'm reading the sensor curve correctly and therefore the max function would return this as the max hs temp...

I found in the source that .set() is commented to set output High, so perhaps it's not inverted as I thought. Which means that case0 and case3 appear to be swapped, or maybe a mux chip with an inverted input has been used in the past?

Fortunately this is all academic as S1 never goes high for me! I'll pull the inverter off the drive tonight and confirm the logic states.