Form 3 much slower than form2

Hello, I am using beta 3.1.2 to be able to print at 25 microns with the castable wax material with the form3, but I am experiencing very high printing times, 23h to print 1600 layers, is it normal?


1 Like

exactly the same 1600 layer work, with the form2 at 25 microns, 9 hours

1 Like

I can confirm that the Form 3 printer is considerably slower than the Form 2. When printing the same collection of models (3194 layers) on both printers in parallel, using the same parameters (25 µm, PreForm 3.2.2) and resin (Black Resin V4), the Form 2 (Firmware 1.19.13-80 from 2019-11-12) needs 24 h and Form 3 (Firmware 1.4.4 from 2019-12-11) needs 33 hours.


This seems most unfortunate, I wonder if it’s a byproduct of their new LSF technology. Considering all of these issues both in print times, reduced print quality in cases, lack of features and resin support I believe I may stick with a Form 2 for the time being. It’s sad because this machine has amazing potential. Maybe with future updates it will live up to the name.


I think that one of the reasons for the slower prints is the new, smaller size laser dot.

The Form 3 has a 85 micron sized laser spot, 40% smaller than the Form 2, which has a 140 micron laser spot. It makes sense to me that the Form 3 would be substantially slower, as it now has to make 40% more passes to fill in the same area it used to fill using the larger dot of the Form 2.

For example if you have a 1.4mm wide path, it used to take 11-12 overlapping passes to fill it, not it would take at least 18-19.

24 + 40% = 33.6. Pretty close eh?


That would’ve been true if the laser power was also lower, but it’s not.
At 250 mW, the amount of light delivered is the same per unit of time, regardless of spot size.

Meaning you can compensate for smaller exposure site with faster scanning.


Bu therein is the issue. IS THE SCANNING ANY FASTER? I don’t think so.

Even if the they managed to get the galvo governing the X axis to go faster, the LPU movement in the Y axis is substantially slower than what any galvo can achieve, so if anything, that slows things down even more (unless you’re drawing a single line in the front to back direction).

Let’s face it, the Form 3 is mechanically slower than a Form 2 (mechanically challenged :wink:).


Yes, but it can go much faster than that - the leadscrew isn’t insanely fine pitch, there are no steppers higher res than 0.45 deg/step (and i doubt they’re using 0.45, more likely 0.9), and even with 128x microstepping, it can still go LOTS faster than it does currently (there’s an ARM CPU powering it, after all) while exposing resin.

So no, i doubt LPU movement is in any way the limiting factor at the moment / i doubt it’s current speeds are the fastest mechanically achievable. In fact, if you look at the motions during exposure, you’ll see it moves about an order of magnitude slower than when it parks, unparks, or even travels through areas with no objects to print.

It seems like you’re assuming the LPU moves continuously. I was under the impression that it doesn’t, ie it’s not moving while the laser is scanning. If that’s the case, the maximum speed of the stepper doesn’t matter, as it will be limited by acceleration, deceleration and settling time.

1 Like

At the speeds involved, there is really no discernible difference between “moving continously” and not.

It moves in steps. If individual moves are fast enough, the motion is smoothed out into a continous travel by the mechanics involved (it’s pretty similar, if not exactly the same, as “ironing out” a half-wave rectifier output with a capacitor).

When I print with Durable, the biggest addition to time is the Z stage travel before a mix occurs. For some reason it moves up incredibly slow. This is after the peel event and the build platform is simply moving up to allow the wiper clearance to mix. I flagged this to support - they mentioned that they think that’s happening on all resins, but I’ve only noticed it for Durable (I haven’t looked that hard for other resins though).

Also, they acknowledged that their time estimates currently aren’t accurately taking into account the mixing times which is why they’re wildly off in some instances (particularly bad for Durable).


Thanks for sharing the info! We also noticed that Durable takes especially long to print, much longer than the estimates.

1 Like

It moves in steps. If individual moves are fast enough, the motion is smoothed out into a continous travel by the mechanics involved (it’s pretty similar, if not exactly the same, as “ironing out” a half-wave rectifier output with a capacitor).

But if the LPU must be stopped when the laser scans, this isn’t the case. I’ve never seen a Form 3 work other than in short clips in videos, so I don’t know whether the LPU moves continuously or discretely. The rasterization of the model would be a bit simpler if the LPU is static when the laser is scanning, but it would be pretty simple to correct for a < .01 degree rotation of each layer due to the < 25um offset from beginning to end of the laser scan.

I had been previously been thinking that the LPU stops to scan the laser, but after the above thought exersize, , unless there is some other reason to, I don’t think the LPU would stop to scan.

I’ve never seen a Form 3 work other than in short clips in videos, so I don’t know whether the LPU moves continuously or discretely.

Again, at the speeds involved, and steps/mm involved, you can’t tell if it does or not. It’s completely irrelevant. To the naked eye, it looks continuous.

unless there is some other reason to, I don’t think the LPU would stop to scan.

There is no reason to, no, the only reason would be to avoid having a very slight skew in the exposure. Since the laser can’t expose both edges (let’s call them Y+ and Y-) at the same time, as the X axis moves away, whichever side came first will be slightly skewed. Not in the sense that the printed object will be skewed, the skew would be very very very fine, and way smaller than distortions peel and other effects cause. But you’ll end up with some unevenness in the exposure. That can be fixed through modulating the exposure as the laser travels, though, and might even be completely negligible at the scanning vs. travel speeds involved.

Or, in other words, the LPU would have to travel WAY WAY faster for the galvo scanning to become a noticeable artefact.

1 Like

I haven’t seen one work either, in person, so let me ask a question. If you were to print a large circle, does it do it like an inkjet prnter wood, or like a plotter.

By that I mean, if it worked like a plotter it would start at the left, then the LPU would move to the right with the laser scribing the upper half of the arc, then when it reaches the apogee at the right it would return, while scribing the lower part of the arc.

If it worked like a printer, it would start on the left, and incrementally move to the right while drawing on both upper and lower halves of the arc. When it reaches the apogee to the right it’s done, and simply retracts.

Which way does it work?

Like a printer, to a point. A cathode ray tube would be a slightly better analogy.
But it’s like neither, as the galvo is free to make multiple passes “per line”.

If we simplify things to one Y pass per X “line”, it looks something like either the red or the blue function here:

Ignore the bit about interlaced fields and driving center vs. side pixel.
(image nicked from the excellent Karl Guttag’s blog here:

This is how laser scanning is done with non-vector, pixel displays - one axis scans continously, while the other one oscillates. What this results in is lower light exposure/density on one side than the other side, alternating between one and the other side. Regardless of the number of passes “per line”, on average, the middle of the “screen” or in the case of a printer the object, will receive the lowest exposure. You can counter that with exposure correction as the laser scans in the oscillating axis, either by changes in galvo speed, or changes in laser brightness, or perhaps both if you can modulate the laser fast enough [1]

Compare and contrast with Form 1, 1+ and 2, which functions more “as a plotter”, and laser is free to make multiple passes on both axes, in whichever order and way is necessary.

Slicing for Form 3 must be a veritable nightmare - it must essentially be almost like 2 onion layers of slicing, almost as if you slice once to transform 3D into 2D, then one more time to get 1D out of 2D.

[1] NOTE: there are other artifacts to it, like galvo overshoot, that also need to be ironed out by various bodges and hacks. Not that it’s important here, but AFAIK Formlabs compensates for overshoot, for example, by turning the laser off at just the right moment, and then turning it back on when the galvo has finished overshooting and gone back to “baseline”.


Actually, I think it’s simpler. It no longer works by creating a toolpath, rather it works like a line printer. Each layer is an image, which is painted 1 line at a time, with the galvo doing the line painting in the x axis (one or more passes), while the carriage moves across the “paper/vat surface” in the Y direction. So it’s no different than a laser or inkjet printer where the print driver breaks the image up in horizontal strips, whereas here, Preform breaks the layer image in vertical strips, each about the width of the laser spot.

1 Like

Yes, it does work like a line printer to a degree, but there’s nothing simple about the path produced.
It’s not a simple black and white map.
Different areas of the print need to receive different exposures, with different power.

Inside of a print is different than the perimeter.
Parts of perimeter that will end up as overhang need to receive even more exposure.
Any unevenness in the exposure due to effects mentioned in previous posts need to be corrected for.

On top of that, what you’ve described pretty much is slicing a 2D image into 1D strips.
The thing is, those 1D strips are also toolpaths, and they’re non-trivial, even though they’re just 1D.

PS: Keep in mind that a line printer prints bitmaps. It prints pixels (well, dots… the shape is unimportant, it’s the fact they’re discrete). You’re dealing with a galvo here - it’s no different than the galvos in Form 1 / 2, except it’s limited to 1 axis. This is why i’ve said it’s more similar to a cathode ray tube (as opposed to, say, an LCD screen).

The thing it’s really the most similar to is a scanning laser display (like Microvision).

Here’s a screenshot of my dashboard for two identical prints started at exactly the same time on the F2 and F3 in black resin.

The F3 is almost twice as slow which is really terrible…!

1 Like

Hmm. a little disappointing to hear this as the original billing boasted significantly reduced print times with better resolution. I have a form 2 and am experiencing tray binding (something is slipping internally sounds like gear noise??) so I will be contacting tech 3-16. I was just about to throw down and get a form 3, but fixing the form 2 is probably a better plan some of these issues are resolved.