Solar car races

Post and discuss your VESC projects here.
lizardmech
Posts: 171
Joined: 19 Jan 2016, 10:54
Location: Australia

Solar car races

Postby lizardmech » 24 Jan 2017, 02:28

Since others were also interested in solar car racing.
This is my VESC controlled inverter that uses isolated SI8920 current sensors as well as isolated gate drivers. The highest voltage I have ran it at was 67V but in theory it would scale up to 500V or so with only minor changes.
Image

Thor
Posts: 16
Joined: 22 Jan 2017, 11:32

Re: Solar car races

Postby Thor » 24 Jan 2017, 23:19

Oh that is looking really promissing. I also did some work in this direction. Lizardmech which solar team did you do that for? A schematic would be helpfull as well.

This state can be called preliminary at best. I am just picking up work again to finish this project.
Endresultat.jpg
Endresultat.jpg (606.7 KiB) Viewed 2641 times

lizardmech
Posts: 171
Joined: 19 Jan 2016, 10:54
Location: Australia

Re: Solar car races

Postby lizardmech » 25 Jan 2017, 04:35

It wasn't for a solar car, just one of my test boards while making higher voltage VESCs. In the end 150V wasn't much use to me because 80V mosfets have much lower RDSon. It only became worthwhile doing all the isolation if you can run 300V+ through SiC or GaN mosfets. But I don't have any high voltage vehicles at the moment.

The original schematic for that one is unreadable, I rebuilt everything so it is sort of modular, I just delete the high or low voltage schematic parts and export a netlist then close the schematic without saving. The HV section isn't 100% finished, the one in the photo has analogue devices ADUM4223, which worked very nicely, however they are hard to hand solder. There are a few other things I would change, rather than bootstrapping the driver high side I would have been better of giving them isolated supplies and using the same supply for the current sense as well via a LDO. I didn't finish building a high voltage buck converter for the 12V supply yet.

You can have a look at the kicad files. The isolated one is named september and the board housing the MCU is called stm32riser. The other one Si3-2 is the latest version with reduced component count but an 80V limit. https://drive.google.com/open?id=0B9A_C ... FJPTXFoMm8

I could probably sort of merge the two designs with minimal effort I was thinking I could add 650V 60A gansytems GaNFets and try to sell them to Solar car teams, its about $300 worth of FETs but I could still make it far cheaper than those tritium controllers. If you ran 500V 20A you get 10KW or so. The VESC software needs some changes to be able to do 100khz+ PWM so I'm waiting for the newest version to come out before I make any changes.

tecnologic
Posts: 15
Joined: 26 Feb 2016, 08:23
Location: GER / Hamburg

Re: Solar car races

Postby tecnologic » 25 Jan 2017, 08:36

hi together,

if u really want to use GaN with switching freqencys of about 100kHz with an efficiency of about +99% u need more PWM resolution. Thats why i' am waiting for the STM32H7 series. These chips have internal HRPWM and delta sigma demudulators. With delta sigma modulators such as the AMC1304L05 u can use very small shunts and u can use ur isolated gate supply as powersupply for the modulator.

and next do not only take the RDson into account when searching FETs. With 500V and only 20A the conduction losses are small compared to 100kHz switching losses. Even if u except dV/dt values of more then 500V/µs u have 100kHz*20A*500V*1µs*0.5=500W of switching losses for each phase! So u are only by that at 75% efficiency. So why do u want to use such a fast switching time? Are the motors running faster then 200kerpm? I dont think so. Do u need a fast transient response of the current controller, as in CNC milling machines? i dont think so too. So whats the benefit?

regards

Tec

lizardmech
Posts: 171
Joined: 19 Jan 2016, 10:54
Location: Australia

Re: Solar car races

Postby lizardmech » 25 Jan 2017, 09:00

I'm looking at the Ti C2000 F28377 as an option. It already has an open sensorless FOC solution, all you have to do is port the motor detection portions and the supporting code for interfacing with BLDC tool and other functions. It's only 200mhz however it has hardware based logic specifically for computing math used in FOC. It has the best PWM as far s I know, 150ps resolution. The VESC observer might be better you would have to test which portions of code are worth porting.

tecnologic
Posts: 15
Joined: 26 Feb 2016, 08:23
Location: GER / Hamburg

Re: Solar car races

Postby tecnologic » 25 Jan 2017, 11:10

Uhh C2000 this arch is a piece of crap. "one byte is 16bits long" .....

200MHz would be enough. Even with an EKF as Observer u have idle time at 20kHz contol frequency. But this data is from a Cortex M7 maybe the antique C2000 without caches needs some more optimizing to achieve this.

lizardmech
Posts: 171
Joined: 19 Jan 2016, 10:54
Location: Australia

Re: Solar car races

Postby lizardmech » 25 Jan 2017, 12:42

The c2000 Delfino is mostly a DSP though, you're not supposed to use the CPU for running motors. You have the CLA and TMU subsections for that. Does the STM32H7 have high res PWM? ST have all their best motor control hardware on the stm32f3 which lacks ram, flash and cpu power.

tecnologic
Posts: 15
Joined: 26 Feb 2016, 08:23
Location: GER / Hamburg

Re: Solar car races

Postby tecnologic » 26 Jan 2017, 07:06

its my personal experience with those c2000 cores. I had to use them at work and i hate the arch. But this maybe because i am mostly working on ARM Cortex cores. And i see pure calculation power much higher than some dedicated coprocessor for which i have to adapt control code. And then comes the point of having to port this code do a different processor....

Yes STM32H743 shall have HRPWM and DFSDM and it has enough TCM to get position- and parameter observers into. The only problem is i have not seen real silicon yet except some on dev boards at a fair here in germany.

rew
Posts: 926
Joined: 25 Mar 2016, 12:29
Location: Delft, Netherlands.

Re: Solar car races

Postby rew » 30 Jan 2017, 18:23

C2000 is annoying. It is a 16-bit architecture with a few 32-bit features tacked on. The address bus is 16+8 bits. You have to know in advance what code is going to run often. You have to manually place that in some special RAM for it to excecute fast.....

lizardmech
Posts: 171
Joined: 19 Jan 2016, 10:54
Location: Australia

Re: Solar car races

Postby lizardmech » 04 Feb 2017, 12:12

rew wrote:C2000 is annoying. It is a 16-bit architecture with a few 32-bit features tacked on. The address bus is 16+8 bits. You have to know in advance what code is going to run often. You have to manually place that in some special RAM for it to excecute fast.....

I'm not sure if you would have to worry about any low level code, they provide all the software for motor control in modules all the assembly level things for the accelerators are already done for motor control. You just call them in standard C code and activate a compiler option. If you tried to use it for other purposes it might be annoying, but motor control looks easy enough.

The NXP KV5x would be ideal except they didn't bother to include sigma-delta support.


Return to “VESC Projects”

Who is online

Users browsing this forum: No registered users and 1 guest