Sunday, April 29, 2007


Aluminum Barrel Refactoring

After fighting with software and firmware all day, I decided to take a break and tackle something I'm comfortable with, something real. I had a couple incremental improvement ideas I wanted to try out on the aluminum barrels.

First off, I have a step bit, which is a triangular bit with varying diameters on it as you go up. I wanted to enlarge the inlet hole on the extruder barrel. My thinking is that it if the filament fed into the barrel comes in slightly wrong, then it might get jammed. if we make the orifice a bit bigger, then it should just guide the filament in properly.

Also, I noticed that the bolts werent going all the way into the acorn nut, creating a cone shaped gap between the bolt and the nut. This was causing the plastic to pool. So, I used my dremel and ground off the first 'row' of threads on the bolt. that way it sticks out a bit further and should cut down on that pooling.

Finally, I wanted to try a different strategy with wrapping the barrel. Previously, I was using tape to mark how far the nut goes in. This was not quite optimal because the marking could be off and then the heater wire may be too far up the barrel. Also, i wanted to concentrate the wire closer to the nozzle. Instead, I wanted to use the nut itself as the guide to where the wire should be wrapped. To accomplish this, I wrapped a plastic bag around the tip of the bolt, then threaded the nut onto the bolt, then pulled the plastic around the nut and taped it down. The plastic allows me to use jbweld without worrying about permanently attaching the nut to the bolt. After it dries, you simply unscrew the bolt, then peel the plastic off the jbweld. Nice, huh?

See more on Flickr...

Saturday, April 28, 2007


Memories of Duco Cement

I finally got down to the hardware store and bought some JBWeld like Zach suggested for tacking down nichrome. I looked at how long it takes JBWeld to set, however, and also bought some Duco Cement. I hadn't seen that since my childhood when I used Duco and Testor's cements to glue together plastic model airplane kits. I was surprised to see that they even made Duco any more.

I set the screw terminal board down with a brass screw and set the comms and power wires down with a paper stapler. The fit on the encoder magnet is perfect. I tested it and it pulses smoothly. There doesn't seem to be any problem with interference from the gearmotor.

Now gravity works to keep the encoder chip on the magnet instead of pulling it away.


Fixing the z-axis encoder

The AS50305 chip that does shaft encoding for the z-axis is in a very awkward location. The little board that holds it comes out of alignment quite easily and when it does I lose z-axis control of Tommelise's positioning system. When I was tapping the z-axis controller once per layer this was no big deal. Now, however, I have to use the z-axis every time I jump from one line segment to another.

As you can see it is under the x-axis gantry and behind the xz positioning plane. That means that if it gets out of alignment you're working with a flashlight and your fingertips and praying that you don't pull one of those fine little filaments that connects the AS5035 to the stripboard fragment that it's mounted on.

(Read the rest of the story)

Friday, April 27, 2007


Ready to print

It's a long story and a story for another time, perhaps. I finally got to the bottom of the problem with the x-axis drift. It was a firmware issue with how I structured the interrupt routine. Here you see a HDPE raft print of an x-axis raft after I figured out how to deal with that problem.

As you can see the x boundaries of the raft are both straight and parallel to each other. As well, you can see a y-axis raft.

Again, no problems. There are no fudges or correction factors in the code any longer. The firmware/hardware interface is sound.

We are ready to do some serious printing.

Thursday, April 26, 2007


Had some 16f648's laying around, so I used em..

I thought I had some free samples laying around and they were 16f648a's instead of 16f628a's. Since I'm getting ready to potentially mess one up with my reprap pic programmer (coming soon TM) I thought, why not use one of those. Then I had another thought, how about just programming it with my test firmware and see if it blows a circuit or something. Well no smoke. I was able to actually communicate with it. It seems to run my firmware. I was thinking that it was running it more slowly but I didn't set the speed to 255, so it may be functioning 100 percent in place of my 16f628a (as far as I'm using it with the programmer anyways). Just thought I'd post this and see if this surprises anyone. I was :) Oh, I just checked, set speed back up and ran it again. Looks like a go!


Serious printing

My first large order of 3 mm HDPE arrived today, 15 lbs worth @$4/lb material + $2/lb shipping.

This was good, because I have just about half a pound of HDPE continuous filament left. I also dropped by the art store and got two more sheets (12 ft^2) of foamboard at $0.75/ft^2). I shouldn't have to restock for several months.

Now I can do some serious debugging and printing.

Wednesday, April 25, 2007


JBWeld Nozzles: Busted

In my recent quest to find the 'perfect' extruder barrel, I've tried a variety of things. I've recently become obsessed with JBWeld, a widely available high temperature epoxy. I first starting using it to attach the nichrome (heater) wire to the heater barrel. It works really well for that. So, I thought.. what else could i use it for?

One of the things I really want is for the barrel to be as easy to make as possible. I'm using a 3/8" bolt with a 1/8" hole drilled through it for the filament to pass through. This allows the nozzle to screwed on and be replaceable / interchangeable. Borrowing the from both Forrest's and Vik's nozzle design I came upon an interesting idea: use JBWeld as either the nozzle itself, or use it to attach a thin metal piece to at nut which would then form the nozzle. For the thin metal piece, I simply drilled a hole in an old penny.

As it turns out, this doesn't work at all. The barrel heats up, the JBWeld loses its 'gripping' strength and it falls apart. Note that the JBWeld on the barrel holds up very well: theres no stress on it. Its the JBWeld that holds the penny to the nut. On the plus side, the plastic melted quite well, so I'm confident in the barrel design.

Another problem I had was that I was using steel bolts and they were RIDICULOUSLY hard to drill out. I basically trashed two 1/8" drill bits trying to drill through them (and i was using lubricating oil too.) So, basically the whole method needs to be changed.

Looking Forward

Seeing how steel was hard to work with, not a very good heat conductor, and quite heavy I had to come up with an alternative. I did some internet searching for a softer metal and found some VERY interesting facts. I found a metal that is very soft (easy to work with), very good at conducting heat (melts plastic well) and is decently cheap. It is also something that one could cast in their backyard if they were so inclined. Can you guess what it is yet? Aluminum is the answer!!

I did some further searching on and found that they sell hex head aluminum bolts. They are a tad pricey (~$0.75 / bolt) but they will work well for me and only one is needed per extruder. This was nothing revolutionary, but rather a rethinking of the working material for the extruder barrel.

The tough nut to crack was the nozzle tip. Basically, I need something that can screw onto the extruder barrel, and also cover it. Nuts are a logical solution, but they are open on two sides... you basically have to weld/braze/solder a cover on it. After more searching, I found an AMAZING substitute that works so well it made me jump up and down. It is called an 'acorn nut' and it is basically a nut that is closed and rounded at one end! Infact, McMaster sells them in aluminum, and they are also similarly cheap (~$0.25 /ea) I ordered some and they should be here by the end of the week. I'll definitely be posting the results of this excellent find this weekend.


Tweaking the x-axis

You'll recall some time back that I discovered that I could get rid of a lot of the curling and distortion when I was printing something out of HDPE if I first put down a single layer HDPE "raft" like this.

What you may not have known about this raft is that the extrusions run in the +/- y direction. Notice how even the x boundary is. Heretofore, I've been noticing x-drift when I tried to build something vertically. This last weekend when I was upgrading the control panel I wrote another button on the control panel to generate a raft where the extrusions ran in the +/- x direction.

This morning I got around to using it.

(Read the rest of the story)

Sunday, April 22, 2007

Well Forrest, here's me being stupid..

Anybody notice a systemic logic problem here?

void toggleP12v()
if(P12v == 0)
P12v = 1;
P12v = 1;
void toggleP5v()
if(P5v == 0)
P5v = 1;
P5v = 1;
void toggleClock()
if(Clock == 0)
Clock = 1;
Clock = 1;

Cut and paste can get you! Lol they turn on just fine by the way.


What to do next

Just figured out that the polymer pump contains about 24-25 cubic centimeters of polymer and that it's going to take about 9 hours to print it. Keep in mind that I'm running this thing about as slow as I can while I learn how to run it. It's going to be going a lot faster as I get more confident with what I'm doing.

Right now a single axis speed is running 1.7 mm/sec and I'm using single axis prints so that I don't have to upgrade the control software to regulate polymer flow rates. Using diagonal infills will crank that speed up to 2.36 mm/sec. Shifting from 3/8-24 to the much more common 3/8-16 threaded rod will kick that speed up to 3.5 mm/sec.

Shifting over to 14:1 gear ratio hack GM2's, something that I don't intend to do with this model of Tommelise will push that speed up to 8.38 mm/sec. The Mk 1 AEM, as it stands, can pump polymer at an upper rate of about 12-14 mm/sec, so there's not a lot of redesign needed there.

So, for Tommelise 1.0 running at 3.5 mm/sec I can get a production rate of about 6.3 cubic centimeters/hour. Tommelise 2.0, should be able to achieve 15 cubic centimeters/hour.

One big thing I want to try in Tommelise 2.0 is to see if I can spin the thrust collar nut with a gearmotor rather than spinning the whole threaded rod. That should give the system a lot better response to changes in speed and direction and should save beaucoup current. It will also let me dispense with the flanged bearings which cost about $3.80/unit and make up over $20 of Tommelise's parts cost, which is just ridiculous.

Coming back from pie in the sky by and by, the new bushing in the Mk 1 AEM extruder seems to be working fine. I had to get used to the extruder putting out a lot more polymer for a particular setting than it was during the past week because the polymer pump is working a lot more efficiently than it was

I've discovered that I have three distinct vertical sections in the polymer pump instead of two. Again, that's no big deal. I've got the first two working with infill and I should have the third layer, for which I already have the perimeter working with an infill in the morning.

I've apparently got the drift problem sorted out. I ran several layers of the first two vertical sections of the polymer pump and that's looking good.

Now to look at problems I'm having and deciding which ones I should work on next.

First, I need for the control panel to pull the extruder head away from the print when it's finished printing.

Second, I need for the extruder head to lift a few millimeters when it finishes one line segment before proceeding to the next one, proceed there and then set back down to extrude. This should be fairly easy.

Third, I need to see if I can get away without pausing when I get to the end of a line of extrusion. This is a firmware issue. I think that I will try this first in the morning.

Fourth, I've noticed that features in a print job that consist of a bunch of short extrusion lines tend to melt down when they're printed simply because the hot extruder head perforce stays over the same area much longer than would otherwise be the case. I need to see if I can confect an algorithm which will pull the extruder head away for a particular length of time so that.

Fifth, I need to figure out a way to interrupt a print job, recover the absolute coordinates where it was interrupted and then save back the remaining print instructions so that the job can be completed at a later date. That would allow me to shut down the printer at night and start it again the next morning. I'm presuming that it's going to be some time before I'm comfortable leaving Tommelise to its own devices unattended for long periods of time.

Saturday, April 21, 2007


More on the Mk 1 AEM extruder

A few days ago someone asked me for more detail about how the Mk 1 AEM extruder goes together. I'm a little hesitant to say a lot because it's not really finished in any meaningful sense. All the same, some more pictures and description would probably be useful. As I'd said before the Mk 1 AEM is basically a RepRap Mk 2 with some hacks and upgrades that I fancy are a bit easier to do and which enable it to pump HDPE at 150-200 degrees Celsius rather than CAPA at 80-100.

One of the problems that I faced in building up the Mk 2 kit that Adrian sent along so long ago was that I had worn out the gearmotor that he was kind enough to send along with it. While I was able to slot in a Solarbotics GM2 into the mounting the drive shaft of the GM2 was both shorter and more robust that that of the gearmotor it replaced. I tried to alter the coupling to accomodate the GM2 and ruined it in the attempt.

As well, I did not have ready access to M5 studding and had to replace it with 1/4-20 American threaded rod. Finally, I didn't really have the skill to mill down the threaded rod to seat properly in the bushings that I needed to make. In order to get around these challenges I adapted an early design that Vik Olivier had sketched out.

(Read the full story)


Doing a maintenance breakdown on the Mk 1 AEM extruder

Over the past few days I'd noticed that it was getting harder and harder to get the Mk 1 extruder to start pumping HDPE. I'd also noticed that the gap (at the top of the polymer pump between it and the filament guide was almost closed (gap indicated with red arrow). This is where most of the force is applied to the filament.

To my knowledge about the only way that that could happen was for the brass bushing (circled in blue) that keeps the threaded rod that does the pumping in place to have been taking some wear.

I've reported before seeing a black dust that might have been brass powder in the top of the polymer pump. When I broke down the polymer pump and removed the brass bushing I knew for sure then exactly where it was coming from.

(Read the full story)


The world class retard finally "gets it"

It seemed for the past few days that just about anything I did to help sort things out with Tommelise's x-axis problem just made things worse. Today, I even started having trouble with the y-axis.

It was the y-axis starting to have troubles that finally tipped me off a few minutes ago. I fixed the problem and both x and y axes started performing beautifully, not perfectly, but beautifully as you see here...

You're not going to believe what it was. If I had any shame, I'd be embarassed to admit what the problem was. Fortunately for you, I have no shame.

(Read the full story)

Thursday, April 19, 2007


Hammering Tommelise

Today, I finally got disgusted and took apart the firmware for Tommelise's microcontroller. I'd reported a few days ago that when I was printing out an HDPE raft to print on that a few of the tracks for the raft were mysteriously missing.

Microchip's 18F family of microcontrollers have a bigger interrupt stack and some registers that monitor how much loading is happening in the stack.

(Read the full story)

Wednesday, April 18, 2007


Nozzle Updates

Well, last night I went to the shop, and I was completely disorganized. I think I was just a bit frazzled from nuking my MacBook. However, I did find some interesting things out:

1. JBWeld holds up damn well under heat. I ran the 30awg insulated nichrome / jbweld barrel for an hour with straight 12v power and it held up fine. The coil was 14ohms, so it didnt get super hot, but it was definitely enough to melt the HDPE I had with me.

2. JBWeld is super easy to machine. I easily drilled 0.5mm holes in each of the JBWeld nozzle attempts. The only question remains is whether they can stand up to the forces of being an extruder nozzle. My gut feeling tells me no... but hopefully I'm wrong.

3. The mechanical pencil nozzles were super easy to grind down, giving a very nice tip. I had to make them nice with the top of a pushpin, but that was really easy. The end result is a very nice looking nozzle. I haven't tested it yet, but I have high hopes for this.

Now, those were the good things I learned. Here are the bad things:

1. keep a nut on threaded rod before you cut it (especially if you use bolt cutters + dremel to smooth the cut) its impossible to get a nut back on once you do that. if you keep the nut on, then you can force it off, and hopefully restore the threads.

2. i need a cross slide vice for my drill press. otherwise it is impossible to drill a 1/8" hole down a 1/4" bolt. twice i ended up coming out either side. either that or move to a higher size. i'm particular to 3/8", but i bet 5/16" would work too. i chose 1/4 because that is the size that the pencil tips screw on to, but if it is nigh impossible to drill correctly, then it could still be possible to jbweld it to a larger nut, then attach that nut to a drilled length of threaded rod. the pencil tip fit best in the 3/8" nut, so I'll probably try with that.

Tuesday, April 17, 2007


One possibility for ambient environment printing

I saw a video clip produced by the fab@home crowd over at Youtube that got me thinking. It was a rather simple thing, really, and at first blush it didn't sink in on me that I was seeing an opportunity.

The videoclip is of a fab@home printer making a watch band. The first thing you notice in this time lapse is that their syringe extruder that puts down the goo that they make their watchband out of hasn't got the capacity to do the whole job without being refilled. That distraction tended to make me miss the interesting bit that happened shortly thereafter. At 1:20 seconds into the time lapse clip they very briefly laid down the watch that they intended would fit into the band. It was only there for a flash and then the printer went on with making a socket for the watch. There wasn't much contrast in the clip so you had to watch the motion of the extruder head to be sure this was happening.

At 1:40 they dropped the watch into socket and the printer then proceeded to print around the watch's perimeter thus making it an integral part of the band.

(Read the full story)


Getting used to the sound of gearmotors

I have Tommelise running something like 18 hours daily now with about a 50% duty cycle. Those little Solarbotics gearmotors are a part of the background noise of my life now. While they're not particularly loud I can already see that acoustic covers for them is going to be a priority before too long.

When I shifted the control panel to a standard 0.1 mm resolution there was a bunch of fall out in tweak parameters in the microcontroller firmware. I suddenly started getting a mismatch between the outer perimeter of a layer and the infill. You can see what I am talking about here...

Notice the gap on the upper long edge closest to you. Getting away from cross-hatched infill had an unexpected benefit. It let me look at problems with the x and y axes in isolation from each other.

(Read the full story)

Monday, April 16, 2007


Mechanical Nozzle: Promising

I made a trip to Staples today after work, and was pleasantly surprised when I found some mechanical pencils with *just* the right tip. It is an all-metal 0.5mm tip with ~1/4" female threads. I couldn't find the right thread exactly, as they didnt have a good selection of screws. The 1/4" coarse threads did jam in there just fine... nothing a bit of force and jbweld cant make work. ;) However, I'm pretty sure that fine-thread 1/4" screws will work just fine... which means easy interchangeability.

One other thing that might pose a challenge is that the 0.5mm part is quite long... I'll be attempting to cut it closer (or flush) when I get to the shop tomorrow. I think with a dremel it should survive pretty well. My plan is to dremel it down flush, use a thumbtack to de-burr it / clear it out, and then bolt it onto the end of a 1/4" jbweld heater barrel. from there i just attach it to the extruder and its smooth sailing!

Also, one final experiment I did was to try and make interchangeable nozzles with jbweld. The process I used was this: Wrap plastic over threaded rod. Thread nut onto rod, until the tip of the rod is *almost* flush with the nut. Then, spread some jbweld into the small depression, and let it sit. (its important to let this setup vertically, as jbweld slowly settles and may run). The results look pretty good. I'm going to attempt to drill them tomorrow too. Hopefully if the mechanical pencil works out, this technique can be abandoned. The molding process was pretty easy, and simple. The upside of these is that they are very cheap and can be easily *made* at home. also, they are naturally flat on the bottoms, as opposed to the pointy tip of the mechanical pencils. I dont know if that will matter at all.


Walking before running and Zach's fudge factor

I've never had a lot of patience. I want to make parts and Slice and Dice is not quite up to the job, not so much because it doesn't do what it is supposed to but more that what I imagined that it needed to be doing isn't a whole lot like what Tommelise is prepared to do.

Fortunately, thanks to Zach Smith's kind assistance, I have an XML format file that Slice and Dice generates that the Tommelise Control Panel. It's pretty easy to hack an existing XML file into making something quite different. That's how I created the shot glass and that's for, for the moment at least, how I am creating replacement parts for the Mk 1 AEM extruder.

Following the principle that I must learn to crawl before I learn to walk I took a hard look at the polymer pump. When you take away the holes and the RepRap logo the RepRap Mk 2 polymer pump really has only two different vertical sections. This one...

(read the full story)

Sunday, April 15, 2007


Tommelise Mk 1.1 AEM polymer pump printing begins

I rescaled the Control Panel to a 0.1 mm base resolution, cobbled up a perimeter XML file for the upgraded polymer pump and had a go.

The measurements come out right save for the fact that measurements on Tommelise are from the centerline of the extrusion thread and not its outer boundary. For now, though, that's not something I'm going to worry about.

Changing the resolution gave me some x-axis drift again. Maybe this time I can see something systematic that I can do with the control parameters instead of just tweaking them till it comes right. I'll settle for a good tweak that works, mind. I doubt I'll be changing the resolution again. anytime soon. :-)


The adventure begins

Tommelise's Mk 1 AEM extruder looks great in print, but the visual reality leaves a bit to be desired. It's basically a hack built up of a RepRap Mk 2 parts kit with pieces of wood added to accomodate a rather different polymer pump design.

I did a lot of thinking yesterday about what I should do next. Indeed, I even set up a priority list in the Tommelise 2.0 Blog. This morning, however, it seems clear that if I really want to be ready for the Maker Faire and make good progress along the way, the first thing I should do is to apply the kaizen principle to the Mk 1 and at least get rid of the wood bits.

So that's what I'm going to do. I've sketched out the filament guide on my whiteboard.

This morning I'm going to cobble together an XML file to print it and see how it works out.

The way I look at it is this, HDPE is dirt cheap, Tommelise costs less than 5 cents per hour to run and I can't get back time I wasn't running Tommelise. Rather than spending a lot of time agonizing over 3D CAD sketches in Art of Illusion I'm going to rough out things, print them and see how they fit and then repeat the cycle till I have something that works well.

In short, for now at least I'm going to use HDPE the same way I used poplar wood when I was building the bootstrap Tommelise that prints for me now. I like to work with my hands. I like to visualise where I want to go with things I can look at and hold rather than squint at on a CAD screen. Prototyping has just got cheap for me and I intend to take advantage of that to boost my creative productivity.

Saturday, April 14, 2007


Getting started...

This is my (David McNeill) intro, and first blog post. I've recently read about Rep Rap, and am really excited by the possibilities. I came across Rep Rap while surfing on Open Source Hardware. I live in rural Franklin district, just south of Auckland, NZ. I've got strong skills in computers, Linux, software development. I've also got lots of spare computer and printer parts, a reasonable metal workshop, plenty of space, and no time.

So I'm going to build one! This is a real project to fire the imagination, and also to draw together many skills and resources that computer people have. Being open source is excellent, it just feels right. I've decided to build a Rep Rap 1.x, rather than a Rep Strap, since I want to build on the shoulders of others, and contribute what I find to the overall project, and to help make 2.0 better. Along the way I'll learn about all the associated technologies, and discover opportunities for using Rep Rap for useful things.

I've had a reply email from Adrian, so that's a good sign, the project is approachable and inclusive. And there's a few rapper's in NZ already, so that's good too.

So far I've read lots of web pages, downloaded (checkout) the plans with subversion, installed Art of Illusion (lot to learn in this package), had a look at how to install KiCad. My server is Kubuntu Dapper, and this workstation is Kubuntu Edgy. My assorted MS Windows PCs have been banished to VMWare jail, where they belong. KiCad is in Fiesty repositories, but that's not out yet, so I'll have to install manually.

Looking at the various posts it seems Rep Raps & Straps create a fairly big tangle on your workbench, look delicate, and need precise measurements. So I've decided to build a dedicated work space and trolley for it first. I've named it Rep Rap Roller (RrR). I can have all the gear together, and wheel it out of the way when stuck. Attached is my first rough drawing of what it might look like. I'll have to learn KiCad & AoI to make a proper drawing of it. I'll make it out of welded angle iron and plywood I've got lying around.

Until the bench is ready I'll be reading more, learning and preparing, and looking at material lists. I'm really looking forward to this project, something interesting, creative, inspiring and exciting!



Testing the envelope of the Mk 1 AEM extruder

I've been running Tommelise pretty much continuously (12-18 hours/day with about a 50-75% duty cycle). The controls are getting more reliable by the day making it possible to do things without hovering over it 100% of the time.

A few days ago I decided that I wanted to run the extruder barrel on the Mk 1 AEM extruder a lot cooler than the 200-odd degrees Celsius that I'd been running it previously. Here is the performance table that I developed.

Gearmotor Heater Temp (C) Amps Flow Rate
12% 60% 132 1.2 ~1 mm/s
25% 75% 152 1.5 ~2 mm/s
36% 75% 152 1.5 4 mm/s
50% 75% 152 1.5 6 mm/s
75% 75% 152 1.5 11 mm/s
100% 100%1952.0 16.8 mm/s

The important thing to remember about these tests are that they were run for periods of just a few minutes. Since then I've been running at settings for hours. Here is what I've discovered.

I started out with a setting of 40% max for the gearmotor and 75% of max for the extruder barrel heater. After about 18 hours of that you start hearing infrequent, intermittent clicks out of the GM3 gearmotor as the clutch slips that drives the Mk 1. The clutch slips at 0.4237 Newton metres (60 oz-in).

The motor for the GM3 will have heated to about 50 degrees Celsius and the 754410 that powers it to about 40. By way of comparison the axes motors run at about 30 degrees as does the 754410 that drives them.

From a cold start the Mk 1 could handle a 75% extruder barrel setting and a 50% gearmotor setting for about an hour and a 60% gearmotor setting for about 10 minutes.

The extruder cooled down completely in about 20 minutes and was able to repeat the cycle.

I tore down the Mk 1 to make sure that nothing untoward had happened inside to confuse matters. There was much less wear on the 1/4-20 threaded rod polymer pump than I expected. The pump was, however, liberally peppered with a fine black dust which I took to a fine powder of either brass or steel. I blew that out insofar as possible. My breaking down the pump, however, disturbed the dust and got it into the extruded plastic thereafter turning the next HDPE raft that I put down grey.

It's worth mentioning again that HDPE and HPP even more resist being cut by the polymer pump's threaded drive rod much more than polycaprolactone. This forces me to use a much tighter setting for these plastics. The tighter setting causes the polymer pump gearmotor to draw more current.

Friday, April 13, 2007


Don't go real fast, but I go pretty far...

I'm in the process of hammering down the software/firmware/hardware interface on Tommelise to where things are becoming routine rather than amazing. Putting down a HDPE pad seems to be the way to go for now for printing on Tommelise. I don't get peeling when I do that.

Here is a YouTube clip of Tommelise 1.0 putting the pad down on the foamboard.

It's hilarious that it's taking as long to get Tommelise to print something as it did to build it. I should have known that from previous experience, I know, but it always comes as a shock to find how difficult it is to get a built system to actually perform as opposed to just look nice.

I've got the square perimeter sealing well enough to hold water on a routine basis now. The drift problem has also become a tithe of what it was. That, of course, happened after I went to considerable trouble in the Control Panel software to record all of the feedback from the PIC that drives Tommelise onto a hard file that I could browse at my leisure.

Here is a YouTube videoclip of Tommelise 1.0 building the cube perimeter.

Anyway, I've got to take a closer look on how I'm handling corners on that cube. The extruder is both putting down too much plastic there and hanging around too long thus causing a problem with the corners becoming a bit melted.


Thursday, April 12, 2007


Hacking the GM2

I hacked one of my new GM2 gearmotors from 224:1 to 14:1 and replaced the existing GM8 on the y-axis of Tommelise 1.0 with it. With a 12v motor driving it the GM8 drives the famously warped y-axis at right at 70 rpm, which when you account for the thread pitch of the 3/8-24 threaded rod that I use gives you a measured traverse rate of 1.2 mm/sec.

On paper the hacked GM2 with a 12v drive motor should give us a traverse rate of roughly 12 mm/sec given the differences in gearing ratio. Unfortunately, the much lower torque that the lower gear ratio implies meant that when I replaced the GM8 with the hacked GM2 I was able to achieve a measured traverse rate of almost precisely 5.8 mm/sec.

I expect that if I were to break down and order some really straight threaded rods I could do considerably better than that. As well, if I were to shift to a coarse-pitched 3/8-16 threaded rod I could crank that up to 8.7 mm/sec.

I'm not going to replace my GM8s on Tommelise just yet. That would simply throw another variable into the mix that I don't want to have to deal with right now.

Doing a 45 degree traverse using both x and y axes running at full tilt gives you, of course, 2^.5 times those figures or roughly 8.2 mm/sec and 12.3 mm/sec respectively.


Can a shot glass be far behind?

I discovered that the shaft encoder magnet on the x-axis was not glued to the end of the shaft but only stayed there through the magnetic attraction of the encoder magnet to the steel rod. I glued the magnet back on and then ran another raft and square perimeter exercise.

While the resulting 20 layer square perimeter still had a drift, to the +x (right) this time, it was pretty continuous and the + and - x walls to the square perimeter were solid. The square and raft assembly was also rigid.

I looked at it a while and then got to wondering if it was also waterproof.

I poured some water into it and ... it was.

The Tommelise Mk 1 AEM extruder can make waterproof vessels in HDPE.


Gearmotors for Tommelise 2.0 arrive

I've been designing Tommelise 2.0 and have got to the point where I've been ordering parts. The gearmotors for her have just arrived from Solarbotics. They are GM2's with a 224:1 gear ratio and 12v motors. That's great for the the Mk 1 AEM (ambient extrusion modelling) extruder. A simple gear hack changes the ratio to 14:1 which, I hope, will have enough torque to let me extrude at a maximum speed of 15 mm/sec.

The price for the motors? US$ 22.00. Eat your heart out stepper motor people! :-)

I'm designing Tommelise 2.0 to have a large enough working volume to print all of its own parts.


Building things with popsicle sticks...

This print reminds me of the sorts of things we build out of popsicle sticks (the wooden stick in an ice lolly) at Bible school in the summers when I was very small. The upper left hand corner of the square was trimmed off a little when I used side cutters to remove a blob of plastic that the extruder, which I just raised after the printing, dripped on it.

Printing a raft of HDPE before you print the object pretty much gets us away from peeling and warping problems, it would appear.


A little better all the time...

I tightened up a few parameters and brought the x-axis drift down some more.

It's drifting to the left (-x) about 2.25 mm now in 20 layers. Fortunately, I can see that on the absolute positioning system measurements out of the firmware but I'm not quite sure how it's happening, though. At least I can measure it. That is infinitely better than what I had before.

The "blobbing" at the corners is largely happening because I'm stopping the extruder head there for an appreciable amount of time and melting down the corners. I'm going to comment out the 200 millisecond pause at the end of each extrusion segment and see if that helps.


I have to admit it's getting better...

I've shifted over to absolute positioning throughout and seem to be getting better results.

I've still got some x-axis drift but it seems to be consistent for a change.

Wednesday, April 11, 2007


Monitoring the x-axis drift

A few days ago I put in firmware code that kept track of the absolute value (in pulses) of the end points of the line segment extrusions that Tommelise was putting down. The idea was to get a look at this x-axis drift that I had been seeing.

It's important to know that I am not attempting to recalibrate the start point for the layers with each new layer, so errors are additive to an extent.

The pulses are right at 0.007 mm. I got up to layer #4 before the the line last line segment forming to square on the -x side fell off. This meant that the extrusion thread was 0.13 mm off centre. At first blush, considering that the thread as put down is right at 1 mm wide one wouldn't think that this should have caused a problem. I am, however, putting down a bit more polymer flow that I probably should. That seems to create a side thrust in the extruded HDPE from observation. It would appear that that makes alignment a bit more critical than one would ordinarily expect.


Note also the sudden shift to the left at layer 8. I'm still trying to figure out what that is about. My first impression is that it is a firmware coding problem associated with trying to figure out when to stop the gearmotors when one of the xy axis gearmotors exceeds its boundary goal.

Both the shifts and the degree of shifts match what the print layers actually look like.

Tuesday, April 10, 2007


An object lesson

I decided to see if I could get rid of corner curling by laying down a single layer of HDPE and then printing the 20 layer perimeter on top of that.

That appears to work. The object lesson is to never leave a hot extruder barrel parked next to something you've just printed. You see the result here.

I carefully sliced the whole platform off of the foamboard to see if the print was indeed almost perfectly flat on the bottom. While the HDPE raft on which it sits curled up on its edges the printed object did not.

Monday, April 09, 2007


Fix that drift!

I thought I'd throw these in just for fun. I have an x-axis drift that I'm pretty sure is in the firmware. I haven't been working on that problem but have been working with the flow rate and temperature of the extruder instead. This morning just for the hell of it I decided to see if I could stack 20 layers. Indeed, I can.

What's amazing about this is the tensile strength of the HDPE laid down in this slipshod manner. There is only a single thread of 0.8 mm HDPE making up that perimeter and the +/- ends of the x side of the print are sort of dare-me-and-I-do.

In spite of that I can NOT tear that flimsy looking thing apart with my hands. Mind, I have very strong hands, too.

Anyway, nothing serious. Just thought you all might like a laugh and a look at what REAL R&D looks like most times. :-D

Saturday, April 07, 2007


Controlling the Mk 1 Extruder

Heretofore, I have been running the Tommelise Mk 1 Extruder at fixed settings, viz, roughly 2 amps of heating current and a 25% setting on the gearmotor that drives the polymer pump. Now that I'm doing a lot of printing exercises I decided that I needed finer control of the extruder than that so today I upgraded both the firmware and control panel to allow me to set control parameters for current and polymer pump speed.

Once I had that done I ran a series of calibrations to give me an idea of what the performance envelope looked like for the Mk 1.

You can see a table of the values that I've measured for HDPE over at the Tommelise Blog.



Darwin RepStrap

Here is my weekend project the frame for Darwin made from MDF. I still have some to go like straightening it out. Also I found the mdf is fragile and I would not transport it to show it at all I plan on just using it to make an extruded machine..

Friday, April 06, 2007


Stopping corner blobbing

I added some code to the firmware which reverses the direction of the polymer pump for 25 milliseconds at the end of an extrusion path. The good news is that that seems to greatly reduce the corner blobbing. The bad news is that I now seem to have a drift from layer to layer on the x-axis.

I'm going to disconnect the new firmware code and see if that goes away. I'm beginning to think that using the WaitMs command when you are running interrupt driven code causes problems. It that is the case it will be no big problem because I have a 1 millisecond pulse written in to the interrupt code that I can count instead of using the WaitMs command.

Thursday, April 05, 2007


Getting the drift under control

I went back and programmed the PIC 18F4610 to do the things it ought to have been doing instead of depending on the overshoot hack like I was before. This is MUCH nicer than before.

The drift pretty much went away. I put in the maths to keep track of the absolute position of the extruder head and I'm down to a 5 pulse overshoot correction now. I set the first layer to smear, which seems to provide enough of a footing for the perimeter I'm printing to keep it from peeling.

I'm getting blobbing at the corners, but that is because I'm not turning off the polymer pump when I reach the end of an extrusion path and I'm pausing for 100 milliseconds between line segments. Otherwise the layering is looking very good.

The extrusion thread as put down is right at 1 mm wide and 0.5 mm high which is just about right for a 0.8 mm extrusion thread. The HDPE is pigmented white and looks just like toothpaste being extruded from the Tommelise Mk 1. It's a pleasure to watch in operation.

I'm getting to feeling really good about HDPE as a polymer for 3D printing.

Wednesday, April 04, 2007


Going over to continuous filament

Tommelise seems to be operating reliably enough that I don't have to watch it every minute so I shifted over from using 18" HDPE rods to continuous filament. I bought a pound coil from New Image some time ago. It should last me till the ten pound coil I ordered this morning arrives.

I was a little worried in that the coil was white rather than translucent. I did a work run on it, however, and it pumps and extrudes quite as easily as the HDPE rod that I have. I've got about sixty metres of the filament. That should last me for a few days at least.



I did a three layer print of the perimeter of the cube with no real problems. I then decided to go for a 10 layer of the perimeter to see what would go wrong.

Several things did, but none that were really troubling. First off, I got the thickness of my layer just a bit too big. For the first 5 layers or so there was no real problems, but by the 6th layer the additive errors got to the point where the extrusion thread began to dribble onto the previous layer instead of being laid down properly. You can see that at the upper right of the foreground perimeter. I should be able to diminish the layer depth and get that problem to go away.

I also observed warping at the corners of the perimeter. I'd seen that before and Vik has reported the same with polycapralactone. I'm wondering if that can be cured by cutting back on the extruder flow as I approach the corners.

Finally, I finally began to get some backlash problems with the y-axis after a while. You can see that happening on the left and right perimeter margins. I noted this before and will attempt to cure that by shimming the sliding joint on the y-axis.

Tuesday, April 03, 2007


Cracked the perimeter mismatch problem!

Once I corrected for the overshoot I did some counting and calculating and discovered that the number of pulses being recorded for the perimeters was the same whether I described the perimeters in a bunch of short lines or a few long ones. The only problem with that was that the perimeters made of short lines were much bigger, 40% bigger than the ones made with long lines.

There was only one possible explanation for that. Some pulses weren't getting counted. I hypothesised that perhaps I was turning off the interrupt monitoring too fast when I got to the end of a line extrusion. Instead of rewriting all the code to leave it on a lot longer I decided to back into the number of pulses that had to be missing. The number turned out to be about 30. I substituted that number for the overshoot estimate and then tried to print the first layer of the shot glass.

Bingo! The perimeter mismatch went away. That was the problem. Thank you Lord for watching over this retarded person.


So far so good

I put Tommelise back together Sunday and backtracked on the firmware to a point before a lot of the confusion began with control of the cartesian positioning system. I ran a bunch of cold extrusion exercises and finally decided that the gearmotors for this configuration were overshooting about 7 pulses. I hardwired that into the firmware and got 98% of the overshoots within a range of +/-1 pulse.

The square that I've printed is right at 30 mm on a side. I've got it scaled up with a pretty fast extrusion flow rate. It's easier to see what's going on that way.

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

Subscribe to
Posts [Atom]