Paul & Sabrina's cheap DIY 144v motor controller

[MisterSid]

Originally Posted by Astro

I had to look up a wiki to find out what >> did in C programming language.
(One day i will have to learn C.)
A bitwise shift. Very neat.
Sometimes readability needs to take a back seat to performance. After all that's what comments are for.

Bitwise shift is a much faster alternative to multiply/divide by factors of 2 and a good complier will usually look for places where it can switch out actions like multiply by 32, divide by 1024 etc. and replace them for you. When looking for speed improvements it is often wise to decomplie the .hex file to check though.

Paul: Feel free to email the dsPIC code at any stage.

[MPaulHolmes]

I'm confused. must check my work. hahaha. Ya, sticking with scaling by 1024 is a good idea, since it matches the A/D resolution.

[MisterSid]

Originally Posted by MPaulHolmes

I'm confused. must check my work. hahaha. Ya, sticking with scaling by 1024 is a good idea, since it matches the A/D resolution.

Sorry, did not intend to add confussion. Take home point was meant to be "Just do what you do, the compiler (should) make it faster for you."

The easier it is to read the better. I hate to think how many 100's of hours I have spent looking for minor little bugs in code the scientists and mathmaticians have made "faster" (what I term "permature optimisation").

[MPaulHolmes]

No it isnt you. I had written a dissertation about convergence rates of that method of taking the running average, and had made a mistake, so I erased all the stuff and just wrote "im confused".

You are absolutely right about premature optimization. Premature code optimization is the root of all programming evil.

[MPaulHolmes]

Rather than a specific current ramp rate in amps, I'm doing a "time to max current" sort of thing. You can pick any whole number between 1 and 8. 1 means 8 seconds to max current. 2 means 7 seconds to max current, ..., 8 means zero seconds to max current. 1 means 1 second from 0 to max motor amps, ... , 8 means 8 seconds to max motor amps.

EDIT!: THAT'S A LIE, i'M NOT DOING THAT ANYMORE!!!

Since motor amps are programmable, this seemed like a good idea. Motor amps can be anything in 1 to 1500 or so. I'm using 16 ounce copper. The 500amp controller uses 4 ounce copper, so I thought it would be OK.

By the way, I don't see a good way to keep perfect compliance with rtd explorer. For one thing, the max motor amps variable doesn't exist. I know Adam is planning on including max motor amps in his controller, so maybe it will be added to rtd. Also, the battery amp limit now needs 4 digits instead of 3. Well, in practice, there aren't too many batteries that will do over 1000 amps, but it would be a shame to have to limit what the controller could do.

I'm feeling really happy with the software rewrite. Everything feels pretty clean and understandable. The serial is just about done. Then I'll just add the eeprom write and read routines.

All the current variables in the saved data are in the units of amperes. So, no weird translations necessary to real amps from virtual amps.

[duncan]

Hi Paul

I don't like your 1 - 8 scenario
Zero may be a little tricky and 1 second is a looong time when playing on the track,
I really don't see anybody wanting more than 1 second

How about going from
Zero to 8/10ths of a second?

I would want mine with 1/10th of a second to full current,
I think 8/10th would feel like a moderately slow response

[flores]

Originally Posted by duncan

Hi Paul

I don't like your 1 - 8 scenario
Zero may be a little tricky and 1 second is a looong time when playing on the track,
I really don't see anybody wanting more than 1 second

How about going from
Zero to 8/10ths of a second?

I would want mine with 1/10th of a second to full current,
I think 8/10th would feel like a moderately slow response

I agree, why build such a powerful controller if you can't burn any rubber..

[MPaulHolmes]

It was how the loop worked out. I'll sacrifice some cycles and make the 10ths of a second (or some variation thereof) work. Still, the PI loop isn't going to converge instantly, so I would certainly try a delay of zero, since it won't really be zero.

EDIT: OK that didn't take long! It will just add a single multiply and single bit shift. This will be no problem whatsoever. Then I can do it in increments of, say, 1/32sec, or whatever. Lots of flexibility. I was in thinking jail, and needed to be free. LOL.

[flores]

by the way:

I've bought this: DC 0 200V 0 1000A Combo Meter Current Voltage Battery Monitor Charge Discharge | eBay measures voltage and amps, good value at 43$ I say. Good for monitoring the controller performance.

[MPaulHolmes]

This is how it is now. Only one extra bit shift compared to before:

Let's say we have an instantaneous throttle step of 0 to MAXIMUM. Time to maximum current is as follows (c-rr is current ramp rate):
c-rr 1 means it reaches maximum current in 16 seconds. (16/1)
c-rr 2 means it reaches maximum current in 8 seconds. (16/2)
c-rr 3 means it reaches maximum current in 5.33 seconds. (16/3)
c-rr 4 means it reaches maximum current in 4 seconds. (16/4)
c-rr 5 means it reaches maximum current in 3.2 seconds. (16/5)
c-rr 6 means it reaches maximum current in 2.7 seconds. (16/6)
c-rr 7 means it reaches maximum current in 2.3 seconds. (16/7)
c-rr 8 means it reaches maximum current in 2 seconds. (16/8)
...
c-rr 16000 means it reaches maximum current in 1mS. (16/16000)
c-rr "anything bigger than 16000" is still 1mS, since the loop is executed at 1kHz.