Home made controller for a MARS BLDC

[patallen]

I just added a second NAND gate from this century, still 4011 so not the best choice ever (a 74xx is on the go with 4ns prop delay), on the low drive so all the pwm outs are now "equalised" in time. There is a 30ns propagation delay from mcu to driver on all pwm outputs, low and high. So now the Ir driver wont know. This will just alter the whole timing from the hall signal to the final gate drive of 30ns+the hardware filtering process+ the software filtering process. Will measure that tomorrow to see how retarded it is. Shouldnt be much to worry about for now.

I also removed the pre-10 ohm resistor which seemed to reduce considerably the dt (now only 10 ohm per fets instead of 20).

Lets see how it gets to the fets...tomorrow...

This should make the whole thing more stable, plus it simplifies my code since i dont have to care anymore of the inverted logic of the low drive
well....4 less instructions...lol...

 

[patallen]

Hi, its been a while since i made a post.

I recently changed my fets again, with genuine 80NF70. I run the whole thing at 24v for now.
Changed a lot of things in my code.
Works very very great, no issue so for, no melt and plenty of fun.
For now maximum current i read is around 250amp on a step hill at full speed/max pwm. thats about 6000w. I guess its the maximum the motor can take/eat at this voltage considering its internal impedance. Not bad for a prototype...i think. I am proud of myself i would say. My son is completely nuts about his toy.

Still have some high current commutation pikes induced into the circuit at start-up from a rest, that bother the adc/current readout. i guess with a real PCB and a good ground plane it would be better. For now i do more consecutive adc sampling before assuming the current reached a predetermined limit. I temporarily patches the problem, and i am not worried too much about going over 350amp since there is 8x 98amp mosfet per arm. Shouldnt be a problem.

BUT....the thing is very noisy, due to the trapezoidal commutation mode. Still very archaic, but works damn good this way.

So, my question would be....is there a way to reduce trapez noise commutation?? (other than going full svpwm/dspic30,...)
i read somewhere something about overlap method,...

Video to come with me doing some wheelies with the tiny Willys. :wink: