geert_2

I would *not* recommend using my salt method for such huge but thin models. The salt method works very well for PLA, for my typical flat and 100% filled models, which ohterwise tend to curl up if I use no bonding aid or with gluestick. But this salt method seems to be a bit more susceptible to shocks caused by the nozzle banging into model edges when there are overhangs. I am not sure why, maybe because real PVA-based glue has some flexibility to absorb shocks, and salt doesn't? Just a guess... But of course, feel free to try. Just stay with the printer to see what happens. If I had to print such a thin but tall model, I would probably add sideways support columns, similar to those used in old cathedrals and castles for the walls. Or similar to the way bosses are designed for injection moulding. Just to be sure. My typical sort of models. The lower one is as printed. The curled one on top is from a heat test in the oven at ca. 80°C after printing, to see how heat would affect it. It shows that there is quite a bit of stress in the models. While printing, this sort of models stick very well to the build plate using the salt method. But only for PLA. For the manual, see: https://www.uantwerpen.be/nl/personeel/geert-keteleer/manuals/ This one got knocked off the build plate, due to the overhangs curling up severely (more than 1mm), and the nozzle banging into it really hard. And then all of a sudden, the model popped off the plate. This is the reason for not recommending the salt method for high but thin models...

Out of curiosity, what is the technical background reason for this?

I am still thinking in the direction it must be something mechanical, or basic electrical. Like the little ventilator sometimes getting stuck or sometimes having a bad electrical contact (but not always); filament on the spool getting stuck under itself; the spool not rotating freely and getting stuck; partial nozzle blockage due to debris picked up or due to burnt filament particles; kinks in the filament that increase friction too much in the bowden tube or nozzle; bowden tube pinched; or something similar along that line? That is where I would search. But I have no UM3, so this is all "educated guessing". Bad settings and bad gcode should show up always the same. And bad heaters or temp sensors should or be compensated for, or go totally wrong and give errors, due to the closed-loop circuit trying to compensate. A slipping feeder wheel (if that is possible at all?) should not grind filament. This all provided that you print from the USB-stick. If via network, I have no idea what a slow network or slow host computer (doing defrag, playing a Youtube-video, or running antivirus) would have as effect on the print?

Yes, that could well be a reason for the observed difference. I had to really saturate the moulds with silicone-oil spray, to get a reasonable life out of them. Silicone rejects water, which is why it is water-tight. But a silicone mould is porous to oils, liquid parafines, and solvents: these leak through it. Don't ask how I found out. So any solvent-like casting material like methylmetacrylate penetrates into the silicone mould itself, polymerises in there, and destroys the mould quite fast. Hence I needed to saturate the moulds with silicone oil prior to casting. A light puff of oil wouldn't do, it really needed to be a good shower. And then of course the mist goes everywhere else too.

Good for you. Maybe the hairspray particles are bigger, heavier, and/or sprayed more directionally than my silicone oil sprays?

Such a hood will help, but it is not enough to prevent the hairspray from getting everywhere in the printer, in the long time. Just try spraying a bright red (easily washable!) paint. Do that ten times. Unless you have extraction, you will see the paint goes everywhere. Way further than you expect. So, preferably do this test on the glass outside of the printer, on a big white sheet of paper. Before the age of 3D-printing, I used to spray silicone moulds with silicone oil as release agent. And even though I had an extraction system, and I used a cardboard box of similar concept like yours (with the extraction at the back of that box), the oil spray went all over the place. Not immediately, but you would notice after a couple of weeks. Silicone oils are non-stick. But sticky stuff like hairspray might cause problems in the rods, bearings and belts, I think. So, spraying on a tissue above the sink, and wiping the glass with that, like gr5 says, seems a better way to me. I was about to say that you could use a vacuum cleaner at lowest power setting for an extraction system for your hood. But this is *not* a good idea. Most vacuum cleaner motors have brushes, creating huge sparks. And most spray cans contain highly flammable pressure gasses. So that would be a recipe for fireworks... If you want any extraction, you definitely need an explosion free system with brushless motors. And a fine filter to stop the glue particles from reaching the motor and fan blades (the gasses will pass the filter anyway, unless you have an appropriate chemical filter).

This may not be the answer you are looking for, but the following method is extremely simple and it works well in an ethical environment with capable people. Put a logbook and balance next to the printer(s), and require each user to weight his printed objects (including any waste), and let them write that down in the logbook. This is similar to the logbooks used for expensive equipment such as electron microscopes in universities. Columns in the logbook could include: username; date; time of starting the print; time of finishing; material brand, material type (e.g. PLA, PET,...); color; weight of used material (including any waste, supports, skirts,...); which printer was used (if multiple); remarks/feedback. Do not forget a column for feedback and remarks, this is very useful. The logbook has to be filled in when starting a print, so that when a print is running, and you notice something is odd, you can immediately identify and contact the appropriate user, without first having to start up your computer, make a connection, and search through lots of files. If required, you could then write this info into an Excel spreadsheet, to automatically calculate costs based on material brand, amount, printing time, fixed costs, etc. If the users typically make long prints, and they need to plan ahead, a list for reservations for the coming weeks could also be usefull. Also similar to electron microscopes or other big equipment in universities, where projects need to be planned ahead. Logbooks are a very common practice in universities, research labs, for any expensive equipment use, parts shops, chemical plants, in lots of other industries, aircraft, ships, etc... So most people are going to be familiar with it anyway. It is a low-tech, very simple, fast, and very stable solution.

I do understand the problem, as I had that too in the very beginning, when printing on bare glass without any bonding aid, or when printing on a glue-layer that was too thick. But if you try to solve this kind of bed adhesion problems in the slicer, it would create ugly lines through the bottom of your prints. I think it is best to optimise your bonding method, to make the small circles stick by themself. It might be a good idea to post a few close-up photos of the underside, to see what exactly is the problem. For example: build-plate too far away, glue too thick,... Since I started using my "salt method" (=wipe the glass with a tissue moistened with salt water) I don't have this problem anymore. But I think gr5's method (10% white wood glue in water) or neotko's hairspray method, or 3DLAC, or a thin layer of glue-stick wiped with a wet tissue afterwards, should also work well on the condition that the build-plate is at the correct level. For the manual on the "salt method", see here: https://www.uantwerpen.be/nl/personeel/geert-keteleer/manuals/ When using a new bonding method, always stay with the printer for the first prints, and carefully watch if it works well, and what happens.

Yes, this is the right question. Upon seeing that photo, I simply assumed it was ABS; I hadn't even noticed the word "PLA" in the title.

For this particular model the dark photos look best, I think. They look like well preserved fossils you sometimes see in museums., or like the Oetzi-man they discovered in the Alps some years ago in the Oetztaler Alpen, but then a bit polished up. I am still amazed at the photographic quality you get with that setup.

Maybe try pulling it apart? If it is caused by poor layer adhesion combined with warping (which I guess it is indeed), then bonding should be so weak that the model can easily be pulled apart. Not only on the visible splits, but also on other parts without visible defects. It should sort of delaminate. If it is a Z-axis problem but good layer adhesion, pulling it apart should be much harder. At least, I think so...

Before starting a next 20hour print, I would suggest you first get your bed adhesion better, and try it on only one column, or on a very simple test piece. If it fails, you will lose less time and material. Long thin models like these act like powerfull levers, so it would be a good idea to use a really wide brim, especially if your bonding is not optimal. Even better, make the brim so large that it connects all pieces together, so that it forms "one plate covering the whole bed". Or design an appropriate thin "base-plate" in your CAD-program, instead of using brim, but similar and with the same purpose. If you make that base-plate 0.5mm thick, or if you give it a few strengthening ribs, it will better withstand levering actions. And the models will wobble less, and thus quality near the top will be better.

I haven't had to diagnose network issues yet (and I am not a software engineer either), so I am afraid that I can offer very little practical help. Usually I use Sysinternals Process Explorer to see which programs slow down my system by eating up my CPU-cycles, RAM, or GPU. I don't remember if I added the graphs for all these manually, or if they were enabled by default on the newer versions, but these graphs are handy. And then I use Process Explorer to suspend or kill hung-up or suspected programs. Or I use Autoruns to quickly disable unwanted auto-starting things. Maybe in your case TCPView or Process Monitor might also help? TCPView shows the network-connections: which programs are doing what on the network? And Process Monitor shows all registry- and file-access of each program (but this is way too much for me).

I once looked at the indents in the filament under a microscope. If the feeder wheel has square "pyramids" on it, the pits these pyramids leave in the filament get deformed (stretched) into diamonds-shaped pits instead of squares, in the feeding direction, due to the high force exerted on them. You might call it a sort of deformation, or "pit-stretching",or "partial slipping" indeed, but without grinding yet. Similar to car tires partially slipping, and starting to make sound, when braking hard but without locking the wheels up yet. The result is that the feeder wheel turns a little bit faster than the filament, due to this stretching-deformation of the little pits. I also guess this is why flexible filament needs a higher "flow rate" percentage than hard filament like PLA, to get the same correct extrusion? To compensate for more stretching of the pits? If you would calibrate the feeder with nozzle absent (and thus with very little resistance), the result would be different from with nozzle present (and thus higher resistance). Also, higher and lower printing temperatures, and higher and lower printing speeds (and thus higher and lower required forces to push the filament through the nozzle) would have an effect. This might actually be a good test: extrude 100mm of filament (=defined as 100mm), both with and without any resistance, and measure how much difference there is? (But of course there could still be other causes for deviations.)

Have you checked if the little fan to cool the nozzle is running (when above 40°C)? It could get stuck if hairs of filament get into it. If this fan is not running, heat may travel up too high in the filament, soften it before the nozzle (which causes it to expand due to the feeder pressure), and thus preventing it from entering into the nozzle. This could explain all other phenomena too (although there might still be other causes of course). In the UM2 you can see that little fan from the back, but in an UM3 I don't know.

It could be a Z-screw issue too, but then it should always be at the same height in all prints. One of my UM2 has this too a little bit, the other has not. Both have a different board: the affected one has a newer board with controlled nozzle fan (switches off below 40°C), the unaffected one has an older board with nozzle fan always on. But there doesn't seem to be any Z-issue. So I have been thinking in the direction of the bed heater too, although from a different viewpoint: if the heated bed (or the nozzle heater) draws a lot of current through the board, this might shift the ground level a few millivolts. And this ground-shift might influence the temperature measurements, which are probably in millivolt too. So the temperature regulator would adjust the temperature incorrectly because of this ground-shift, although it would still display the correct temperature value. This is a very wild guess as I don't know the board layout, but it does not seem impossible? Are the ground-lines of measurement circuits (temp sensors) totally separated from the grounds of power-circuits (heaters and steppers)? Or do they use the same printed circuit traces? And are these circuit traces really thick for lowest resistance (e.g. 5...10mm), or are they rather thin? If this would be the cause - if - then soldering thick wires parallel to the ground circuits involved might reduce the ground resistance and thus this effect. However, that could lead to the wires acting like an antenna and picking up or radiating HF-noise. Like a simple transistor-amplifier with long open input wire sometimes acts like a radio, playing all stations together (had that long ago when trying amplifier circuits on a breadboard for the first time). If this would be the cause - always *if* because I don't know - then another option would be to never heat or move steppers and measure temp at the same time: switch off all heaters and steppers, let transient currents settle down, measure temperature, and then resume heaters and steppers.

As gr5 said, search for the STL-export settings in your CAD-program. In DesignSpark Mechanical you have the options course, medium, fine, and custom. Fine is a good balance between file-size and smoothness, although the triangles are still visible in balls. Course and medium are too rough to my taste. For really smooth surfaces, a custom setting is required. But be aware that a finer setting (=more triangles and thus a smoother curve) can soon make the file-sizes really huge and unmanageble. Instead of 2Mbyte it may easily jump up to 20Mbyte. And this might create other issues: the slicer slowing down, printer slowing down, wasting 95% of hard disk space for little improvements,... So I would suggest that you experiment with all settings to compare the results, and then go for a good balance between quality and file size.

I don't know the Anet A8. But for printing PLA on a heated glass bed, you might find a few tips on build plate adhesion in my manual here. It is the manual on the "salt method": https://www.uantwerpen.be/nl/personeel/geert-keteleer/manuals/ The text is written for an UM2, but should apply to similar concepts too, I think.

Have you tried the Sysinternals Suite tools? Then you can monitor which programs are active and how much resources they eat up (CPU, GPU, RAM, IO,...), both in numerical and in graphical format, with the Sysinternals "Process Explorer". And you can kill or suspend programs that you suspect create problems. They have lots of other diagnostic tools too, to see which files are open, etc... The tools are safe, I am using them for +10 years now. And Sysinternals has now been acquired by Microsoft.

I have no experience with low-quality settings. I think most people go for medium to good quality, and then they often adjust the settings to their preferences, models and materials, by trial and error, for a better quality. I would suggest you try the same model in 0.1mm layer height, 25mm/s print speed, 195°C nozzle temp, 60°C bed temp. And see what that gives? That should be quite good quality, similar to the photos on the website you mentioned. Or even better: try a smaller testmodel first with various settings, before doing a big model? This will let you gain more experience in a shorter time. It's very hard to see in the photos and on a black model, but the underside seems to be a little rough. So I think your nozzle might be a little bit too far away from the glass. When printing the skirt (=outline around the object), inspect how thin the first layer is, and maybe try adjusting the three screws for bed calibration a little bit closer (I don't know by head in which direction, you need to look that up). I my models, the underside is normally quite flat and really glossy.

Het zijn niet de onderranden die omhoogtrekken, wel de bovenranden. Zie deze foto. Dit is een extreem voorbeeld voor het printen van extreme overhangs zonder supports. In dit geval was deze voet nog niet groot genoeg. Door de hoogte en de harde botsen van de nozzle tegen de omhoogkrullende delen, was de krachtarm te groot en kwamen ze los. Dit soort dingen is het enige waarvoor een brim of een nog bredere voet bij mij nodig is. Voor de rest niet. Onderaan is er soms een hoekje van 0.5mm tot 1mm dat iets loskomt, maar meer niet. Overigens is dit al een beetje teveel zout op de glasplaat, minder is beter.

Ik gebruik bijna nooit een brim. Alleen als het hoge modellen zijn met grote overhangs, dan soms wel om te vermijden dat ze omver geworpen worden. Immers, bij grote overhangs (denk aan een ondersteboven prisma, punt naar beneden), hebben de randen de neiging om omhoog te krullen, en dan bots de nozzle daar met geweld tegen. Dat systematisch kloppen kan de print doen loskomen. Meestal zien mijn modellen er zo uit, en die print ik alleen met een skirt: om de nozzle goed te purgen, om te checken of de flow goed is, en om de afstand glasplaat-nozzle te controleren. Het kromme model is door verhitting in de oven tot ca. 70°C of 80°C (als ik mij goed herinner), om te proberen hoe ze kromtrekken. Bij het printen was hij mooi recht, zoals dat vooraan. Dit is PLA. Om een idee te krijgen van de hechting, kan je best na afloop van de print, terwijl de glasplaat aan het afkoelen is, regelmatig eens aan de modellen trekken (wel voorzichtig, nooit met bruut geweld) om te zien hoe vast ze zitten, en bij welke temperatuur ze loskomen. Dan krijg je een goed idee hoeveel "reserve" je hebt, en hoe groot het risico is dat ze tijdens het printen zouden loskomen. De hechting kan immers verschillen van materiaal tot materiaal, en van omgeving tot omgeving (vochtitgheid, luchtstromen). Ook de bedtemperatuur heeft veel invloed: bij 60°C is het bij mij het beste, bij 50°C en 70°C is het duidelijk minder. Bij 40°C springen de modellen er soms af. Ik zou zeggen: probeer dit allemaal eens met kleine testmodellen, vb. kubusjes van 10mm^3.

1. Most of the time (80%): A: What is the filament brand: colorFabb B: What is the type of filament: PLA/PHA (various colors) C: What is the diameter of the filament: 2.85mm D: What is the weight of a full spool: ca. 1kg (=750g filament + empty spool)? (I have no balance available for this weight) E: What is the inner diameter of the core? 53mm F: What is the outer diameter of the core? 105mm G: What is the outer diameter of the whole spool? 200mm H: What is the weight of an empty spool? ca. 300g ? (my balance goes to 250g, and it is slightly more) I: Width of full spool (only when it is completely full because it can expand) 56mm (= empty +1mm) J: Width of an empty spool (only when almost or completely empty) 55mm 2. Occasionally (10%): A: What is the filament brand: Ultimaker B: What is the type of filament: PLA (pearl) (old model of spools, no chip yet) C: What is the diameter of the filament: 2.85mm D: What is the weight of a full spool: ca. 1kg ? E: What is the inner diameter of the core? 52mm F: What is the outer diameter of the core? ca. 104mm? (no empty one available now) G: What is the outer diameter of the whole spool? 200mm H: What is the weight of an empty spool? (no one available now) I: Width of full spool (only when it is completely full because it can expand) (no one available now) J: Width of an empty spool (only when almost or completely empty) ca. 54mm 3. Occasionally (9%): A: What is the filament brand: ICE B: What is the type of filament: PET (various colors) C: What is the diameter of the filament: 2.85mm D: What is the weight of a full spool: ca. 1kg E: What is the inner diameter of the core? 52mm F: What is the outer diameter of the core? 105mm G: What is the outer diameter of the whole spool? 199mm H: What is the weight of an empty spool? -- (don't have one now) I: Width of full spool (only when it is completely full because it can expand) 53mm J: Width of an empty spool (only when almost or completely empty) ca. 53...53mm ? (don't have one available now) 4. Seldom (1%): A: What is the filament brand: colorFabb B: What is the type of filament: NGEN C: What is the diameter of the filament: 2.85mm D: What is the weight of a full spool: rest of the data: see colorFabb PLA/PHA (same spools) E: What is the inner diameter of the core? F: What is the outer diameter of the core? G: What is the outer diameter of the whole spool? H: What is the weight of an empty spool? I: Width of full spool (only when it is completely full because it can expand) J: Width of an empty spool (only when almost or completely empty) Main spool differences: - inner core diameter 52mm (Ultimaker and ICE spools) / 53mm (colorFabb) - Ultimaker (old) and ICE spools have spokes / colorFabb spools have a free ring on outer edge (=easier to make filament clamps slide around edge) - rest seems fairly "standard spool"

Those shifts might come from the edges curling up, and the nozzle brutally banging into these edges, and missing a few steps. Just a guess, but not unlikely. I have occasionally had models being knocked off the build plate, if the edges curled up 1 or 2mm, at a layer height of only 0.1mm. This causes brutal crashing of the nozzle into the edge. For small models I use a dummy tower to print such small things where extra cooling time is needed. But for a large model like this, it would waste quite a lot of material. Unless you really hollow-out the dummy tower at places where no extra cooling is needed. Depending on the purpose, an option might be to cut the model in half using a vertical cutter plane, print both halves flat on the glass, and glue them together and sand the seams. Then you need no support at all, and there will be no overhangs (except for the little hole in the center).

I am not sure if the following is good advice for the UM3, so it would be best for an UM-engineer to comment. For cleaning my UM2 nozzles, and for scraping off the ashes, I have rounded and sanded the tip of a long brass M3-threaded rod (=3mm outer diameter). So the tip is now round and smooth, and not so likely to damage the inner core. When inserting it from above into the nozzle, I can very gently (!!!) scrape the side walls of the core and of teflon coupler, to remove loose debris and ashes. I used a brass rod, no steel, because steel is too hard, and more likely to cause damage to the brass nozzle. And I never brutally bang the rod into the nozzle, always scrape very gently. If the residu baked in the nozzle is shiny and hard like glass (like some PET materials do), then this is not going to work. But for rough, rather loose and carbonised material, it works well for me. Also, for cleaning the nozzle opening, I have sanded down a needle to 0.39mm, cut off the sharp tip, and rounded it. So I can insert it from below into the opening in case there would be a partial clog, to gently push it out. However, if done brutally, this could very easily damage the brass nozzle opening (since the needle is steel, even though soft steel). So I normally don't do it, unless I suspect there are some particles partially blocking the nozzle opening. And you definitely have to cut off the sharp tip, and round it, otherwise this is a guarantee on nozzle damage. Loose debris can then be removed with compressed air first, and atomic pulls. But as said, this is not official advice, I don't know if it is applicable to UM3-printers too; and it might be a good idea for an UM-engineer to comment. It requires common sense and being very gentle. So, for example if I had to run school printers, I would not let every pupil do it...