geert_2

On my UM2 I have succesfully printed PET, which I believe is a sort of co-polyester? Its brand was ICE from Trideus in Belgium. I had to download the file "materials.txt" from the UM2-printer to the SD card, and then make a new material-profile in that file, and upload it again to the UM. That profile lists the default temperature and material flow, and fan or no fan, and a chosen name. So I can now select PET from the printer-menu as material. Usually I print PET at 25...35mm/s, and 215...225°C, no fan, bed temp ca. 80...90°C, layer height depending on the model 0.1 or 0.2mm. For me this works well, although most specs list higher nozzle temps. Expect a "frosted glass" look when printing transparent materials. It survives summer days in the car, contrary to PLA. It prints reasonably well, except for overhangs, due to the "no fan": it is very difficult to bridge and fill gaps. PET is more rubbery than PLA: while PLA tends to pull a nice string when bridging, PET rather tends to curl up on the nozzle, like a rubber band snapping back, and that blob gets deposited onto the next wall upon arriving there, instead of making a bridge. I print it on bare glass, cleaned with salt water. If I would print with fan, then I would get better bridges. But then I also would need dilluted wood glue for better bonding, otherwise it warps. But then I risk chipping the glass plate (I had this once), so I don't want that. On the long term PET is more likely to clog up the nozzle: it leaves a thick sort of very hard and glossy "varnish" inside the nozzle: hard to clean with cold pulls. Contrary to PLA that leaves a coating of dry powder-like ashes, which is easier to remove with cold pulls. PET or co-polyester might be a good choice for printing accessories like knobs or enclosure shells. But I would not use it for structural elements: it is a bit flexible (think of PET bottles), so for a CNC machine it will not provide good accuracy. The structural elements should be absolutely rigid, thus made of thick steel, brass or aluminium, depending on the wear and loads they will see. Personally I would also make this consideration: would I print materials unknown to me, for an occasional external customer, and take the risk of damage to the printer? No. I would refer them to services like Shapeways, Materialise, Melotte and various other. Would I take the risk for myself, if there was a need for it in our laboratory, and we needed multiple models? Yes, but then I would invest time to search it out well. The models below are in PET, printed on my UM2: key chains (with hollow watermark text), test blocks in various layer thicknesses (0.06 ... 0.4mm) and in various speeds (10 ... 50mm/s) with a hollow watermark, and carabiners (only the green is PET, the cream are PLA).

There are lots of different formulations, and lots of additives, so I think no one can give good general advice. You will need to look up the specs of each individual filament, or ask the manufacturers. Or do tests yourself: cut off a piece of both filaments and keep them in a flame. And see how it burns and how the fire propagates, or does it extinguish by itself? Also melt it with a very hot soldering iron, and see if it catches flame? Of course do this in a safe environment, e.g. outside, or in a kitchen in a metal sink, with plenty of water available. And with safety glasses.

This sounds like general underextrusion, and may be not related to the type of filament (PCL). I think you would first need to sort this out with a well known standard filament like PLA, until that prints smooth. It could be a (partially) blocked nozzle, incorrect feeder tension, incorrectly mounted bowden tube, worn out teflon in the nozzle (not sure if this also applies to an UM S5?), too much friction somewhere in the feeding traject, etc... Search for "underextrusion". Personally I have no UM S5 and no experience with PCL, so it's hard to give more suggestions.

The printer can bridge quite a lot, but then the first layer over the gap sags a bit. And I need to slide the other part of the keychain into that hole, so I want it to be more accurate. However, I don't want the supports to fuse with the underside of the model. So that is why I use these dimensions: they give reasonable accuracy and not too much fusion. But it differs from model to model, trial and error... So, I would suggest you try the concept on small test plates, to find out what works best for your materials, your printer (and printer settings), and your typical models. Also, bigger models may need different dimensions than small models.

@gr5: thanks for clarifying the degree-thing: although I knew the setting, I didn't know what it refered to in the physical world. I took a look in DesignSpark Mechanical, and indeed, it defaults to 10°, and 0.5mm max deviation for STL-export, in "Fine quality" settings. See the picture below: the carabiner as designed, and as exported to STL (Fine quality, 10°, 0.5mm). For reference: the whole hook is 60mm long x 6mm high, and the text caps height of the watermark is 3.5mm, text legs are 0.5mm. The watermark is present but not visible in the STL-file: I made its color opaque to show the facets better. So, for the rocket-engine, I would rather go a bit more accurate, maybe 5° or 2°? Unless the facets would be no functional problem.

I just stumbled on a photo on my harddisk. This is why some supports have extensions beyond the area to support, so I can grab them and wiggle them loose. This makes support removal very easy. Otherwise it would be a nightmare with these small dimensions (see ruler in mm and cm). Also note the ugly defects in the Z-direction in the print: at that time I did not yet print these models with an extra dummy "cooling tower" to increase printing time per layer for better cooling. Sudden changes in total layer printing time, and thus differences in cooling, show up as ugly horizontal defects in tiny models.

In the beginning when I started 3D-printing, my collegues of the department of Product Development suggested me to keep STL file-sizes below 20MB for normal models, even for their industrial polyjet printers. If you are in doubt which quality of STL export-settings to use, then cut out a small part of this design, export it with various STL-settings, and print these little test pieces next to each other. So you can compare and find the best balance between file-size and quality.

Very hard to say from the photos, but at first glance it looks like the first layer is rather thick. So, could it be that the nozzle is a bit too far away from the build-plate? My first layers look way thinner, and they are usually defined as 0.2mm. PS: Beautiful letters by the way...

I don't know the technical name for this in English. What I mean is that at the moment at which the outside of the heater reaches its temperature, inside there could already be so much more heat accumulated, that even when switched off, this heat when traveling towards the edge makes the temperature overshoot. This is also a common problem in heating systems in homes. If the heater shuts off at the desired temp of 20°C, then there might still be so much hot water in the pipes and radiator that it overshoots. Or the thermostat might be in a location too far away from the radiator, so it doesn't sense the changes fast enough. This could also happen in 3D-printers if the heater is too powerfull, and/or the sensor too far away from the heater, or if the sensor has bad thermal contact. Further, you should also look at the Watt of the heater (not only the voltage), if that Watt is according to specs? Because that is what determines the heating. A 24V, 25W heater will generate far less heat than a 24V, 100W heater. If the system was designed for the first, then the latter will cause overshoots. I am not saying this is the cause in your system (which I don't know), but it is one of the many possibilities. It could of course also be the sensor itself: wrong type of sensor, poor batch, poor thermal contact with the heater block, bad electrical contact. Or it could be the hardware- or software-settings that do the regulation: too much or not enough amplification. Like when turning a car into a corner, and correcting the turn: you could correct too much and start to zigzag, or correct not enough and go off the road. The correction has to be the exact right amount. In turning it is the driver, the "software", that is the problem, not the car. But I don't know any real practical values, that is something you should ask the developers of the printer.

Maybe this? Weight of the (half-) full spool, minus the weight of an empty spool, and this sum divided by the weight of one meter of that same filament? You would need a balance and a ruler.

I also think this is not the right direction for printing these models. Like in wooden matches: the grain of the wood should go in the same length direction as the match. Otherwise the match will snap if you try to light it. So this wing is likely going to be very brittle. It should be designed so that you can print all beams flat on the glass, and assemble these parts afterwards. So that it can absorb the forces and flex, like real wings. Meanwhile, for best strength, print rather hot and slow, but with fast travel speeds (so the nozzle doesn't leak while traveling through the air). I don't know if switching off retraction would help, or make it worse? It would surely cause more strings and be ugly, but I don't know about strength? Maybe you could cut out a small part of this design, and do a few testprints with various settings? Also, PLA may not be the best material: it is going to break upon the first crash landing. And it is likely to warp in the car, in a hot summer, while on-road to the meeting or airport. But I have no experience with 3D-printed planes myself, so it's hard to give good recommendations.

Maybe the heater is too powerfull, or has a too big stored heat capacity, so it jumps over the set value? Or the sensor is too far away from the heater, so it senses the temperature changes way too late? Or the thermal connection between heater and sensor is not good enough (maybe try copper paste or the white paste like in CPU's for better contact)? Or something else along this line of thinking?

Is the feeder *gear* not moving? Then this would indicate a slipping feeder wheel on its axis, maybe a screw that came loose in the feeder. Or is the gear moving but grinding the filament, and thus the *filament* not moving? Then I would suggest that you disconnect the bowden tube and manually insert a piece of filament and try to extrude manually. Directly into the nozzle. Clean the nozzle if necessary, and inspect the teflon coupler on deformation. Also, separately inspect every part in the whole feeding traject on too high friction: spool, feeder, bowden tube, nozzle. Also check filament diameter.

How I came to the conclusions is simple: by printing thousands of models during the past 4 years, and by trying lots of different things. I wouldn't say I am an expert, but "advanced user" would be more appropriate. The experts are those who invented and developed the 3D-printers themself, and who worked with allmost all available technologies. Which I didn't. Note: you shouldn't just take my word on this topic, you should verify it by carefully testing. :-) But I do understand your concern. Working at a university myself, I know the rule: "A publication is only considered scientific if it has lots of references to other publications that are already known established scientific publications." However, I think this is not how reality works: the really big developers, scientists and inventors of the past, did everything by themself. Think of Newton, Keppler, Volt, Watt, Tesla, Einstein, Mendelejev, and so many others. They were the first to do it, so they had no list of scientific references to refer to. They invented or discovered it all. Which in my view does not make them less valuable or less scientific, on the contrary. Info on new technologies these days is often only available on specialised forums, or small specialised developer groups, before the rest of society jumps on the bandwagon. So there are no "scientific references". This is a problem the "established scientific world" will need to address. In Asia there is a saying: "An animal or plant does not exist, unless a learned white man has given an incomprehensible latin name to it, and until he has published this in an established scientific journal." It doesn't matter that millions of indiginous people see the animal every day, or eat the plant daily. It still "doesn't exist"". :-)

I won't have time to do tests, but from experience I can say a few things. For PLA, which is the most common material, bonding to the glass is affected a lot by moisture, when printing directly on bare glass plate without any bonding aids like glue. In a super-dry environment (e.g. freezing cold winter weather), bonding on clean bare glass is reasonably good. On very wet spring days, thus high humidity, bonding is very bad in my experience. Wiping the glass with *salt water* greatly improves bonding when printing on bare glass: it gives good bonding when hot, but no bonding at all when cold (=easy to remove models). The use of glue (dilluted wood glue, glue stick, hair spray, 3D-lac, etc...) also eliminate or reduce the effect of moisture on bonding. Printing quality of PLA is not, or very little, affected by moisture; I haven't seen any difference. I haven't seen much influence of environmental temperature, which in winter usually is 20...22°C in my lab, and in summer up to outside-temp, usually 20...26°C, occasionally 30...35°C. For moisture-sensitive materials like nylon however, quality will be greatly affected by moisture, but I have no experience with these, so I can't comment on that. I believe the quality of your test model is going to depend far more on the print-settings than on the environment, at least when printing in PLA, PET, or similar. For example, quality will depend a lot on: bed temperature, bed bonding technique, glass cleaning method, nozzle temp, layer height (especially for overhangs), printing speed, material flow, cooling air flow (fan percentage), infill percent, support settings, cooling time per layer (with or without additional dummy models to increase layer time), and various other settings. And on the stiffness of the material itself when cold, and on the viscosity of it when molten. For printing nylon and similar materials, some sort of dry-box to keep the filament dry during printing is required anyway, just like drying methods are required in nylon injection moulding.

Good that you found a solution. If the flowrate is too high, from observation, also check if the filament diameter is correct (both the settings and the actual diameter), before you change the hardware-settings. For example if the filament would be 3.0mm instead of the standard 2.85mm, or 2.0mm instead of 1.75mm.

For my small models, and my single-nozzle printers (UM2), small ribs on top of the support worked best. Also, and very important, I always design enough features into the supports so that I can grab them carefully with pliers, pincettes, other tools, or I can get in there with a knife. And I split them in small chunks which I can wiggle loose easily. So, for these reasons, on delicate models I usually design custom supports into my models, instead of the automatic supports. See the red and orange supports in the first picture. The ribs on top are usually 0.5mm wide, and 1mm separated. The gap between the ribs and the underside of the model is usually 0.2 to 0.3mm, depending on the model. For some models, thin plates floating in the air, also work well as supports: they can be peeled off layer by layer (see the pictures). It depends a lot on the model. So I would suggest you make several test models with different dimensions, and try which works best for you.

To me this also looks like overextrusion, but it is difficult to see on the white. If you watch closely while printing, does the nozzle sort of "wade through the mud"? If yes, it's overextrusion. Probably you will also see accumulation on the nozzle, which then sags and gets deposited on various parts of the print, thus worsening the effect. Try printing a test plate of let's say 20mm x 20mm x 10mm, at 100% infill. And then play with the settings while printing: change flow rate, nozzle temperature, fan percent, printing speed, on the fly. Until you get it right? I usually print PET at 215...220°C, 30mm/s, no fan if the model allows it, 90°C bed temp. If I need fan for overhangs, then the temp has to be a bit higher, 220...225°C. My brand of PET has a tendency to overextrude when changing from printing long trajects to small details at high speeds: it seems the pressure in the nozzle can not change fast enough. The molten PET seems to be more rubbery than PLA, like chewing gum, and react slow. So normally on my UM2 the overextrusion happens on very short trajects only, not on long. It is when the printer has to frequently slow down to change direction. Then the built-up pressure in the nozzle can not immediately go down, resulting in temporary overextrusion. At least, that is how I understood it.

For complicated designs, you can always design custom brims, custom supports and other custom structures in the CAD file itself. And then switch off the automatic brims and supports in Cura. Then you have full control. For example if you want brim on some areas, but not on others. Or if you want special supports with holes suitable for your pliers and tools, to make removal easier from the inside of complex models. It will take some trial and error (do this on tiny test plates or so), but once you have found a concept that works well for your application, you can re-use it quickly. For example in this tiny model (key chain), the orange and rose custom supports have some brim, and the green dummy tower also. The dummy tower is to provide extra cooling time, otherwise the top of the yellow model does not solidify well; it is too small. I don't want to risk a model being destroyed because a dummy tower or a support comes off, so they get brim. But the rest don't need any. Further, these supports need that extended shape, otherwise I can not grab them with pliers and wiggle them loose, and the space is too small to get in there with a knife. So I provide all this in the CAD file. For reference: text caps height is 3.5mm, and text legs are 0.5mm wide.

Note that the pastel blue and green models shown above are not mine, but from neotko. But they look great indeed. My dimensions are in this pic below. The character sets (RSDOC-format) and testplates (STL) can be found here: https://www.uantwerpen.be/nl/personeel/geert-keteleer/manuals/

In my very old Cura, it goes like this: Let's say these are the settings: - first layer = 0.3mm - layer height = 0.1mm - bottom thickness = 1.0mm Then it does: bottom = 1 x 0.3mm + 7 x 0.1mm = 1.0mm in total I would guess it is the same in Cura 3.6? You should be able to see this in layer preview, if you make a small test object?

I always thought that this alternating was the desired behaviour? When the slicer encounters the first edge it turns the material on. At the second edge = material off. Third edge = on, fourth edge = off, etc...? Otherwise I think it has no way to know whether it has to fill a model or not, since the STL-file only consists of triangles, if I understood that well? Maybe software-engineers like smartavionics or ahoeben could shine a light on this? Anyway, I use this feature to make hollow text and watermarks inside a model. I design the watermark separately, outside of the model, and then I just move it inside the model. After export to STL it is automatically sliced correctly, thus Cura automatically hollows-out the text. Like here in this example, where I have copyright notices, a ruler, and some decorations as watermarks inside the model, a keychain. Or in the transparent testblock below. (For reference: text caps height = 3.5mm; text legs are 0.5mm.)

Have you tried printing this in PLA? Then you know if it is related to the material, or to the model/slicing? Or else, design a couple of test pieces, with only the faulty parts, plus a dummy block to provide enough cooling time per layer. In some of those test pieces, model them as shown here, as separate models. In the other test pieces, do connect both edges of the openings with a thin plate to each other. So the printer does not have to stop printing, and does not retract, but can keep going. Then you would have this sequence: dummy block --- connection to the test piece --- test piece itself --- connection to the next test piece --- test piece itself --- connection back to the dummy block. So, no gaps, no retractions (theoretically and hopefully).

Have you checked if the fans are running freely? Not (partially) stuck by sucked-up filament hairs and strings? Or no filament or other debris is covering the exit or inlet holes, or something mounted incorrectly, or something else along this line? What could help, is putting a desktop fan in front of the printer, at its lowest speed setting. This provides a huge amount of fresh air at low speed. But of course, this does not address the source of the problem. What also helps for overhangs, is printing in thicker layers, e.g. 0.3mm instead of 0.1mm. But that may not always be desirable for the rest of the model. Printing slow and cool also helps.

Just a thought: is there no way to overrule that in the gcode files, or via gcode files? Or is it executed even before the gcode files are read?