Re: experiments with EMC smoothness

["Les Watts" <leswatts-at-alltel.net>]

Hi Dave

Thanks for running the tests!

Pity you can't monitor the current as that pretty much relates to
torque.

I think you are fairly heavily loaded as far as inertia as my calculations
show the 0.18 oz in s^2 rotor inertia of the SEM is about the equivalent
of 1000 lbm load with a .2" pitch screw as I use. The flywheel would give
even more.

With my tests I used 0.001", 0.01" and 0.1" long series of moves at various
speeds.

The scope indicated that I was going to the equivalent of DEFAULT_ACCEL
current with each G01 move. Since I had it set as 30 in/s^2 that is about
5 amps or so with the SEM. That includes friction.

Could you post or e-mail me your emc.ini please?
I would particularly like to know velocities. What is smooth at low speeds
can be nasty at high. DEFAULT_ACCEL is a clamp value so lower speeds
might be less than that. But I would think hard corners would demand max
value at any speed. The pulse width would vary though.

I think the damped oscillations are due to relatively low I and D values
that you mentioned.
Is 10.16 the actual number in the .ini for the accelerations? That would be
10 times
what I am using in meters/sec^2. If you are in inches the units are
in/sec^2.

I found out some further information from the archives by the way. A long
time ago a source file was created to replace emcmot.c that did real
blending
all the time rather than a simplistic connect-the- dots thing. It is called
emcsegmot.c.
It has a preprocessor directive to #include segmentqueue.h.
It was written by a guest researcher at NIST- Rogier Blom and modified by
Fred.
 It seems that this routine was needed to run smooth blended moves in a
Mastercam
auto body carving program. I understand the original (and current?) emcmot
was too jerky
for smooth shapes like auto bodies and molds.

Apparently it worked. However, Fred described it as about "90% there". What
was the
other 10%? If the math worked it might be something simple. And it's sitting
there
still in the distro unused.

Thanks again for doing the test. Please get me your .ini so I can try to
figure at
least part of this out. Connect the dots motion planners are not good for
those
of us going at high speeds typical of wood, plastic, and plasma. Actually in
this
day and age I think it is not good for anything. We could all use a 10x
improvement
in finish quality!

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
Surplus cnc for sale:
http://www.alltel.net/~leswatts/forsale.html
----- Original Message -----
From: "Dave Engvall" <dengvall-at-charter.net>
To: "Multiple recipients of list" <emc-at-nist.gov>
Sent: Tuesday, May 20, 2003 4:00 PM
Subject: Re: experiments with EMC smoothness


>
> List:
> Les's comments on "jerkiness" caused me to do some playing with my bench
> experimental system.
>
> System:
> SEM MT30H4-44 servo motor
> SD 1525BR servo amp with a 70 v supply.
> a 2.5" dia flywheel 2.5" long mounted directly on the shaft with a helical
> 25 x 6 mm coupler
> to a heidenhain ROD 426.000-2500 encoder
> controller: ppmc with encoder and dac cards running ppmc_dionull in place
of
> ppmc_dio
>
> This system is not very loaded so the settings are fairly agressive.
> Emc:
>
> P = 1020
> I = 0.11
> D = .1
> FF0 = 0.0
> FF1 = 0.0
> FF2= 0.0
> Input scale = 40000 0 simulating a .25" pitch ball screw
> max and default accel = 10.16 m/sec**2
>
>
> With a short program that makes successive 1" moves...from x = 0 to x = 10
> and then a single return to zero move.
>
> G1 X 1 F n00
> G1 X 2
> ........
> G1 X 0
> M2
>
> If one logs this move using drive voltage vs. position (since there is no
> acceleration vs position) then
> at low velocities the voltage curve swings in a dampened oscillation from
> +10  V to -7.5 V on the first swing and then dampens out to about 4 V +/-
1
> V all the way to 10" where one sees the same thing for the reversing of
the
> axis and then again as it decelerates approaching zero. Following error
> maxes at about 0.005" and decays about the same as the drive signal.
>
> With a single move to 10" or n = 1 to 3 there is curve that looks just
like
> the slow speed moves. Only when I tried n = 4; therefore 400 ipm which is
> max speed for the system do I start to see small dampened oscillations in
> the drive voltage at the start of each new block of code. The oscillations
> are about twice the voltage swing for a move in equilibrium.
>
> I was expecting much larger transitions but at least for my system they
> don't happen.
>
> Results from high speed systems with real loads to push should be
> interesting.
>
> OK guys it's open season!
>
> Good luck!
>
> Dave Engvall
>
>
>
>
>
>
>
>
>