Tommelise 1.0

I hit a technical problem back in September and you haven't heard much from me since then. I've been able to print in HDPE and am well on my way towards sorting out the problems with using brushed gearmotors to run both the extruder and the Cartesian positioning system of Tommelise. I'm publishing in the main blog rather than the builder's blog this time in that the issues that I am addressing impact on both Darwin and the various repstrap machines that are abuilding.

As you are all aware by now, I am sure, HDPE and to a lesser extent CAPA tend to curl when extruded at ambient temperatures.


This curling tends to happen at corners when the aspect ratio of object being printed approaches 1. When you have long objects you tend to get curling in the plane of the longest dimension. Nophead published some really compelling photos of this effect over in the builders' blog last month.

I got very discouraged about the warpage issue and resolved not to share my misery with the rest of you if and until I came up with a viable solution to the problem.

Some months ago, I printed an updated version of the polymer pump for Tommelise. I was able to defeat, to an extent, the tendency for the pump to warp during printing by using a lattice infill rather than a solid one.


It wasn't until I separated the pump from its HDPE printing raft and cleaned it up a bit that I noticed that the back side of the pump had warped all the same.


You can see that the curling is in the plane of the longest dimension in this case. You can also easily see a delamination occurring between layers roughly a quarter of the way from the bottom of the pump to the top. I will talk about this in more detail a bit later

Also notice that the top surface of the pump, the critical one, is perfectly flat.


Here's what I think has happened.

• I've printed the pump onto a HDPE raft much larger in area than the pump itself.
  
• The pump can't peel off of the foamboard simply because the contact area is much larger than it would have otherwise been.
• Printing each successive layer creates a situation where contracting plastic is creating a moment arm which tends to bow the accreting pump down into the foamboard. The foamboard is not mechanically strong enough to resist this downward thrust.

All that doesn't explain why the top surface is flat and the bottom one bowed, however. In that regard, here is what I think is happening. I've noticed that if I position the extruder nozzle too close to a surface the rate at which plastic can extrude is constricted.

Now whether or not the pump is trying to bow down into the foamboard, the extruder still tries to extrude in a flat plane. When the extruder head reaches the left and right extremes of the pump it is not able to put down as much plastic as it is in the centre simply because the distance between the extruder head and the last layer is smaller at the extremes and bigger in the middle.

If I were trying to print a solid infill this situation would have probably caused more distortions in the print than it did.

The top and bottom ends of the pump are square with the top simply because the cartesian positioning system keeps returning to the same perimeter with each layer regardless of the fact that the previous layer tries to shrink inwards a bit as it cools.

The delamination mentioned earlier happened at exactly the layer where I stopped printing one evening and picked up again the next morning. I had a bit of difficulty repositioning the extruder to start up again. This likely weakened the weld between the new layers and the old ones.

Proposed Solution

It would appear that we might well be able to cancel out the downward bowing of the print by printing on a solid piece of HDPE instead of foamboard. Solid HDPE has a high thermal resistance which should preclude extruded plastic from cooling off too quickly and not welding properly. Keep in mind that this delamination developed over several days and was not apparent when I originally separated the pump from the foamboard and cleaned it.

Such HDPE planks are easily acquired in the form of cutting boards at any hardware store. They are usually at least 3/8 - 1/2 inch thick and should have more than enough resistance to bending, especially if I use a loose infil in my print.

Of course, getting the printed object off the HDPE plank will be considerably more difficult than getting it off of foamboard. All the same, using a heated wire to separate the printed object or, failing that, a hand jigsaw should do the trick. As well, printing the piece on top of a cross-hatched rick strong enough to resist downward thrust, but less solid than solid HDPE should make separation a bit easier.

The HDPE baseboard should be relatively easy to resurface with a small belt sander. I checked and those are quite inexpensive.

Keep in mind that constraining the printed object in HDPE in this manner does not mean that it will not warp after it is separated from the HDPE baseboard. I'm hoping that it won't to any great extent, however.

Just got through with my first 7-toothed gear perimeter. The print layer has bad settings for height and flow rate, but it's pretty obvious that the work I put in on improving the positioning system, never mind making a new extruder barrel is paying off.  The teeth on the gear all look the same regardless of their location on the gear perimeter.  I made them really big so that I could be sure and not fool myself.


A lot of the problems are coming from the routine that I use to generate the perimeter.  It does the teeth twice at the same level for each layer.  I've got to fix that.

BTW,  I've discovered that you don't really need an LED to tell you that your extruder is operating.  The length of #32 Nichrome 80 from the heating coil back to the connector glows bright orange whenever the heater is active, so why duplicate functions?  :-D

I got the extruder barrel finished this morning and ran it in for a few hours. After the last coats of paint were done I rebuilt the extruder barrel/PTFE block/filament guide assembly.

When I rebuilt the barrel correctly on the second try I used the recycled PTFE block flange from the old barrel. That was warped, of course, and when I tried to reassemble the pressure system that holds the guide, block and barrel together, nothing wanted to fit right.


I solved this by putting the whole assembly in the vise and giving it a squeeze. While I was at it I tightened the nuts and bolts. That was handy.

I had a little trouble getting the alignment of the assembly right so that the filament went through the whole assembly as easily as it did just through the extruder barrel. Using a piece of #4 threaded rod let me get by that one fairly easily though. Some more thought is needed in this part of the design.



One thing that I haven't mentioned which wants saying is that you should never try to thread a bent piece of filament into the pump. As well, you should taper the filament as shown with sidecutters so that it finds the filament path easily and doesn't snag on the transitions between plastic and metal. That is important!

I reassembled the whole Mk 1 and mounted it in the vise for a test run of an hour or so.

It lit right up at a setting of 90% for the heater and 25% for the drive motor. I ran it a bit hot and let the filament fall free to the floor to see what I got. Here is the extrusion exercise starting up.



The Mk 1 did a very smooth extrusion that fell about 20" to the floor and coiled nicely, as you can see.  It looks like nothing so much as high quality fishing monofilament.


My son Adriaan tried to do a Tarzan on the coil to see if there were any faults in the extrusion.  There weren't.  It measured out at a diameter of 0.69 mm.  This is less than one usually encounters because, I presume, of the stretching the extrusion did on the way to the floor.  I was running it a bit hot, too.

I finished up the painting of the extruder barrel about dark yesterday and let the six coats cure most of today. Late this afternoon I designed a 2 amp heating coil of #32 glass fibre and Teflon insulated Nichrome 80 wire. It is rated at 0.32 ohms/foot. Using Ohm's law...

R = V/I = 11.5v (measured off of the power transistor)/2 amps = 5.75 ohms


LENGTH (inches) = (5.75 ohms/10.32 ohms/ft)(12 inches/ft) = 6.686 inches

(Read the full story)

I extended this to 6.75 inches to account for contact losses with the connector.



I wrapped the wire on the last inch of the extruder barrel as before and tied it down with very thin copper filaments taken from a piece of #20 stranded connection wire.



At that point I began to apply six more coats of BBQ paint to secure the heater coil to the extruder barrel.

One word of warning.  Braising the copper tube makes it quite soft.  Very little force is required to bend or kink it after braising.  It's also the Devil's own job to straighten it out without kinking it afterwards.

I checked and had free travel all the way down to the extruder orifice with 3 mm HDPE filament.  No kinking or binding. 

This morning just after midnight I woke up and had a "left the faucet running" moment, viz, had I put the support plate on my new extruder barrel before I braised the copper PTFE flange onto it. I decided that I really didn't want to know right then and if I got up and looked at it I'd be tempted to try to fix it. I'm not that wonderful at midnight work sessions.

This morning, I got up and sure enough I'd left the brass support plate off.



I used the recycled copper PTFE plate this time to cut down on the amount of sawing I'd have to do. Having had a trial run at it last night I avoided some of the mistakes I'd made then and got the whole thing done, with the 0.5 mm hole drilled too, in about 5 minutes. I wasn't hurrying, either.



Here it is.

I woke up on Saturday at 0100 with a gonzo case of food poisoning thanks to my letting a few New Zealand mussels get past their best before date when I was making an Udon for supper on Friday night. After a day, a couple of litres of IV fluid, some Vikadin and muscle relaxants I was ready to do on Sunday morning what I had intended to start on Saturday.

Given my "warmed over death" state of health I decided to stick with with making another copper heated extruder barrel rather than attempting to make one out of brass with a MAPP torch.

I used measurements off the old one and reused the bottom mounting plate. Since you will be clipping and grinding everything that isn't right on the bottom of the extruder barrel you need take no great care with measurements.


Recovering the plate was a simple matter of simply cutting the old extruder barrel and sliding the mounting plate off with a pair of pliers and my vise. After that I had to cut a .005 inch blank for the extruder orifice.



As you can see 0.005 inch copper plate is pretty easy to work with.


After that it was a simple matter of pounding a small dimple into the 0.005 inch copper blank so that the orifice would be a little below the extruder barrel. Mind, the barrel is only 5/32 of an inch in diameter so there is not much that overhang in any case.

Once that was done I used needle nose pliers to bend the copper blank so that the dimple would seat over the bottom of the 5/32 inch copper tube.


Then there was the matter of bending a piece of phosphor copper braising alloy around the copper pipe and seating it as shown. You need to braise phosphor copper in a well-ventilated place. A cheap 10 amp clamp will hold the assembly together quite nicely. I attach the vise to the clip rather than the 5/32 inch tube in that the copper is quite soft.


The braising was done with about 10 seconds of the propane torch. Keep in mind that the hottest part of the torch flame is just a few millimetres beyond the tip of the combustion cone.

I cut a new PTFE flange out of copper before I realised that I would able to drill out the old one quite successfully. All the same, I used the new one, if only for the look of the thing.



At this point I made a mistake, the same one I made last time I buit one of these. I let the flange rest on the vise rather than using the alligator clip to space it well above the vise. Copper is conductive enough that I had a hard time getting the 5/32 inch tube to a high enough temperature for the braising melt to spread out properly. Once I realised what I'd done I let the whole thing cool off and then used the alligator clip to keep the flange well above the vise. There was little trouble getting the braised joint to work properly after that.

Please note that the 5/32 inch copper pipe achieves a pressure seal with the PTFE block by sliding into a 5/32 inch seating in the block rather than depending on threads. I use this strategy on either end of the block. In this way I depend on neither the strength of the PTFE threads nor its resistance to hoop stresses over small amounts of PTFE to achieve the extruder barrel operating pressure.

While this all looks rather intimidating, I did everything that I show pics of in a little over 30 minutes. Documenting it in the blog, however, took about two hours.

If I have time in the morning I will grind off the flash around the extruder nozzle and apply three coats of high temperature BBQ paint. Applying the 6 inches of insulated #32 nichrome wire and sealing it with three more coats of BBQ paint will probably have to wait till Tuesday in order to let the paint cure properly.

I haven't run the Mk 1 in over a month and before that I'd run it pretty much 24x7 for several months. When I started testing the new xy_move routine I noticed that it was having a bit of trouble keeping a steady extrusion rate and seemed to require a much higher operating temperature than before.

It sounded like time to schedule a maintenance workup. Usually, rising demand for power in the Mk 1 indicates that the nichrome heating coil has begun to separate from the barrel. Oddly, though, this time the coil was tightly bound to the barrel.

I then checked the wear on the top bearing and the threaded polymer pump shaft.  There was virtually no wear on the polymer pump shaft and the top bearing, which is under the most stress, had worn less than a millimeter, a situation that the spring-loaded pump should have easily accomodated for.

I then took a look at the output voltage of the power transistor checking to see if I'd got some sort of firmware problem into the mix during the rebuild.  Again, no problems there.

Finally I took a very close look at the extruder barrel.  It seemed to be a good time to refurbish the nichrome wire the insulation of which was little more than ash.  Mind, there was no shorting as a check of the ohm rating for the coil quickly revealed.

After stripping the nichrome I sanded off the many layers of bbq paint and discovered something that I'd worried about for some months.


The 5/32 inch thin-walled copper tube used for the extruder barrel is relatively easy to bend, especially when it is heated.  I've had various mishaps with the extruder barrel over the months and found that it can easily straightened by hand.  Ordinarily these little bends matter very little, but occasionally you get a slight kink in the barrel which restricts the cross-section available to the polymer filament. 


When I cleaned the barrel I discovered that I had two of kinks that needed to be worked out.  You can see the two kink sites circled on the pic.  While I was able to straighten the blue circled one acceptably the red circled kink shattered rather than straightened. 

Time for a new extruder barrel.

I am not at all happy about the service life that I am getting from the copper heated extruder barrel.  I guess that I need to go ahead and get a MAP torch and braise one out of brass and see how that goes.  A brass extruder barrel would certainly be simpler to make.  Unfortunately, I haven't been able to get good braised joints on brass with my propane torch.

I got the new xy_move routine working with the rest of the Control Panel and firmware ensemble this morning. There were a few false starts until I tracked down some code in the z_move routine that turns off the interrupts that was left over from an earlier version of the firmware.

Aside from that I hadn't run the z-axis in some time and it was a little sticky from a slight, humidity driven expansion of the wooden slide assembly. Once I ran it a half-dozen times, however, I had no further problems.

From the looks of it I can keep 95% slewing errors well below +/- 0.1 mm for line segment lengths and closer to +/- 0.1 mm for 0.1 mm line segments.

Now, I've got to hook up the code that controls the Mk 1 extruder and I should be good to go.

My grown "kids", both on holiday,  slept in this morning leaving me to my own devices so I got the four paths for the condition (Y2-Y1)  > (X2-x1) done and working. 

Now I am going to see if I can print the outline of a proper involute profile gear properly.

I've had my youngest daughter here for the past week or so on holiday, so I haven't been devoting as much time to Tommelise 1.0 firmware development as I might.  All the same, I have been able to make some progress.  Here are a few highlights.

The first big advance is keeping two separate but overlapping coordinate system storage units.  The first is what I call "ideal" because it keeps track of where the extruder head ought to be.  The second I call "absolute" because it keeps track of where the extruder head actually is.  Mind, both are in absolute units as dictated by the rotary encoder.  The units are pulses from the rotary encoder.  Given that I am using 3/8-20 threaded rod, you get...

    (128 pulses/turn)(24 turns/inch)/(25.4 mm/inch)  =  120.9 pulses/mm-turn

This works out to just under 12.1 pulses/0.1 mm of linear travel along an axis.  I'm averaging about 5 pulses slewing error or about 0.04 mm slewing error.  The 95% error rate is closer to 0.09 - 0.10 mm.

The development of the firmware routine to carry this off has become a major project.  Basically, I have to guide the extruder head from one point on the xy plane to another along a path determined by the 2D linear equation...

       y = mx + b

As a practical matter I use that equation for values where (X2-x1) >= (Y2-Y1) where m <= 0.5.  Where (Y2-Y1)  > (X2-x1), I use the alternative linear equation...

       x = ny + c

This is a pretty tricky thing to accomplish with a compiler, viz, Oshonsoft BASIC, which has neither true integers nor real number operations.  At first I tried to do all of the real number operations within the PC control panel.  While that worked it meant that I had a huge amount of information going back and forth between the PC and the controller board.

For me the situation was esthetically unacceptable.  The PIC 18F4610 is quite a powerful controller with a lot of flash memory available.

I got around the problem with cardinals rather than true integers by keeping track of when numbers went negative.  To handle real numbers I resorted to the ancient Greek trick of expressing real numbers as a pair of integers in the expression.

          N(real) = A(integer)/B(integer)

That left me with 4 conditions to keep track of depending on the signs of the x and y values of the coordinates of the point that the extruder head was being moved to and twice that for a total of 8 conditions when the state of the slope was taken into account.  So far the routine doing all that takes 10 Kbytes of flash memory on the PIC.  A lot of that will go away when I pull the serial output that monitors what is happening within the processes.

Right now, I've got everything working for the four cases where (X2-x1) >= (Y2-Y1) and am working on debugging the code dealing with the condition where (Y2-Y1)  > (X2-x1).