Re: Higher performance systems

Les,

If you have questions or suggestions please feel free to e-mail me.
I've got data you might find interesting.

Mariss

--- In CAD_CAM_EDM_DRO@yahoogroups.com, "Leslie M. Watts"
<leswatts@a...> wrote:

> Mariss,
>
> The topic is near and dear to me as well as we are quite
> hard at work trying to implement FUNCTIONAL third order
> splining in the EMC trajectory planner.
>
> And yes I would describe the math as "a little busy"!
>
> I will get to your group and read about your system.
> If it is all digital that moving average must be
> a rectangular convolution that might be done with
> an FIR digital filter with constant coefficients-
> otherwise known as a nested loop.
>
> Your method must go beyond the second order (arc, parabola,
> or ellipse) rounding that automatically happens when
> acceleration (torque) is limited to a particular value as
> it of course always is.
>
> Integrating velocity over a finite (moving)interval would seem
> to make accel dV/dt tend to look more like the original
> unfiltered V(t) since it would be the integral of a derivative.
>
> Anyway, interesting. I will follow up on your group to
> avoid too much math gibberish here.
>
> A note also for Josh... Doing 3-d contouring can be very machine

time

> intensive so yes routers and such typically can have quite high
> feed rates and correspondingly high accelerations so corners can be
> sharp at speed. In the big machines cutting at several FEET per

second

> is common.
>
> But to use those speeds you have to have a LOT of spindle

horsepower at

> high rpms. Many of those fast machines have 25 HP+. If you are using
> a small router motor you may not need such high travel speeds

simply because

> you can't push that router through the work that fast. An exception

might be

> cutting foam or something that requires little power.
>
> High rapid speeds might not be such a big issue either if your code
> is not cutting a lot of air.
>
> You might try some manual speed/feed tests to see how fast you can

really

> go with the spindle you use. If you can go a lot faster than your
> current stepper tables then it would justify something with higher
> performance.
>
> Very high speeds and accelerations tend to be very expensive though!
>
> Les
>
> Leslie M.Watts
> L M Watts Furniture
> Tiger Georgia USA
> http://home.alltel.net/leswatts/wf.html
> Engineering page:
> http://home.alltel.net/leswatts/shop.html
> CNC surplus for sale:
> http://home.alltel.net/leswatts/forsale.html
> CNC carved signs:
> http://home.alltel.net/leswatts/signwp.html
>
>
>
>
> -----Original Message-----
> From: Mariss Freimanis [mailto:mariss92705@y...]
> Sent: Thursday, August 28, 2003 7:25 PM
> To: CAD_CAM_EDM_DRO@yahoogroups.com
> Subject: [CAD_CAM_EDM_DRO] Re: Higher performance systems
>
>
> Leslie,
>
> Ah, a topic near and dear to me just right now! The conundrum is
> vector path fidelity and a constant vector velocity are mutually
> exclusive.
>
> To accurately follow a path means you have to decelerate to a stop

at

> the end of every line segment which is irritatingly slow and noisy.
>
> You can "round" line segment nodes to keep a constant contouring
> velocity. It is a computational nightmare though, having to insert
> arcs between nodes whose start and end angles and radii depend on a
> given axis set acceleration and velocity. The "on the fly"
> trignometry and path look-ahead schemes get real ugly real fast.
>
> I have come up with a method that neatly sidesteps those problems
> entirely; it is computationally trivial and acts automatically. In
> effect, the faster you go, the "rounder" the nodes are.
>
> The method outputs a moving average of velocity, effectively
> integrating velocity with time. A step change in velocity gives a
> ramped change, a ramped velocity results in an S-shaped velocity
> profile.
>
> As you probably know the G2002 is up and running using this
> algorithm, now on 3-D moves (XYZ). The results are very pretty

indeed.

>
> Mariss