make plastic 3D models from CAD designs using an automated "milling machine"

A good way to make 3D plastic models easily.

Step 1 Buy stuff

you'll need:

-- an EGX-300 engraver machine (we will use it for milling)
(http://www.rolanddga.com/products/engravers/EGX-300/) I saw it on ebay for $2000 at some point. If you buy from Roland, costs $5000 unless you're using it for educational purposes (then it's cheaper)

-- a 3D CAD program that can make .3dm files (I use Rhino 3D)

-- VisualMill software program (http://www.mecsoft.com/Mec/)

-- a computer with a serial or parallel port (to send instructions to the EGX-300)

-- a vacum cleaner

-- milling bits (and drill bits if you need to drill holes too)

-- delrin plastic

-- a hacksaw to slightly modify a piece of the EGX-300

-- mechanical timer

-- glue gun

Comments

[smoothie]

Roland DG EGX-300: Alarm over questionable resolution claims No, this is not confusion of software/mechanical resolution figures. I understand those perfectly clearly. The EGX-300 printed and Japan-website specifications both claim: "Software resolution 0.01 mm (0.00394in.)/step or 0.025 mm (0.000984 in.) /step (XY axes only)" and "Mechanical resolution X, Y and Z-axis: 0.00125 mm (0.0000492 in.) /step (micro-step control)". Typographical errors are not the reason for my alarm. I have assumed that "0.00394 in." is a typographical error and "0.000394 in." was intended. However, the smallest mechanical step I can achieve using the X or Y axes under any circumstance is 0.01375mm (which is significantly larger than the claimed resolutions). I have tried issuing feed commands (constant speed), traverse commands (with acceleration and deceleration) and single steps under software control and I have also used the keypad on the machine. The observed movements have been validated under a microscope. No, the discrepancy is not caused by backlash on the X or Y axis drives, as they are both tensioned-wire drives with zero backlash. The motions I obtain are validated both when moving unidirectionally and when reversing the direction of motion It appears that the mechanical resolution is not eight times higher than the software resolution as claimed, but is actually 1.375 times lower than the software resolution! The claimed software resolution of 0.01mm is achieved by firmware interpolation. This can easily be verified using the keypad on the top of the machine. Each key-press for the X or Y axes increments or decrements the position display by 0.01mm, but only 8 out of every 11 key-presses cause any movement of the axes! For my work, which relies on the claimed 10µm step size (the reason I purchased the Roland) I have increased the mechanical resolution to match the software resolution, by reducing the diameter of the aluminum capstan pulleys so that each whole motor step corresponds to 0.01mm of linear motion. Then, by using a software algorithm to compensate for firmware interpolation ‘steps’ which do not cause any movement of the axes, I privately manage the phantom steps. It’s a very complex matter, which is not cheap or easy to do. No, I can’t simply send a program to another user, which will “fix” his own software. Your software has NO IDEA which 8 of the 11 steps to believe, or which 3 of 11 to ignore! Also, I have some reservations about the Roland brochure claim of "micro-stepping control". I can see from the board and the Roland service manual schematics that it is implemented in hardware, but I have made measurements under many circumstances and I don't think it's actually used. Even if 1/8 micro-stepping was used, it would only increase the actual mechanical resolution by 8 times to 1.71875µm, so it still would not meet the claimed mechanical resolution of 1.25µm. Perhaps the design engineer at Roland DG Japan could not get micro-stepping to work, was afraid to tell his boss, and devised the cheating technique of moving only 8 steps for every 11 commanded steps? Any EGX-300 owner can validate one aspect of my alarm, just using the position display and the motion keypad. Each key-press for the X or Y axes increments or decrements the position display by 0.01mm, but only 8 out of every 11 key-presses cause any physical movement of the axes! If I am correct, most US states have laws that require a buy-back. Just imagine advertising a “5-speed manual” car fitted with a gear-shift that has 5 slots, but 2 of them are extra neutral positions and there are only 3 gear ratios!

[bopeep]

I have a belly button.

[eapprenti]

Some similar features are shown here, eapprentice
they don't provide everything but good stuff as well.

[TheMadScientist]

the premise of "modifying" my fancy 5k piece of tech with a hack saw, it sounds frightening.

[arzie2000]

it wouldn't be "frightening" if you really know what you're doing. I've modified my own EGX300 & PNC2300A, i've changed the metal table to a solid T6 Aluminum table with T-slots. and the spindle holder from bended metal sheet to a solid T6, the bearings of the Z-stepper motors to a sealed type. Maybe someday, I'll be modifying the whole body. The whole casing is thin and when you're engraving, it makes a lot of sound! I think changing it to a solid cast steel might do the trick. but i'm worried that the stepper motors might not be strong enough to move the "heavy" cast steel.

[alan2525]

That sounds like an interesting modification. I've got an EGX-300 and have been thinking of doing something to the spindle and the mounting to something more rigid. Do you have any photos of your own conversion? Alan

[RaisedByRobots]

cool. personal 3d printers can't be far off now... I wish kinko's could do 3d printing, that would be awesome...

[smoothie]

nelrdoamus: It's not friction drive that allows homing to hard stops, it's current sensing of the drive motor circuit as it hits a hard stop. The current in the Vcc line to the motors is monitored and anytime it rises above a set threshold, the circuitry says "Oooh I must have hit the stop, or something else must have got in the way while I was homing". If you look inside EGX-300 and other Roland wire-drive machines like MDX-15, you will see the stops are sheared sheetmetal striking across sheared sheetmetal (!!!) Each time you home, there will be a small change in the edge shape of the sheetmetal stop, so your home position (a) will drift slowly and (b) there is nothing you can do about it. This is not a comprehensive reply to all your queries, just bits I can help with. Re the buffer size you need, this depends on the part program length, which in turn depends on the elegance of the coding language and the skill of the programming software (or a living, breathing programmer!. Roland RML-1 is NOT an efficient language. It has no loop/repeat functions, it cannot switch on external devices like coolant/airblow/vise, and in their engravers there is no facility for tool offsetting, so your part dimension will depend on the accuracy of the tool dimension. If you break your 1/8" tool and want to substitute a 3mm tool, you need to compose a fresh part program. Engravers do good TOILET and EXIT signs.

[nelrdoamus]

Very interesting discussion. I've been running an EGX-300 for about a year making prototype robotics parts. Maintaining dimensions within 0.001 inches has not been a problem, though I have taken to frequent homing of the machine due to apparent "drift" discussed above. I am surprised to hear that the unit uses friction drives as I plunge cut with a 1/4-inch diameter 90 degree V-cutter for some part features. The depth increment is only 0.2 mm and the XY feed rate is kept at 5 mm/sec--however, at approx. 2.5 mm depth, a fair amount of material is being removed and the tone of the machine lets you know it. The material is ABS plastic. Couple questions and comments: 1)Does the friction drive allow homing against hard stops? I've found the home position to be perfectly repeatable. Homing against a hard stop would be difficult to impossible with a lead screw. 2)Questionable resolution claims aside (a pretty serious issue)I would be interested in opinions on a more cost effective machine for a)the approx. EGX-300 work envelope, b) ability to upmill/downmill and c)ability to buffer 1 MB and process files, thereby freeing up my computer for other tasks. 3)Is the manufacturer of the controller board visible--I've been trying to find a cost-effective embedded controller so I can make my own milling machines (the Baldor NextMove ST--model no. NST001-501 seems a good candidate though I'm having a hard time getting pricing info on it).

[smoothie]

Roland DG EGX-300: Alarm over questionable resolution claims No, this is not confusion of software/mechanical resolution figures. I understand those perfectly clearly. The EGX-300 printed and Japan-website specifications both claim: "Software resolution 0.01 mm (0.00394in.)/step or 0.025 mm (0.000984 in.) /step (XY axes only)" and "Mechanical resolution X, Y and Z-axis: 0.00125 mm (0.0000492 in.) /step (micro-step control)". Typographical errors are not the reason for my alarm. I have assumed that "0.00394 in." is a typographical error and "0.000394 in." was intended. However, the smallest mechanical step I can achieve using the X or Y axes under any circumstance is 0.01375mm (which is significantly larger than the claimed resolutions). I have tried issuing feed commands (constant speed), traverse commands (with acceleration and deceleration) and single steps under software control and I have also used the keypad on the machine. The observed movements have been validated under a microscope. No, the discrepancy is not caused by backlash on the X or Y axis drives, as they are both tensioned-wire drives with zero backlash. The motions I obtain are validated both when moving unidirectionally and when reversing the direction of motion It appears that the mechanical resolution is not eight times higher than the software resolution as claimed, but is actually 1.375 times lower than the software resolution! The claimed software resolution of 0.01mm is achieved by firmware interpolation. This can easily be verified using the keypad on the top of the machine. Each key-press for the X or Y axes increments or decrements the position display by 0.01mm, but only 8 out of every 11 key-presses cause any movement of the axes! For my work, which relies on the claimed 10µm step size (the reason I purchased the Roland) I have increased the mechanical resolution to match the software resolution, by reducing the diameter of the aluminum capstan pulleys so that each whole motor step corresponds to 0.01mm of linear motion. Then, by using a software algorithm to compensate for firmware interpolation ‘steps’ which do not cause any movement of the axes, I privately manage the phantom steps. It’s a very complex matter, which is not cheap or easy to do. No, I can’t simply send a program to another user, which will “fix” his own software. Your software has NO IDEA which 8 of the 11 steps to believe, or which 3 of 11 to ignore! Also, I have some reservations about the Roland brochure claim of "micro-stepping control". I can see from the board and the Roland service manual schematics that it is implemented in hardware, but I have made measurements under many circumstances and I don't think it's actually used. Even if 1/8 micro-stepping was used, it would only increase the actual mechanical resolution by 8 times to 1.71875µm, so it still would not meet the claimed mechanical resolution of 1.25µm. Perhaps the design engineer at Roland DG Japan could not get micro-stepping to work, was afraid to tell his boss, and devised the cheating technique of moving only 8 steps for every 11 commanded steps? Any EGX-300 owner can validate one aspect of my alarm, just using the position display and the motion keypad. Each key-press for the X or Y axes increments or decrements the position display by 0.01mm, but only 8 out of every 11 key-presses cause any physical movement of the axes! If I am correct, most US states have laws that require a buy-back. Just imagine advertising a “5-speed manual” car fitted with a gear-shift that has 5 slots, but 2 of them are extra neutral positions and there are only 3 gear ratios!

[smoothie]

Nearly forgot! EGX-300 has a wire drive, running on friction drums. If you do very long programs there is always a little slip between the wire and the drum, so it gradually forgets where it is. There are no position encoders so you can come back after a 3-hour milling job and find the EGX-300 is several millimeters away from where the program thinks it is. Very sad when that happens (always) - it is important to realize this is an engraver, designed to make EXIT signs and TOILET signs etc, which it does quite well. But a precision mill? No no no no ....

[leevonk]

I milled out a large number of very accurately cut designs. So has another lab down the hall with their own machine. my designs usually don't take very long to mill, but some of my longer jobs have ended in weirdness (so you might have a point about the build up of error over time due to no encoderS).

[smoothie]

Sorry, this machine claims to move in 10-micron steps but it does not! It moves in just-under 13-micron steps and omits every 7th and 11th step. The total distance of motion is about right but the vectors are jagged! If you are looking for smooth vectors, fine resolution and high precision, push a fully-insured EGX-300 out a high window and buy another brand!

[kinetic_elite]

The timer is a nice touch. How long is the process?

[leevonk]

I make really small stuff (the design usually fits within a 1 inch by 1 inch square of plastic), so usually takes under an hour. The thing that takes the most time is when the machine has to shave off a lot of plastic if the plastic sheet you're using is significantly thicker than your project is tall. One thing I forgot to mention about this machine is that it has very fine resolution and precision.

[brk]

This is a handy tutorial. I bought an EGX-300 a few months ago, but haven't done as much with it as I would like. How have you dealt with it's minimal z-axis travel? The 1" travel seems like it could be a limiting factor at time.

[leevonk]

I mostly make really small things with the machine (electrode holders, etc). But when I do need something taller, I just make two seperate pieces (top and bottom half) and then spot melt the pieces together at the seems with a soldering iron (I use delrin plastic). You could use 'zap gel' to glue together other materials (look it up, it's -awesome- stuff: http://zap.supergluecorp.com/pt26.html)

[cartertools]

Great example, I'll be adding a link to my pages.

It can be ultimately cheaper to buy a mill and software than to send work out, if you plan on making stuff for many years...

There is a large learning curve, but learning is fun!

I have a similar page showing the general process here:
http://www.cartertools.com/3Dpath.html
(Rhino>Bobcad/cam>Mach3>Taig CNC mill>part)

I wish Visualmill wasn't so spendy though...I use Bobcad and hate it. Software can add up!

[syntaxinflux]

Interesting, but why not use a e-machine shop or make friends with a local machine shop? For 4K (price on Roland's website), you could buy a used benchtop milling machine from the various machine shop auctionhouses that has horsepower or a lot of orders to a machine shop.

[leevonk]

you're right, this isn't for everyone cause it's too expensive for what most people would use it for. Our lab needed the capability of rapidly prototyping different designs for different projects (milling out most designs takes less than an hour), also it gives us the flexibility of cheaply producing and selling the designs we mill out to other labs.

[leevonk]

I just looked up an emachine shop (http://www.emachineshop.com/) and saw that they charge $140 for ten, 1 inch square, aluminum milled pieces. This would mean that you could get 28 orders before it added up to the price of the educational priced EGX ($4000 as far as I remember). So think about that before you get an EGX :) For some people this will be cost effective, for other not.

[Beanwaur]

nice, the only problem i can forsee with this would be the cost of the machine and the need to cut it up right away to use it for milling

[leevonk]

the part you cut is on a piece of plastic attached to the vacuum hose adapter, a new hose adapter can be bought very cheaply. Or you could just keep the piece you cut off (just a small piece of clear plastic) and glue it back on later if you want to (all it does is extend the vacuum hose thing closer to the plastic surface).