MakerBot Industries

Inspired by Nophead’s excellent work, I decided to do my own testing. I had the folks up at Najet make me a prototype and pitted it against the existing timing belt pulley from SDP-SI. As with Nophead, I found the worm-gear style pulley to be much superior.

My process was pretty simple. We’re testing the pull strength of the extruder only here, so it involved hooking up the filament drive portion to a force meter (fish scale) and running it until it either stopped or failed. Here’s my setup.

Force Meter
A fish scale 50lbs / 23 kg. I couldn’t figure out how to make it do metric. FAIL! It still worked reliably, and I got a rough sense of what forces were involved. Now that I know my range, I am purchasing a much nicer digital force meter. (Its held down by capa / shapelock… I know, I should really have printed that part)

Filament Harness
Since we’re measuring how hard the extruder pulls on the filament, we need a way to hook up the filament to the force meter. I ended up bolting the filament between two washers and then looping that over the hook. I only rarely had problems after that.

Extruder Drive System
No need for the hot end or anything like that for this test. I laid the electronics to the side. Then I just clamped it to the wood, making sure to clamp in the middle of the dinos and not too hard. The bolts dug into the wood after a bit and made it snap right into place.

Then I just did a bunch of test and measurements. The fish scale did not have a very fast update rate at all (5 seconds?) It also did not have a peak weight or anything like that. I just watched it and tried to remember the highest number. This was a big deal because sometimes the filament itself would either break or elongate and deform. There readings could be slightly low because of that. It was also a bit chilly in the Botcave (its late december but I’m not going to go that deep on ya.

Here are the results as a google doc:

As for the extruder pulley, yes we will have these eventually. No idea on the timeframe, but having to wait is the price we pay for being as open as possible. I also made the silly mistake of making the diameter of the threaded parts different from the diameter of the current pinch wheel teeth. Its not much of a difference, but I’d like it to be a drop-in replacement. Its a really daunting thing to go into production on a part without a prototype, so it means I’ll probably need to get a new one made. I might have a few other variants made and see how they match up with the results from this one.

I love the open source hardware community. MakerBot operator Cid Vilas just posted these amazing pics to the MakerBot Operators group. Check them out!

I saw this photo in the MakerBot photo pool. Bas Piljs used the scrap lasercut part that holds the brackets as a sorting tray to keep the nuts and bolts for assembly separate. So elegant! I’m definitely using this idea in the future.

Ever since I got the Frostruder working on air, I haven’t been able to stop thinking about the possibilities of air-powered plastic extrusion. Today, I was able to put together a test rig and get some preliminary impressions. The verdict is that yes, air power seems to be a viable path forward, but there are definitely some hurdles to overcome.

My test rig was really ghetto, mostly because I didn’t have the right pipe fittings to make things easy. I managed to cobble together a monstrosity that did indeed hold pressure and work. The basic setup was: quick connect -> compression fitting -> PTFE tube -> compression fitting -> nozzle. I used some stainless steel nozzles that I got from McMaster. I used these because they have 1/8″ NPT connections on them. I have no idea what the nozzle profile looks like, and I had to modify them to get better extrusion. More on that later.

I used a standard extruder setup: nichrome wire + a thermistor + kapton tape. Since the hot end now had lots more metal in it (a bunch of pipe fittings connected to the nozzle) it took quite a while to heat up (2-3 minutes) but it did indeed get up to 220C. The entire assembly got really hot to the touch, so of course i ended up with lots of tiny burns. The PTFE tubing held up great in the compression fitting. We’ll need to do a long-term test on that to see just how well it actually holds up at temperature though.

As for the nozzle, I had to modify it to work. It comes with a wicked looking 2-inch blunt needle tip. I cut it down to ~2mm with a dremel. I tried to extrude with this and it just barely went. I got the dremel out and cut it flush. It extruded better, but still not so good. I started hacking into the nozzle body and cut away a good chunk of it to shorten the nozzle orifice length. That made a big difference the shorter I made it. If this proves to be a viable path, we’ll get custom nozzles machined with a short orifice distance for the best extrusion speed.

The extrusion itself was relatively slow. I didn’t measure it, but it looked like 10-15mm/sec @ 120PSI w/ 0.8mm diameter. Unfortunately this is much slower than the current MK4 which extrudes at ~26mm @ 0.5mm diameter. I’m going to try and tap a pipe plug to M6 and see how it compares when using the exact same nozzle as on the standard designs. That will be some excellent data. It definitely seems like we’re going to need to use much higher pressures if we’re going to extrude plastic with pressurized air/gas.

I started researching various options for higher pressure sources and came across a very interesting possibility: compressed CO2 cylinders. These are used all over the place, from paintball to welding to beer pouring systems. CO2 itself is really cheap and ubiquitous. C02 tanks also put out a steady 800PSI at room temperature! That is definitely going to be more than enough for our uses, and we’ll need to put a regulator on it to get the proper temperature out of the tank. The nice thing about CO2 is that it would be silent as well, compared to an air compressor which is typically very loud.

Another interesting challenge is delivering the feedstock. The air pressure does not just automatically push the plastic out the nozzle. Rather, it tunnels through the plastic and when the air pressure reaches the nozzle, it blows the plastic out and then jets air out the nozzle tip. If you push on the plastic to fill that gap, it will continue extruding and will extrude nicely for a couple minutes until you add more plastic to the mix. I was using filament in my setup and was pressurizing the entire PTFE tube that had the filament in it. This also meant that I had to de-pressurize and push the filament down in order to plug/fill the tunnel that formed. With air powered extrusion, we’d need to find a way to either feed filament into a pressurized system (some sort of o-ring?) or find a way to load pellets in at regular intervals. Perhaps a hopper with a servo-operated valve that would let the chunks of plastic enter the melt chamber?

A few of the hurdles / things to be researched:

* what pressures give us useful extrusion speeds (150, 200, 250?)
* can we find solenoid valves that handle those pressures and aren’t unobtainium?
* how much CO2 does a print need / how often would you need to fill a 1LB tank? (volume usage of a print = volume extruded + pressurized chamber size * extruder starts
* method for refilling plastic for printing (filament or pellets)
* can we use an o-ring to seal off the PTFE tube and thus be able to feed in filament continuously?
* can we create a hopper system that can sporadically fill the melt chamber with plastic pellets?
* can we do this safely?

The joy of R&D is that moment when everything comes together and there’s a shiny bit of goodness in front of you, gleaming with potential. This is not one of those stories. One of my research goals is to ‘close the loop’ on a few of the MakerBot subsystems. This is tech-speak for adding feedback systems so that you can know if your motors (or whatever) are doing what you tell them to do. In particular, I want to close the loop on the XYZ motors so that we know the exact position of the bot at all times.

My first crack at this is based on a cool chip by Austria Microsystems called the AS5304/AS5306. These tiny little chips are magnetic linear encoders (measures movement on a line). I’ve used the Austria Microsystems chips before and they’re really nice. In particular our magnetic rotary encoder board (measures rotation) uses the AS5040. The linear encoder chip has some really nice specs: resolution down to 15microns (0.015mm or about 1600DPI) As you can see, this chip would be a really nice addition to our MakerBot arsenal. What that means is that you put down a strip of adhesive backed magnet and then run this chip over it. It will tell you your exact position over the strip down to 15 microns. Hawt.

So, I modified the rotary encoder design for this new chip, including creating a new part in Eagle. I looked over the datasheet and everything looked good. I placed the components, then slept on it. I routed it, then slept on it again. I saw a few flaws and re-placed and re-routed it and then pronounced it ‘good’. I sent it off to the fab and moved onto the next board. Well, unfortunately I made some mistakes. First, the AS5040 chip which I copied for the footprint is wider than this chip, even though they have the same pin spacing. CRAP! The pads dont line up. The chip definitely still shows great promise, so I’m going to fix this mistake and send it off to the fab soon.

Moral of the story: always print your design and test the footprints against the real chip. I usually do this, but I was in Portugal at the time and wanted to have a bunch of goodies when I came back. Guess I learned my lesson.

Oh, and if you’re interested in this design… we always do our dev out in the open and this one is located in subversion.

We’ve released an updated version of skeinforge. Nothing major has changed with the script itself, but we have updated configuration settings for all bots from Batch 5+. Actually the new settings apply to anyone with a MK4 plastruder and/or the new nozzles from Najet.

You can download the 0006 release here.

If you’ve used skeinforge before, you need to move/delete your old preferences directory before the new ones will take effect. This is something we’d like to change, but for now thats just the way it is.

Under linux/mac you can find the folder in your user directory at: ~/.skeinforge
Under windows, you can find the folder in your user directory at: C:/Documents and Settings/user/.skeinforge (or something like that)

If you mess up or want to use the old settings, replace the defaults directory in skeinforge with the prefs/foo directory of your choice.

These past two weeks, I’ve been honored to run an intensive MakerBot workshop in lovely Lisbon, Portugal. The goal was simple: build two MakerBots, and print print print print print. We achieved goal #1 over the course of two evenings. Grand total of about 8 hours for 10 people to build 2 makerbots. It was really awesome.

Of course that left us with 12 days in which to design, print, and learn all there is to learn about how to do 3D printing on a MakerBot. Many wonderful things were designed, printed, and shared with the world. In no particular order, I’d like to present some of the interesting results of the workshop

Radiohead’s Thom York
This one was done by Tiago Serra and turned out great. He also managed to get reblogged on a bunch of other sites too. Click the image to view the model on Thingiverse.

Experience
This piece, entitled ‘Experience’ by Sónia Moreira was a great bit of art and very interesting to hold in your hands. I believe this was her first time using 3D modeling tools and was made using Wings3D and Blender. Click the image to view the model on Thingiverse.

Servo Suspension System
This design was very cool and very practical. Created by Guilherme Martins, it is a servo mount with integrated suspension system intended for use in a robot. Very cool! Click the image to view the model on Thingiverse.

PS3 Eye Support
Another really nice, really practical item was this custom support for a camera. It was designed by Sergio Ferreira for use in an art installation involving ants (also really cool!) Click the image to view the model on Thingiverse.

Pink Panther Woman
Last, but not least, we have this provocative model by Pedro Januário who is a very talented 3D modeler. He did many interesting models, but this was by far the most impressive. While printing this piece, he shattered many records, most notably the Longest Print, clocking in at 4+ hours. Click the image to view the model on Thingiverse.

Of course there were many, many other things designed and printed during the course of the workshop, and these are just a sample of the cool stuff that was done. Congratulations to all the participants and keep on rocking the 3D world.

One of our goals when we started MakerBot and designed the CupCake CNC was to automate one of the most tedious things of all time: frosting cupcakes. This incredibly difficult task has plagued mankind for centuries, but at long last we have found a reliable way to automate the process.

Our original design for a frosting extruder followed in the footsteps of Fab@Home’s paste extruder. They have a fairly elegant solution that uses linear actuators. Unfortunately, the linear actuators themselves cost more than our target price for the entire kit, so we attempted to come up with an emulated design that uses standard motors, lasercut gears, and some threaded rod to create a motorized plunger.

Well, this worked up until a point, but it was a pretty error prone and bulky solution. The gears were difficult to attach. The threaded rod based plunger required a design at least 2x the height of our desired syringe (60cc) and basically was heavy and didn’t work well. It was certainly an interesting design challenge for Bre, Will, and I that saw about 7-8 iterations and a failed appearance on TV, but ultimately it was fruitless. There’s a reason we never released the MK1 for sale.

Not only that, but the motorized piston based solution is an inherently flawed approach: The extrusion of a material is based primarily on the pressure, viscosity, and nozzle diameter. There’s not much you can do about viscosity and nozzle diameter, so we’re mucking with pressure. In the motorized piston approach, the pressure builds up gradually as you push the plunger down, and releases as the material either leaves the syringe or you back off the plunger. The result is that you either have extremely slow start/stop times or you have to deal with massive ooze problems.

Which brings us to the MK2. I was musing over the design failures one day when I had the idea that instead of trying to create the pressure in the syringe mechanically using a plunger, what if we directly applied pressure using air. I reasoned that we could use a commonly available air compressor and electrically controlled solenoid valves to push frosting out of a syringe tip. I had this idea for about 6 months, but it had to sit on the backburner for a while until I had a chance to work on it.

My first experiment was with some thick, chocolate frosting that you can find in nearly any grocery store. I simply wanted to see if it was possible to use air pressure to extrude frosting, so I wired up a solenoid to a switch and used that solenoid to turn the air pressure on and off to the syringe. I was using a 21GA (0.53mm) needle and a standard 60cc syringe. I hooked it up to the air pressure and opened the valve. Nothing happened right away, but I gradually turned the pressure up until about 50-60 PSI I started getting a frosting extrusion. I kept turning up the pressure to about 80 PSI where I got a really nice, very fast frosting extrusion that was about 0.5mm wide. Success!!!

Well, it wasn’t a total success: when I closed the valve, the syringe was still pressurized and continued to extrude. Obviously that is a problem, so I went back to the drawing board. I came up with the idea of adding a 2nd valve that would act as a relief valve and release the built-up pressure to the outside world. I grabbed a second valve and went back to the garage to experiment. The result was phenomenal! I was able to start and stop the extrusion at will, with zero oozing problems. This was excellent news. I soon had an Extruder controller wired up to the solenoids and a tester gcode script that would cycle the valves every 10 seconds. It was amazing to see a stream of frosting coming out and stopping every 10 seconds. I ran to get Bre and Adam to celebrate and we danced a frosting dance.

So: fast forward a few weeks and we’re gearing up for the Yahoo Hack Day in Times Square. We really didn’t have a solid game plan for what to do, but we knew we wanted to have fun and stay up all night hacking. I was really into the new frostruder design, so I brought it with us along with a portable air tank and a bike pump. We spent all night hacking on the frostruder and trying a variety of edible materials (frosting, peanut butter, and jelly). We ended up winning the Best Hardware Hack category with our New York Toast entry. It was a fun, fun hacking adventure.

For more info, check out Thingiverse and the MakerBot wiki.

One of the nice things about being an open source hardware based company is that we don’t need to keep our R&D efforts under wraps. Infact, we do all of our R&D out in the open. Our public facing subversion repository (http://svn.makerbot.com) is where you can find the latest developments that we’re working on. As all of our hardware designs are based on digital files, its easy for us to post them for all to see.

The reason behind this post is two-fold. First, I’d like to start blogging regularly about new developments. Things like new prototypes, new ideas, and stuff like that. Some of these things will probably be failures and would otherwise never see the light of day. Others may work but not be practical or useful. Some lucky ones will work great and fit with what we want to do and get turned into products. All of this information is useful and valuable the community and should be shared.

The second reason is that I’d like to give a bit of a ‘behind the scenes’ look into MakerBot, how we operate, what we’re thinking about, and perhaps show ways in which the community can participate in the development of our shared technology. If it shows that we’re not some faceless MegaCorp of just 5 people, well thats a nice bonus too. To start, I’d like to give a tour of the various resources where development is taking place:

MakerBot Subversion
URL: http://svn.makerbot.com
This is the main repository where most of our development takes place. This is where our electronics, cupcake cnc, plastruder, frostruder, and many other things are stored. Each of us also has our own personal directory where we do more experimental work. Here’s a few useful directories:

http://svn.makerbot.com/trunk/ – this is where the main development happens. we try to keep it stable, but dont assume things from this folder work. this is where the latest and greatest stuff is though.
http://svn.makerbot.com/tags/ – this is where we ‘tag’ the releases once they’re ready for public consumption. you can get it from here or our google code project.
http://svn.makerbot.com/users/ – these are our own personal directories where Bre, Adam, and I do our hacking. We usually put stuff in here that is highly experimental that may or may not work. Sort of a playground for hacking, if you will.

Github
Url: http://github.com/makerbot

We are trying out Github as our repository for the Firmware and ReplicatorG. Github and git in general are really great for distributed development. Since people are more likely to collaborate with us on the code side of things, we wanted to make it as easy as possible for them. If this experiment works out, we may switch over our entire repository to Git. Who knows?

http://github.com/makerbot/G3Firmware – this is where the latest 3rd generation firmware resides.
http://github.com/makerbot/ReplicatorG – this is where the latest ReplicatorG code resides.

Community
Of course all the sourcecode in the world is meaningless unless you have a community. Most of the Makerbot community resides on the MakerBot operators group. There’s also a growing dev list for ReplicatorG. If you’re interested in helping out… just dig into the source and see what you think. Hit us up in the makerbot operators group with any progress you make!

A few weeks ago, we attended the awesome Yahoo Open Hack Day in Times Square. We brought along a couple makerbots, some prototype extruders, and some half formed ideas. Over the course of the 24 hour hackathon our ideas solidified. We had a prototype frostruder, various things to frost with, and we set forth to print ‘All the news thats fit to eat.’ Along the way we bumped into Richard Morris, a producer for CNN that shot some video of the bot. Somehow we even made it into the final cut. Yahoo hack day was a really great experience and we had a blast. Yahoooooo!