Friday 30 April 2010

Flash bang bed

As my MK3 heated bed on HydraRaptor has been working well I decided to scale it up for Mendel.

Buying aluminium that is flat seemed to be a hit and miss affair until a friend told me that what I need is tooling plate and put me in touch with a company that sells it. They recommended C250 cast machined tooling plate. It wasn't cheap (I got 5 pieces 200 × 200mm for ~ £140) but they are all flat.

I can't find a geometric definition of flatness. It is given as +/- 0.4mm for a 6mm sheet of C250 (I would have preferred 5mm to reduce the mass a bit but that is +/- 0.8mm). I take it to mean that all the points on the surface of a metre square plate will lie in a volume 0.8mm high. For a 200mm piece I expect the deviation to be about 1/5 of that, i.e. 0.16mm assuming it is a single curve rather than wavy. Since the bed can be levelled at the corners the deviation in the middle should be about half that again, 0.08mm, just about acceptable for raft-less printing.

When I tried levelling the bed I ran into a problem though. With my Dibond bed I could level each corner because it can flex a bit. With the rigid aluminium bed I can only level three out of the four corners at a time. When I move the nozzle to each corner in turn it behaves as if two diagonally opposite corners are lower than the other two. That would imply the plate is not flat, but I know it is when I put a straight edge across it. I think this means that the two y-axis bars are not quite level with each other at both ends, causing the bed to twist about the y-axis as it traverses it. I expect it could be corrected by adjusting the frame but I haven't got my head around what to adjust and in what direction yet.

Given that I am using 188W on a 150mm bed on HydraRaptor, to get a similar warm up time I would need 335W. That seems a lot to get from a PSU, so I decided to make it mains driven. I found that I could get 47Ω TO220 resistors cheaper than other values. Five in series across the mains gives about 250W, so I used two strings of five to give 500W. That gives a warm up time of about three minutes.

Equally spacing four or nine resistors on a square is easy but placing ten is an interesting problem. I used the solution to packing ten circles in a square that I found here. This is my layout with 16 magnets as well.

And here it is wired up: -

I used wire with PTFE insulation rated to 300°C. I have an earth connection of course. It would be a good idea to have a second earth in case the first one breaks due to the constant bed movement. I also fitted a 150°C thermal cut out that came out of a microwave oven. With 500W it would get very hot indeed if the control circuit failed.

I intended to mount the magnets the way I did before, by drilling holes not quite through, leaving a rim to retain them. I didn't tighten my drill stop enough and went all the way through so I decided to glue them in with JB-Weld.

I placed the bed onto a sheet of glass with some cling film on it. I then dropped in the magnets and glued them. When I turned it over the next day I found the magnets were sticking up from the surface. The glue must expand as it sets pushing the magnets down and lifting the plate!

I tapped them down with a punch but, unsurprisingly, they fell out the first time the bed was heated. In the end I jammed them in with PET tape. Drilling part way through is a much better solution.

I mounted the bed on top of the Dibond bed with nylon stand-offs.

Not an ideal solution as a lot of z-travel is lost, but the thermal cut-out is quite deep.

I used chocolate block connectors to wire up the mains. To make them safe and provide strain relief for the cables I RepRapped some plastic covers.

The lids just clip on with some tabs that fit into small slots. They didn't fit very tightly, I need to make the tabs bigger and a tighter fit. A boss and a screw hole would have been better I think.

For safety all the wires should be inside the cover as everything accessible should be double insulated. I will make it wider at my next attempt.

The bed worked well for the first few objects I made. Simple bang-bang control gave about 10°C overshoot initially but settles down before the object build starts so does not really matter. One thing I have realised is that the nylon pillars expand about 0.1mm when they warm up so I give them some time to do that otherwise the first layer has varying height.

I got some new ABS from that turned out to be white, I was expecting natural as that is easier to work with. It seems to need higher temperatures to get it to stick to itself and the bed. I am extruding at 240°C with the bed at 140°C for the first layer and 110°C after that. I built one object like that and then disaster struck. The bed heated to 140°C and levelled off. While the extruder was heating I heard a few pops and crackles. When I looked at the temperature graph I saw the bed temperature soaring. Before I had time to think what was happening there was a loud bang and flash from underneath the bed and the 5A fuse in the plug blew.

What happened was one of the resistors developed a short between its tab and one of the connections. That caused a path to earth which increased the power on the remaining four in the chain. Several of those went short circuit as well in a chain reaction which ended up shorting the mains.What I couldn't explain at first was why the firmware did not turn it off and why the thermal cut-out did not cut the power. It turns out that I had swapped the live and neutral connections in the IEC connector, which meant that the solid state relay and the cut-out were in the neutral connection. As soon as the first resistor shorted it had bypassed all the control, not good!

I had originally chosen the resistors when I was making a bed for PLA at 60°C. Looking at the datasheet they have a maximum operating temperature of 155°C but they are de-rated to zero wattage at that temperature, so by putting 50W into them at 140°C I am grossly over loading them. I have abused AL clad and vitreous enamel resistors in this way and not had any problems but the TO220 seem far less robust. I don't know what they use for the tab insulation but I wouldn't be surprised if it was epoxy. The high voltage may also have been a factor as the ones on HydraRaptor have survived a similar overload so far. They have the same de-rating curve, but are made by a different company.

I rebuilt the bed and changed my firmware to stay inside the power curve by reducing the PWM ratio as the temperature increases. Unfortunately , I found I could only get to 130°C so I had to change the zero power point to 200°C to get to 140°C in a reasonable time. Even then it takes 400 seconds instead of 175.

So far it is holding up, but it is nowhere near as fast as I wanted. A shame because I had bought 50 of the 47Ω resistors, but I think I will have to scrap them and go back to AL clad. The smallest ones that I have used before are not rated for mains voltage so I will need some bigger ones. PCB or stick on silicone heaters are starting to look more attractive!

Saturday 24 April 2010

ABS on PET tape

I find ABS sticks to Kapton very well to start with, but as it ages, it seems to stick less well. Corners start to lift and eventually builds are ruined. I have tried cleaning it with isopropyl alcohol and with acetone but it makes no difference. Charles Pax has reported that sanding with 220 grit paper makes it stick better. I cannot reproduce this. In fact, I find the opposite effect. It always sticks well when new, and if anything, sanding it makes it worse.

Somebody pointed out a while ago that you can get PET tape that is rated to 250°C. That is not as high as Kapton, but just about adequate for a heated bed when extruding ABS at 240°C.

I bought some and when my Kapton stopped working I decided to give it a try. It seems to work well. The first layer goes down perfectly :-

and the objects stay flat: -

I do the first layer at 240°C with the bed at 120°C and subsequent layers at 220°C with the bed at 110°C. I have made all the parts for an extruder on it so far and it has performed perfectly. The extruder will be on eBay this evening.

It is too early to say if it better than Kapton, but it looks promising.

Tuesday 6 April 2010

Dibond bed

I had been making Mendel parts with my Mendel, using PLA on blue masking tape, as it didn't have a heated bed . When I made a frame vertex on its own it came out completely flat. Larger parts like the z-base brackets warped a little at the corners, but were still acceptable. However, when I made a bed full of parts the warping was much worse. Frame vertexes warped a little and z-base brackets curled up several millimetres and jammed the y-axis, ruining a bed full of parts. I think the reason they warp more is that it takes so long for each layer that the parts are completely cold when the next layer is deposited. The odd thing is that Adrian Bowyer manages to print trays full of parts on blue masking tape without a heated bed. I have added it to the growing list of things that work better in Bath than they do here: AOI and PTFE being another two.

I had some aluminium plate on order but I wanted to knock something up quickly. I figured PLA on blue tape would only need 40-50°C to stop it warping. My bed is made from Dibond, which is 3mm thick and has the following characteristics:
  • Thickness of aluminium layers 0.3mm.
  • Core polyethylene, type LDPE.
  • Surface: lacquering - modified polyester lacquer system.
  • Temperature resistance from -50 ° C to +80 ° C.
  • Aluminium grade premium A1Mg aluminium alloy.
The great thing about it is that it appears to come pretty flat and is strong, light and easy to machine. I wondered if the aluminium layer was thick enough to spread the heat. I didn't think heat would flow though the LDPE very well so I mounted 10 47Ω 50W resistors around the top edge. I have found that for some reason 47Ω are cheaper than the 12Ω ones I used on HydraRaptor's bed. I wired them in pairs in series and then all the pairs in parallel giving 18.8Ω. I connected them to my 48V AC transformer with a small solid state relay. The total power is about 120W. Not as much as I use on my aluminium beds, but plenty of power to get to 50°C quickly. In fact, it warms up faster that my extruder does.

An initial test showed that the middle was about 10°C cooler than the edge. Not a big surprise considering how thin the aluminium is and how far the heat has to travel. When I measured the other side the difference was only about 5°C, so I decided to mount it upside down with the resistors on the bottom and the thermocouple on the top.

It works very well, and the objects stay flat. The first multi-part build I did though failed after the first few layers.

The extruder jammed because the top of the thermal insulator got hot enough to allow the PLA filament to go soft before the entrance. The extruder was finding PLA very hard to push anyway and the maximum speed I could get was about 24mm/s of 0.5mm filament. This is because the thermal transition zone is too long. The extra heat rising from the bed must have pushed it over the edge, literally!

The insulator is a combination of PTFE for slipperiness and PEEK for strength, but I think PEEK conducts too much heat. It doesn't help that my heater is not insulated yet and the Mendel carriage traps any rising heat.

I am quite happy with with Wade's drive mechanism but decided it was time to try another hot end design, coming soon ...

I think that for PLA, Dibond and blue tape / Kapton is a good solution. It won't handle the temperatures for ABS on Kapton though, but it might be good for ABS on PMMA or PC.

Friday 2 April 2010


Vik Olliver asked for a volunteer with a heated bed to see if we can extrude onto copper clad board. I didn't think it would stick, but gave it a go anyway.

I first tried ABS onto double sided copper clad FR4 taped to a bed at 120°C. The ABS stuck well enough to extrude the first layer of a 20mm square, but when it cooled down it had no adhesion at all.

PLA at 55°C did exactly the same, but PLA at 130°C stuck very well, so well in fact that I can't get it off with my fingers (the blob was where I aborted the print after the first layer).

Maybe ABS would stick in the same way at an even higher temperature, but maybe not as it is less like glue than PLA. The 120°C / 55°C temperatures are what I use for Kapton, which is why I used them as the starting point.
An interesting aside: I had to measure the PCB to work out the z-height. It is only 1.4mm thick, whereas a standard PCB is 1.6mm. You can also see the grains in the FR4 showing through the copper. This means the board I bought in Maplin for home PCB use is actually the same stock material that they use for the first part of a commercial production process, but when they plate thorough the vias they increase the thickness of the copper all over to get the standard 1oz/inch2. I don't know if this is always the case, i.e, that all home made PCBs have less copper than a production one, or whether you can get bare board with 1oz on it already.
Anyway a good result, assuming PLA will resist PCB etchant. Also, it seemed like a potential bed technique. I.e. do the first layer onto hot copper and then cool it to about 50°C for the rest of the object. I tried it with this butterfly: -

It worked perfectly. After the first layer I blew it with a fan to cool it down to 50°C. It took about four layers to get down to that temperature. Since I added the insulation under the bed it takes longer to cool it than it does to heat it.

After it had finished and cooled down to 40°C it was still firmly attached, so I removed it by flexing the PCB.

The base of the object is perfectly flat.

I think for PLA this might be a better technique than Kapton. I can't imagine the PCB wearing out. It could also be self heating with a serpentine track on the other side. I don't know that just taping it down would be strong enough for making large objects. I could solder fastenings on the back if not.

I don't know if there is anything special about copper and PLA, or whether other hot metals and plastic would work . I tried similar things with ABS on AL, but may not have had it hot enough.