This is a computer model of a flexible, general purpose CNC machine.
Overview:
For many years I have wanted to have a machine which I could use to
build things under computer control somewhat similar to the many
machines I worked with while employed. I partially realized this
dream when I got my 3D printer about a year ago. Due in large to
the widespread popularity of 3D printing, the components for this type
of machine have come way down in price, making it affordable to expand
beyond 3D printing into such activities as CNC (computer numerical
control) laser engraving and cutting, CNC routing, CNC vinyl cutting
using a drag knife, and a variety of other similar activities.
The common factor in all CNC operations is that they need computer
positioning in two or more axes. Many hobbyists convert their 3D
printers to be able to remove the plastic extruder and mount other
devices. This approach works with varying degrees or
success. There is essentially no load on the movements of a 3D
printer, and as a result they can be built with very lightweight
components. Operations using a laser also succeed using
relatively weak components. When trying to drive a router or a
milling cutter however, the loads can be very heavy, and the weak
structure of a 3D printer just will not hold up.
I decided to take advantage of the super-low prices available and
design my own desktop CNC machine.
Design
Considerations:
I had some general requirements in mind when I started this project and
quickly added more. Some of the major items on my list were:
- It should be sized to fit on a small table or desktop.
- It should have a workspace of about 8 x 10 inches by several
inches of height.
- It should run from a low cost computer or controller such as
a Raspberry Pi or Arduino single board computer.
- It can be made using wood, MDF, or similar, but the design
must support possible changing to aluminum or other metal later with a
minimum of
modification.
- It should be as strong as practical using low cost materials.
- It should readily accept a variety of work heads.
- It should support easy modification
- It should be self contained, but could possibly require an
external computer or laptop.
Of course there are a multitude of additional finer details which
will be dealt with as required.
The first task was to decide what the basic construction should
be. Should it be a gantry machine where the entire mechanism
moves back and forth over a stationary work table, with upper
portions moving at right angles and up and down?
Should I make a split axis machine where the work table moves in one
direction, while above it a smaller assembly moves in the other,
supporting
a mechanism which then moves up and down?
After sketching several possible configurations, I decided that the
simplest and most rigid configuration would be a split axis
machine. The next step: model it on the computer!
My
Computer Model:
I fairly quickly decided on the overall dimensions and basic
construction. I then started modeling it in Fusion 360, my
favorite 3D CAD and CAM program. I decided that I would model all
the basic structure, all the tables or carriages, and all the linear
rails. I decided NOT to model simple, fairly obvious things like
the motor/belt drives and axis end indicators (endstops). I did
model the Z axis motor/leadscrew assembly as it had to pass through
some pretty crowded real estate. Parts like the power supply and
the control box were either not modeled or modeled after the actual
construction.
This is the final result of modeling what I now call CNC-RSM. The
computer model has been maintained to show
all actual frame, table, and linear drive dimensions. The 3 axes
have all been set up so they can be moved within the
model along their respective bearings with realistic travel
distances. This allows virtually moving all parts throughout
their range of motion to check for clearances, travel distances, and
possible interference.
This is the XZ plate. It is probably the most complicated and
congested part of the machine. It is the X axis and moves
left and right along the 12 mm linear rails driven by a lug
belt. In addition to the X bushings, the plate mounts the Z
bushings
which are spaced out from the plate so the 8 mm Z rods will clear
the X rods. On the Z axis, the linear rods move
up and down with the Z axis mounting plate (turned off in this image
for visibility) along with the driving motor and leadscrew.
The leadscrew nut is fixed in height to the XZ plate.
I was able to output a detailed drawing of the XZ plate from Fusion to
make it easier to build.
Electronics:
At the same time I was doing my computer modeling, I was also
evaluating control systems. Out of the many options available,
there are two which seem to make the most sense:
1. An Arduino driving a Ramps 1.4 (or
later) shield running Marlin software for the operating
system.
In Arduino terms, a shield is a custom board
which plugs directly into the Arduino.
The Ramps 1.4 board is used by a large percentage of current 3D
printers, and plugs directly into the Arduino (Mega version)
and provides all the inputs for endstops, an LCD display, and an SD
card reader along with the outputs for up to 5 stepper motors and the
LCD display. The Arduino runs a copy of Marlin software, which
was developed for 3D printer use as open source software. This
software is VERY configurable (and very complex!), The source
code not only has the provisions for many options such as automatic
board leveling, but also specific code sections for many, many specific
machines and can be adapted
to just about
whatever configuration machine you have. This combination is a popular
approach for small CNC machines and results in a machine which
is totally stand alone, but which can be also be connected to a
computer for its control.
2. An Arduino driving a
GRBL shield running GRBL
software as the
operating system
GRBL is another open source software package which was developed
and tailored
especially for CNC use. It is a much simpler and less flexible
package than Marlin since basic CNC machines typically do not have the
degree of complexity found in 3D printers. GRBL fans cannot see
why anyone would set up a CNC
machine using Marlin, but those who do say it will work just as well -
and besides, they are already familiar with it after using it on their
3D printers.
I have decided to start with option 1 since I am familiar with that
approach. For less than $6 in parts I can change over to GRBL and
try that
out also, but my original system will use the Ramps 1.4 board and
Marlin software.
As I received the electronics, I was able to set them up on my desktop
to try out. First I set up a single motor, then when I figured
that out, I did all 3.
I was now ready to start making parts! Between my shop and
my 3D printer I started cranking out pieces which quickly went together
to create my basic machine.
Building
the Machine
I started with a melamine clad panel, cut the base and two supports to
size, then edged the cut edges with iron on strips.In machining the
supports, I positioned the X
bar holes and took a light cut across the bottom in the same setup to
assure accuracy. The side supports fasten to the base with
through screws using a large washer plate on the
bottom of the base with the screws threading into the round nut
cylinders set into the brackets. This combination seemed to give
good accuracy and rigidity.
The right picture shows my box of mechanical parts from Amazon and Ebay
waiting to be added.
The main table which moves in the Y axis rides on 12 mm fully supported
rails and open bushings.
The XZ plated which moves in the X axis is shown on the left with only
the X bushings and the lead screw nut and mount. The right shows
the completed plate. I decided to leave this plate
made from a tempered MDF hardboard, but I did later paint it.
Here is the machine with all three axes assembled. The Y table
which mounts the work piece and the Z table which
mounts the tool are both temporary made of 1/4 inch hard board.
Both will be replaced before completion.
I now started designing and 3D printing many, many pieces. These
included the motor mounts, pulley mounts, belt mounts,
endstop mounts,
endstop flags, housings, and various spacers, nut plates, etc.
This picture is only a sampling!
Axis
drives and accessories:
I decided to use belt drives for the X and Y axes, and a leadscrew
drive on the Z axis. The stepper motors I started with are NEMA
17 motors, which are quite small at about 1 5/8 inches square and about
1 1/2 long plus the shaft. I used a 20 tooth pulley on each
driving a lug belt which formed a loop, with one side driving the axis.
The Z axis is driven by a leadscrew directly driven by a stepper
motor. This was done to prevent the Z axis and whatever tooling
is attached to it from dropping when the motor power is off. I
mis-judged this on the original leadscrew as once I added any weight to
the Z table, it dropped slowly until reaching the end of its
travel. I then changed the screw from one with an *8 mm lead (the
amount it moves in one revolution) to a 2 mm lead. I don't think
this shallower screw angle will back drive no matter the load!
I met my design goal axis travel of 8 X, 10 Y and several inches of
Z. The final unit has approximately 8.3 inches of X travel,
10.5 of Y, and 4 1/4 of Z.
I initially used optical sensors with a black printed flag that
would interrupt the sensor when the axis was at the end of its
travel. These sensors were used for homing the machine and
killing the power if the axis drove too far in the direction with the
sensor. I put sensors only at the 0 end of X and Y, but both ends
of Z.
*Note: I refer to the leadscrew's lead instead of its more
common term of pitch, as there is a difference here. The pitch is
the distance between adjacent threads, while the lead is the distance
the thread will progress in 1 turn. Usually these are the same,
but
my first leadsccrew had 4 independent threads spiraling up its length
(called 4 starts) which gave it a pitch of 2 mm, but a lead of 8.
Controls
and Wiring:
Now that all the axes have drives and endstops, it is time to wire it
together with the control boards. The control consists of an
Arduino computer board and a Ramps 1.4 shield that plugs into the top
of the Arduino. This shield then has sockets for the stepper
motor drivers and connection points for all the motors and limits,
along with many features used for 3D printing that are not used
here. I designed and printed a box to mount these boards along
with a cover that mounts an LCD display and SD card reader. I
then ran the wiring between this control box and all the electrical
parts on the machine.
I printed a control box which mounts the Arduino and the Ramps 1.4
shield in the base and the display panel
in the lid.
The display and front panel controls allow stand alone operation.
There is an SD card slot on the left side of the cover and a USB port
on the right side which allows operation while
tethered to a host computer. The front panel of the control
box has a small key shaped cutout which when pressed operates the reset
button inside. The small grill openings allow
sound from the beeper to be heard, and the knob is a press and
turn resolver which controls all the menus and operations.
The power supply is mounted to the side support using a 3D printed
adapter plate.
To do simple testing I printed an adapter which holds a pencil or pen
with a slight spring load downward, allowing the machine to
print patterns or text. As one of my first tests, I hand coded a
gcode file to print out "HELLO WORLD". It worked great!
As I learned how, I started using the proper tools to create other
files of patterns and phrases.
The machine is fully operational at this point. It will home all
the axes, then run a program entered as gcode. This is a format
used by automatic machines since their beginning where simple human
readable lines of text move the machine point by point. There
are often hundreds of thousands of lines of gcode in a single program.
Initial
evaluation:
As I started using and testing the machine, I discovered
several
weakness of trying to use the complex Marlin 3D printing software to do
the less complex job of running the CNC machine. The Marlin
system is necessarily more complex than I need as it has to support
several additional functions for 3D printing. These include
controlling the feed of the extruder, verifying the filament is still
present, monitoring and controlling temperatures of the extruder
and the bed, and several more subtle functions. It is also set up
with a massive selection of the exact equipment for it to
control. For the simple CNC machine I am building, I just need
control for the 3 primary axes and a small amount of control for
accessories.
- Even though the program is very well organized and well
commented, I was never able to get the Z axis motor to run consistently
the way I wanted it to. Initially I was unable to get the speeds
I wanted until I discovered an additional parameter where values needed
to be changed. I then thought I had it under control until I made
a fairly benign change and it "blew up". It would run,
stall out, then sound like a whooping siren as it rapidly changed speed
repeatedly. I then had to struggle to find values where it would
again run smoothly. I had none of these problems with X or Y.
- I was bothered by the fact that the operating system, Marlin
supports only one coordinate system. Most machine tool systems
have at least two: One is the machine coordinate system, usually
with the home position at or near the limit of travel of the X and Y
axes, and either all the
way down or (usually) all the way up for Z. It is important for
the machine to always know where it is within its own limits of
movement.
Software in most machines will not let it be commanded outside its
usable travel. The machine coordinates are the only way to do
this. The other system is the work coordinate system where the
zero is usually a prominent feature on the part being made.
This is always needed as the part designer never knows what machine
will be used to build his part, and it is important to be able to
locate the part anywhere on the machine. Sometimes there are
several other systems used to define the locations of certain fixed
vises or fixtures on a machine. Marlin fans claim there are
work-arounds, but they seem clumsy and non-intuitive to me.
- The way Marlin is designed, any changes in the machine
constants such as travel length of the axes, maximum speed,
acceleration, and a number of other factors that need constant tweaking
as you are setting it up, require editing the source code, re-compiling
the code, then uploading it to the Arduino. This process is well
automated with the Arduino special software, but it is quite time
consuming with a program the size of Marlin.
I ran into a couple other minor problems, but the 3 above are the
major ones.
I then started to study the GRBL system. It appeares to be a
much simpler system, and was designed to be consistent with the
automated machines in industry. GRBL shields for the Arduino vary
in price from about $4 to about $150. I realized that I could try
the GRBL system for almost no cost at all. (I thought!)
I change
to using the GRBL system after minimal testing:
I decided to give GRBL a try. I bought 2 GRBL shields for about
$7 (that was the way they were
offered). They accept the same stepper drivers as I already had
in the Ramps 1.4 board. I figured that I could change the system
over to GRBL in about an hour's time. I was almost right!
This is the GRBL shield. I plugged in the 3 driver modules from
the Marlin setup and was ready to go. I did add the
short row of pins on the right to supply +5v to the endstop boards
which were somewhat more elaborate than this shield
was designed to use. This shield really has almost no active
circuitry on it. It is just a very convenient way to hook up
all the various wires and parts to the Arduino. You can
accomplish the exact same functions if you just do a lot of
jumper connecting to the Arduino.
In a little over an hour I had the new GRBL shield in place, connected
to all the motors and endstops, and loaded with the GRBL
software. I powered up the machine and all 3 axes worked,
and
the Z axis worked over a wide range of speeds just the way is should!
However, the endstops were wonky (is that a word?). I discovered
that GRBL has inputs for two endstops per axis, but it only has one
circuit each with the two inputs wired in parallel. I am using 2
for the Z axis and having the two optical sensors wired in parallel
does not work. I tried several wiring changes, but they did not
work either. I groaned and decided that I could change the
two Z endstops from optical to microswitch sensors fairly easily and
did so. Now I discovered that the polarity had to be set
differently between the two types of endstops, and the GRBL only has
one setting that changes all of them at once. I now changed the X
and Y to microswitches also. I now discovered that the way these
particular microswitch endstops are wired on their small mounting
boards pulls the
common lead to either +5 volts or to ground. This is a good noise
reducing technique, but when you have two switches in parallel, the
power supply is shorted out if one switch is activated (and of course
the other will not be). I made a simple wiring change on the
board (I cut the normally closed lead off the switch) and now it all
works fine! (Whew!)
The opto-coupler endstops that I originally used are my favorites, but
complications changing to GRBL forced me to replace them with the
standard microswitch type.
Now after the above changes, everything is working. I found a
gcode sending program (the computer operating panel for running the
machine) that I like and learned to use it. In GRBL, all the
machine settings that required a re-compile/upload in Marlin are
available on the computer screen. There are 34 parameters, known
as the $ codes, which
once set, remain set through power on off cycles. These are all
set by assigning a value to the $ code, for example by typing $132=110,
you would set the Z axis maximum
travel to 110.000 mm. Typing $$ displays a list of all 34
settings. I did a lot of adjusting of these $ codes and really
like the performance I can get.
I tried running some of the text printing programs I used for the
Marlin testing, and with some minor changes in the gcode they worked
great. I am not happy about having to use a separate computer to
control the machine, but it does allow better control and visibility of
what I am doing. I also generated a couple new programs
designed to run on GRBL initially, and they run just fine. In
spite of the separate computer needed, I think I will stay with GRBL!
Drive
changes:
Along with my change to GRBL, I decided that the small steppers I
started with would be marginal when I have the machine loaded with
heavy tooling and workpieces. I changed from a 12 volt motor
power supply to a 24 volt one and increased the size of
the stepper motors for the X and Y axes. To drive these I
replaced the tiny low
current stepper drivers with stand alone units with a higher current
capacity. The Z axis has a
very high ratio of motor to axis motion, and therefore should be
fine with the original stepper and driver. I also noticed some
jitter in some of the lines during my initial testing and attributed it
to a dynamic belt stretching. I changed the belts on the X and Y
axes from 6
mm wide to 10 mm wide belting and pulleys. This should stiffen
the drive up considerably. Since I had to mount the external
driver assemblies somewhere, I redesigned the control box to make it
larger which also gave me room for several controls on the front
panel. The larger box also eliminated the need for a separate box
I had previously made just to hold excess wire lengths. My policy
has
been to use any supplied cables as is, unless they are too short, in
which case I will extend them. Overly long cables merely have the
excess coiled up somewhere.
The photo on the left shows the comparison between the tiny original
stepper drivers and the new stand alone units. The larger control
box on the right mounts two of the
larger stepper drivers in the rear, along with the Arduino controller
and the GRBL shield (including one small driver) inside. The
large emergency stop switch is standard
issue for automatic machines in industry. It looks out of place,
but is easy to quickly hit if something goes wrong. The large
stepper drivers contain a couple of status LEDs
inside the top of the box. To be able to see these, I pressed a
couple pieces of plexiglass into printed openings in the control box
lid. These transmit the lights very well
to the outside.
The laser
arrives!
About this time the laser which was ordered from China arrived. I
quickly printed a mount for it and installed it. I had been doing
research on how to control a laser and was ready to try it. It
works great! I ran a couple of brief tests as soon as the laser
arrived, but knew that I would have to address the smoke and fume
issues. Just about a 1 minute run really smelled up my shop and
made it uncomfortable to breath. That was the end of testing
until I had a ventilation system!
The laser is mounted and ready to go to work! This was my test
image, designed to correctly label this part. It takes about 45
seconds
to complete this pattern. I ran tests on paper and on fiberboard
trying many combinations of speeds and power settings. This was
the
culmination of my testing (actually after the venting was done.)
Venting it:
I was not going to run the laser after my initial test run until I
installed a suitable venting system. Home Depot had a dryer vent
kit and was the basis for my design. This and a fan should give
suitable performance. I found a Muffin fan, which is a 115 volt
more powerful version of what has become know as a computer fan.
I 3D printed a couple transition pieces to adapt dryer vent hose to the
size of the fan. I installed the vent through the outside wall
under my work bench and connected it all up. Just laying the hose
behind the laser a few inches away worked wonders. I could barely
smell any burning at all, and there was no noxious smoke hanging over
my bench as there was before this system.
I then designed a very convoluted pickup nozzle and connected it with
a length of RV sewer hose (new out of the package) to a printed
transition piece to connect the 3 inch sewer hose to the 4 inch dryer
vent hose and I had a working system.
The fan is sandwiched between 2 printed transitions to connect it to
the vent hose. On the right it is connected to the vent and the
hose leading up to the machine. The vent comes
out on my patio with a flat plate containing 3 louvers that open when
the fan is running.
The pickup nozzle was a challenge to design, but it came out
great. It attaches just behind and under the laser as shown on
the right. The rear of the nozzle is held up by a wire coming
down from the top of the Z table and supports the weight of the hose.
There was a problem trying to keep the hose off the rear of the build
plate. It was solved by printing a hoop which slips on the front
of the shelf. This holds the hose is a position where
it doesn't contact the machine, and has enough play to accommodate the
full X and Z travels of the machine.
Additional
AC power control: (May 2019)
With the additional accessories to the CNC machine such as a separate
computer, a laser or routing spindle, and a fan, I needed additional
switches and outlets. I decided to print a small box to attach to
the top of the control box and place the required switches in it.
I also needed to print a larger base box for the power supply with
enough space for the outlets. The power supply box would have
been impossible to service using the same concept as the original one,
using only the narrow end opening to wire all the internal
components,so it was designed to have a 2 sided cover to attach to the
3 sided
base box. The base half contains the master switch, the power
input connector, and 3 AC outlets. The control top box contains 3
switches. All the switches are illuminated rocker switches.
The power supply bottom box is connected to the switch box with a 5
wire harness
The added switches certainly make it more convenient to operate the
system. The MOTORS switch controls power to the 24
volt power supply, the LASER and FAN
control their
respective outlets, and the computer can plug into a
dedicated, always-on outlet. The master switch on the power
supply controls everything but the computer.
On the right is the almost final configuration of the CNC
machine.The "almost" is because I have a piece of aluminum jig
plate on order to use for the table.
The MDF table just is not flat enough for precision work.
Mods
to Power Control (July 2019)
I have noticed a recurring problem with my new switches and power
control. Often when I switch one of the switches, MOTORS, LASER,
or FAN, the system goes into alarm mode. This shuts me down until
I do an alarm reset and usually requires me to re-home the
machine. Obviously, the switching transients cause enough
disturbance of some logic circuits to do that. It is usually in
the form of an "out of bounds" alarm resulting from an erroneous limit
switch trigger. I can avoid the problem by turning off the hard
limits function. I am not willing to do this long term, as it
leaves me vulnerable to crashing the machine in the case of an error in
my gcode, or if I neglect to reset the zero in some special cases.
To solve the problem, in initially considered putting a number of
ferrite cores around various wires to try and suppress the
surges. This is pretty much a hit and miss proposition, although
my targets are narrowed considerably by the fact that it is the hard
limits causing a problem. Instead of trying to supress the
surges, I decided to eliminate them in the first place. I figured
that if I use solid state relays (SSRs) which use zero crossing
technology, switching only as the AC power passes through the zero
voltage point, there should be no transients generated.
Doing an online search, I found a series of boards available at a very
reasonable price which contain 2, 4, or 8 small solid state
relays. These do truly switch only at zero crossing and are
capable of switching up to 2 amps AC each with only a logic level
input. I purchased a 4 relay board and totally rewired both the
small switch box on top of my control unit and the extension on the
rear of the power supply where I added the outlets for the laser and
the fan, and controlled input power to the power supply itself.
To implement this change, I needed to rewire the switches to control
only very low current 5 volt signals. These activate the solid
state relays located near the outlets and power supply which in turn
switch the same items the switches used to switch directly at 115 volt
full current. I also decided to provide a totally separate 5 volt
power supply to drive only these switches and relays. This gives
me true isolation which I would not have if I used the existing 5 volt
logic supply. There were a couple of nitpic items to handle such
as the switch lighting. Originally the switches were selected
because the toggle lights up when the switch is on. This is very
handy to quickly check their status. The lights were tiny neon
bulbs which work only above 65 volts. In the new configuration, I
did not wire to these lights but installed a small LED above each
switch which truly do light at 5 volts.
The largest problem this change caused was that the new SSR board and
the 5 volt power supply would not fit in the existing housing, or even
one where I just lengthened it some reasonable amount. I ended up
leaving the length the same, but increasing the height and width by
about an inch each. As it still had to attach to the existing
power supply, one end is the original size then it tapers out and up at
45 degrees to the larger dimension. This gave enough room for the
added SSR board and the 5 volt power supply in its own 3d printed case
holding the board from a cell phone charger.
These changes seem to have totally solved the problem!
This is the enlarged case (without its cover) showing the 5 volt power
supply at the far left with the 4 circuit SSR under it. The 3
wires hanging out
attach to the 24 volt power supply which is switched by one of the
relays. The right view is of this box as it will be seen after
attaching it to the 24v power supply.
Here is the (semi-) finished machine. It shows the new aluminum
table, the laser and vent, and the latest switch changes.