Perfactory Machine: Can we make one?

Little backstory: My Sister is Friends with a Guy who owns/runs a foam shop. He uses hot wire machines to cut simple 2d foam shaped objects for his clients needs. He's built ever damn machine in that place. He is building a 3 axis machine.


He knows how to work/build with the large scale stepper motors, servos, motion controllers- the works. He uses Autocad to cut his patterns out and thats all been 2d until just recently.

I'm a 3d CGI artist guy. He asked my Sister to ask me if I could make him some 2d DFX files from any type of 3d model.

Trying to figure out how to do this. I found out about Meshlabs. It has a slice function that saves all the slices from the 3d model as 2d profile SVG "Scalable Vector Graphic". Theres another program out there called "POV-Ray" which might do a better job with this. It was used to do the same thing on a homemade sugar/heat based Rapid Prototype machine from the Evil Mad Scientist LAbs. That would probably be the better program to seek out if this Garage Perfacto Machine is possible.

Initially He was just going to wire cut out a few slices made from these 3d models and glue them all together to make the part, but now he's doing the 3 axis route. I need to get to the point.

Being exposed to all this stuff really set a fire to my backside.

I really want 3d prints of my work with excellent detailing. I want an RP machine of my very own.

I've seen all the machines out there and what they can provide and the Perfactory is the one thats standing on top of the mountain when it comes to Level of Detail and Speed. I make my model with 6 million polys most of the time. I want that kind of Detail!

Not only that. The machine seems to be way more simpler than what you Guys are creating. It only has one Axis controlled motor on it. That motor pulls the arm with the top of the build tray up before each new layer is created. No more 3 axis needed! Most models made in the PErfacto don't even need Support Post work added to them- Thanks to the way this machine builds the model.

The first layer naturally glues itself down too the Z Axis arm's roof. The roofs surface is made of simple glass. After each layer is created it rises up 25 micrometers. The next layer is built below the first layer, but somehow it is separated from the lower glass tray using a "Vacuum Diffuser Tilting Mechanism (Haven't looked that up yet- it sounds a little complex)


The real money for the machine has to be in the Texas Instruments DMD chip. And maybe the Projection Bulb.

The way it works is like this.

The DMD Chip is a rectangular array containing millions of microscopic sized mirrors. Underneath these mirrors are tiny electrodes. Voltage sent to these electrodes cause the mirrors to tilt towards or away from the light source (projector bulb)

These mirrors when tilted into the On position redirect that Projector Bulbs Light out onto the Observation Area. In this case the Observation Area is the Glass under the vat of photoactive resin.

The mirrors that don't turn on, dont send any light to the observation area.
Each mirro/tile represents a pixel in the image.

Basically Perfactory is using this DMD chip to project a simple 2 bit Black and White colored image into the vat of photoactive resin. The White color becomes the lighted up tile/mirrors. Theres no light projected in the black area. The Black image information is ignored by those mirrors. They remain off. Only the White image information is translated into a "turn/tilt little mirror" command where that mirrors position on the chip is relative to it's position in the image. All of the white areas in this image is projected from the projector under the table through a clear piece of glass into the photoactive resin. It cures into this resin at 25 microgram depth, and this layer is cured into the exact shape of that white image/light shape information. All you need is a program like Meshlabs or POV-Ray to cut up your model into these several thousand 2d 2color scale information images. Might have to reverse black to white/white to black. since these program like to make the model information black.

It's like 436 layers to make half an inch. Each layer requires 8 seconds worth of static image projection time. And then it takes 3-5 second for the Vacuum Diffuser/Tilt Mechanism to separate the model from the lower glass tray (not the top one) and then the Z axis arm goes up 25 micrograms to start the new layer.

This machine sounds painfully easy to build- well not really. It will make way better models then what you guys are toying with right now. I've included an image to help you visualize it. Please tell me that it's possible to build this garage style for way less money.

[img377.imageshack.us]

/\ If the HTML code doesnt work. This is a link to the image to explain things and help you all visualize it better.

Just my 2 cents..

A guy here called Fernando is the guru on photopolymers and has done some mixing. Look around the threads and the blogs, there is tons of cool material.

For the projection system, hm 1280x1024 seems familiar. Could it be that they are using a chip used in projection systems in ordinary home-cinema equipment. If so, this is good news, since these chips are manufactured for a consumer mass-market, nudge nudge, know what I mean...

So I don't think the chip needs to be the biggest expense.

Perhaps one could take a normal video projector and then just render each layer/slice to the VGA framebubber like a slideshow.

my biggest question is how you maintain a very precise depth of polymerization (layer height)? I'm certainly no expert, but I think Fernando has done some experimenting around that issue.

25 microns isn't very much! You're getting MUCH more Z resolution than X or Y. Even so though, with only 5 seconds per layer my rough calculation is that you could print a 10cm high model (of any x and y dimension) in under 4 hours. That's VERY good.

>how you maintain a very precise depth of polymerization
I presume its just the height the object is lifted. Above that the object is already solid. The liquid layer is between the base of the object and the glass.

---

[www.hydraraptor.blogspot.com]

What is the polymer liquid in the tank? Cheap? Readily available?

nophead Wrote:
-------------------------------------------------------
> >how you maintain a very precise depth of
> polymerization
> I presume its just the height the object is
> lifted. Above that the object is already solid.
> The liquid layer is between the base of the object
> and the glass.

Sure, but what happens if you want to make a layer larger than the last one? Will it get fuzzy edges?

drvanthorp Wrote:
-------------------------------------------------------
> What is the polymer liquid in the tank? Cheap?
> Readily available?

Well, judging from Fernando's descriptions of what's currently used probably neither of the above. However, this approach has an enormous advantage over the "conventional" method of UV-polymer printing that you don't need to flood the build area with the stuff, so you only need enough to cover the volume of the part plus a little extra to keep a thin sheet of it stocked across the entire build plate. Clever.

Since the polymerized monomer sticks to the glass plate the is being raised, I guess I'm left to wonder why it doesn't stick to that glass plate that the polymerizing light is being shined through?

Forrest Higgs Wrote:
-------------------------------------------------------
> Since the polymerized monomer sticks to the glass
> plate the is being raised, I guess I'm left to
> wonder why it doesn't stick to that glass plate
> that the polymerizing light is being shined
> through?

The diagrams says "Vacumm Diffuser Module separates each new layer bond from the glass after it's cured using a tilting mechanism."

I don't know what a Vacumm Diffuser Module is or how it works.

JohnWasser Wrote:
-------------------------------------------------------
> I don't know what a Vacumm Diffuser Module is or
> how it works.

Me either. Sounds like some sort of sky hook to me.

[www.pixelprecision.net]
Software to control the DMD chip to make it do what you want

[www.dlinnovations.com]
Sells the DMD chip with control board included.

heres a video of how the Digital Micro Mirror chips work------- [www.dlp.com]

Notice in the animation there is a projection bulb it light passes through two lens before it hits the DMD chip (don't worry about color wheel- not needed) this light is redirected to the projection lens. This is what we are trying to do to the DMD chip. We need to get control over this chip, tell it how to display the image sent to it, how long to prject that image, when to stop and when to turn back on to displays the next image. We want to work with 2 color black and white images. Turn either that black or white pixel information into mirror tilt input and that titlted mirror reflects that light info out. Each mirror on the chip equals one pixel in the image information, One of the colors in the 2 color image will be ignored and that mirror will not tilt toward the light source.

I'm pretty sure that Perfactory uses home consumer DLP sytled projectors, but they are pretty expensive. Even then you can't be sure how well those premade units could be altered and reprogrammed to do what you want.

Like the reprap, we have to go after specific parts. That way once we figure this stuff out, anyone can repeat our steps with very little hassle. The DLP projector body and lens housing might be worth purchasing a used or broken one on ebay. Since those parts seem to remain unchanged in the Perfactory machine.

In regards to eliminating the layer from sticking to the glass in the vat tray. It might be possible to coat that area of the glass with a thin cured layer of optically clear silicone to prevent the resin from sticking to it.

The resin is dispensed into the tray before each layer. This might account for the level/volume of cured material between each layer. If the volume in the tray represents the layer thickness. That could be the solution to repeating the layer thickness for each new layer. It is also possible that the position of the light source from the resin and the time it's left directed at it causes that amount to cure.

Remember I'm not a mechanical guy. I haven't really gone through this with the guy I mentioned- beyond a few explanations.

I'm really happy that you guys see the possibilities in this. If anyone can see this through. It's you guys. I say we try to use the same resolution of the perfactory used chip. Why shortchange the final result,

Kyle- If there was a larger layer between a thinner layer. It would cause a stair stepping effect to the model in that area.

The layer system must be percise, but I don't think thats too much of a problem, since the layer will cure from the projected light image at a timed rate for each layer. So these layers should all be equal to each other. I'm guessing the only issue will be in refilling the vat/tray with the same volume needed for the next layer- since the lost resin from the vat from each newly made image layer wont be the same, due to the layers shape and area coverage. A weight scale or volume sensor, or a wire that detects the liquid when it touches it- Some system like that could be employed to know when to apply a stop sending material command to the pump system when the tray is/or needs to be filled.

Even then we don't know how percise that refill tray/vat fuction is.

Biggest issue after seeing what layer level is cured is making the arm go up the exact distance of that layer so the new layer thats built under it will attach itself to the bottom part of the previous layer.

Edited 1 time(s). Last edit at 07/14/2008 03:56PM by makeit.

Just to give you an idea of what kind of detail these perfactory machines can do.
And heres the original CGI 3d model. Not mine by the way. This model required no post work to support it. Another bonus against many other rapid proto machine. Some parts will still need support posts, but many do not.



I also believe this system eliminates the need for low poly count models- since the model is made from 2d image information converted out of the model from another program like POV ray or Mesh Labs.

Edited 1 time(s). Last edit at 07/14/2008 10:56AM by makeit.

So is this made starting with the head or the feet?

If the head I would expect a flat spot on the top. If the feet, how are the arms supported until the shoulders appear?

---

[www.hydraraptor.blogspot.com]

feet/base first. You want the greater flatter area at the start of the build. That or you have support posts in that starting area, and those are cut off.

I think the real trick to this machine is going to be getting a consistent 25 micron layer of liquid across the build area. Surface tension/capillary action and viscosity are going to be relatively powerful at that scale and will tend to cause either dry areas (via non-wetting) or thick areas (via wetting meniscus).
The "Vacuum Diffuser Module" is probably not easily replicable, but I wonder about the possibility of just using some Teflon and a transducer to vibrate the plate.
I think support structures will be needed a much as with any system that does not use a filled bed. You can often plan a build to avoid support structures, if you look at the monster model everything seems to flow from the head down, even the joint spikes start out with horizontal connections to the body.
The real problem for us as regards this technique however, is going to be patents. I am sure there are a number and I am sure the company will defend them vigorously. One of the main attractions of FFF for this project is that the basic patents are expiring soon and the holder seems OK with what we are about.

So if this was built feet first, upside down, it must have had support material to connect the hands to something in the same plane as the base. Otherwise when it does the first layer that includes the finger tips there is nothing to lift them to do the next layer. Or am I missing something?

---

[www.hydraraptor.blogspot.com]

What are the characteristics of the resultant polymer? Is it engineering-grade plastic, or is this only good for visual prototypes?

I'm not very familiar with UV polymerization, or organic chemistry in general!

[www.envisiontec.com]

Look under the "Materials" Tab on the display bar going across the top of the webpage. They offer several types of photopolymer resins. Most seem to have a Shore D hardness of 75 or better.


As for the patents. I don't see how we are imposing on them. I haven't looked at any of Perfactorys patents, but now that you mention it, it might be a good idea to give them a look over.

Patents are not top secret documents. Your allowed to look at them. Plus we are not building these machines and selling them as complete systems to others. The objective here is to learn/teach others how to build one for our/their own personal use.

A car company can't sue you for building a scratch built replica of one of their cars. When your intentions are for personal use only and not commercial resale.

I don't think thats an issue at all.

nophead Wrote:
-------------------------------------------------------
> So if this was built feet first, upside down, it
> must have had support material to connect the
> hands to something in the same plane as the base.
> Otherwise when it does the first layer that
> includes the finger tips there is nothing to lift
> them to do the next layer. Or am I missing
> something?



Edit: I see what your saying. Yeah the hands must have had post work to support them. since those layers would not be connected to anything. If it was built from the head down with the posts supporting the top of the head. It might not have needed posts near the hands.

Edited 1 time(s). Last edit at 07/15/2008 04:23AM by makeit.

makeit Wrote:
-------------------------------------------------------
> A lot of people are confused by this building
> system.

Yes I think I am one of them!

> Each new layer will extend from the
> previous layer. Think building with legos. If the
> fingers shape is changing. This change is slowly
> achieved in each new layer added to the previous
> one. With these layers reducing or growing or
> moving away from the previous layer to make that
> shape.

Yes but if it was built upside down the fingers are not attached to anything when they first appear. My understanding is that everything has to be connected directly or indirectly to the top platform to be lifted upwards for the next layer to be built underneath. When upside down, the fingers have nothing above them. They only attach to the arms which have not been built yet.

---

[www.hydraraptor.blogspot.com]

makeit Wrote:
-------------------------------------------------------

> Patents are not top secret documents. Your allowed
> to look at them. Plus we are not building these
> machines and selling them as complete systems to
> others. The objective here is to learn/teach
> others how to build one for our/their own personal
> use.
>
> A car company can't sue you for building a scratch
> built replica of one of their cars. When your
> intentions are for personal use only and not
> commercial resale.
>
> I don't think thats an issue at all.


Right, and there was a longish forum discussion on this topic a little while ago. IIRC the conclusion was that many of the issues brought up by cheap and plentiful RP machines have no precedent, and it will take a court case or two to sort it all out. I would be hesitant to brush off the whole concern as irrelevant, however. You can bet that whether they have a right to or not, companies will do all they can to defend their sources of profit. And hey, who can blame them?

That's why I like FFF - as BDolge said, the patent will be up by the time RepRap begins making an impact.

Here's the manual on how to use perfactory, it should clear things up a bit:
[digfablab.wikispaces.com]

It is interesting to note that a deck of playing cards is used to clean the build area after a build.

It is also worth noting that perfctory is gravity fed and doesn't use pumps to get the photopolymer into the basin.

Here's the technical details on the materials it uses:
[digfablab.wikispaces.com]

Apparently some builds have to have some sort of support structure. Some plastics can also be burnt out for casting.

Found a movie here:
[video.google.com]

It made a lot clear for me, while it's not really that good a good movie at all. Interesting technique, for sure! How environmentally friendly are such polymers? I know that most plastics can (theoretically) be infinitely recycled or like PCL / PLA are biodegradable... If every future household has a RepRap based on this technology it would be nice if it's Eco Friendly as well

---

Regards,

Erik de Bruijn
[Ultimaker.com] - [blog.erikdebruijn.nl]

Kyle Corbitt Wrote:
-------------------------------------------------------
> makeit Wrote:
> --------------------------------------------------
> -----
>
> > Patents are not top secret documents. Your
> allowed
> > to look at them. Plus we are not building these
> > machines and selling them as complete systems
> to
> > others. The objective here is to learn/teach
> > others how to build one for our/their own
> personal
> > use.
> >
> > A car company can't sue you for building a
> scratch
> > built replica of one of their cars. When your
> > intentions are for personal use only and not
> > commercial resale.
> >
> > I don't think thats an issue at all.
>
>
> Right, and there was a longish forum discussion on
> this topic a little while ago. IIRC the
> conclusion was that many of the issues brought up
> by cheap and plentiful RP machines have no
> precedent, and it will take a court case or two to
> sort it all out. I would be hesitant to brush off
> the whole concern as irrelevant, however. You can
> bet that whether they have a right to or not,
> companies will do all they can to defend their
> sources of profit. And hey, who can blame them?
>
> That's why I like FFF - as BDolge said, the patent
> will be up by the time RepRap begins making an
> impact.


I understand. I hope I'm not coming across wrong or uppity. I know I've said quite a few incorrect things. I'm still learning about the machine myself.

ErikDeBruijn Wrote:
-------------------------------------------------------
> Found a movie here:
> [video.google.com]
> 001332541&q=perfactory&ei=4VB8SJKBDYXyiQKE_N2VCw
>
> It made a lot clear for me, while it's not really
> that good a good movie at all. Interesting
> technique, for sure! How environmentally friendly
> are such polymers? I know that most plastics can
> (theoretically) be infinitely recycled or like PCL
> / PLA are biodegradable... If every future
> household has a RepRap based on this technology it
> would be nice if it's Eco Friendly as well


Cool video. It looks like the arm goes up and then back down again. Is that how it releases the new layer from the lower tray before it resets itself the next build?

... i think this is for 'sucking' new material in the gap and 'squeesing' it again down to 25 microns ...

Viktor

makeit Wrote:
-------------------------------------------------------
> Kyle Corbitt Wrote:
> --------------------------------------------------
> > That's why I like FFF - as BDolge said, the patent
> > will be up by the time RepRap begins making an
> > impact.
>
>
> I understand. I hope I'm not coming across wrong
> or uppity. I know I've said quite a few incorrect
> things. I'm still learning about the machine
> myself.


No, no, I wasn't trying to be derogatory, I'm merely saying that patent obligations are still worth keeping in mind. I agree that it's good to explore this avenue.

Carry on with the interesting discussion! Does anyone know how this compares to Z-corp printers (for instance) resolution-wise?

Gene Hacker Wrote:
-------------------------------------------------------
> Here's the manual on how to use perfactory, it
> should clear things up a bit:
> [digfablab.wikispaces.com]
> terature
>
> It is interesting to note that a deck of playing
> cards is used to clean the build area after a
> build.
>
> It is also worth noting that perfctory is gravity
> fed and doesn't use pumps to get the photopolymer
> into the basin.

A brand new projector lamp has an
exponential power drop in the first
100 hours during its "burn in" so that
it is highly recommended to check
the Gray mask and carry out the 48-
point measurement before each
build job.

Before starting the 48-field
measurement observe that the
projector brightness is 10 - 20% over
the nominal value 550mW/dm

Edited 1 time(s). Last edit at 07/15/2008 09:05PM by makeit.

makeit Wrote:
-------------------------------------------------------
> thought all of that information was pretty
> important. They have a calibration routine to
> insure the projector and chip is working right.
> "mW" stands for "Molecular Weight"

You sure? Looks a lot like milliwatt to me, and the capitalization makes more sense that way too.

> I don't know
> what "dm2" stands for. I am wondering if this
> terminology is just a branding system- or if it
> really is relevant and unavoidable.

decimeter^2? Although I've never heard of anyone using decimeters to actually measure something... generally people skip straight from centimeters to meters. But it's certainly conceivable, and it would make sense that they would have a power/unit area calibration test of some kind.

Kyle Corbitt Wrote:
-------------------------------------------------------
> makeit Wrote:
> --------------------------------------------------
> -----
> > thought all of that information was pretty
> > important. They have a calibration routine to
> > insure the projector and chip is working right.
> > "mW" stands for "Molecular Weight"
>
> You sure? Looks a lot like milliwatt to me, and
> the capitalization makes more sense that way too.
>
> > I don't know
> > what "dm2" stands for. I am wondering if this
> > terminology is just a branding system- or if it
> > really is relevant and unavoidable.
>
> decimeter^2? Although I've never heard of anyone
> using decimeters to actually measure something...
> generally people skip straight from centimeters to
> meters. But it's certainly conceivable, and it
> would make sense that they would have a power/unit
> area calibration test of some kind.

I think your right. I did a quick google search on the combined term and found that "molecular weight" listed on a page, and it just didn't sound right to me.