Paul & Sabrina's cheap DIY 144v motor controller

[MazdaMatt]

Adam, perfect. Probably cheaper than the one i was looking at, but i didn't price the one i was looking at. It was just a very quick example from that supplier. The guy at work here that gave me the company name says they are very cheap for enclosures.

[patx]

Hello !!

WOW what a great thread !!!

I am in the process of building my first EV (bmw 318i 1993), I got the ICE out and ordered my ADC FB1-4001A 72-144VDC, I was a volks mechanic 15 years ago now i am an electronic technician...

I would like to start building mine... Where do I start ?!?

Is there a release yet ?!?

Is the schematic on page 76 is the final one ?!?

Where do I get the Atmel code ?!? ( I am using AVR studio 4)

Thankx !!!

[MazdaMatt]

woah, easy there champ

Things are moving along. We're still in alpha testing, so to speak. The first revision is at the board house being etched, then the boards and components will be delivered to a friendly engineery-type who will be testing them and changing components as needed. When those three first revision boards have been tested, they will go to three lucky EV owners to test out in their already-functioning cars. If all goes well, 15 boards will be produced and paid for by 15 lucky beta testers... some pages back there was a survey asking who wants to be a beta tester (can we get a fresh link, somebody?).

Classy EV you're building PLEASE start a new thread about it and PLEASE take a million pictures and post them all. Everyone drools over the littlest details here.

[JayC]

Whew, made it to the end. Didn't want to skip ahead and miss some of the banter but I had to once I got to page 100 Sure wish I could have seen this back in Jan.

Anyway, Paul congrats you are a huge inspiration. You are fearless and your actions speak volumes. I have the following skills to add:

AVR programmer - I too have been reviewing your code (have to check the wiki for the latest and greatest) and have been impressed with your straight forward no frills approach ... I envy that your "just do it" attitude. No excuses, you just do it, and you share so very openly.

I hope to be able to contribute even though I won't be converting anything to electric in the near future (2010 Prius should be here next week ).

Jay

[JayC]

No joy on finding the code on the Wiki. Paul, if I PM my email do you mind sending it to me? Also, I implemented PI on my personal robot. Honestly I can't see why you would need a PI controller in this instance as the human is the controller. In my robot I use it to slave the two motors together so that when I want it to go straight, it will. I use the Mega128 so I could have expansion but as you already know moving from one chip to another is very easy.

Jay

[MazdaMatt]

JayC, the purpose for the PI controller is the preference to control current, rather than duty cycle. Current is a function of duty cycle and RPM, so at 0RPM a 1%duty has a HUGE current draw. At 1000RPM a 1% duty has a small current draw. Since current is directly proportional to torque, we felt that we should be relating the pedal position to current so that the control is a) more intuitive and b) you don't burn out at every stoplight.

[JayC]

Originally Posted by MazdaMatt

JayC, the purpose for the PI controller is the preference to control current, rather than duty cycle. Current is a function of duty cycle and RPM, so at 0RPM a 1%duty has a HUGE current draw. At 1000RPM a 1% duty has a small current draw. Since current is directly proportional to torque, we felt that we should be relating the pedal position to current so that the control is a) more intuitive and b) you don't burn out at every stoplight.

From my viewpoint the PI is used to servo a system response to a desired output. For instance, I want my left wheel to spin at 20rpm. That is my goal. However, my wheel sensor reads 15rpm. Thus the PI controller multiplies this "error (20-15) by the proportional constant Kp and adds the previous error + the new error*Ki (integral constant). Because you add in the previous error each time you are integrating the errors.

What you are talking about is acceleration "slewing" or ramping. You are controlling how fast the system changes based on input and limiting it. In RC they call it expo (short for exponential) control and it used to desensitize the system with small input but let you have your "burnouts" if you floor it.

Jay

[MazdaMatt]

My foot is at 20% throttle and the current sensor reads 15% of max current. It then takes the error and multiplies it by the proportional constant Kp and adds the previous error plus the new error *Ki, adding the result to the current PWM output in an effort to generate a 20% reading from the current sensor. Hense, PI control.

PI, PID, P Control does not have to specifically relate to speed or position control. You could use it in business to set prices based on buyer demand (Kp) and buyer backlash (Ki). You could use it pressure control systems with valve actuators... anything really. You take in two values, scale them, subtract them to get an error value, then modify your output to try to get that error to zero. It is a very generic term.

edit: the system would not require PI if it had a speed sensor and a prior knowledge of the relationship between speed and current per dty%.

[MazdaMatt]

Jay, i just looked up what exponential control is in an RC controller and it is not what we are doing here - it actually isn't even a "control system" in the same sense, it is just an input modifier. What we are trying to do is relate pedal position directly to current, but we don't have control over current, we only have control over PWM duty. We do not have an exponential go pedal, it is linear. Whatever % of throttle is used, is the % of CURRENT being pushed through the motor. The duty is controlled via PI to achieve the required current.

[JayC]

No expo is a modifier, a remapping of input to output values.

I'm with you on the PI for the current now. There is concern that the rate it takes for the system to go from 0 to full. As I remember from looking at previous code from Paul he is incrementing once per cycle. 16-bit PWM and it could take "too long" and give a mushy feeling.

Jay