DIY Electric Car Forums banner

101 - 118 of 118 Posts

·
Registered
Joined
·
730 Posts
Discussion Starter #102
With CANBUS one can try to reduce the total length of wiring in a car.
The first I/O satellite board design had too much I/O in a single location.
It had to be split up.

INSTRUMENT CLUSTER
So here's the next CANBUS teaser. This duino compatible board goes behind the instrument cluster.
Outputs pseudo (i.e. filtered PWM) analog to the legacy instrument cluster: speedo, revcounter, fuel, temperature and boost.
Yeah, that includes the ancient turbo boost needle.
Other I/O close by for instance: BRAKE (pedal), DRIVE, REVERSE, THROTTLE. And specific EV: PRECHARGE and MAIN CONTACTOR relay driver outputs.

DISPLAY MODULE
Damien Maguire (Jackbauer) made his version of a CANBUS display module a couple of years ago: https://www.diyelectriccar.com/forums/showthread.php/can-bus-lcd-displays-178585.html
Something similar can replace the ICE version that is usually found somewhere in a central position in the dash.

Aren't those duino compatible boards great?
 

Attachments

·
Registered
Joined
·
730 Posts
Discussion Starter #103
Delphi is ready for VW/Porsche/Audi (?) and the next ten years: 800V SiC inverter in an EV package:

https://insideevs.com/news/371247/delphi-800v-sic-inverter/

Such a nice, tidy, little grey box. Does it need liquid cooling? No visible signs of it in the picture.

I've done the prep bits for my 10kW DIY power stage (for instance evaluation of the alu pcbs in the precharge circuit thread) .

It takes a little bit of programming and testing before I can move on to 100kW+.



And there's no sign of SiC inverters on the scrap yard yet.
 

·
Registered
Joined
·
23 Posts
i've designed SiC inverters, the first one was in 2007 with IGBT + SiC diode, was 540Vdc, 50kw for motor drive application. that was expensive back then. i had access to a fully military qualified lab (bench test, EMI chambers, shake & bake, pcb assembly, you name it, we had it). and a near infinite budget too....

going high in switching frequency, higher than 21kHz, is only useful in certain applications. for your typical motor drive in an electric vehicle.. with no AC output filter .. what's the point? standard Si can do 21kHz no problem, be it mosfet or igbt. if you go high in switching frequecy you will generate a lot of low frequency harmonics which creates a new problem you were not anticipating and takes away from the advantages you were seeking. yes you can solve this in firmware but you lose DC bus utilization. i solved it in hardware (patented) without lose of bus utilization which was nice.

personally, for typical EV application, i'd rather have lower loses that SiC offers and thus allow more amps through to get the same average temperature. you're still going to need a heatsink either way.

gate driver is no problem, and also desaturation protection in the standard way is no problem IF you chose your gate driver IC carefully and spend time on the bench to tune it.

so at the end of the day.. you spend 3x the $ to get SiC mosfet which is LESS reliable than Si IGBT (bad for EV application) and you get what benefit? slightly smaller packaging... that's cool. but is it ultimately really needed in an EV? you have lots of room there...

$0.02
 

·
Registered
Joined
·
730 Posts
Discussion Starter #105 (Edited)
Yeah, 2007. Great year. IGBT or SiC mosfet?
I never liked IGBTs with their tail bump, high losses, big snubbers (high dV/dt at turn-off, IGBTs have a built-in serial killer).

There are big advantages attached to an output filter if you're going to use a "standard" (DIY) ACIM.

Why 50 kHz: much smaller filter, low phase inductances of low voltage high power motors, less wear and tear on insulation and bearings with a filter.
Not so long ago, a member here had some serious issues with a low switching frequency in a Tesla drivetrain.

EMI, climate chamber: hey, this is a DIY forum. Let's aim for compliance by design.

What do the pros indicate? Well, Tesla has already moved on to SiC mosfets in the model III. Or Musk has given the green light for it.

VW is ahead of Tesla on the voltage: Porsche Taycan 800V, higher phase inductances, lower I^2R losses, faster charging ...

Five years later since my first post, maybe 20 kHz is OK now for 800V.





BACK to programming.

Wow, this bit banging of a dedicated uC works fast.
No overhead of schedulers, interproces pipes, semaphores and issues with latency.
Certainly less than 10 pages of code (with comments).
 

·
Registered
Joined
·
730 Posts
Discussion Starter #107
No more details about my projects from now on. New privacy policy here.

KiCAD has improved. Version 5 is more user friendly.

IGBT and DC are still dominant here but really: it is technology from the EV-1 period.


No worries, I've taken you back there in a few threads. It has been a lot of fun.
 

·
Registered
Joined
·
730 Posts
Discussion Starter #108 (Edited)
INEXPENSIVE motor controller basics. BOGS style.

1. No PID and no torque control. Why not? For a PID and torque control you have to set control and motor parameters correctly.
That approach creates a lot of overhead, especially in software, when a controller is supposed to be "universal".
Motor parameters not set correctly? BOOM! or best case scenario: OVERCURRENT FAULT.
In my motor controller designs the driver is part of the control loop.
Have a look at my hardware DC and ACIM controller circuits.
Just hit the throttle and away you go. That is DIY IMHO.
No need to set a dozen parameters (or more) via a web interface or even worse: a serial terminal.

You are the driver and you provide the feedback path for speed and acceleration.

So 2. No need for a web interface or serial interface (WiFi, RS232: whooaaa!, really?)
Throttle stuff etc. is handled in the dedicated I/O controller.
Those system parameters (like type of throttle) can be set via a touch display unit.
Very easy to get that up and running with a cheap duino board.
And so here's:

3. CANBUS topology. Car system design basics. Implies many dedicated inexpensive controller circuits.
You can even hook up reverse engineered auxiliary units from the scrapyard. Charger units for instance.
ARM processors? Digital signal processors? (Embedded) programmers love them.
But hey, this is not IT (WHAT????! USB and ethernet in a motor controller??, you're kidding, right?) project
and not an industrial automation project (so no CANopen).

4. This is a SiC project, but it is possible to use those old 600A+ IGBT bricks again.
Recycling is great. But you do need a different gate driver and cooling system design.
 

·
Registered
Joined
·
730 Posts
Discussion Starter #109
A SCRAPYARD build with Tesla parts has become very popular recently.

Here is a teaser image of the ACIM controller hardware that is needed for a Tesla model S front or rear drive unit
as an inexpensive controller board replacement.

This board is the ACIM version of my inexpensive DC controller design, but with a little bit extra: CANBUS!
A tiny 8 bit controller has all the hardware stuff on board, but the Tesla drive units do have an awful lot of NTC thermistors.
So one extra piece of hardware has been addded: an analog multiplexer (US$ 0,50).

- full hardware (hardwired) desaturation power stage protection
- full (power) overload protection provided by a single current sensor on DC bus.
No interface hardware needed, no special construction:
inexpensive LEM sensor directly connected to uC via simple passive network.
- full status information via very inexpensive duino based CANBUS display module

I'll let you know where you can find the details as soon as possible,
but I want to be the first to build it. So no details yet!
 

Attachments

·
Registered
Joined
·
730 Posts
Discussion Starter #110
Cascode SiC JFETS are now available. They are short circuit rated for eight microseconds. The build can start. Please see update in first post.
 

·
Registered
Joined
·
255 Posts
Hi Tony and Steve and anyone else active on this thread. This thread seems to be one of the few alive on this forum on controllers. I am having difficulty finding a commercial controller for the Emrax 188/208 PMAC. I need 250V and about 125A continuous. Since this is a multirotor airplane application weight is critical. I can only find heavy EV controllers rated at 400V 200A cont, and weighing 8 kg or more.

Would you recommend a DIY homebrew for this project? If so, where do I start? Any pointers or advice is greatly appreciated.

Than you so much.
 

·
Registered
Joined
·
317 Posts
What's the gate driver du jour, given this thread's links on the ones bandied about for SiC have expired?

And, of course, any further details on the ACIM's & instrument panel generic's designs? Squarematics (sic) would be nice.

thanks!
 

·
Registered
Joined
·
730 Posts
Discussion Starter #114
@Stevem08: The working isolation voltage can be found in the datasheet.
@Solarsail: I'd recommend my design. Coming soon. Suitable SiC devices are available and there is progress on the controller HW and software front.
Not just another homebrew inverter (I'm a BSEE system designer), but part of a CANBUS conversion kit.
Images of the first KiCAD renderings of the MCP ATMEGA32M1 and TI ISO1050DW CANdriver prototype boards will be posted very soon in my "inexpensive AC controller" thread.
OK, both threads are sort of merging, so here they are in the attachments.

First image shows KiCAD rendering of the prototype PCB for an isolated Rapberry Pi CANBUS interface.
Based on MCP2515 and ISO1050DW. The big DIP is actually a SSOP to DIP piggyback PCB for the MCP2515.

The interface circuit uses the GPIO hardware PWM0 as clock signal for the isolated power supply.
The second image shows the wiringpi gpio utility based shell script that starts the power supply.

The third image show the KiCAD rendering of the ATMEGA32M1 prototype PCB. The 32M1 is on a TFQP32 to DIP piggyback PCB in the centre.
These boards are very important for all my projects. They add (isolated) CANBUS connectivity and uC control.

Both boards have been built and are up and running. Wireshark is used on the Pi to capture CAN communication.

@remy_martian: The cascode JFETS should work just fine with the current gate driver designs.
Actually, even much better: they have a fourth kelvin pin and much they are easier to drive (higher gate resistor values).

raspberrypi_isolated_canbus.png wiringpi_script_turns_iso_power_on.png dev_board_ac_drive.png
 

·
Registered
Joined
·
255 Posts
I'd recommend my design. Coming soon. Suitable SiC devices are available and there is progress on the controller HW and software front.
Sounds great. Can't wait to learn about your project. I'll check your thread.

600V would be nice but 420Vmax is good too. 150Acont I would need. 200Acont even better. Water cooling is OK. Supply is 28V. Resolver/encoder good too.

Running 4 rotors 40kW each. So need 4 of everything.

Thanks
 

·
Registered
Joined
·
317 Posts
Nobody makes 420V transistors. 600V, then 1200V.

Lots of repurposing of salvage inverters these days, up to 600V which can be had for the cost of the power transistors themselves, or less, and they have the caps and cooling circuit.

Has the world blown past this homebrew effort, as with many of us left holding the bag with $5,000 traction motors when you can buy a motor now for $1,000 with gearbox?

Maybe you should move up to 1200V devices and enable up to 800V packs? That is something that's rare and worth building from scratch, imo.
 

·
Registered
Joined
·
730 Posts
Discussion Starter #117
Well, my contributions here have been all about evaluating circuits for DIY conversions. For my personal use.
And having some fun doing it. Blowing up a few mosfets, trying out active filters in an all hardware design are a few examples.
A thyristor based precharge circuit is nothing more than just a bit of "back to the 70s" fun.

But 800V is definitely the way to go. And CANbus of course. There's no way around CAN. Proven reliable communication is essential. As many DIY salvage yard visitors have found out.

My system design: CANbus backbone with multiple cheap and easy to program automotive grade micro-controllers. I have selected the ATMEGA32M1.
I'm not going to do FOC for ACIMs, but it is feasible with three ATMEGA32M1 devices. Now I need two. One handles the low frequency stuff ( for instance thermistors).
The other one generates the high frequency gate drive signals. The KiCAD schematic is ready. A rendering of the prototype PCB can be found here in a couple of weeks.

In my other "inexpensive ACIM controller" thread there's an image of the two uC PCB for a popular scrapyard ACIM drivetrain.
 

·
Registered
Joined
·
730 Posts
Discussion Starter #118
Dev PCB layout powestage controller is ready. Rendering is in the attachment. Software: just finished customizing Optiboot for CAN flashing of the atmega32m1 using avrdude.
@Solarsail. 400V means: 40kW is like a "walk in the park". Supports for encoder up to 32 pulses per rev per channel (SW framework ready). Water cooled.
@remy_martian: Basic config for the instrument panel: 2.8" touchscreens.The VMA412 has a -40 to 85C temp range. Wireless link to a tablet is possible.
uc_dev_board_powerstage.png
 
101 - 118 of 118 Posts
Top