150KW VESC

What purpose this VESC has been designed for ?

Can't tell., but I hope it can also run a car

Did you have to adjust the deadtime to deal with the IGBT switching at lower speeds?

Yes, I calculated 1.2us but its running 1.4us. The original vesc is something like 70ns.

Benjamin exposed the deadtime setting, line 204 of the configuration file of this board

https://github.com/vedderb/bldc/blob/master/hwconf/hw_palta.h

I wrote a .c code to calculate the actual register value, its not a simple calculation and a wrong dt setting can blow the inverter, so I sent it to benjamin a while ago, hopefully he can integrate the code in the GUI. Ping me if you need that code.

This board also has some hardware fault protections in case anything goes wrong, so it also needed some custom code to unlatch the protections if they are triggered

Joke apart, be careful with the trademark now.

Yes, no problem on my side if vesc requires me to change the name, I had to build this with an impossible deadline so the name was the least of my concerns at that time. However, I do wish this is as official and vanilla as possible, I spent a lot of time doing things the right way so the hardware is fully supported by a vanilla code.

I mostly provide engineering support, this niche stuff doesn't sell in quantities.

How do you receive email notifications from this forum?

I'm gonna leave this here

https://hackaday.com/2017/09/11/open-source-high-power-ev-motor-controller/

I found ChibiOS quite easy to deal with. What is HARD is the motor control part of the code, I barely had the courage to tweak minor things in benjamin's code to suit my needs. A single misplaced C code line can wreak havoc and destroy a very expensive system.

Having said so, I'd encourage you to stay away (far, far away) from a MCU port. Your code base will branch away from benjamin's and you won't be able to keep up with his changes and improvements. Its not easy to get it running to begin with, but maintaining and keeping track of benjamin will become impossible.

Thats why my only chance to approach a digital sigma delta is using an external FPGA, thus avoiding MCU changes, and integrating security and fault tolerance into the FPGA. Such an expensive drive unit can't blow up with a firmare bug, thats why the FPGA.

Therry asked for the code, I'll post it here when I get home.

Here it is, you set the  deadtime_ns variable to your needs. VESC operates by default at 168mhz.

It will print the register value you need to set in the firmware

#include <stdio.h>
#include <stdint.h>

int main()
{

float deadtime_ns = 1400;
float clock_freq_mhz = 168;
uint8_t DTG = 0;
float timebase = 1/clock_freq_mhz*1000;

printf ("timebase_ns = %.4f\n", timebase);


if (deadtime_ns <= (timebase * 127) )
    DTG = deadtime_ns / timebase;
else
{
    if (deadtime_ns <= ((63+64)*2*timebase) )
    {
        DTG = deadtime_ns / (2*timebase) - 64;
        DTG |= 0x80;
    }
    else
    {
        if (deadtime_ns <= ((31+32)*8*timebase) )
        {
            DTG = deadtime_ns / (8*timebase) - 32;
            DTG |= 0xC0;
        }
        else
        {
            if (deadtime_ns <= ((31+32)*16*timebase) )
            {
                DTG = deadtime_ns / (16*timebase) - 32;
                DTG |= 0xE0;
            }
            else
            {
                printf("Cannot represent this deadtime setting");
            }
        }
    }
}

printf ("DTS[7:0] = %d \n", DTG);
printf ("Or... 0x%0x\n\n",DTG);
}

I want to drive a VESCed Sportscar in 2018! Frank

Me too!

I am interested in making VESC6 firmware work with something like AMC1303. But to get 12 bit accuracy you need sample 8us. How current looks trough one pwm period on phase shunt. Is there a lot of change?

Besides accuracy you need FAST response times. You either use the AMC1301 and convert from differential to single ended like I did on this build, or you convert the delta sigma to voltage using analog filters, or you design a digital filter to get a useful adc value. I don't think the STM32F405 alone would be able to compute that delta sigma at 20MHz clock.

In the first paragraph of its description it says 80MHz cpu. Are you willing to cut your clock frecuency by a bit less than half, reconfigure *every* clock (timers, prescalers, deadtime, usb, pll, etc)?

Who knows what else is not compatible with vesc firmware, and even if chibios supports it. Usb libraries are a nightmare on their own, even when chibi's are one of the best written out there.

In my case I want full sigma delta signals, so I need 7 filters and those MCU have only 4. I think an fpga is the best choice.

Good point, I got one FPGA board from school times so maybe I'l give it a try.

Here I posted what I think it needs to be configured in the fpga. Let me know how it goes, when I get some idling time I want to try that verilog code.

https://endless-sphere.com/forums/viewtopic.php?f=30&t=89056&start=25#p1...

Wouldn't be easier putting STM32L476 and use it only for filters? 

In the first paragraph of its description it says 80MHz cpu. Are you willing to cut your clock frecuency by a bit less than half, reconfigure *every* clock (timers, prescalers, deadtime, usb, pll, etc)?

Who knows what else is not compatible with vesc firmware, and even if chibios supports it

You want use STM32f405 + FPGA for filters. I am talking about using STM32F405 + STM32L476 for filters. 

Oh, sorry, and sorry if that came out rude, I'm lazy writing on the phone.

That is a doable approach for sure and would require minimal changes to the vanilla firmware. Its especially good if you never used an FPGA, they have quite a learning curve and its more common to know how to program a mcu.

I'd still go for the FPGA/CPLD because I can process all 4 voltages, 3 currents, and monitor for PWM shoot through and several fault conditions using HARDWARE. Firmware is often not considered rugged/reliable enough for some applications. You would develop a simple and reliable HDL that makes sure nothing blows up by standing in the middle of the PWM itself (cant do that with an MCU), and it makes your main mcu firmware testing easier. Without this a single-line firmware error can destroy your build, which I suffered last year.

Hi Jens, PM me anytime, its the only chance I get a notification. By sheer luck and curiosity I checked the forum, its been like 2 weeks since I polled this thread.

Do you want to add an fpga or remove the STM32 and implement *everything* in the fpga? One is a month away, the other is more like a year away ;-) 

The benefit of the AMC isolator is the differential signaling, which provides higher noise immunity and the option to use digital signaling (changing part#).

Other than that, yes, its just a current->voltage transducer like any other.

I'll leave this here and walk away

Sorry, I dont understand. In my boards the AMC1301 has its own isolated dc/dc converter that is always powered. You sound like using bootstrapping, I'd avoid that for >10kw.