Re: Universal Stepper, Formerly: Re: [CAD_CAM_EDM_DRO] Re: Brushless servo drive - Cheap Solution?????

"Paul R. Hvidston" wrote:

> Jon,
>
> I'm confused ( no surprise there ;-) ). Does your board simply output
> independent channels of step pulses, possibly with trapezoidal accel/decel?
> If so how does it handle curved paths? Hopefully I'm missing something here.

Yes. It just refines the ability of the CPU to produce pulses at extremely
fine timing resolution. (It has a bunch of other features, too, ie. encoder counters,
digital I/O, Estop, etc.) The CPU tells it how many pulses/sec to deliver on
each channel, at, say, 1000 times a second. Since the CPU is not directly
aware of the step pulses, the encoder counter will count the step pulses
produced, if there is no encoder on the motors.

The CPU handles velocity ramping, curves, etc.

> On a slightly larger topic, what must such a device be capable of doing to
> handle the capabilities of G-codes? Is it limited to linear and arc
> profiles? That's all I'm seeing in the G-codes (granted helixes are just a
> third axis in synchronized linear motion).

Way too much to do with a little bit of hardware. Use a Pentium for what
it is good for, ie. sequential programs and numerical calculation, and use
hardware for what it is good for, ie. doing things in parallel, putting out a
pulse with nano-second resolution, etc.

Yes, G-code is generally capable of describing arbitrary n-axis straight
vectors, and arcs in any of the 3 cartesian axes. (When you put in rotary
axes, it gets more complicated, and the G-code interpreter is just doing
things literally, without really understanding the actual path. (In EMC, and
other higher-end control software, layers BELOW the G-code interpreter
MAY understand the real kinematics going on, but the G-code, itself,
without a detailed description of the positioning of the axes to accompany
the G-code, don't really carry enough info to make the motions -relative
to the workpiece - completely clear.)

There are enhanced G-code dialect options that allow arcs in any plane,
NURBS surface definitions, equations representing a surface or path, etc.
All of these can be done with zillions of short vectors, but the programs can
get VERY large.

Jon