Tuesday, 28 September 2010


My Mendel has been very reliable and consistent running virtually 24/7, but about a week ago, after putting on a new reel of plastic things started to go wrong. The initial symptoms were that small parts built fine, in fact I printed a mini Mendel or Huxley that came out well: -

It took just two full Mendel beds, plus a few parts on HydraRaptor. I did the gears on Hydra for accuracy and the Bowden clamps at 100% fill because they look weak to me for the job they are intended to do. The plastic weighs 335g (including a Wade's extruder), slightly more than 1/3 of a Mendel by weight but the print time is about 1/2, because small parts need finer filament. I printed most of these at 0.5mm whereas I do a lot of Mendel at 0.6mm.

But getting back to the problem, the quality of large parts had started to fall off a bit. They were coming out with blobs on the outside formed by the nozzle oozing as it moves from one object to another. These were not well bonded, so they could be simply scraped off with a fingernail, but something I had tuned out ages ago. Another change was that it was not doing 45 degree overhangs well, so it left filament hanging down in the tops of tear shaped holes. Again, not a big problem as they just get drilled out anyway.

I started to suspect the temperature was too high so I pushed the thermistor well into the heater block. Then the filament started jamming after the first layer (which I do very slowly). After a few attempts the extruder drive gear broke where the captive nut for the grub screw is. This seemed more like the temperature was too low, so I suspected the thermistor was no longer reliable. I decided to rebuild the heater assembly as my last one was put together in a hurry from parts left over from an experiment. It had been in the wars as well, being entombed in ABS and hacked out again, not to mention running almost continuously for about 2500 hours. Originally the thermistor was glued in with RTV silicone, but that was long gone and it relied on the wires holding it in place.

Since my original heater hack using a vitreous enamel resistor I had moved on to a smaller resistor on Hydra and found that worked better. The surface area of the block is a lot less and that is where most of the heat is lost from, so the amount of power required goes down. It also warms up faster of course, both due to less heat being lost and also less thermal mass. The resistor I have settled on is a Vishay / Sfernice RWM04106R80JR15E1

The thermistor is drilled as close as I dare to the thread for the nozzle and then counter-bored so that the entrance is wide enough for the PTFE sleeving. The wires have PTFE insulation to withstand the temperature and the resistor is soldered with 300°C HMP solder. I think I could also get away with ordinary unleaded solder as well because of the length of the resistor leads, but I didn't want to chance it.

After a tip from Giles I used Rothenberger high temperature glass rope adhesive to glue the resistor and the thermistor. It sets in only half an hour, which is a big advantage over other things I have tried. I also used it to stick ceramic tape on the outside of the block to insulate it.

When I first heated it up the adhesive bubbled causing a downward slope in the temperature graph. I thought at first the thermistor had been dislodged by the blistering, but I think it was just temporarily cooled by the out-gassing. I should have heated it much more slowly the first time I think.

The new heater works much better than the old one. The warm up time to 255°C is about 280 seconds, whereas the old one took about 400 seconds (the bed takes about 350 seconds to get to 140°C). It also runs at about 70% to maintain 240°C while extruding, whereas the old one needed about 90%. The bang-bang control cycles much faster and only deviates by one degree. That is because of the close proximity of the thermistor to the heater. Because it is mounted between the heater and the barrel I can be sure the swing at the barrel is even less. I calibrate against a thermocouple inside the barrel, so any temperature difference across the block is calibrated out. It should be negligible though because the thermistor is also very close to the barrel and aluminium is a very good conductor. The extra power needed to heat the ABS when extruding 0.6mm filament at 32mm/s is about 10%, i.e. ~2W.

The new improved heater didn't solve any of my problems though. While reassembling the extruder I tried pushing filament through by hand. It was much harder than I remembered it was when I first built the extruder. At this point I was beginning to suspect the plastic was different in some way although it looked identical and was part of the same purchase.

I noted that the filament was coming out very curly. That was something I had noticed happening on both my machines when I do a test extrusion, but I had ignored it. I measured the diameter though and found whereas it normally swells to 0.7mm this was coming out oval and about 0.5mm by 0.6mm. It all fell into place then. I have read that the difference between straight hair and curly hair is whether it is round or oval. The only way the filament could be oval is if the nozzle aperture is no longer round. I put a 0.5mm drill bit through it and it started to extrude round, straight, 0.7mm filament again. The hole must have been partially occluded by the burnt plastic that tends to glaze the end of the nozzle. That caused the plastic to come out thinner and faster. It was fine when making objects with 0.5mm filament because it was still being stretched but when building with 0.6mm filament it was being compressed, so would hang loose if given the chance. The smaller hole increased the barrel pressure, which is why it oozed. The plastic would be compressed more, so require more backing up to release the pressure and stop the flow. Also the extra pressure was too much for the pinch wheel when extruding at the top flow rate I use, which is 0.6mm at 32mm/s. I think the M8 hobbed bolt is below the ideal diameter for softer plastic like ABS.

I also re-bored HydraRaptor (with a 0.4mm drill) and that stopped the filament being curly as well. It seems nozzles need occasionally re-boring. I had assumed that the hot flow of high pressure plastic would have kept the hole clean, but not so.

So a simple fault had my machine out of action for days because I didn't recognise what the symptoms meant collectively.

Sunday, 12 September 2010

Some corners like it hot

Large objects with sharp corners, such as the Mendel z-leadscrew-base, produce enough stress to form a blister in the PET tape on my heated bed. These can only be flattened again by pricking the tape. I can't understand how air gets in and cannot get out again, but that is what seems to happen.

The blisters leave a small indentation in the object's base. It is only an aesthetic problem because the base remains flat, i.e. it doesn't rock on a flat surface.

Sometimes the blister allows the corner to peel from the bed towards the end of a build, allowing the corner to curl upwards a little. Generally I can avoid that by cleaning the bed with acetone before problem builds. I also use hexagonal infill on those parts and only two solid layers rather than three in an attempt to reduce the stress. When I design my own parts I round the corners, where possible, to prevent such problems.

A solution may be to use a sheet of PET rather than PET tape, but then you need to find a way of holding it down. One thing I have noticed though is that when I build a bed with four of the z-brackets closely packed the corners on the inside don't blister or lift. That must be because the air around them is hotter. As an experiment I added some little plastic walls to the build to act as baffles to keep the heat in as the bed moves through cooler air.

These have a 5mm thick base to help keep the tape flat and are 1mm away from the edge of the object. They work well and stop the blisters forming at the corners. They are very similar to Forrest's apron technique but their primary function is thermal rather than mechanical. A more general technique would be to build a thin wall all the way around the perimeter of the objects to cocoon them. I expect that would only need to be one filament thick and perhaps might give a similar effect to having a heated build chamber.