Andrew1

Cura seems to prefer locating the z-seam at corners even when set to "random". Corners can concentrate physical stress, so irregularities/artifacts occurring at corners could compromise the strength of some models. It would be useful in some instances to have the z-seam positioned on the flatter areas of a model or the areas of least curvature.

One final suggestion: When printing atop a support with a z-gap, the print lines end up being more cylindrical rather than flat. It might benefit layer adhesion if the part of the next layer that is directly above this area had an option for increased flow. Assuming a .4mm nozzle and that the lines atop a support are cylindrical, and touching but not overlapping on the sides, the additional (cross-sectional) area corresponding to one print line would be the difference between that of a .4x.4mm square and its inscribed circle, divided by two. I've not done any experiments with this though. I've been using a 3d printer for about a year now and I've found that getting smooth, dimensionally accurate supported surfaces is in general very challenging. Its sometimes difficult to work around this limitation depending on the model to be printed.

The problem in this case would be that the first layer printed atop the support would have to "fall" twice its layer height. A .15mm layer taking up a .3mm gap would probably give you a similarly low-quality surface as a .2mm layer taking up a .4mm gap, albeit on a smaller scale. It would change the model, and the resulting support would also be different. Consider a 10x10x10mm cube floating 10mm above the built platform. If it is printed atop a support with a .4mm z-gap, then you end up with a 10x10x10.4 cube. (I tested this using cura's "standard quality" profile, the cube actually ended up 10x10x10.57, perhaps due to irregularities in the support surface.) If the model were to be adjusted so that it's a 10x10x9.6 mm cube, located at 10.4mm above the build platform the cube should be dimensionally accurate, in principle. Incidentally, a variation on adaptive layer height could be used to achieve sub-layer-height dimensional accuracy for horizontal surfaces. (both upward and downward facing) Any time a horizontal surface is not at a multiple of the layer height, the height of the last few layers before that surface could be adjusted so that the surface is at exactly the correct height. Another thought, which would really only be practical for horizontal support surfaces, might be to have the printer pause before the final layer of the support surface. Then the user could manually swap a dissolvable filament for that last layer of the support, with the printer pausing again after that part of the layer has been printed and the user swapping in the original filament. Then one could print without a z-gap (which I presume is partly the point of a dissolvable filament), at least for horizontal surfaces.

Yes, the underside takes up the space and rests on top the support roof. The smaller the z-gap and denser the support interface, the harder it is to remove the support. I just made a feature-request thread about this topic: https://community.ultimaker.com/topic/42320-support-settings-feature-request/ Presumably, the real solution is to have dual extruders so that you can have a separate, dissolvable support material. I wish that more machines besides just high-end printers came with dual extruders. An extra hotend and stepper motor is no huge expense. As you've discovered it's otherwise hard to get a decent surface where supports are involved. In my experience, a higher density support interface will help somewhat. Make it as dense as possible without it permanently bonding to the part. When designing a part, I will usually make an effort to minimize necessary supports.

Can I ask what sort of filament and print surface you're using, and how it's prepped? In my experience, when the first layer fails to adhere to the bed, it's always one of two things: the nozzle is too far or too close to the bed (i.e. the bed is not trammed/leveled properly), or, the bed is not prepped correctly (cleaned, and a thin application of gluestick depending on the filament material). Now I can usually spot the problem right away and very rarely do I have adhesion failures.

With certain materials, particularly those with strong layer adhesion and poor bridging (e.g. tpu, nylon), it is challenging to achieve dimensional accuracy and a decent finish in areas that require supports. Relatively high support density (or rather, high support interface density) is a must. I have two suggestions: 1) Arbitrary support z-distance/support top distance. Presently, the setting "support z distance"/"support top distance" is always rounded to a multiple of the layer height. For 0.2mm layer height, a z-gap of 0.2mm frequently results in support structures that are strongly bonded to the model. On the other hand, a 0.4mm z-gap introduces more dimensional inaccuracy and the underside contacting the support is quite rough. I tend to favor the latter option but strongly suspect a 0.3mm gap would be a good compromise. 2) A setting to vertically "erode" the planar surfaces (or even just those that are horizontal) that would contact the top of support structures, by a certain distance (presumably defaulting to "support top distance"). This would make room for the extra space in the z-gap that gets taken up by the first layer built atop the support structure. To be clear, this would be a pre-processing step, with the support generated as normal afterwards, based on the adjusted model (not the original). Edit: For implementing the first suggestion, the support could be built as normal up until the last few layers of the support top. Then, for example, if we have a .2mm layer height and want a .3mm gap, it could print two .15mm support layers and then skip the final support layer for a total .3mm gap , using a z-hop for moving the head to and from the support within those layers. Then the rest of the print would proceed as usual. What I mean by "arbitrary" is a reasonable range of values, say between 0x and 2x the layer height.

Hair spray with glue stick applied on top (after the hairspray dries) is what I've been using lately.

I agree with GregValiant's comment that it might be easier to just treat the problem as a general adhesion issue. I'm sure there's a reason that cura prints the walls first, though I'd still be curious to know whether or not the line ordering I suggested would work. And yes, you can probably reduce the complexity of the comparison function by organizing all the individual coordinates into a suitable data structure, perhaps a k-d tree. This depends on the adjacence function though. If you only check for every coordinate in a remaining sequence whether or not it's within a certain distance of at least one coordinate in the already-processed sequences, then yes, I think it would work. More generally, you'd still be doing a pairwise comparison between all sequences (and presumably all coordinates). Anyway, I've never had great results with PEI. What material are you using? For PLA and PETG I've had much better luck with a glass bed and either gluestick, or an "extra firm hold" hairspray, optionally topped with a light gluestick application. Printing ABS on PEI was very sensitive to bed temperature and there was hardly any middle ground between adhesion failures and permanently bonding the part to the bed, especially with larger models. ABS doesn't seem well-suited for 3d printing in general and I don't use it anymore.

I had the same thought while I was having adhesion issues with ABS. I've only just now had a cursory look at the gcode format, but it seems like you could separate out the G1 sequences, pick the longest one to start with, and then choose the remaining sequences in order of "greatest adjacence" to all the prior G1 sequences. It's an N^2 algorithm but approximating the degree of adjacence (perhaps by only taking every nth point of the current G1 sequence) could save some time. It might not be too bad if it's just the first layer. This would probably maximize the degree of contact, but I should think the G0 sequences would have to be re-generated once a suitable order of G1 sequences is computed and I'm not sure how cura does that off the top of my head. It would really make more sense to do this during the slicing process rather than after the fact. A basic definition of the adjacence measure between the already-printed lines and a remaining sequence could be the average number of individual coordinates in the remaining sequence that are within a certain distance (presumably the line width) of at least one coordinate from the lines already printed.

Looks like support for multiple middle raft layers was added in 5.0 (didn't see it at first because the setting was disabled), so I guess that solves my problem.

Thinking about it more, the raft and support settings could probably be unified so that any part of the object not touching the build plate (even if it's the entire model) is supported, or just the lowest part of the model (which would be similar to enabling a raft but without having supports). In this case, it would be convenient to combine the features of the support/raft so that rafts and supports can have separately adjustable bottom layers, top layers and any number of middle layers with adjustable fill, fill pattern, etc. That's probably how I'd do it, anyway. P.S. z-fade is a marlin setting. I assume Ultimaker printers have a similar feature. It's my understanding that if the z-fade is set to N millimeters then the Z corrections applied by mesh leveling are gradually reduced during the first N millimeters of the print and are no longer applied after Z > N millimeters. I'm sure Ultimaker makes a fine product, but at present I only have an ender 3 whose bed is only slightly flatter than the swiss alps.

I'm using cura 4.8 and there does not seem to be any way to adjust the density of a raft without the top layer being uneven. The density could be controlled with the line spacing but there's no way to make the topmost layer(s) solid while having many lower-density middle layers. This seems important because unless the print bed is very flat I don't see how to make a dimensionally accurate part without using mesh leveling and a raft whose height is greater than the z-fade setting. If your raft is higher than the z-fade then the flatness of the bed no longer matters, just that the printer's x/y/z axes are orthogonal. However, if all but the first two layers of the raft must be solid then a 5-10mm raft for a large, flat part is not very economical. Wouldn't it make more sense for the raft to have a structure like that of supports, at least for the middle raft layers? Perhaps the easiest way to implement this would be a setting to adjust the number of "middle layers" so then many low-density middle layers could be used to achieve dimensional accuracy and take up the unevenness of the bed without using too much material.