small Servo (DC-Brushed-Motor + Slider-Pot) needs control

[aaaaamartin]

Dear readers,

I have been trying for weeks with no acceptable results to
control a small servo, I screwed together.

I have 3.219d / 16F877 /
Circuit is L293D H-Bridge connected to 2 Pins of PIC + PIC's PWM.

Servo is a small 12 V DC-Brushed Motor coupled to a Slider Potentiometer with 7 cm travel and a gear. The motor turns 15 times along the whole distance.

I'd very much like to see this thing positioning at 7 bit resulution with an
error +-2 or even 0.


Some things I've tried so far:

PID: directly working on the position error, no movement profile or what
so ever. Hard to implement, and results were not good.

TRAPEZOIDAL: just rising up PWM until half the move is done then down again, when tuned a bit, works very well, can achieve zero error, but works sluggishly and overshot can be large.

One of my probs is that I don't have a clue, how PID and a movement-profile do interact. I've read a lot so far, but still no idea.
Until now I'd explain it like this:
The movement profile always estimates the point where the motor is to be decallerated. PID therfore adjusts the output to stay within the limits of the profile.
You see, no idea.

So please could someone make it clear.

The other thing, which way to go for a working loop ?
What can I expect with this kind of Servo, will it ever work ?

I really hope someone can help.

Very unhappy Martin, thanks

[Mark]

Use a stepper motor or you'll need some sort of positioning sensor, optical encoder, ect...

[Eugeneo]

I've built a simular app based on a H bridge and a DC motor. Instead of a sliding pot, it used a worm gear reduction and a rotational conductive plastic pot. I first used a PID loop. It worked, but not as well as I expected. Then I used just a PI loop and started to get better.

I finally just went with a simple look-up table that was based on the distance from the target(your analog sliding pot) vs. voltage(PWM percent). This worked really well for its simplicity. I was going to put a PID correction on top of this, but it worked so well even under load or no load, it wasn't needed.

I figured that with such small moving parts with so little inertia, the response for a dc motor is so fast that a PID would only be good when used as a minor correction or compensation due to load or no load.

This project was built around a windowed PIC16C74 so it only had an 8 bit a/d resolution with very good results. It hits the target everytime.

So I would suggest starting with a very simple table. You could even extrapolate the values in between if your using the 10 bit converter.

[Ttelmah]

Email me directly, with a return address ([email protected]), and I'll send you the code for a PID servo system I did a while ago.
This is a PID control loop, that is fed with a trapezoidal motion 'profile', based on an optical quadrature encoder (you will need to replace this code with the ADC code - remember that your servo 'loop' must not stop for the ADC reading, so you will have to use the 'read_adc(ADC_START_ONLY)' in one loop of the servo, and then a few loops latter take the reading with 'read_adc(ADC_READ_ONLY)', otherwise the ADC will interfere with the servo loop - this may be part of the problem you were having with PID). Also the servo factors will need to be tweaked for your motor/inertia. You don't say how the system is told 'where' to go to?. The code I have accepts a serial command for position, or to move by single steps etc..

Best Wishes

[sseidman]

[Mark]

[aaaaamartin]

Hello,

thank You all for quick suggestions.