Saturday, June 17, 2006



If you scrape away all the idealism what we seem to be doing here is as simple as making a CNC positioning system that mounts a polymer extusion tool instead of a more conventional tool head.

We have a brilliant design for a room temperature polymer extruder. It wants to use filament which is not a terrifically difficult technical problem to solve. The rest of the stuff that we're doing is, when you get right down to it, nothing more than the same problems that the rest of the hobby CNC community are addressing every day in a million different ways.

Cool! :-D

How did I get here? A few days ago one of the people who occasionally comments on what we do on this blog noted that what he sees CDC machines doing most often is making parts for for other CNC machines. That thought has been rooting around in my mind like a kanker ever since.

Shortly after that in responding to a comment on one of my blog postings I did a recap of a quick quantity survey on the poplar that had been used in Godzilla. What jumped out at me is that if I had a proper workshop I could have made it out of rough poplar that cost maybe US$30 rather than the milled, expensive pieces costing maybe US$100-150 that I bought that suited my hand saw and mitre box.

What didn't hit me until this morning is that if I mount my Dremel tool with a cutting head on it on Godzilla instead of a Mk II extruder... I have a proper workshop...

Letting the other shoe drop the natural question arises that if you can replicate the structure of Godzilla out of U$30 of poplar why should you want to insist on doing it instead with maybe US$75 worth of caprolactone?

I think the melts I did in my toaster oven started me thinking in this direction. It is a simple matter to melt HDPE into quite a respectable ingot (this is even easier to do with caprolactone) whereas making the same HDPE into filament that will run in something like a Mk II is a lot more trouble. Why not mill the ingot into what you want and save the swarf instead of making everything into filament and trying to extrude everything.

Similarly, Simon, if I recall correctly, was talking about setting up a furnace capable of melting aluminum. Those aren't hard to make. Why not use aluminum in a RepRap? You can make a cupola furnace that will melt steel into ingots for not a lot more. Why not use steel in a RepRap?

After all, we do already anyway, don't we?

If you think of a RepRap as a CNC machine that has a bunch of tool heads, an extruder being only one of many, the whole concept gets a lot more flexible and interesting to me. :-)

Well, it's even more general than that actually - reprap is an arbitary 3D positioning robot...that has a whole bunch of applications - you see them all over the place doing everything from loading CD duplicating machines for small scale record producers to moving test tubes around in DNA sequencers.

However, overgeneralizing will detract from the final goal of building an ADDITIVE machining device rather than a SUBTRACTIVE one - the idea being that in principle an additive device can make absolutely any 3D shape that fits into the machine - where a subtractive milling machine can only cut shapes where the tool will fit.

It's a very fundamental difference in ultimate goal - although we should certainly recognise that there is some generality in what is being done.
I've been thinking for some time a mill is easier. It is a different machine, more limited in many ways but more precise when doing shapes suited to a mill.
As I said in another blog. I did not find HDPE melted well. I think it does need and auger or something.
To make a bubble free block you could also spin it in a crude centrifuge.
For a mill head I was thinking model aircraft parts.
I'm using a motor from for an RC project.
Plenty of power and only 27 gram.
The also have bearing tubes and shafts. It only needs some way to mount a tool bit.
You could make a milling head for plastics at less than 100 grams.
Constructive fabrication allows for much more versatility in the long run. You can make much more complex objects than with destructive fabrication. That's particularly true when you want to seamlessly mix multiple materials, such as integrating 3D electronics and so on directly into things.

Still, a CNC machine is probably a good start for constructive fabrication too. Just too expensive for most.

Also in the longer term, we would hope to make a machine that produces the raw plastic, so that will bring the cost down further than poplar. And it is more quickly renewable than poplar.
One significant design difference between the platform for an additive manufacturing machine versus a subtractive one is that in the additive machine the only forces that the positioning mechanism has to deal with is gravity acting on the tool and inertia during accelleration. With a subtractive tool you have to apply forces to get the cutting bit to push against the workpiece.

It follows that a RepRap machine that could do subractive milling would be over-engineered for additive deposition work.

So it's not true to say that you get a CNC milling machine 'for free' - you paid for it in over-engineering your deposition platform.

Of course it might turn out that this is a good trade - but we shouldn't come away with the idea that it's a total freebie.
Not to be brutal or anything, but steel, even aluminum, are just dumb ideas for this project. You aren't thinking about cutting tools or their hardness, or the precision you neeed.

You need to take a look at what the real cutting forces are for mill work; Cutting poplar for your parts is maybe just barely in spec for your reprap chassis, though I'm betting not; you've got lots of long arms waggling around.

One of the things that makes reprap plausible is that by doing additive machining you only need precision, not precision and strength. Further, you don't need a whole forge infrastructure to complete the replication.

If you want to build machine tools from scratch then look to Gingery, who will sell you cheap instructions to get from wood+scrap iron to mill + lathe.
"Further, you don't need a whole forge infrastructure to complete the replication."

I would just like to point out that casting aluminium is only moderatly difficult. I suspect it is a very good choice for making heavier duty parts beyond the capabilities of plastics.

There is a wealth of very good info on the web about simple home built blast furnaces, and some are getting spectacular results with their castings. My results have been more modest!

I would still suggest keeping the RepRap project focused as is for the reasons stated above by others, but aluminium is bound to play a role in things if this machine starts to spread out into the population at large.

Just think of the fusion of RepRap and Gingery concepts ...
I would also suggest we stay focussed on additive machining - as has already been pointed out, subtractive machining is best done using the Gingery machines and a backyard foundry. If you want CNC subtraction, it'd be easiest to add stepper motors and sensors to the axes of the tried and tested Gingery machines.

That said, of course, deposition machines do have the ability to make certain parts that subtractive machines cannot - one option with the existing apparatus that I recently suggested would be to use the reprap to make low temperature wax shapes which can then be surrounded with plaster of paris or green sand or whatever and used in lost-wax casting. Of course, I'm still interested in cutting out this step and adding a molten metal deposition head directly to the RepRap machine - the melting part's easy, you just need a high current transformer, say from an arc welder. The tricky bit is getting the layers of metal to adhere properly without oxides causing problems. For that, I was considering an ultrasonic welding head to follow behind the deposition head to finalise the layers - driven by magnetostrictive iron transducers as opposed to piezoelectric, so no exotic materials are needed. Unfortunately that branch of development is on hold until I get my foundry finished, which is necessary for melting the raw material and casting it into feed rods. If anyone's interested in the progress of the foundry itself, there are new photographs on my weblog.
Been lurking quite a while, but just had to register to comment on this one. Sorry Zach, I know I said I'd behave myself, but I can't stand by and watch the baby get thrown out with the bathwater.

The structural engineering differences between a CNC "additive" (+) machine vs. a CNC "subtractive"(-) machine is actually quite negligible. If one is after similar results as a (-) via (+) the positioning resolution requirements increase simply from Nyquist, drive system dither effects, and material 'slop/slump'. However, if post-fabrication machining is going to be done, you don't need the resolution and just 'microsoft' away the problems after initial fabrication. The big mechanical difference between a (+) and a (-) machine is that a (+) does not need the metal ripping torque of a Bridgeport. However it does need the positioning resolution and structural stability of one.

The 'real' [big-money patented] jobs use a 2-D magnetic positioning drive that has a theoretical resolution limit in the micrometer or smaller range, even though the filament extruded is .6 mm or smaller. There are NANOMETER (1*10^-9 meter?) resolution linear stepper motors out there. I've seen 2*10^-5 inch resolution in shop-made single axis linear positioning systems. I used to routinely set inch hand-tooling to three decimal places. You just need a really good shop and good measuring tools.

If anything, the basic techniques used in the design of a (-) are entirely translatable across to a (+) as long as one accepts the inherent limitations of your tooling. I know where I can find stock end-mills in the X*10^-3 inch range and can order smaller. The MkII probably would limit out at .4mm filament using current the hand manufacturing techniques. Anything smaller would require custom work but could be done or modified ['blown' pyrex tube, photo-etched brass/steel, EDM modified hypodermic tubing, laser drilled orifices, 'Pin-prick' hydro machining, or similar].

If the reprap project does change to a wood-based "first-rap", might I respectfully suggest looking at the jgro 'Gantry' router design available in the CNCZone forums? It is extremely capable and sturdy, constructed out of MDF, ENCO ACME threads, uses 'skate' ball bearings [$12/dozen] and gas pipe for rails. Hundreds of them have been made, and the well made ones are repeatable to .5mm or less.

It is the tooling heads, and the software that differentiates a (-) design from a (+). The MK2 extruder is an extremely capable design, that with modifications [mostly dealing with head design and temperature control] could be used for eutectic metals deposition, thixotropic plasticized fluids, E-20 polymerized ICW, even melted Butter Caramels [an earlier 'sweet' success of mine using techniques similar to the RepRap.]

The generalized hardware could be purchased on the open market [ for example], built [Godzilla, Witch, jgro or 5 Bears Research], modified [Sherline, Harbor Freight],it doesn't matter.
XYZ Hardware is commodity now, and any XYZ machine with sufficient resolving capability could be used.

It appears that you are failing to grasp the two things that the RepRap project has going for it.. The Tooling Head and the Software/Hardware. That is where your efforts should go, because that is where you will see the greatest rewards from.

The cheapest CNC machine project I was ever involved in was $37 US. Single purpose 2.5 axis, 6" cube, .001" repeatable using 20TPI all-thread, bit-banged parallel port controlled ULN based drivers with $4 Jameco steppers.

+4 from GMT
seagrol: I'm very glad that you decided to misbehave. Thanks for the commentary. :-)
One approach I'd been wondering about was if one ought to think of putting down a layer as fast as one could and then switch to a much slower cutting tool to trace the perimeter of the layer much more accurately and slowly.

Similarly, it seemed to me that you might want to map the surface of the extrusion periodically for uneveness and either fill in the dips or cut down the high points.

The basic concept you are describing is similar to a early Wax FDM machine done by a company that is no longer in business. I'm sorry to say but I can't remember who. Essentially what they had was a "two-step" extrusion process. The "hot" extruder head would lay down a fairly precisely shaped layer of wax, and then a large rotary cutter bit would "mow the lawn" across the entire surface of the working area at a fixed height to give a flat "donor surface" for the next layer of wax.
It's not all that difficult to modify the procedural code that creates the extrusion path to regenerate the path a cutting tool would follow to create a 'final' "milled" surface finish using a second "milling head". Unfortunately I can barely comprehend Java, so I don't know if I could even supply psuedo-code that would fit the language. Possible implementations would be a "flip flop" tooling head that rotated between tools depending on need, an automated tool changer, or a seperate trapeze mount. For mechanical speed and simplicity the 'flip-flop' head is best. A 1/4" endmill as the "big" cutter and a small (multi-thou) endmill for the "small" cutter would be big enough for a small worksurface. Just use a double ended sealed bearing spindle with a integral shaft motor on a 'bang-bang' axial joint. Using a reversible motor you wouldn't even need to spring for CW and CCW cutter heads. One hint on the tooling head: Axial TurnTable. Might want to consider a vaccuum or cold air chip removal process.
CADSoft EAGLE (a very nice Electronics CAD package) has an ulp (User Language Program?) called outline.ulp that takes a color board file (~ HPGL'd Gerber?) and processess it to a 2-D outline. The RepRap STL viewer code could probably be modified to generate the requisite data. Using a "variable" tooling offset, you could then use a "fine line" wax extruder at a precise viscosity/temp setting to draw the outline just slightly oversize in X and Y axes. Wait for the wax to cool, then "flood-fill" the enclosed area with a different viscosity/temp setting. Allow to harden. 'Mow the lawn' and run the outline procedure again while compensating for the radius of your cutter.
Extruder temperature control code with material specific settings could be used for other styles of material.
I have a HPGL interpreter [not my code] written in VB somewhere in my harddrive. One could also bypass the necessity of an interpreter and generate the tooling paths via software directly from the STL file (or mathematically process a .DXF/DWG file). This data could then be "drip fed" to the controller.
As a side effect, this technique also allows "multi-stage" machining of the wax envelope, where a slab of wax could be 'bulk' machined to 'donor' dimensions and then have material added to create a 'finished' form.
If any of this sounds a little too well thought out, Zach knows why.

Any of this make sense?

+4 from GMT
//Why not mill the ingot into what you want and save the swarf instead of making everything into filament and trying to extrude everything.//

I had wondered about that to. I think I mentioned it to someone and they indicated that, while you can mill down blocks of plastic, there are limits to the shapes possible with that method. I agree, though, that a large number of parts could be milled and maybe a few would need to be printed.
It would be possible to compromise and design a Reprap that could be made from recycled HDPE sheet - or chopping boards as I have seen elsewhere. We could even print the plans so they could be stuck down on the boards to be cut out.

Vik :v)
There is another reason for staying on additive machine exclusively. When you don't have to deal with a lot of variables inherent in machining (different materials, different cutting tools and their sharpness, cutting speeds, etc) it makes a lot of software-based shortcuts possible. When you have an additive machine you have very few variables; your software could easily interpolate and extrapolate to improve quality beyond your sensors and motors precision, so you'll get more from your cheap (or imprecisely replicated/assembled) hardware.

Oh, and don't forget that any extra tool/machine would have to be made, programmed, debugged, and then replicated, so making multiple different machines in one makes the ultimate self-replication goal more difficult. The same time spent on the additive process would help more in the long run, I think.
re wax/milling systems. I think you're thinking of the old Cubital 'Solid Ground Curing' machines.

Cubital went bust a few years back, and most of their engineers were re-employed by Objet.

On the subject of support material, I was wondering why we can't use FDM for the build material, but powder material for the support structure? If we laid down a build layer, then filled the rest of the space with sand and levelled off, we could keep re-using it and have free support material? Important for some of the applications I have in mind which would have very large overhanging areas, and would use a hundred times as much support as build material
Sand... Hourglass... Hmmmm... Nice...

No lateral support though. Vik is already having problems with the extrusion thread dragging his extruded object back and forth across the work surface. :-(
The great thing about RepRap (if I may say so) is that, once we get our first release out there, anyone can take it in any direction they like, including down a subtractive route.

But as others have said, there are two big advantages to additive manufacture over subtractive: 0 force, and no tool collisions. As someone involved in automated path planning, I can tell you that avoiding tool collisions is a very very very difficult problem in all types of subtractive manufacture.

I sometimes wonder where we would be now if, at the start of the industrial revolution, people had taken something like a Jacquard loom (additive, of course) and made it cut out and lay down thin veneers on top of each other, gluing them together...
At the extrusion temperature of the MkII head, can you describe the viscosity of the material? If it's anything approaching thick honey, cold treacle or real maple syrup [Sorry. It's near dinner time, and I am absolutely famished.], one might consider using what I've always refered to as a 'magnetic-reed' valve to add a secondary control to the material flow. The best everyday example I can think of for this valve is the spring blade valve on an table syrup dispenser. Energize a small electromagnet as a micro-actuator to pull a iron/steel U-channel 'blade' back against an expansion spring [a cutdown "BIC" stylus spring is a suitable 'scrouged' source.], extrude, de-energize the magnet, HDPE coated blade snaps back, cutting the extruded filament, and stops the 'dribble'. It's a bit of a sticky wicket getting the clearances proper, but should not present any great difficulties. Only problem is keeping the 'chop' stuck to the blade from either gumming things up or falling into the work. It can be done with some finicky small parts [been there.] to automatically scrape the blade as it is being retracted at the beginning of each cycle. A poppet or needle valve would also work, but would require unduly complex and ugly modifications to the MkII extruder. Such modifications would also unduly complicate the fabrication processes one needs to follow to build an extruder.

The 'Sand fill' technique works better if the entire work surface is filled with sand instead of just the internal areas of the 3D shape. Makes for messier cleanup, and you'd need tooling to put the sand where you want it and not on the extruded material. Essentially what you are doing in this technique is building a 'loose' sand mold around the 'pattern'. This is a "sand table" SLS without the laser.

The MkII type design, and the capabilities it allows, are endless as long as small changes are made to adapt it to the specific material being used. Without some serious (real) engineering, a different suite of fabrication materials, and a plethora of safety issues [Thermal, Power, Pressure], I believe that it would be best to not even dream of seeing a 'hot-melt metal' capable RepRap. It would be doable, but should be best considered more an exercise in perverted engineering than of any practical value at this point in time.

Obligatory Project Management:
Focusing on 'killing' one problem at a time always results in a high-speed 'incremental' evolution of a design that is more 'robust' than an entirely new design every time. You can change horses in mid-stream and only get a little damp, but you can not rebuild your aeroplane whilst flying it.

Well, we are going for cold-melt-metal in the form of Wood's metal of course; that will allow us to build 3D electonic circuits...
On the subject of melting metal, have a look at the Stovetop Metalcasting site.
Wood's metal? Humm... that contains about 28% lead. Over the years I've researched lead and it's infamous neurotoxicity a number of times. Great stuff for permanently reducing the IQ of children. And all this at a few microgrammes per -decaliter- of blood.

I hate to say it, but I would think it is a very risky choice for anything other than supervised lab use, never mind third world application. There is also the possibility of legal issues in other countries where they are on the ball about polution and public health, take for example lead in car battries. As things stand right now, there is also a push to gradually ban lead solder in electronics...

I would suggest seriously investigating the new alloys being proposed to replace lead solder. Yes, there are a number of vexing downsides, such as higher melt temps, but none of them are as substantial as brain damage in kids.
Some form of Field's Metal or similar (Bismuth-Tin alloy with or without cadnium, indium, or silver) aka 'Lead Free Solder' or 'fusible metal') would be a better choice than Wood's Metal. Nowhere near as nasty neurologically or systemically. Readily available in bulk formats (billet, rod, wire, thread) already just about anywhere. Alloy dependent liquidus temperatures between the 180F to 300F range, highly eutectic with sharp phase transition points. Most of them would need an extruder head with forming capabilities.
For specific alloys, might I suggest looking at Canfield Technologies ( alloys 174 and 255. Cerro MetalProducts ( has an extremely broad collection of alloys that are not listed their website.

+4 from GMT
The problem that I've seen with low melting temp alloys is that if you want them non-toxic (no lead or cadmium) and you want them to melt at temperatures lower than about 140 degrees Celsius you've got to include a fair fraction of Indium. That runs the cost of the alloy way up. :-(
If you want to go readily available, I at least can get an alloy of 95% tin and 5% silver for a few dollars US for a half-pound over at the hardware store. I can't quite find that particular mix's melting point, but tin with 3.5% silver and .9% copper melts at 217 C and is supposed to be eutectic.

Might be worth working out an extruder-like device that can run at 250 C.
Polyaniline-like polymers?
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to
Posts [Atom]