Print layers drifting towards rear

[DDLLC]

I followed your suggestion and ran a range of exposures: 20, 40,80, and 160 seconds. i measured at the corners and mid points of each rectangle element. The 160 second print did have a ridge along the midpoints of each side of the rectangle and from the ridge corners to perimeter corners. I've only attached a picture of the last one. The results in the chart for that group are measured close to the outer perimeter to avoid the ridge. The height of the ridge was about 0.25-0.30 mm more.

All of these films are very thin and cleaned on the FEP after draining. They were left on the film and then subjected to curing from both sides. The twenty second film was only post cured for 2 minutes which might not have been enough as it curled shortly after removing from the FEP. The others had 6 mins of curing on each side. They curled slightly but the 20 second one was curled bad enough that I really had to squeeze the caliper to flatten it and that might have induced some measurement error.

From the chart it really doesn't seem that front edge is being over exposed relative to the rear edge. The differences in the average thickness of the front measurements to the rear measurements is mostly less than 0.01 mm and half the exposure groups are smaller at the front edge. It's hard to conclude anything since this is a single trial set and differences are very small relative the measurement tools/ technique. I'd probably have to run at least 10 replications to analyze the data mathematically with a reasonable confidence interval.  The average of all data within an exposure group for both 80 seconds and 160 seconds are the same: 0.3675 and standard deviations of 0.016 and 0.014 mm. Ignoring the ridge for 160 s, this means the data population look the same. Pulling believable data from the noise on this technique will not be easy.

Hopefully Elegoo will respond with some suggestions as I don't have any great ideas of what else to look at.

[cliffyk]

Good work! Though I see nothing that would explain the rearward shift that has troubled you. One thing that is of interest is that the 20 s exposure cured to a thickness of nearly 200 microns and 40 s to 250 microns this indicates to me that bottom layer exposures of > 20 to 30 s may not accomplish a whole lot. My usual bottom expopsure has been 60 s; I may reduced that to 30 s and see what happens.

FWIW i printed a 44 x 87 x 10 block with my Mars--it came out with nice square sides and a thickness variation of 10.15 to 10.26 mm from "left to "right". IT does have an odd central depression on the "bottom" side, some 22 x 4 x 1.25 mm. I'll grab some photos later this evening/in the morning...

[DDLLC]

The shorter exposures are not even close to being "fully cured" so adhesion is likely the gating condition for minimum exposure time. I have often had my center rectangle not adhere through the whole print process.

Yesterday I made a small modification to my rectangle models and that is to put a small "drain" hole (1/2 half of a 1 mm diam hole) in each of the rectangles. The thought process is that theoretically the resin area (ie wet not crosslinked) against the build plate inside the larger rectangle could create a force equal to circa 140 pounds (neglecting the center rectangle). I don't know why such a force would be concentrated at one edge to force a tilt on the plate, but desperate times require desperate measures.

The short story is it was the closest print with a height close to the nominal but there is still about 0.5 mm difference between front and rear edges. Since the overall thickness is smaller (5.59 +/- circa 0.3 mm) the slant is less obvious. Conditions were zeroed on 120 micron paper, 60s base layers and 8 seconds on other layers. I'm repeating this now as it could just be random (at this point I'd say luck) but I've zeroed on 180 micron paper. The reason for doing so is that I've often noticed that I usually (maybe always) don't hear the FEP "snap; off the first print layer. I also didn't hear it on the first layer in today's trial either but we'll see what happens. I do hear the subsequent layers and with my double rectangles I hear both.

Another question I have is does your system on the first print layer move down to zero and then raise up a few mm before moving back down to perform the first print layer?

[DDLLC]

The reprint today with a zero on 180 micron paper gave the same results as before: rear average thickness is 5.42 mm vs 5.77 in the front. The standard deviations are large but there are only 3 data points, so I don't think the that data is too relevant.

[cliffyk]

Another question I have is does your system on the first print layer move down to zero and then raise up a few mm before moving back down to perform the first print layer?

Yes, it moves down to a position "1 layer thickness" above the limit switch controlled zero. If a logical "zero" position has been set (using the Z=0 function) it will move the a position "1 layer thickness" above the logical "zero"...

[DDLLC]

Today I printed my hollow rectangles at a 30 degree slant. The average thickness back and front are respectively: 5.14 and 5.02 while left and right are respectively: 5.14 and 5.04. I've tried to avoid measuring at a support location, but there does seem to be some undulation between supports.

The length average is 95.095 and width average is 65.025 mm.

I should mention that plate lift after tightening the set screws was 0.178 mm on the left side.

These are very close to design but the print time is 6 times longer.

I still want to find out why I can't print this without support and achieve good dimensional control.

Attached is the support side of the rectangle

[cliffyk]

I think that to large extent your current model is indicative of that we are mucking about with what is basically a "happy-homeowner/hobbyist" level machine; from a field of 3D printers (both resin and filament based) dominated primarily by users who play with dolls (judging by the majority of builds posted to Thingverse, Repaables, GrabCAD and their ilk).

In that field of endeavor the discrepancies we have been discussing would be barely noticed, and if noticed be empirically irrelevant to their goals.

I bought mine as a tool and could care less how well it produces wizards, trolls, castles or HO scale feed bins--or devices intended to save my ears from the ravages of face masks (I have one I carry in my pocket and fish out if the business/whatever I am visiting seem to feel the silly thing is needed)...

[DDLLC]

I think that to large extent your current model is indicative of that we are mucking about with what is basically a "happy-homeowner/hobbyist" level machine; from a field of 3D printers (both resin and filament based) dominated primarily by users who play with dolls (judging by the majority of builds posted to Thingverse, Repaables, GrabCAD and their ilk).

In that field of endeavor the discrepancies we have been discussing would be barely noticed, and if noticed be empirically irrelevant to their goals.

I bought mine as a tool and could care less how well it produces wizards, trolls, castles or HO scale feed bins--or devices intended to save my ears from the ravages of face masks (I have one I carry in my pocket and fish out if the business/whatever I am visiting seem to feel the silly thing is needed)...

I understand your point and I have no interest in making those items, although I admire the parts that were originally designed due to the intricacies that were modeled (clay or CAD: either would demonstrate great skills). I made a flat trial late yesterday and experimented with a 6 mm lift height, 60 mm/min speed and adding light off time which seemed to help. Thicknesses ranged from 5.03 mm to 5.34 mm. That would have been the best of all excepting the one with supports early yesterday. I will set up a test matrix with various lift distances, speed and off time. Something is causing a bigger build up on the front of the build plate and forcing that plate out of parallel to the screen. I at least have a hint that the mentioned parameters can be optimized to minimize that error. It will take awhile, but maybe I can get it done in the next 1-2 weeks.

[DDLLC]

I'm only midway through my Design of Experiments screening test and ran into an odd failure. At what should have been the end of that print, I found the Z axis "stuck" about 2 inches above the vat with the ball screw turning. I'm not sure it really completed the print, but I didn't hear anything odd (like the Stepper loosing steps or making sounds like it was jammed (again the screw was rotating), so I cancelled the run without noting how many layers were actually printed. Once I cancelled the print, I could manually move the Z axis up and down repeatedly. I repeated manual testing this morning again and everything seemed normal so I decided to tear down to the split nut that rides on the ball screw. I was able to easily drive the Z bearing off the ball screw. I didn't see anything odd on that nut assembly, but I also didn't remove the Z arm before driving off the bearing so I didn't see how it was riding on the ball screw (a mental lapse on my part). Upon re-assembly I tried to hold the top and bottom as close together as possible but I couldn't minimize the gap any better than what you see in the picture. Has anybody been able to achieve no gap and have it thread onto the ball screw? If so, how?

Also I have a bit of frustration with Chitubox that perhaps somebody can provide some clarity about. I have a experimental design that examines lift height, Lift speeds (base and interlayers), Exposure times (base and interlayers), and Light Delay times (base and interlayers). After creating the specific conditions and saving a cbddlp file; I can't find a way to see all those variables if I reload that file into Chitubox. You only see a subset of the conditions and except for exposure time you can't distinguish between base and interlayers parameters since only one is show. How can I review the full parameter set? If you go back and open the printing parameter window, you only get the last saved print parameter values (usually not correct for that cbddlp file).

Also I can report an observation that might help somebody. First to review this is specifically for Elegoo Grey Resign; 20 micron layer thicknesses; plate zeroed on 180 micron paper. The best print so far had an average thickness of 5.155 mm (design 5mm) with a standard deviation of 0.071mm.  For comparison the other trials have an average that is larger by 0.150 to 0.500 mm and deviations 2-3 times greater. The parameters that achieved this were:

Base layer numbers: 25     Base Exposure 60 s  Interlayer exposure 8 s
Lift Distance: 7 mm
Base lift speed: 100 mm/min    Interlayer speed: 50 mm/min
Retract Speed: 60 mm/min
Base light on delay : 15.2 seconds      Interlayer Light on delay: 16.6 seconds.

This is a screening experiment so these parameters are not optimized. Nevertheless it could be helpful to somebody who may be struggling with this resin.

The other observation is that I think a base exposure time of 40 seconds is too little. I started with that for a few trials and noticed corners lift or my inner rectangle  releasing from the build plate. Fifty seconds seems minimal.

[cliffyk]

That is an "anti-backlash" nut, there should be a gap of 0.5 to 2.5 mm between the upper and lower bits--with no gap they would quite surely bind and jam up.

This is shown in Jackson Products' instructional video:



Jackson also includes a heavier spring with their kit as they have found the OEM spring's tension to be insufficient.

I am confused as to how the lead screw could have been turning with no motion of the Z-axis trolley and undamaged anti-backlash nuts.The topmost nut is quite positively fastened to the trolley with four socket cap screws--it seems to me that they would have to be broken or/and both halves of the nut stripped for the screw to turn with no trolley motion...