L. M. Watts cnc router

CNC PAGE

Our biggest project has been to design and shop build a large format moving gantry cnc mill for wood

and metal. We have previously built moving table laser carving machines and high- speed voice coil servo manipulators.This page will detail some of the construction issues for a commercial duty mill fabricated in

a small shop environment on a tight budget.

There are a lot of pictures and diagrams here. I tried to make the files as small as practicable but
it still may take a while to load them. Please be patient!

The machine with the way covers removed

The machine was designed using Autocad for drafting and Grape for Finite element analysis of the

structure. Construction is a space frame design using A36 steel structural tubing and hot rolled 1018

MIG welded. The ways are 40 mm high profile THK linear ball bearing type bolted to precision scraped mounting surfaces. Ballscrew drive is used with a dual screw for the heavy x axis. Hiwin preloaded 25x5mm ballscrews are used throughout. The screws are cam and software error mapped.

The drive is dc brush servo (SEM MT40H4-44) driven by Copley 412 amplifiers. The e-stop system
is a hardwired custom design. The controller is the NIST Enhanced machine controller running in Linux
output to a Servo to Go ISA pc card. The control pc is a 200 MHz K6.

Cam corrected leadsrew pitch

The first detail I will overview is the cam matched twin ballscrews driving the 1000 lb gantry.
Such a structure really must be driven from both sides to prevent racking. Since the ballscrews
are rolled to an E300 of .05 mm (about 0.003"/ft lead error) and driven by a jackshaft with
spiral bevel gear transmissions a mechanical pitch matching system was used.

The way I did this was to borrow from a very old technique used in the early 19th century to make
very accurate screw pitches for lathes (Maudsley?). The traveling nut is mounted in a thrust bearing
that allows some limited rotary motion as it moves along. This rotation enables it to add or subtract
a little from its position to compensate for errors. An arm with a ball bearing follower is spring
loaded against a linear cam that has been profiled by mapping the screw errors. Here is a sketch
ofthe design:

Only one of the twin screws is set up with the moving nut. It is matched against the master screw on
the other side (which is itself software mapped). The resulting mismatch after correction was 0.0005"
maximum over 60" at exactly 25 degrees c. This is 12 times better than the original screw specs!
Here is a picture of the slave screw:

The spring loaded rotating nut engaging the linear cam

Ballscrew end Fixity

The ballscrew ends are terminated in shop designed and built preloaded angular contact bearing units.
These can be had commercially for about $300 each (ouch!) so we chose to do a run of them and
buy them with shop sweat equity. The units use DB configured ABEC-1 7205 40 degree angular
contact bearings in a precision ground housing. One for each screw is fixed; the other can move axially
to allow for thermal expansion while still resisting bending moments.

The ballscrew ends are shop machined with conventional locknut threads and keyway. This is a demanding
operation because the ballscrew lands are not always good mechanical datums. Also,the units are usually
casehardened. We take off most of the hardened skin with a hand angle grinder, clean up and thread
with carbide on the lathe, and finally cylindrical grind to plus or minus 0.00015" TIR.

Here is a picture of one of the bearings on the Y axis along with the US Digital encoder, Ace constant
force adjustable shock absorber, and limit switch:

Y axis end bearing, shock, encoder, and limit switch

Way covers

I don't particularly like bellows type way covers. They cost waay too much, use up precious
axis travel, and wear out. But ball bearing slides and screws must be kept very clean to have a long life.
On this machine I used neoprene impregnated nylon cloth material in a window shade arrangement. The rollers are kept in tension by a "traveling spring" block and tackle system. The spring moves along at half the axis rate and as such is not unduly stretched even though it is small. The cables wrap around small drums on the cover takeup roller shafts. It is simple and cheap.
Here is a view of the system for X (cover removed):

Traveling spring way cover tensioning system

Motors

This machine uses the SEM MT30H4-44 dc brush motor for x and y. A medium inertia servo (0.18 oz in s^2), the SEM is a good mechanical impedance match to a gantry or vertical mill. The maximum torque is
over 1400 in oz and it has a built in tach for good slow speed tracking in velocity mode. I use these
for many projects.

For the z axis a smaller Magmotor dc brush servo is used.

SEM HT40 servo directly driving the 250 lb y+z

Tandem drive transmission

To save a control axis and give robustness a mechanical jackshaft drives the two tandem x axis
screws through shop designed and built spiral bevel right angle gearboxes with pressurized lubrication.
The gear maximum backlash translates to only 0.0001" of linear travel. The transmission uses
Boston Gear spiral bevel 1:1 units on 7203, 7205, and 6203 bearings in preloaded thermally
compensated DB and triplex configuration. These mount on ground angle plates:

X axis spiral bevel low backlash gearbox

Control system

This machine is using the free NIST Enhanced Machine Controller with a Servo to Go II i/o card.
The software runs under Linux with a special real time kernel on an ordinary pc.

The Emergency stop system is hardwired and not under computer control. Here is a low res
schematic; it is not fully readable but a higher res .dwg can be emailed.

The system uses limit end emergency switches in series (normally closed) to actuate a latching
5v control relay which in turn actuates a high current relay that will disconnect the servos and
short their windings through a low value resistor for self braking. The spindle is disconnected
as well. An arm switch and a limit over ride (to allow jogging off the limit switch) is provided.
Not shown is the diode and Transorb voltage protection devices across the motors and relay contacts.
Three front panel center zero meters show current or voltage of the servo amp outputs.
Another addition planned but not shown is a logic level fault signal to the i/o card to inform
EMC of the estop event. Right now EMC indicates a following error if the system is actuated that
disables the amps and stops program execution anyway, but a direct fault signal would be better.

As I am new to EMC and Linux I have tried to document my experiences with the system. The software
was installed on the computer by using Paul Corner's "Brain Dead" Install CD 2.11 that automates
the sometimes formidable task of getting Linux, the realtime kernel, and compiled EMC on the computer.
I can report that the process was very easy and pretty much worked right out of the box. I did find a
minor bug in the EMCMOT file that we think is now fixed and will be distributed on the cd soon.

I think That EMC is a viable fairly high performance control system. Over the years I typically used
Tech80 LM628 based cards and wrote my code in c. The EMC can easily outperform the LM628
in most areas.

The main task is to write the proper parameters in a text file (emc.ini) to tell EMC about the machine.
The copy I set up can be found here. This is just as I use it with the Servo To Go II except the pid
gains... I have remmed out the ones I use and put a safe low value in for others to keep life from
getting too exciting when they power up!