Form2 doing laser calibration before *every* layer?!


#1

I haven’t seen this one until now, so no idea what triggers it.

Printing white resin, fairly big object (small cross section, thin shell, but spans the build volume diagonally). 13+ hour print.

About 3.5 hours to go and i’ve just noticed the printer does a laser power calibration run before every single layer. I’ve sat in front of it and waited for about 30 layers - every single one gets the four extreme corners flashed, than a second or so of pause, then the actual layer.

Any ideas what the purpose of this is, and why on earth it’s necessary?

I’m worried it will create half-cured gunk at best, or complete mess of things at worst over the span of the next 3.5 hours…

Video coming up tomorrow. Gotta get some sleep.


#2

It just went nuts while dispensing resin - it started opening and closing the bite valve quickly about 10 times in a row within a timespan of seconds (zt-zt-zt-zt-zt-zt-zt-zt-zt), and continued dispensing while the vat was already in motion.

It closed the bite valve as printing of the next layer was starting.

I’ve never seen erratic behaviour like this before.

It’s almost like the longer it prints, the buggier it gets.
Formlabs, do you have an integer or buffer overflow of some kind in the latest firmware?


#3

My machine does it as well. There are 4 photodetectors in the corner.
However, I’m not sure if all machines do this as I haven’t seen them on a newer machine when I disassembled it.

Maybe this is a way to calibrate the galvo’s? Not sure.

This is new behavior since newer firmware releases. I think this is done to stimulate the rubber of the bite valve so it closes better and doesn’t leave a big resin drop at the rubber.


#4

Yeah, i know they all do it, i’m more worried about it doing this for every layer of a 3200 layer print. I don’t remember that happening before.

Additional source of worry is the fact i can clearly see it diffused through the white resin. If i can see it through the resin and through two layers of acrylic, the amount of energy dumped into the resin cannot be negligable. Couple that with 3200 repetitions and it could be resulting in semi-polymerized resin crud in the tank at best, and blobs of solidifed resin in the corners of the tank at worst.

Can’t say i see the point in it, the drop is still there (in my case at least), and i have a (completely subjective and non-quantifiable) feeling it reduces the life of the bite valve significantly through sheer increase in the number of cycles.

Formlabs always stresses the cartridges have limited lifespan and are not reusable because of the bite valve MTBF. In fact, they suggest NOT to dump resin back into the cartridge because of precisely this.

This just makes it worse, doesn’t it?


#5

Hey there @Ante_Vukorepa,

Thanks for bringing both of these issues to our attention.

For the laser testing at the four corners behavior, that’s totally normal and really shouldn’t impact your prints in any way. The post above from @fantasy2 is exactly right about this behavior and there’s nothing to worry about.

Now, for the dispensing issue you’re seeing, frankly, that’s a new one for me, so I had to ask around the office a bit, and turns out it’s new to most people, so we’d love to help you investigate that problem ASAP.

If you’d like to reach out to our support team at the link below we can start to figure out what’s going on! Typically you’ll get a reply within 1 business day. While you wait to hear from us, grabbing a video of this behavior happening would be hugely useful. Much easier for us to hear weird noises for ourselves than get them described to us. :slight_smile:

https://support.formlabs.com/s/contact-support?language=en_US


#6

Unfortunately, didn’t catch it with a phone on hand last night.
I’ll try and catch it in action on the next print (if it happens again).


#7

I know the test fire is normal, but is testing before every single layer normal behaviour, though? I don’t remember noticing that before.

Out of sheer curiosity, if that’s normal, what’s the reasoning behind doing it for every layer?

Edit: vid for reference


#8

Additional observation that might or might not have something to do with the overzealous calibration - the estimated print times with recent firmwares and preform versions are way way off.

Not slightly off.

50% to 100% off in extreme cases (e.g. i’ve just finished printing a print that was estimated as a 7.5hr, it is now 5am, and i started it at 5pm - so it took 12 hours). And the amount of error rises sharply with the ratio of time spent calibrating before each layer vs. printing the layers, i.e. gets more pronounced the smaller the surface area of an average layer.

Now, i realize i don’t have all the facts (not even close), but the huge error in estimated print times, coupled with the fact the printer didn’t do calibration runs before every single layer before, and the (subjective) silliness of the idea of calibrating before every. single. layer. leads me to believe something’s definitely not quite right here.

Is it true the new printers don’t have photodiodes in the corners like @fantasy2 says? If so, did you test the firmware on old rev hardware?


#9

Hi @Ante_Vukorepa ! I’m an engineer on Dan’s team here at Formlabs. I wanted to get back to you on some of your questions about the printer’s laser scans. We refer to the process as the “fiducial scan” and the panel with the photosensors as the “fiducial board” so I’ll use those terms throughout.

First, just to clear up some confusion—neither the fiducial scans, nor the sensors that are used for them, are new or temporary, and they’re not going away either. Every Form 2 ever made has had a fiducial board and (apart from a few beta releases around launch) every firmware we’ve ever released has included fiducial scans on every layer. I’m not sure what @fantasy2 saw, but I’m afraid they must have been mistaken when they said a new printer was lacking those sensors; that’s just not the case. Every print you (or I!) have ever done on a Form 2 included these scans.

Fiducial scans are an important part of the printer’s optical system, and doing them every layer really does benefit the results. For a variety of reasons (thermal expansion of the printer components and frame, wear and tear on the galvos over time, electrical fluctuations, and more) the galvo mirrors that aim the laser will drift away from their initial position. Fiducial scans let us detect and compensate for that drift and keep the laser on target, to ensure that your part prints as designed. If the fiducial system breaks down, layers won’t always line up from one to the next. This might result in a strange “wobbly” quality to the part, or in extreme cases, parts will start skewing across the platform into a curved or parallelogram shape.

The fiducial sensors are located at the corners of the optical window—or more specifically, outside those corners. When we aim the laser at the fiducial sensors, the laser spot itself is well away from the window or the tank. Although some scattered stray light is visible coming up through the tank, it’s a tiny fraction of the laser’s total energy, so it shouldn’t be an issue for your resin. If you do notice any half-cured gunk in your tank (with no print failure to explain it,) then please let us know—it could mean that your particular printer is having an issue or aiming at the wrong spot when performing a scan.

As for your print time discrepancies: although fiducial scans do add a tiny bit of print time—and, as you say, the time cost will be a greater percentage of the whole the smaller the surface area of each layer—that time is only a few seconds per layer and should really be negligible in almost every case. We haven’t changed fiducial behavior lately, so the scans are only adding as much to your print time as they always have. There are a lot of things that can extend print time, but by far the most common reason a print might be longer than PreForm’s estimate are pauses for tank level sensing and for resin dispensing.

Because these parts of the print are non-deterministic and can’t be calculated beforehand, they’re left out of PreForm’s estimate, and will always result in a print time slightly longer than the prediction. On a healthy printer with a full cartridge of resin, the time difference will be about 5-10% (and always has been). If the printer’s level-sense system is having trouble (for instance, if there’s resin on the sensor panel) or if the cartridge is getting low and dispenses slowly, the discrepancy will grow. We’re currently working on revising our estimates to include an approximate allowance for levelsense (based on averages of similar prints) which you can try out by opting in to our beta program, but since it’s based on global averages, it won’t help if your particular printer is having trouble. PreForm simply has no way to know that that will happen when it calculates the print time.

I see we already suggested opening a support case to check on the strange dispense motor behavior you saw (which definitely isn’t normal)—if you did so, then I’d definitely encourage you to check with your support agent about what might be affecting level-sensing or dispense on your printer and causing these extra-long prints. I can tell you that I’m confident that fiducial scans aren’t the source of the trouble.

I hope this helps! Please let me know if I can explain anything else and I’ll be happy to keep the discussion going.


#10

Thanks for the detailed explanation!

Learn something every day. :slight_smile:

The form 2 teardown has some photos and info on the photodiodes.


#11

LOVE the response from @Declan!

Thanks!


#12

Awesome response, thanks a lot!

Regarding the fiducial scans - i’m noticing them recently, where i’ve never noticed them before (in years of ownership). While my memory is, admittedly, pretty fallible, i did notice that for layers with a very small cross-section (a mm² to a few mm²), the scans light up the resin almost as much as the layer curing, at least viewed from the front, and very subjectively. In a dark room, they’re visibly diffusing through the resin itself as well (and are visible from the top).

If the scans were always there for every layer, and are unchanged, then i can only offer two explanations:

  1. either something is not quite right with my printer and the fiducial scans are hitting the wrong spot, or
  2. the light from the scans diffuses through white v4 much more strongly than through other resins - since i never used v4, and the last time i’ve used white in particular was v1, long time ago, that would explain why i’ve never noticed it before

I guess 1 would also cause significant degradation in print quality or dimensional accuracy, which i would’ve noticed, so 2 seems more likely. Unless, again, it’s just my memory that’s off :slight_smile:

Regarding real print times and estimated print times… Things are much more murky here. The deltas i’m seeing are definitely not 10% or 15%, but 50%-100%. Prints that are estimated at 7 or 7.5 hours finish around the 13h mark or 14h mark. Additionally, these are not happening under resin-low conditions. In fact, the reason i haven’t contacted the support yet about the weird bite valve actuator behaviour is - because i haven’t been able to catch it in action. The one time i did hear it happening (in 4 prints or so), i wasn’t fast enough to get from the office to the other room. The prints i’ve been doing lately have fairly small cross-sections (even though the prints themselves are fairly large) and have not required frequent resin level checks nor refills. The cartridge is about 2/3 full.

To add an additional odd data point - i’ve been printing black and grey (v4) for the past several months, and have not noticed any huge discrepancies between estimated times and actual times. Could the white material profile be the culprit?

While i’m already wasting your time, i’ve got one semi-unrelated question about the resin level checks - why is there always a resin level check before the very last layer of the print?

Oh, and back on subject - i still don’t quite get the need for fiducial scans / calibration on every layer. Do the galvos really drift that much? We’re talking seconds between layers in some cases. How much thermal drift are you going to get in a span of, say, 10 seconds? The scans themselves take longer than the layer itself in some cases. Also, just out of my own curiosity, how come Form1/1+ could get away with not having them, yet still not have significant layer drift for most prints?


#13

Sorry for the wall of text, i thought splitting it into separate posts would just make it worse.


#14

It’s definitely not impossible that your printer is having trouble hitting the sensors for the fiducial scan, and is somehow hitting the tank instead—but as you said, if that were true, the results would be very visible in your parts! If your parts are coming out the right shape and size (and don’t have any rough surface texture or other signs of a blurry laser) then I’d be confident that the scans are going correctly. I think option 2 is definitely your guy. The visibility of the laser in general (both for scans, and for actual curing of the layer) is drastically different from resin to resin and White definitely shows it off more than almost any other resin—even Clear won’t make it as visible, since Clear won’t scatter the laser away from its destination and toward your eyes as much. (Think flashlights in air vs. flashlights in fog.)

100% increases from the print time estimate to real print time are a surefire sign that something is going on with the printer, and there are definitely ways we can help nail down the cause—if it’s not a low cartridge, then something else is doing it. Since you’ve only seen it in one resin, something about that one cartridge would absolutely be my first guess, although we could also check on any possible interactions with the material settings for White 4. A support case is really the best option here. I can only do so much in a forum post, but a case will allow our support agent to check your printer’s diagnostic logs and get exact durations for each layer of the print, data on dispense amounts and levelsense readings, etc. to start narrowing down possible causes. Whatever’s hurting your print times, I’m sure we can help you sort it out.

As for your other questions: We do a final level-sense check just so that we can verify the level in the tank and, therefore, the amount of resin dispensed from the cartridge during the print (so that both those amounts can be tracked accurately on the tank chips and on Dashboard.) IIRC we also top off the tank if it’s not quite full, so that you won’t have to wait for that to happen when you start your next print (I’d have to actually run a print to check).

As for scanning every layer vs. some other schedule—you’re right that we might have been able to get away with scanning a little less often and still gotten acceptable results, but we went with a more definite safeguard against any disruption to the print, and the highest level of accuracy we could get.

As for the form 1/1+, the answer is…it DID have layer drift/jitter for at least a sizable number of prints. Progressive galvo stretch/galvo drift issues were the eventual cause of death for a sizable number of 1+'es, but even before things got that bad, the 1+'s layer registration wasn’t as tight and saw far more positional variance from layer to layer than the 2 because of its lack of a fiducial system. If the drift was more of a back-and-forth jitter than a continual movement in the same direction, the effects would be less apparent, and of course it wasn’t every print every time, but 1+ layer lines were more visible and accentuated in part because each layer might be a few microns off from the one before.


#15

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.