How to not kill a tesla drive unit

[Bigpie]

There's been a few dead drive units lately so Damien has made a video, watch it before trying out your drive unit to avoid an expensive error.

[johu]

Just added Damiens video to the Electronics Basics wiki page and also described how to check a power stage. If anyone asks, just link here: https://openinverter.org/wiki/Electronics_Basics

[P.S.Mangelsdorf]

Just added Damiens video to the Electronics Basics wiki page and also described how to check a power stage. If anyone asks, just link here: https://openinverter.org/wiki/Electronics_Basics

The video was exceptionally helpful in that regard.

[ekohn04]

I took some comprehensive notes on this video because I want to make sure the information is as accessible as possible as I believe it is critically important to understand every part of this when using an LDU for a conversion.

https://docs.google.com/document/d/1jCw ... sp=sharing

I also typed out some chapters for the YouTube video. This will help a long but informative video be much easier to navigate if someone can get these to Damien so he can put them in the description. All he has to do is copy and paste this into the description of the video and the chapters will show up when scrubbing through.

00:00:00 Intro
00:08:22 Test main busbars (H.V. in) on DU with multimeter
00:10:10 Flip multimeter polarity and test busbars again
00:11:07 What do we NOT want to see?
00:11:48 More multimeter tests
00:14:21 Explanation of tests
00:15:25 12v test logic board
00:16:00 12v power in
00:16:51 Application of 12v power
00:17:43 Soldering the board connectors
00:19:73 Flux application
00:21:50 Soldering starts
00:28:37 Checking solder
00:30:29 Second 12v test to ensure proper function
00:33:22 Installation of logic board in drive unit
00:35:46 Encoder test intro
00:36:39 Parameters check
00:39:03 Getting into RunMode
00:40:23 Plotting speed values
00:42:55 Spot value check
00:44:52 Spin test setup
00:48:24 Power supply startup
00:50:33 Spinning the drive unit/mistake
00:52:08 What’s inside a drive unit/how does it work?
00:59:12 How an OVERCURRENT error happens
01:11:46 How to prevent OVERCURRENT during testing
01:16:23 How to prevent OVERCURRENT during driving
01:22:45 Outro

[Bigpie]

I've given him the heads up about the timestamps

[ardillolambo]

guys,
A newbie here, In a few days I'll soldering my SDU board and start testing as described by Damien in the video and summarized by @ekohn04 (kudos to him as well)
So while I'm doing the testing is clear to me how to protect the transistors....sorry about my ignorance in this one... but how this gets protected when everything is fine and dandy...for example when turning off the car using the "ignition switch"

[johu]

I want to ask at this spot, how many people have killed their drive unit after it had been working fine for a while? Most of all mid-drive would be interesting.
One German customer has lost drive after it worked fine for 40.000 km and apparently he's not the only one.

[catphish]

One German customer has lost drive after it worked fine for 40.000 km and apparently he's not the only one.

That's a shame. Do you know the the nature of these failures, just randomly fried IGBTs?

[P.S.Mangelsdorf]

I want to ask at this spot, how many people have killed their drive unit after it had been working fine for a while? Most of all mid-drive would be interesting.
One German customer has lost drive after it worked fine for 40.000 km and apparently he's not the only one.

I've beaten the shit out of mine and not broken the inverter yet. I've broken axles, motor mounts, mounting tabs, etc, but it seems the inverter is robust, even though it has the (much discussed) weird WOT OC error issue.

I do know of at least two LDUs with blown gearsets. Jon Volk broke his, and I bought a sport LDU with a bad gear set that was broken in a stock P85. I have heard rumblings that this is more common than is often discussed.

I would hazard a guess that failures mid-drive are gears, tune, or lost 12V. I did experience a loss of 12V shutting the car off (DC-DC had failed to turn on) and even then, the car stopped safely and restarted once the 12V battery was back up. Essentially as the 12V voltage dropped, it managed to safely shut itself down. I could see how a sudden loss of 12V could drop contactors and cause a failure though.

I also had a weird gremlin that took forever to figure out where my hazard lights / 4 way flashers were triggering the brake light input, causing very concerning behavior.

[johu]

That's a shame. Do you know the the nature of these failures, just randomly fried IGBTs?

He was driving 100 kph on cruise control and then lost drive.
Afterwards the bad old "turn on - see OVERCURRENT message". He did measure the IGBTs and apart from not showing any bubbles or magic debris they also measure ok. So the gate driver board gone faulty?

I've beaten the shit out of mine and not broken the inverter yet. I've broken axles, motor mounts, mounting tabs, etc, but it seems the inverter is robust, even though it has the (much discussed) weird WOT OC error issue.

Right, I reckon that rules out my first suspect: the deadtime. It sits at its default of 63 or ~1us for the LDU which could indicate it was never tuned to fit. But I guess that would have shown earlier, or would it?

[mane2]

I broke my first one, but it was my mistake due opening contactor while regen. Second inverter is still on shelf as it got wet and LDU leaked from both ends. Third motor is in, but throwing now constant OC while phases measure ok. Changing the OI board to V6 this week, an'd well see if the OC error continues. Might be old OI board gone bad with first failure and current spike, so these do not really match the mid-drive error.

[P.S.Mangelsdorf]

the deadtime. It sits at its default of 63 or ~1us for the LDU which could indicate it was never tuned to fit. But I guess that would have shown earlier, or would it?

I'd defer to your expertise on whether that would matter, but I don't believe anyone has tried tuning that parameter. All of the example sets I have saved show it at 63. What would be the benefit/danger to changing that?

[johu]

I'd defer to your expertise on whether that would matter, but I don't believe anyone has tried tuning that parameter. All of the example sets I have saved show it at 63. What would be the benefit/danger to changing that?

Setting it too low means high current flows across the phase legs and potentially destroys it. So I certainly wouldn't advise a lower value. Setting it higher means you loose some output voltage and with it some performance.

I always tuned it by diminishing returns with a current limited power supply. Start at a low value (say 30) and observe the input current with no motor connected and the inverter idling in manual mode or something (it just needs to run the PWM). You should then observe that the input current becomes smaller as you increase deadtime. There will always be some residual current (say, 50 mA) that you can't get rid off even with the highest deadtime setting. So you're tuning for the point where only that is left.

Thing to note: deadtime is only changed on starting the inverter. So when tuning
1. Change value
2. Start
3. Test
4. Stop
5. back to 1

[jrbe]

Temperature & current (voltage & rpm too?) will all change the dead time.

https://www.infineon.com/dgdl/Infineon- ... aefc41005b

Would it make sense to use a very sensitive current sensor to tune these and modifiers?

[johu]

Thanks for posting, was new to me.

Fortunately and surprisingly more current means quicker turn-off or less dead time requirement. So what is being tuned at no-load is already worst case. Only temperature is indeed not taken into account, so a safety margin should be added (in their example 20% for 125°C vs. 25°C)

Would it make sense to use a very sensitive current sensor to tune these and modifiers?

The display of even a cheap lab supply is the right order of magnitude

[jrbe]

The display of even a cheap lab supply is the right order of magnitude

I wasn't very clear, meant an in the vehicle sensor to monitor 0 accel to low motor current to watch current at pack voltage. I didn't mean extreme resolution, just a decent resolution in the low amperage window to watch this (not just say a 1000A shunt giving vague low current info.)

I suppose the big shunt could then be used to tune dead time at different currents in the useful measurable range under some to full load.

Temperature could be fine tuned as well by purposely heating components up / keeping the coolant fan off below the target temp.

[Arlin]

Setting it too low means high current flows across the phase legs and potentially destroys it. So I certainly wouldn't advise a lower value. Setting it higher means you loose some output voltage and with it some performance.

I always tuned it by diminishing returns with a current limited power supply. Start at a low value (say 30) and observe the input current with no motor connected and the inverter idling in manual mode or something (it just needs to run the PWM). You should then observe that the input current becomes smaller as you increase deadtime. There will always be some residual current (say, 50 mA) that you can't get rid off even with the highest deadtime setting. So you're tuning for the point where only that is left.

Thing to note: deadtime is only changed on starting the inverter. So when tuning
1. Change value
2. Start
3. Test
4. Stop
5. back to 1

Hi Everyone, I have spent extensive time developing 3 phase inverters.
I feel its needed to comment on the dead time settings.
Dead time is not to be adjusted after setting it the 1 time.
And Dead time is to be set to worst case for all components involved.

So dead time is set as you add up all the longest off times of all components and compare that to all the fastest on times of all components.
So when you have the IGBTs (or what ever power switches used) in the Hi and Low side of 1 phase you are looking to prevent having both of them be in the on state at the same time.
What you start with is the main IC that sends the PWM Look at the fastest on time subtracted from the slowest off time
Next you add all items in the series from the Main IC to the IGBT it self. So it might look as follows
Note: these are all made up example numbers just to show how you do it.
IC slowest off (20ns) - fastest on (10 ns) = 10ns
Optic isolators (if used) Slowest off (30 ns) - fastest on (10ns) = 20ns
IGBT drivers Slowest off (150ns) - Fastest on (50ns) = 100ns
IGBT (or what ever powerswitch used) Slowest off (1200 ns) - fastest on (300ns) = 900 ns
So that would look like 10ns + 20ns + 100ns + 900ns = 1030ns or 1.03uS

There might be more components in the series. So you need to add worst case of any other components as well
When we talk worst case you need to look at the data sheet and see for all temperature and currents what is the worst case for both the fastest and slowest at all temperatures and all currents.

If we know all the component part numbers I would be happy to do this based on the data sheets.

I will also comment IGBTs are usually quite slow. 1uS dead time is likely too short or just on the edge of what is ok.
My first inverter using Fuji IGBT moduals and good gate drivers needed 2uS dead time for instance.

-Arlin

[arber333]

Hi Everyone, I have spent extensive time developing 3 phase inverters.
.....

I will also comment IGBTs are usually quite slow. 1uS dead time is likely too short or just on the edge of what is ok.
My first inverter using Fuji IGBT moduals and good gate drivers needed 2uS dead time for instance.

-Arlin

Nice to see you here Arlin.

I must chime in to agree with your experience. I would even go further than 2us dead time.
I use Volt inverter with MBB600 IGBT modules viewtopic.php?p=9791#p9791
While i could set deadtime for 2.6us according to this calculus here: https://www.infineon.com/dgdl/Infineon- ... aefc41005b

In the beginning i erroneously used 1.8us which resulted in very sharp torque delivery, but it also resulted in two dead IGBTs despite desat sensing. Allways happened with the car driveoff. Probably because of high amps demand...
Then i read the article from Infineon and it seemed calculus should be 2.2us in best case but like Arlin said i couldnt exactly determine driver delay.
I decided to stretch this a bit to 3us on a safe side and it worked great so far.

[Arlin]

Nice to see you here Arlin.

I must chime in to agree with your experience. I would even go further than 2us dead time.
I use Volt inverter with MBB600 IGBT modules viewtopic.php?p=9791#p9791
While i could set deadtime for 2.6us according to this calculus here: https://www.infineon.com/dgdl/Infineon- ... aefc41005b

In the beginning i erroneously used 1.8us which resulted in very sharp torque delivery, but it also resulted in two dead IGBTs despite desat sensing. Allways happened with the car driveoff. Probably because of high amps demand...
Then i read the article from Infineon and it seemed calculus should be 2.2us in best case but like Arlin said i couldnt exactly determine driver delay.
I decided to stretch this a bit to 3us on a safe side and it worked great so far.

Yeah 2uS ended up being the calculation for my first High Power inverter.

[johu]

Arlin is here!

Found in the teardown thread, IGBT is AUPS4067D1-B . t_off (@4R7) 230us, t_on 60us.

Driver 1ED020I12
Not sure what is important here. Propagation delay is about 190ns . Rise and fall time is gate capacity dependent. Not sure if "input capacitance" is the gate capacity? If so it would be 7.7nF. Since it is 16 gates 123 nF. The data sheet only has rise and fall times for 34 nF, so we're into the unknown. At 34 nF t_rise(min) is 200 ns and t_fall(max) 600 ns.