Re: [CAD_CAM_EDM_DRO] PWM signal to step/direction signals

On Wed, 9 Oct 2002, Alan Marconett KM6VV wrote:

> Hi David, Bailey,
>
> I might be mistaken, but I think we're missing something here. The PWM
> signal that can drive a servo motor does NOT have a direct 1:1
> relationship to the distance (or velocity) desired. The PWM signal is
> changed to control acceleration/deceleration, and to meet load
> requirements. A "cheap" or inefficient motor will require more "juice"
> (current) to achieve the same goals as an efficient one. Hence, the PWM
> signal will be DIFFERENT for these two motors to accomplish the same
> task. The steppers move a finite distance for a finite number of steps,
> and will rotate at a finite rate for a finite step rate.
>
> Otherwise, yeah, a PIC would be a good way to do it!

Alan,

I hear what you are saying, and I think I've abused the term "Pulse Width
Modulation" in my previous response.

I believe that Howard was referring to the PWM-ish output from a common RC
servo controller. This signal is really just a communication to another
closed loop system, namely the servo itself. It's not a true PWM waveform
in the common sense. In an RC's "PWM" output, the apparent duty cycle
isn't important; it's really the ratio of the width of the active pulse
with respect to some ideal width.... not necessarily the *actual* period
that one would perceive in a common PWM output.

This sort of "PWM-ish" control is better named PWPS for Pulse Width
Position Servo (control).

For instance... RC servo controller's PWM output:

____ ____
_____________| |____________________| |_____


This may actually be the waveform of a servo at full 90 degrees
deflection! Even though it clearly has a duty cycle of roughly 15%.

If the pulse widths were ~2.5 mS, you'd see that RC servo deflected its
full range.

The kicker here is that the actual position information is related by
ratio of the width of the active pulse over some accepted maximum value
which *isn't* the perceived period.

The period you see here is better described as the "repetition rate",
which tends to vary between 50 and 400 Hz and really has nothing to do
with the servo's actual deflection.


I believe your PWM reference is to the more commonly used PWM controller
of i.e. open loop DC drives/solenoids/lighting levels and the like.
You're absolutely right; without any kind of positional feedback, the
system would be pretty inflexible... being useful on only one kind of
motor, at a certain temperature, load, etc...

Offhand, http://www.4qdtec.com/pwpm.html has a pretty good summary of the
differences.

You did identify a crucial set of parameters. The speed of the
steppers/system might be a problem. Namely, running them open-loop might
be too slow. It'll almost certainly be slower than your garden variety
DC, digitally-controlled servo.

I gotta' figure that the stepper's output will be more precise though.
It'll only be as good as the servo controller's PWPS output. I don't have
a good feel for just how precise *that* is, however.

-d

> David Kott wrote:
> >
> > On Tue, 8 Oct 2002, Howard Bailey wrote:
> >
> > > Anybody know of a circuit or IC that is capable of taking a PWM signal as
> > > used for RC angular servos and outputs step and direction signals with
> > > output adjustable for scaling? Basically, I would have a PWM signal for an
> > > angular servo and would like to use it to run a stepper through a gear
> > > reducer to increase resolution and torque. I couldn't find much on the web
> > > on this so any information would be greatly appreciated.
> >
> > Eyup... that would be a PIC microcontroller running some custom microcode.
> >
> > The devil's in the details.
> >
> > How fast would the stepper have to run?
> >
> > If all you want is a low-current, TTL-level step and direction signal
> > based on the PWM duty cycle of a servo controller output, it's a
> > straight-forward implementation. Then, you'd take this output, and run it
> > into your preferred stepper motor driver (Gecko, et. al.).
> >
> > Developing the software to run on an $8 microcontroller to do this is
> > about 40 man-hours worth of work. It's not dissimilar to some
> > motion-control projects I've spun in the past.
> >
> > A normal RC servo driver (bar some ringers) runs at a base frequency
> > between 50Hz and 400Hz with a typical pulse width of 0.6 mS to about 2.5
> > mS.
> >
> > This pulse width is proportional to some angular rotation between 0 and 90
> > degrees (or thereabouts).
> >
> > What is the ultimate precision you require for your application?
> >
> > 1 degree is relatively simple to implement. 0.018 degrees would be about
> > the best a 20 MHz microcontroller could generate.
> >
> > Does that meet your requirements?
> >