AW: interpolator

["Till Franitza" <xfa-at-isw.uni-stuttgart.de>]

Hi Les!
You are right. The next thing to do in the interpolator is the quintic
interpolation. So it would be possible to have a limited jerk. Let me see if
i find a student that wants to work on this.
till

-----Ursprüngliche Nachricht-----
Von: emc-at-nist.gov [emc-at-nist.gov]Im Auftrag von Les Watts
Gesendet: Dienstag, 10. September 2002 04:33
An: Multiple recipients of list
Betreff: Re: interpolator



Hi Till
I am kind of doing the same thing... looking closely at cubic.c, emcmot.c,
etc.

EMC is working well for me in a commercial operation... it never locks up
or malfunctions as I use it. I do not excercise all the commands however.

But some tests I have done show that I could run even faster with much
better surface finish with an appropriate interpolator. I sometimes get
choppy
or chattery results  at certain high feeds and accelerations with a typical
cam program generated with many short linear moves. I do not consider
this a fault so much because the fact is EMC can actually run fast enough
to excite structural resonances in many machines including mine.

Sure I could turn accel down and go slow but I would rather go fast!

I understand the theory for the cubic interpolation. But if I hade  finite
bounded
jerk values (acceleration continuity at end points) I think I could get
much better surface finish at high speeds. I really don't see the effects of
the current cubic interpolator but that may be because of my cycle times-
Task and
servo update are 20:1 for me right now. I have to understand if the
interpolator runs at the task or servo update (emcmot) level or both.
It must be the latter.

For all the world it looks like my machine does G01 sequences as simple
linear point to point moves.

I guess I am talking about quintic interpolation... There has been talk in
the past
about it but I think I may have a real need. I could write it but I must
understand
the current code to be able to integrate it easily.. If it was just simple
locally
declared variables etc It might not be a big deal to do.
It is just 6 term look ahead instead of 4. I would prob want to put in a
non-linear
term to limit interpolation at certain hard corners.

So, it would involve solving

At^5 +Bt^4 +Ct^3 +Dt^2 +Et +F= x

dx/dt= 5At^4 +4Bt^3 +3Ct^2 +2Dt +E = v

d^2x/dt^2=20At^3 +12Bt^2+6Ct +2D = a

d(d^2x/dt^2)/dt= 60At^2 +24Bt +6C = jerk

for A,B,C,D,E, and F

when

x(0) , v(0), a(0), jerk(0), x(delta t) , v(delta t), a(delta t) and
jerk(delta t)
are given.

At least I think. I need to read up on efficient algorithms for fifth order
splines.

Perhaps Fred or Will could enlighten as far as the current program function.
Like you I only am following some of it.

Les






Leslie Watts
L M Watts Furniture
Tiger, Georgia USA
http://www.alltel.net/~leswatts/wattsfurniturewp.html
engineering page:
http://www.alltel.net/~leswatts/shop.html

----- Original Message -----
From: "Till Franitza" <xfa-at-isw.uni-stuttgart.de>
To: "Multiple recipients of list" <emc-at-nist.gov>
Sent: Monday, September 09, 2002 10:30 AM
Subject: interpolator


>
> Hi emc!
> I am trying to understand how the path-planning works. I think i partly
> understand it. Could it be that there is some documentation available
about
> it, the structures and how the decoder preprocesses everything and so on?
> Till
>
>
>