Re: Hello

Mike Romine wrote:

> From: Mike Romine <mromine@...>
>
> I just subscribed to this group and have browsed some of the postings.
> Great stuff! I had just dusted off my home made cnc about 3 weeks ago,
> and put a new stepper controller on, so I am ready for some fun. I also
> have been collecting spare machine tool parts, and have more than enough
> to build a new machine. Hence this group is right up my alley.
>
> I am fascinated by the 'NIST' group I have been reading about here. I
> want to ask a question to those familiar with this. Is the NIST group
> privately funded?

NIST is the (US) National Institute for Standards and Technology, the successor
to the National Bureau of Standards. Definitely NOT privately funded, a
US government facility under the Department of Commerce.

> Or is it like the GNU consortium? I am very
> impressed with the NIST software. So you are really able to close 3
> servo loops in real time with 100 micro second sample times over a
> parallel port in Linux? That is amazing-- especially given your are
> doing feed forward velocity and jerk control in addition to PID. I
> think I will bring some spare brushess motors and drivers home from work
> for a weekend and try this thing out. I want to learn more.

Note, that is Real-Time Linux, which is standard Linux with the real-time
patches installed.
No, it doesn't use the parallel port for motion control. The parallel port is
used in some configurations for spindle motor on clockwise/counterclockwise,
spindle brake, flood/mist coolant on/off, and things like that. A dedicated
card (currently ISA) with quadrature encoder counters and servo DACs
interfaces to the motion hardware. But, yes, 100 uS servo loops on 3
axes is quite possible. I'm running 1000 Hz on 3 axes, just because that
sounded like enough! I'm using a 100 MHz Pentium classic CPU, but the
CPU load from the servo is undetectable - no more than about 5%.
I found that feed forward compensation VASTLY improves the accuracy
during high speed moves. I used to get about .050" error when moving at
50 IPM, for instance. With just a bit of feed-forward, I get the error down
to about .002" at 50 IPM! The PID calculation is not really done
correctly, and is basically useless (it doesn't keep error history in a
rolling buffer) as everyone has discovered. I don't know what jerk control
is - do you mean backlash compensation? It does this (it's an easy
thing to do) but I haven't used it.

You probably know this, but ordinary (self-commutated) brushless
motors often don't have 4-quadrant control, which you really need for
servo work. Brushless servo motors with commutation in the servo
amps do have 4-quadrant operation, meaning that the servo amp can
extract energy from the moving load and turn it back to electrical
energy (ie. braking).

Jon