I printed off another plate of Mendel parts the other week, including two more Z-axis drive screw blocks. This time I increased the height of the support layer (7 slices rather than 4) to better accommodate the slight sloping of the glass foam base. I was more careful in removing the large parts this time, and tried first to peel the HIPS support off of the glass foam as the first step, rather than trying to remove the ABS parts from the HIPS right away. I found that using a pocketknife to lift up the HIPS at a corner worked very well, and I was able to remove all the parts with no breakage. The HIPS is still bonding to the ABS more strongly in some spots on the tray than in others (the bonding in the rear right corner still being the strongest), and I’m at a loss as to why.

I decided to try printing one of the large toothed pulleys this time around to see what the resulting quality would be like.  While it’s certainly functional enough for the goals of a self-replicating rapid prototyper project, I think using traditionally manufactured off-the-shelf pulleys when possible is a much better solution – no need to cripple precision in the name of purity.  You’ll also note the helical looking object at the far right – this was the first non-Mendel part I had tried printing.  It’s a screwable jewelry box that I found on Thingiverse. Unfortunately, I was a bit eager when putting the two halves together, as I should have lightly sanded the surfaces first. The fit is rather tight, and now I can’t get the two pieces apart.

With this latest batch of parts complete, I had a look inside the dry box on the Stratasys to see how much filament I had left.  Very little, it turns out – perhaps 7 turns each of ABS and HIPS.  I was expecting my shipment of filament from New Image Plastics to have been here over a week ago, but in doing a little digging, it appears they can be slow to ship to their hobbyist customers.  I can’t blame them – the big industrial orders that actually keep them in business get priority, and it’s very kind of them to take the time to deal with RepRap users at all.  I suppose a bit of patience is in order.

So, what to do with a Stratasys FDM 1600 that’s just sitting idle?  Have a look at the innards, that’s what.  I could find no real information on what is inside the FDM machines other than illustrations in Stratasys patents, and what I can see inside the build chamber.  However, the Stratasys 1996 10-K filing notes that the “sole current supplier of the X-Y stage for the FDM 1650, FDM 2000 and FDM 8000 benchtop systems is Asymtek.”  It was a solid bet that I’d find some Asymtek hardware inside, and likely other off-the-shelf parts as well (as commodity 1/16 DIN temperature controllers were used on the front panel rather than a more integrated system).   The manual cautions against removing any panels, as it could wreck the calibration.  The side panels do look rather beefy, but I’m guessing there’s not a great deal of interesting machinery or wiring behind them.  The upper cover, on the other hand…

The front of the machine is to the left – you can see two of the thermocouple wires that run to the temperature controllers on the front panel.  There’s a DIN rail for power distribution at the top left of the photo, and as best I can tell, the white box underneath the large circuit board on the right is just a power supply.  Just out of view in the upper left is the LCD keypad interface, which is an Intelligent Instrumentation CTM150B-00. The big beige box in the lower left is the Asymtek controller, model A-201 (for which I found the operation manual and the service manual).  Asymtek manufactures fluid dispensing equipment generally used in manufacturing circuit boards, and the A-200 series appears to have been specifically targeted at OEMs to use as a turn-key motion control system. This looks to have been a very shrewd choice by Stratasys – rather than having to build a motion control system from scratch, they found an off-the-shelf system that was extremely well suited to the task. Given the wording of the 10-K filing, I’m guessing that the X-Y mechanics were all from Asymtek as well (looking inside the FDM build chamber, it easily looks like an upside-down A-100 or A-300 for the X-Y).

The big circuit board itself is what I assume to be a proprietary Stratasys board, as there are no company, brand or model names silkscreened onto it.  The two ROMs are labeled as firmware 7.04 (which I think is the version the LCD panel displays on startup).  The large square chip is a National Semiconductor HPC46003 16-bit microcontroller – no internal ROM on this version of the uC, hence the need for a pair of socketed ROMs.

I couldn’t learn a whole lot right away from the circuit board (though if I get a chance I’ll dump the contents of the ROMs), so I started looking into the Asymtek controller.  I came across a paper on fractal fill patterns that used an FDM 1650 as a testbed (the late 90s must have been a great time for grad students to play with Stratasys machines – unlike newer models, these older units have a high hackability factor).  The paragraph that jumped out at me was:

The Stratasys FDM 1650 machine used for the experimental tests is driven by an Asymtek A-201 digital motion controller. The A-201 controls the x-y movement of the depositing head, the z movement of the stage, and the rotation of the two electrical servo-motors mounted on the head that feed the thermoplastic wire into the two liquefiers. The controller uses Automove Control Language (ACL) for programming [7]; Stratasys has implemented a slightly modified version of this language, called Stratasys Machine Language (SML). It is similar to Hewlett Packard’s PCL used to control plotters and all commands are strings of ASCII text.

Another google search, and I found the Automove Control Language reference. Sure enough, the commands detailed looked just like the lines in a .SML file generated by Quickslice. I wondered what modifications Stratasys made to ACL to create SML, as a sampling of commands I pulled from a generated .SML file are all present in the ACL reference. In fact, I have a sneaking suspicion that the “Stratasys Modeler Language Programming Reference Manual” noted in Øivind Brockmeier’s thesis was hardly more than a re-labeled ACL manual (perhaps to hide the identity of a key supplier), especially as Øivind notes that the revision of his copy was 3.4 from May 1991, and revision 3.4 of the ACL manual was released on April 22, 1991.  Sure enough, in tracing the RS-232 cable in the FDM 1600, I found that it runs right into the A-201 – the brains of the FDM are Asymtek, not Stratasys!

When I popped into work on Monday to check on the Stratasys, here’s what was waiting for me:

A beautiful plate of Mendel parts!

Unfortunately, disaster struck when I was a bit rough with trying to remove the two large z-axis drive screw blocks.  As you can see, the bottom layer remained attached to the support layer.  I think this was due to two factors: 1) The back right corner of the build base was higher than any other part of the foam, and I think the ABS was extruded more forcefully into the mesh of the HIPS support layer.  2) I used a loose fill pattern on the entire build, so there wasn’t a great deal of tensile strength between the bottom layer and the interior – you can see the fairly large fill pattern inside the part on the left.  The fact that I wasn’t as gentle as I could have been may have also contributed – now that I know that you can break parts when trying to remove them, I’ll be more careful in the future.

The other issue is one that the seller of the Stratasys had shown me, but I hadn’t yet seen it on one of my own parts.  The photo shows how the bottom 4 layers or so of outline on one side did not get fused to the inner fill.  This can be remedied post-production by dipping or spraying the part with Weld-On 3 or MEK.  Note that the outline pass on the part is a lighter color than the fill – the ‘lobster red’ ABS filament loses its color as it sits in the extruder head and cooks – after sitting idle for a half hour or more, the first ABS out of the extruder is almost the color of the light gray support material. I don’t know if this is detrimental to the mechanical properties of the ABS, or if it is only cosmetic.

Sealing the parts is also a topic Frankie and I have discussed.  I found that Fortus (the high-end division of Stratasys, with Dimension covering the low-end) has a number of interesting application notes available. The appnote on investment casting particularly caught my eye. While the documentation for my FDM 1600 notes that a special wax material (and accompanying support material) can be used to build wax masters, there seems to be little information available on this material and process – I’m guessing that ABS is a much more popular end-user material. Also, given that the head in my FDM 1600 is specifically marked ‘ABS’, I’m also guessing I’d need a separate head for ICW or investment casting wax prints. Given my previous contact with Stratasys, I’ll wager that my chances of acquiring such a head are slim-to-none. Anyhow, the application note indicates that ABS masters can be used for investment casting, given that the burnout process is done at a high enough temperature. I eagerly passed this information on to Frankie, but he was already ahead of me and showed me an aluminum pizza cutter grip that he had just cast from an ABS part from when he had borrowed a Stratasys demo unit last year. The fill texture of the FDM part was apparent, but not too bad. Ideally, filling in the mesh would be needed before creating the mold, but we have a few crazy ideas on how to accomplish this. Well, Frankie likely has ‘good’ ideas while I have the ‘crazy’ ideas.

A nice long holiday weekend is a great occasion to tinker around with a rapid prototyper.  Well, all right, any weekend.  Or weekday for that matter.  There’s something enthralling about watching your creation emerge slowly, slice-by-slice from the build platform.  POV-Ray enthusiasts will know the feeling well – watching your file slowly transformed into an image, one line of pixels at a time.  Although in the case of a rapid prototyper, the transformation is subtly different – you know exactly what the output object will be (unlike in POV-Ray, where you have to alternate between adjusting your scene description file and rendering that file until you achieve the effect you want – enjoyable, yes, but lacking a direct translation of vision into result).  A rapid prototyper on the other hand, allows a leap between the digital domain and the physical one.  Perhaps not as exciting in terms of watching the process, but more profound in concept (as evidenced by the abundance of ‘replicators’ in science fiction).

I’ve done a lot of research into the Stratasys machines and RepRap lately, especially as I now have to feed this beast which I posses.  While I have a decent amount of filament, the cost of getting more has kept my use of the Stratasys to a bare minimum. Bolson Materials charges $250 for a 4 pound reel of filament, which is probably a great price, considering that the HP-branded Stratasys units recently released for sale in Europe have filament cartridges that are (as of this writing) $987.94 each after currency conversion. And you thought buying inkjet consumables was bad.  Given that this is only $12.06 less than what I paid for the machine (yes, I practically stole the Stratasys for a cool grand, and everyone I speak to agrees that I basically won the  Craigslist lottery), I began looking into the filament itself.

The first step was to determine exactly what material I was dealing with. I knew that it was ABS plastic, but I didn’t know if there was anything special about it (I know almost nothing of plastics engineering, but I know that numerous formulations are possible – Lego bricks, for example, are not pure ABS, but a custom formulation). I knew Stratasys wouldn’t be keen on divulging this information (imagine the response at a KFC counter when you ask precisely which 11 herbs and spices are used – okay, not the best analogy, but bear with me). I had discovered many years ago that MSDS sheets are an excellent window into proprietary information, divulging industrial secrets by virtue of safety legislation. So I did some searching and found that the MSDS sheet for the modeling material indicated that the filament was essentially pure ABS, with perhaps a tiny amount of mineral oil, tallow, and/or wax thrown in for a lubricant.  Similarly, I discovered that the breakaway support material appears to be nothing more than HIPS (High-Impact Polystyrene).

Obviously, buying Stratasys branded filament would quickly become prohibitively expensive (the classic razor and blades business model in evidence).  It’s generally a bad sign when you go looking for a product, but can’t find a price for said product.  Yet diligent web searches led me to stumble onto this post. Given that New Image Plastics could apparently supply smaller diameters than the 3mm RepRap standard (I needed .070″ for the FDM 1600), I figured I’d give them a shot. I called them up and inquired about getting a few pounds of ABS and HIPS in a .070 filament size.  Unexpectedly, I was speaking with James Waring, owner of the company.  He told me that a number of customers were using .070 ABS filament from them in Stratasys machines with no problems, and that for newer Stratasys machines that had chipped cartridges, he is aware of someone who is successfully reloading cartridges and will hopefully be revealing the technique once perfected.  Anyhow, I ordered 5 lbs. each of ABS and HIPS, and James said that he’d have it scheduled for production this weekend!  I can’t imagine a more phenomenal response time from a vendor.  If the spools feed through the Stratasys without issue (shouldn’t be hard – the current ‘lobster red’ spool appears to be about a decade old from the tag information, so they’re not exactly prime material), I’m sure I’ll be ordering lots more.

Meanwhile, I also tried my luck with contacting Stratasys to see if any spare parts or other service was still available for the unit (not that it appears to need any, but I wanted to be prepared).  Their tech support department told me that no parts or service were available for the FDM 1600, and they couldn’t direct me to any third parties that did service on the unit.  I also asked about the availability of a document called the “Stratasys Modeler Language Programming Reference Manual” that I found referenced in a thesis I came across.  The thesis covers the building of an automated build platform loading/unloading system, and the document appears to detail the format of the .SML file that gets uploaded to the printer.  Unsurprisingly, the person I spoke to knew absolutely nothing about the document.  Getting updated software required speaking to the software licensing department, which I called next.  I found that the last version of software to support the FDM 1600 was Insight 6.3, and that yes, I could certainly get a copy.  However, that would mean paying $2000 for an annual software support contract.  Oh, and the FDM 1600 is being dropped from support at the end of the year, so it would essentially be $2000 for 6 months of software support.  I declined to sign up.

Friday I decided to try and tackle the offset between the ABS and support material.  The manual explains the procedure, but references a calibration filename that doesn’t exist in my software version.  After opening a few different similar files, I finally found the one I needed – one that would print out a square of ABS and then deposit a single circuit of support material on top.

I’d run one, estimate the offset, then launch the tip calibration program to upload the offset to the FDM 1600.  After a few iterations, I finally was pleased with the path matching and decided to try a larger print to run overnight.  I grabbed the Mendel SolidWorks file and created an assembly with some of the larger parts that don’t lend themselves well to machining easily. This step is needed because Quickslice can only work with a single .STL file (whereas Insight can work with multiple objects). In order to print multiple parts, I have to create a SolidWorks assembly of how those parts are to be printed, and then export the entire assembly as a single .STL file.  Quickslice churned through the slice and road generation slowly – I’m guessing this old software was far from optimized.  Once done, I uploaded the file to the FDM, hung around for a little while to make sure the base layers were being laid down correctly, and left the machine to its business for the night.

I wasn’t sure what to expect the next day – RepRap machines seem to still be at the stage of needing a lot of monitoring during a build, and I worried that I’d come back to find a giant birdsnest of filament occupying the build chamber.  But no such disasters had happened – I had a beautiful print waiting for me:

It took a bit of coaxing to remove from the foam, but I could then admire the work.

I’m not sure why Quickslice decided to run supports up on the left – the teardrop holes on the Mendel parts are less than the overhang support threshold in Quickslice, but perhaps a rounding error tipped the scales.  This print took 14 hours, so the FDM is no speed demon.  I think a Mendel could probably run the job in half that time, although RepRap currently bypasses bases and supports, and has a coarser output due to the 0.5mm nozzle orifice.

I’m pondering whether or not to actually build the Mendel right now, but I’ll probably pass the parts over to Chuck so that he can have a crack at the project.  I’m still interested in the RepRap project, but if I were to build one, I’d probably redesign the system into something that looks more like the FDM, but has a larger working envelope.

I previously wrote about my start on building a Mendel RepRap.  I recently purchased the bearings and stepper motors, and have had the shafts and threaded rod at the ready for several months.  But on Monday, while scanning the local tools section on Craigslist, one of the titles jumped out at me: “Stratasys FDM 1600 Machine”.  Stratasys…  I could scarcely believe that someone was selling an actual rapid prototyping machine on Craigslist.  Not just a rapid prototyper, but a fused deposition modeler (FDM) – the very same type as a RepRap.  For anyone wondering just what in the world I’m talking about, imagine a CNC controlled hot glue gun – a FDM rapid prototyper builds up an object out of many thin layers of plastic extruded out of a tiny nozzle.  This particular machine has two nozzles – one for the plastic (ABS to be specific) and one for a support material (a very brittle plastic that supports overhangs and is broken off once the part is complete).

The asking price was simply too good to be true, and I called the seller immediately, who confirmed that it was still available, and that it was in fine working condition.  I told him I’d stop by around 6pm to have a look at it.  Well, perhaps my eagerness was far more apparent, as I told the fellow to consider it sold (I was actually shaking from excitement).  The size of the unit was probably going to be too big for my truck, so I called my dad to see if I could possibly borrow his.  As generally happens, I wound up borrowing dad as well (which is always a comfort when buying and moving machinery – something about decades of prior experience to keep me from doing something boneheaded).

Sure enough, the machine was just like the ad indicated.  The seller gave me a crash course in using the machine and the software, though I knew I’d have to do a lot of reading as well – running a rapid prototyper isn’t something you can really do with only 10 minutes of fragmented instruction.  The three of us managed to wrestle the unit and accompanying stand into dad’s truck and we were off.  Dad helped me unload the unit at work, where I hoped to be able to keep it indefinitely (as we’d probably find it useful to have for the engineering department – I couldn’t imagine having enough to run on it personally to keep the machine very busy).

It took a few days to get the unit situated in the office area (an office environment is needed as the machine needs reasonably controlled temperature and humidity), but I was finally ready to attempt actual use of the machine.  I had read enough of the manual to understand how to load the plastic filament (about 0.070″ in diameter, fed from a reel) into the extrusion head, so I started with this simple procedure.  I managed to get the gray support material extruding out of the nozzle, and then the red ABS, but when I switched back to the support material, there must have been a jam and no more came out.  I powered down the machine, let the head cool, and cleared the backflow from the head.  I tried it again, but once more I could get the support material to extrude only at first, but not after having switched to the ABS.  Close inspection of the head indicated that the gripper wheels weren’t getting proper traction on the filament (perhaps it was getting a touch too soft from the heat).

A few days later I felt ready to try actually making a part, even if the support material extrusion gave me issues.  I also powered on the air conditioner on the back of the unit for the first time, hoping that the cool air blown onto the backside of the head unit would improve my luck with the support material (this cold air flow seems to be critical in keeping the filament from melting before it actually enters the heating chamber).  Again, I was unable to get the support material to extrude after switching to ABS, so I decided to try just running without support material at all.  I was going to just try making one of the .stl files for Mendel, and then I remembered that the traditional first part made on a RepRap is the shot glass (or ‘minimug‘) to allow you to make a toast to your own creation.  While the FDM had likely made hundreds of parts already, I figured the minimug was the perfect inaugural part for me to try.  I drew up my own version with zero overhang and generous draft angles, not knowing how well things would actually work – best to start simple.

The file I sent to the FDM included about 1/16″ worth of support material as a base – since the support extruder wasn’t cooperating, I was hoping the ABS would ‘fluff’ itself up to form a reasonable support for the build (much like a basket of clothes dumped on the floor would have a height greater than if the items were folded and stacked).

This actually worked quite well, but about halfway up, disaster struck when the part lifted from the foam base.  After making a presentation in my experimentation class many years ago (detailing a blower and heatsink and subsequent failure of the test heating element when we applied too much power for the airflow to effectively remove the thermal energy), my professor implored me and my lab partner that there was no such thing as a failed experiment.  He was so emphatic about this point that it has stuck with me ever since, and I recall it whenever I am disheartened by a setback.  I suppose the old maxim “every cloud has a silver lining” is an approximation, though I feel the Dilbert-esque phrase of ‘learning experience’ is closer to capturing the sentiment of retrieving useful knowledge from mishaps.

While the top portion of the minimug looked pretty good, the underside showed a rather loose fill – the lines of ABS should be much straighter.  There is also very little adhesion of the ABS to the glass foam base.  I knew that I had to remove the dead space that should have been filled with support material, or increase the surface area of the part bottom.  I wound up doing both.  I figured out how to remove the support layer from the part (Stratasys Quickslice 6.4 is not exactly intuitive, dating to the previous millenium), and I added on a hefty plate to the bottom that would hopefully improve adhesion to the foam.  These two changes worked beautifully.

To make things interesting, I added some holes and flanges to the bottom flange of the part.  The very first layer was close enough to the glass foam base that the plastic was injected into the pores.  This worked almost too well to secure the part, and I needed a putty knife to pop it off of the substrate.

I then decided to give the support material one more try.  The ABS had absorbed a good deal of moisture – the manual instructs to test for ‘droop’ by extruding some ABS, wiping the nozzle, and then seeing how much leaks out over the next 10 minutes.  The more moisture the ABS contains, the more the material will ‘droop’ – an acceptable level of moisture is to have 0.5″ or less of ABS dripping out of the nozzle after 10 minutes.  I was seeing several inches after 5 minutes.  I guessed that maybe just running through a foot or so of support material through the head might finally get to a drier section of material (the spools of plastic filament are kept in a ‘dry box’ on the machine with desiccant bags).  Wonder of wonders, this actually worked (or perhaps I was simply just lucky this time) – after feeding support material through the head, I was able to switch to ABS and back again and be able to extrude at will on both nozzles.  I loaded up a small Mendel part and gave it a go.

This support (which appears to be called the ‘raft’ among RepRap users, though they currently have no separate support material) looked to be adhering to the substrate very well.

Once the ABS started being extruded, however, I could see that all was not well – the support material should be perfectly centered between the two uprights.  Still, things were progressing smoothly, so I let the part complete.

I’ll need to calibrate the head so that the support material gets extruded properly in line with the ABS, but I’m extremely pleased with the results thus far.  In short, I’ve leapfrogged well beyond where I would be if I actually had a working Mendel.  Rather than trying to figure out how to make the machine work, I can now decide what I actually want to make.