Re: Low cost Servo Controller

David Howland wrote:

> From: David Howland <dhowland@...>
>
> Peter, Bertho, Andy, Steve, John, Kirk, Jon thanks for your help. I'm tracking down what I can. Do you have a pointer to the Servo to go card?
>
> Some thoughts I had over the weekend include:
>
> To gain the advantages of accurate machining while responding to Step and Direction signals (if that is what motor controllers are presented with), and increasing the speed while keeping the resolution high, several issues come up.
>
> Stepper controllers and motors are not a favorable technology when you want to close the loop at the table (or DRO). What I have heard about is - backlash compensation - however no doubt backlash may be different at different locations because of diference in wear (so ball leed screws are best in this case).
>
> A servo controller when faced with backlash might get a little crazy trying to chance directions at a high rate of speed. So here is the idea followed by the questions. I'll start with the first question.
>
> Can the motor controller pause the step and direction data comming at the motor controller while the motor controller takes up the slack (backlash) or for any other reason which requires time to work out? If the Step and Direction data presented to the motor controller may be held in pause by the motor controller, then (the idea) the servo controller can manage the difficulty of keeping things running smoothly, even when the control proplems become more difficult for specific issues.
>
> If firmware designed to drive steppers has no ability to be suspended or paused, then it might be possible to buffer the step and direction data in RAM while the controller is taking care of operations that take longer because of backlash. As long as the controller can process most of the step and direction signals faster than they are sent to it, the controller will have room in RAM (Memory), when it takes extra time to keep precision in order. Because vibration may cause more problems than changing directions on purpose, an allowance for a certain tolerance may be desired. This might be set by settings on the controller, and where called for you could fake out the system with a tool change and change the tolerance for specfic parts of the program. If you have a really tight machine, you'd just leave it set to a very tight tolerance.
>
> So the concept is forming here of a DC Servo Controller that allows feedback from the tables or the motors and responds to Step and Direction signals as if it were a Stepper system. The two advantages are that it is faster and that it is possible to close the loop to the table mounted position data improving the accuracy by overcomming the backlash in the system. Does that concept make since?

This is all so INSANE, I just can't understand why anyone would want to do things like this.

I mean, it is something like taking the output of a car engine and coupling it to a paddle wheel that throws lead balls
to the back of the car, and then having another paddle wheel that intercepts the kinetic energy of the balls to drive
the rear wheels! Why on earth would you DO such a crazy thing?

If you want to use stepper motors, then use software that is designed for that purpose. But, if you want to
use servo motors with position feedback, why not use software that is designed from the ground up to do that?
The advantages are that resolution of the encoder and achievable velocities are pretty much whatever the machine
can deliver, or that the task requires. Most stepper software will get seriously sick if the step resolution were
equal to .00005", and you wanted to move up to 120 IPM (or 2 IPS). This requires the CPU to output 40,000
step pulses per second, and most programs that control the steppers entirely in software just won't go that
fast. With servos, it is no problem at all.

One other consideration is that noise pulses in combined logic & power systems are inevitable. A single
noise pulse every once in a while could slowly move the machine out of sync with the program, whether
stepper or a hybrid stepper program driving a stepper-simulating servo drive. The quadrature encoders
require two false pulses of overlapping width and the proper polarity to fake the encoder counter into
accepting that as an actual movement. Any single pulse just shows up as a temporary move of one
count, and then a complementary move back to the correct position. Noise on the analog signal
driving a servo amp will just cause a tiny fluctuation in velocity, and the servo system will immediately
compensate.

Anyway, people keep going ON and ON about ever more complicated schemes to make a servo
motor run with stepper-controlling programs, while I've been happily making production parts for
over a year, now, and loving every minute of it! No steppers (motors or programs) here!

Jon