Speed controller design (Based on arlo's Game Changer)

trialspower2

100 W
Joined
Dec 31, 2016
Messages
108
I am in the process of trying to design a speed controller to run a 25kW brushless motor (http://alienpowersystem.com/shop/brushless-motors/120100sw-sensored-outrunner-brushless-motor-50kv-25000w-water-cooled) I have already had a previous attempt which did not go very well, while it did run the motor I had various issues. I therefore went back to the drawing board and did some research. I game across arlo's game changer speed controller on this forum and liked the design. But as the controller has to fit on a bike it needed to be more compact. So I have put together my own version which has a water cooled core.

I have not done the PCB layouts as yet, but it will be a dspic controller with three UCC21520 Ti FET drivers (4A source, 6A sink) with isolated power supplies for the high side rather than a bootstrap. I will be using 36 x IPP110N20N3 88A FET's which have a low gate charge and reverse transfer capacitance. I am looking at switch at a minimum of 100KHz and run upto 100 volts.



capacitor pcb end_03.jpg
capacitor pcb end_04.jpg
The main hub section which bolts onto the zero volts plate to form a hollow section for the water to pass through in the middle
capacitor pcb end_13.jpg
capacitor pcb end_14.jpg
Hub section in place on the zero volts plate
capacitor pcb end_11.jpg
capacitor pcb end_12.jpg
Low size PCB board with 18 FET's, gate leg will hopefully reach to the control PCB which is at the other side of the zero volts plate. The PCB is bolted against the zero volt plate for the FET source.

capacitor pcb end_05.jpg
capacitor pcb end_06.jpg
Phase plates, split into three sections for each phase. These are bolted to the hub using a layer of heat conducting insulation and plastic screws. The FET heatsink (Drain) bolts directly onto these.

capacitor pcb end_07.jpg
View attachment 6
Next we have the high side PCB, this is bolted onto the three phase plates for the FET source. The gate leg is connected to the control PCB via an extra link.

View attachment 5
capacitor pcb end_10.jpg
Next we have the positive plate bolted to the hub again on insulation with plastic screws. The FET drains are bolted onto the rear of this plate.

View attachment 12
capacitor pcb end_01.jpg
Finally we have the capacitor end showing the connection point for the +Volt lead in on the centre plate at the left (between the caps) and the -volt lead to the right hand side of the centre. This PCB connects to the end of the hub (zero volts) and to the Positive plate on spacers.

On the underside of this capacitor PCB I have allowed for 9 x 12 uf polymer capacitors, as I believe these have a very low ESR which might help things? With 50mm high capacitors I think I can fit around 36000uF.

Hopefully I have explained this well and would be interested to hear any comments/ suggestions. I want to try and get things right this time from the start.

Thanks
Danny
 
Its very good to have a symmetrical layout, but you want to keep DC+ and DC- as close to each other as possible. If you have + and - separated, you have a current loop, which translates into inductance, and inductance is what kills mosfets because of the turn off overshoot.
Even if you have great caps with low ESR and low ESL, if you have a suboptimal path between the capacitors and mosfets, you will have inductance and overshoot.

Maybe you can move the - plate from the bottom to the top, so + and - sit next to each other.
 
Hello Marcos,

Thank you for your comments.

I could move the - plate to the space between the capacitor board and the + plate. I would then have to find a way of connecting the low side source fets back to this. One option would be to run then all on individual bolts coming back to the - plate. Do you think this would improve the performance?

the other option would be to put the + and - plates in the middle of the FET's, but this would mean splitting up the phase plates, so we have 3 low side plates and 3 high side plates. These would then have to be linked together by the motor wiring running down the outside.

I guess at the end of the day any arrangement has its pro's and con's its just knowing which things are the most important and designing around them.

Thanks
Danny
 
Just come up with a different configuration which I think might be better. Its slightly larger in diameter (140mm) but shorter in length.

no1.JPG
This time we have an inner positive plate with an outer negative plate around it. The capacitors mount directly behind them

no2.JPG
This shows the inner and outer circuit boards


no3.JPG
Finally we have the phase plates, allowing the wires to exit through the middle at the capacitor end. Not really thought how water cooling would be best implemented to this as yet. The control pcb would go in front of the phase plate with link wires passing through the phase plates to the FET gates. I would look to have a complete ground plane on this board as much as possible.

Any thoughts?
 
Should be incredibly good. Controllers only can scale up to the limit of the devices ability to share current, so whatever it takes to get device packaging geometry to scale is the key to unlocking EV performance.
 
I was planning on getting them laser cut out of 3mm aluminium, they can cut out holes down to 2.5mm diameter without any issues out of 3mm plate. The place we use for laser cutting can also cut copper, so this would be an option.
 
I guess I should ask where you're based out of. Pending location (US?). I'd be interested in sponsoring some machine work for these parts when it comes time. I own a CNC manufacturing firm and have offered CNC work on a few projects around here and am doing some design work (very very slowly, unfortunately) for Paul's AC inverters.
 
Thank you for the offer coleastering, that would be very helpful as the water cooling section in the middle will need to be CNC machined. Unfortunately though I am from the UK, and yes this is unfortunate when undertaking projects as some things which you can get all over the US, you cant get here in the UK. One example of this is 4130 tubing!

Hopefully I will get chance this weekend to finish the speed controller drawing, I have thought of a couple of minor adjustments I want to make.
 
Just a small update, I have tried to get everything coming off the capacitor side with a hole through the controller for the hall sensor wire (will be shielded) and a water pipe. To make my bike design compact I am thinking about putting the controller directly behind the motor and having it all enclosed in a larger cylinder. I am hoping the main PCB will be far enough away from the motor to not be effected by it.


no4.JPG

I have got a current sensor at in the middle of the capacitors with the red connection being batt+, the blue batt- I am not sure if my connection to the rear - plate is a good idea, but I cant connect it at the center as I would like. The three phases are coming out in the middle in orange.

no5.JPG
no6.JPG
 
A little update, I have thrown together a schematic for circuit board. Its a'bit all over the place, but hopefully doesn't have any mistakes on it. Any suggestions/ improvements would be appreciated. I am hoping to get the design finalized and the PCB layouts started soon.
 

Attachments

  • part2.pdf
    63.3 KB · Views: 50
  • part3.pdf
    57.6 KB · Views: 48
Its a personal taste, but I prefer reviewing schematics split in several modular sheets, because these kind of schematics tend to grow bigger when you start to figure out you need way more protection circuits.
I'm building a system similar to yours and so far its more than 20 pages long.

Few basic checks:
* do you have desaturation protection?
* how do you ensure that pwm high and pwm low wont ever be active at the same time?
* i dont see pulldowns in pwm pins. Can you ensure that the mcu while booting up will drive those pins to 0v? They usually start in hi-Z
* no protection on any pin? I usually see tvs or schottky clamps everywhere. Microchip parts used to be rugged, maybe they still are.
* the schematic is hard to read. Try to make a component for the dc/dc, I cant follow that part.

My biggest concern is that I don't see any desat protection... its your last barrier against catastrophe.
 
Cool... I'm glad others are working on this. I have been meaning to get back to it but have been crazy busy trying to get the crx ready for some 1/4 mile stuff. :)
 
Yes +1 for desat.

I will not build another controller without Desat.

I will most likely always use Isolated supplies as well and will always have UVLO as well.

Amongst other things.


Don't feel bad. Keep at it. I was there once only a couple years ago......
 
Back to the drawing board, although I would sooner get it right at this stage!

Would it be better to use a driver like the TIISO5852S which has a DESAT function built into it? along with UVLO?

With regard to the DC-DC converters, I was going to use some TMA1515S, which is a 4 pin module. I had 1 to supply all the low side drivers, and 1 for each of the high side drivers. These supplys are isolated upto a 1000V.

I will put some TVS diodes and the inputs/ outputs from the board and add pull down resistors to the pwm channels. The driver I was going to use was has a lock built into it to stop high and low being on together. I take it I am right in thinking that you don't need dead time for a BLDC application?

I was going to use a current sensor on the battery input cable connected to one of the analogue inputs.

I had started to layout the PCB which is proving very difficult due to the shape and small amount of space. I will hold off on this for now until I get all the design issues sorted.

View attachment PCB vr1.pdf
 
The ISO5852S looks good to me but I see it uses Capacitive Coupling instead of Opto Coupling and not sure if there would be any issues with that.
Not sure about the single DC/DC supply for all the Low sides, its common to have a DC/DC for each low Side.
And you will need Dead Time, If you search the forum there is lots of info about dead time. It was around 1-1/2years ago when I was playing with deadtime
on my controller and I have to look at my notes again. I do recall the Max Dead time of 5% of the PWM Period.
 
with the lebowski controller how will we apply a reset signal if the desat trips?

They have these at Element 14 for around $17AUD not bad considering the features.

Something to consider for a big power controller. if I ever finish my low power controller

Andy
 
Animalector said:
with the lebowski controller how will we apply a reset signal if the desat trips?

They have these at Element 14 for around $17AUD not bad considering the features.

Something to consider for a big power controller. if I ever finish my low power controller

Andy

Im using leboskis chip. I have the desat sending a signal to a arduino then if it trips the arduino holds the reset button on the lebowski brain while flashing the led 1 or 2 or 3 pulses to tell you which phase tripped.
 
If you want 25kw use one dc/dc per gate driver.

ISO5852S is a good chip, but be sure to follow its layout guidelines. Also, try to finish you schematic before starting layout so you dont waste time.

Aarlo's and zombiess threads are very educational, be sure to read and understand all of those hundreds of pages. When you read them you'll figure out the things you're missing, like current buffers, caps, cleareance, etc.

In the layout is a bit terrifying that I see tracks from the igbt to the microcontroller. You dont want to touch that.

Keep it up
 
trialspower2 said:
The driver I was going to use was has a lock built into it to stop high and low being on together. I take it I am right in thinking that you don't need dead time for a BLDC application?
No. You need to calculate your required deadtime based on the datasheets of every component in the signal path. MCU->traces->logic (if any)->isolator->buffer (if any). On the complete temperature range.

Otherwise you're starting a short on every pwm cycle.
 
Yes all the fastest possible on times subtracted from all the slowest possible off times.

And the brain stages needs full electrical isolation from the driver stage.
 
I was under the impression with BLDC you dont need dead time because the change between high and low for each driver is always 1 phase apart. When I had my first controller running I found the following sequences;

position A
phase 1 +
phase 3 -
Position B
Phase 1 +
Phase 2 -
Position C
Phase 2 -
Phase 3 +
Position D
Phase 1 -
Phase 3 +
Position E
Phase 1 -
Phase 2 +
Position F
Phase 2 +
Phase 3-

Then back to the start. I thought because the same phase didnt change from high to low( or low to high) in consecutive events dead time was not an issue.

My board has limited space, so I am struggling to keep the true isolation barrier. Do you think this is an issue when running up to 100V?
 
Back
Top