So, finally I got around to make another test-print - the results of which I will use to flesh out my theory of why the prints turn out like they do.
So in this “episode” we’re gonna take a look at the specific issue of “overcured resin” or missing detail and lack of sharpness on the support-side of prints printed with Gray v4 resin on the Form 3. I’m also gonna give a bit of an explanation as to why I think this is happening.
First off, I’m gonna give a bit of summary on the test-prints that I’ve done, as well as some print of actual models that I had intended for sale, before I became aware of the printers “defects”.
As of this writing, I have done three test-prints, for the specific purpose of finding out exactly how the printer is misbehaving and why it does that. The prints were made up of small “centicubes” that was supposed to act as “simple” geometry. The reason that it had to be “simple” geometry, is that it makes it a lot easier to see and measure the difference between how the model was supposed to be, and how it actually turned out. Now of course a cube is not gonna expose every single type of flaw that the printer might have, and it might not indicate some issues that would occour at odd angles and the like, but it does give a good indication of whether the issues we are facing are related to errors within coding/math expressions and calculations of the actual machine or preform - or if the issues are related to a physical error in concept, design or construction of the machine itself.
The specs of the three prints are as follows;
- 25 Micron Layer Height
- Gray v4 Resin
- Firmware 1.5.17
- Preform 3.4.4
- 25 Micron Layer Height
- Gray v4 Resin
- Firmware 1.6.4
- Preform 3.4.6
- 100 Micron Layer Height
- Gray v4 Resin
- Firmware 1.6.4
- Preform 3.4.6
Print 1 was my initial test-print, and my first post in this forum was based on the subsequent analysis of the printed parts from that print. Two issues were identified from that print;
1: Overcured resin of support-side of model
2: Artifacting/displacement on/of parts of model
As Print 1 was printed long before the post on this forum was made, new software and firmware had been released at the time of posting. The next two prints (Prints 2 & 3) were done after the printer and Preform had been updated to the (at that time) most current version of the soft/firmware.
Print 2 was a follow-up print to the discoveries with Print 1, and to see whether the new soft/firmware had done any difference in the outcome of the print. It could be concluded (which was also documented in a another post of mine in this thread) that issue 2 had been significantly reduced, but issue 1 had slightly worsened or was stagnant.
Print 3 is the first print that was supposed to try and test whether or not my theory - which I will get into at the end of this post - could have any merit. In the meantime between the actual printing of Print 2 and Print 3, new soft/firmware had once again been released. I did not update anything - as it would be changing to many “parameters” for my tests to be conclusive about anthing. So I had to keep Print 3 the same soft/firmware as Print 2 - otherwise improvements or new errors to the soft/firmware might change the outcome of the test and thus give a distorted view of what is really happening.
Now for the actual review of my third print - which I have yet to show:
The print itself has turned out really great. Issue nr 2 is completely gone at this layer height - and even under the microscope I can’t seem to find any signs of this issue.
Issue nr 1 however is still present. The deviation is though significantly less at 100 micron than what the same test showed for 25 micron in Print 2. Now we’re down to rougly 0,35mm in deviation instead of the 0,55mm that we saw in Print 2. This is a good development and also ties into my theory about why this is happening at all. The hole and notch of the cube has also decreased in the amount og deviation. This is down to about 0,25mm from approximately 0,32mm in Print 2.
Take a look at the attached pictures and see for yourselves.
It can clearly be seen that the cube from Print 2 is higher than the one from Print 3. If you look closely, you can also see that the notch is “higher” on the support side on the cube from Print 2 than it is on the cube from Print 3.
Same goes for the hole on the other side of the model. You can clearly see that the hole is more rectangular on the cube from Print 2 than the one from Print 3. Contrary to what some may believe; it is not the “top” of the hole that “sags” down due to lack of support. The part of the hole that is too narrow is the bottom-side, which is taller than it should be - once again a good indication for my theory.
Now then, what is this theory of mine? Well, it’s a theory inspired by people within this thread. I can’t be bothered to sift through the thread to find everyone who has ever mentioned “light control” as a possible issue, but nevertheless, if you have once theorized that “light control” might be the issue then I’ve probably taken some inspiration from you
“Light control” is a very broad term though, so it would help us to narrow down the amount of possible issues that migth lie within the umbrella term of “light control”. When looking back at Print 1, 2 & 3, a pattern of issues and non-issues start to emerge.
First of all, what elements of the print went as expected? Well, apart from some artifacting - and maybe some displacement (which was not consistant across all models) the models turned out fine on all sides. That is, the sides were within spec. The biggest deviation on any vertical side was 0,01-0,02mm, which I would classify as being within what one can expect from a printer of this quality and price range. What about horizontal sides then? Well here there were two different outcomes;
- Smooth and within the same tolerances as the vertical sides
- Smooth but curved surface - Surface curves upward (towards the buildplate) and has a deviation from spec of about 0,55-0,35 depending on layer height.
Out of these two outcomes, only outcome number two is interesting to continue investigating, as it is the only outcome that produces something different than what we expect.
Now all horizontal sides are not affected by this issue, so we must try to understand what leads to some being affected while others aren’t. A common denominator of the sides being affected, is that it is always the top-side of any geometry that seems to have the issue. Contrary to the top-side, the bottom-side of any geometry always seems to be within spec. What do I mean when I say “top-side”? “The top” can be many different things when talking about prints, models etc. Well, when I say that it’s the “top-side” being affected, what I mean is that it is the side facing the buildplate or the “top” of the printer. It is the side facing away from the tank, and facing the support structure.
As only top-facing surfaces of prints are affected by this issue, then we must ask; why is that? Well, this is where my theory comes into play.
When I first started to print on my Form 3, way back in November / December 2019, I printed some models that I’ve designed for the purpose of selling via my company. One thing I noticed at the time was that there was always a “bad” and a “good” side of a print. Usually when I printed I used the “auto-orient” function of Preform, and the only thing I ever adjusted was the supports and touchpoints. As I printed more models, I noticed that it was always the same side that was “wavy” undetailed in its finish - the top-side of the print. All other sides were usually nice and crisp - except of course for the artifacts and displacements that so many other people have already documented at length.
If one takes a moment to think about the way the Form 3 works, one realizes that the printer uses a laser, guided by mirrors (Front/back) and steppermotors (side/side) to direct the laser light onto the right point in the 2D-plane that a layer represents. The buildplate controls which layer is currently being worked on, and this is also facilitated with steppermotors, like the side/side travel of the LPU, though in this case it controls up/down movement of the buildplate.
When printing the first many layers of any model or geometry - this is done directly onto the buildplate. Whether it’s the model put directly on the buildplate itself (Formlabs doesn’t recommend this) or either full/mini-rafts for the model + support-structure. The laser will hit the resin immediately infront of the buildplate - travel through it to some degree and will afterwards be stopped by the buildplate itself. However, when the printer works on layers further down the model, the buildplate doesn’t act as an endpoint for the laser, and the laser will (as much as it is able to) penetrate through the current working layer, and into either previous layers or residual/uncured resin of the tank. Just as lightsabers doesn’t work in real life, due to the fact that you can’t just tell a laser when to stop, the laser in the Form 3 will keep on going until all its energy has been absorbed and/or reflected by other material/objects. While printing directly onto the buildplate the top surface turns out excelent, as the energy of the laser is completely absorbed and/or reflected by the buildplate. But as the buildplate moves further away, and the current active layer finds itself deep within the tank of resin, the laser is only going to be stopped by the surrounding resin. If your model has a top-facing surface, that isn’t directly on the buildplate, the only thing that lies on top of that surface is uncured resin. (As the working layer + a bunch more of the model is always submerged in resin).
When uncured resin is hit by laser light it absorbs some of the energy, but in that process it also begins to cure - which is how this technology works in the first place. Although in this case, the resin recieving laser light is resin that we don’t want to cure - so we end up with cured resin, in a place where we don’t want it!
Why does this happen then? Well, this takes us back to what my theory actually is; Laser-power control being mismanaged, thus causing the laser to penetrate too far beyond the currently inteded working layer, and as a consequence of this, resin occupiyng this space will be cured to some extend.
This theory seems to be supported by the fact that this issue - of overcured resin on the top-surfaces on any model - seems to get worse as layer height decreases. As the only difference between Print 2 & Print 3 was the layer height, the correlation seems to be quite obvious. at 25micron the deviation on the top-surfaces was around 0,55mm while at 100micron the deviation was only around 0,35mm.
Now why would this difference occur? Well, I have tried to illustrate what I believe the issue to be;
If you take a look at illustration nr 3 above, you’ll see an illustration of the laser’s layer penetration. The image is to scale with the real deviation, so you can see just how much resin the laser actually overcures, when printing the top-surface of any geometry. It is important to know that this illustration is for the 25micron layer height. The red area of the illustration, is to show the resin that is within the zone where resin would be cured, if it stays in place. Although as you all probably know, the buildplate moves up/down between each layer, and thus the resin that is too close to the edge, will fall of or be mixed around between layers. This is what I think is the reason for the “curvy” nature of the top-surfaces in my test-prints.
Although I don’t know how much energy is disapated in a cured 0,025mm layer of resin contrary to a uncured layer of resin at 0,025mm - This illustration tries to show how much resin the laser has to pass through, before so much of the laser energy is absorbed, that the laser can’t cure any further resin.
Now back to the question; Why is there a difference between 25micron and 100micron? Well, for this we’ll take a look at illustration nr 4:
On illustration 4, we see how the laser layer penetration looks, when the printer prints at 100micron. This illustration is also to scale, and you can see that the amount of overcured resin is much less than that of illustration 3 - especially compared to the layer height. First of all, much more of the lasers energy is needed to cure the full 100micron layer as opposed to the 25micron. This will mean that the energy-level of the laserbeam, that makes it all the way through the layer, is much lower than that of 25micron. Only after the layer-thickness of 100micron, will the laser start to go into unwanted territory and cure resin that is not supposed to be cured. As the laser has already used up alot of its energy, it doesn’t have to go through as much uncured resin before it is depleted of energy. Also it is worth noting, that since every layer is 100micron, the laser will be moved 100micron further away each layer and get less energy overall wasted on resin that is not supposed to be cured, and thus the amount of unwanted cured resin is less. On illustration 4 it is seen that the laser moves out of the “overcured resin”-zone much faster than on illustration 3 - supporting this idea.
In conclusion to this mess of a theory, I woundn’t 100% say that it’s proven and that no other explanation might be valid or possible. This theory however seems to be describing what is seen across a bunch of different models posted on here. Especially the recent test done by @lmlloyd shows very clearly how the top/support side of the “wheel-thingy” is unsharp and undetailed, while being nice and fine on the bottom.
Why is this issue less pronounced with Black resin? Well, I don’t know, but I would suspect it has got something to do with Formlabs maybe trying to dial in one resin to perfection, before they begin on all the other resins - or maybe black resin just requires more energy to cure, and thus the issue will not be as pronounced. As far as I can tell, from pictures on this thread, it seems that black resin also has this issue - it’s just a lot less than with Grey.
If I get around to do it, I might try printing at an even bigger layer-height than 100micron, just to see if the issue gets even less then. According to this theory, it atleast should Looking forward to seing the further developments on this thread.