Resin gauge on the touch screen display, building time inaccuracy, "deleting" one part during multiple part building cycle, Part Failure sensor

  1. Great would be to have a resin gauge incorporated into the touch screen. A simple visual that shows the amount of resin left.
    Especially when making many different parts one can quickly lose count of the amount of resin used.

  2. This might have been mentioned before, I am not sure but the actual time a part needs for building differs significantly from what the software calculates.The actual time is always more than suggested.

Here are some numbers from current projects:

Claimed 1 hour 5 minutes
Reality 1 hour 22 minutes, 17 minutes longer

Claimed 53 minutes
Reality 1 hour 5 minutes, 12 minutes longer

Claimed 13 hours 13 minutes
Reality 15 hours 41 minutes, 2 hours 28 minutes longer

Claimed 10 hours 52 minutes
Reality 13 hours 23 minutes, 2 hours 31 minutes longer

Claimed 7 hours 10 minutes
Reality 8 hours 50 minutes, 1 hour 40 minutes longer

I don’t know how the time is being calculated but I think there would be a more accurate possibility.

  1. Pausing a building process. While it’s a necessary feature (I have recovered some models that way, using clear resin) the parts always show the layer where the building process was paused - I have never had a smooth and even continuous build after pausing the machine. Not sure why that would be as the building platform moves up stops and comes back down all the time, it’s what the machine does…

  2. I typically load up the building platform anytime I can. Sometimes one part out of a set of multiple parts starts failing and when the machine is paused the resin stuck on the PDMS layer (most common cause of failure on an otherwise proper running machine) can be removed, not always will the machine recover that part but all the other parts on that run would still come out fine. As it is currently, the laser hammers the silicone layer at that area into submission until the rest is built.
    In order to safe the tank surface it’s better to stop the building process entirely - wasting the rest of parts (and time) that otherwise would have come out well.
    If there was an option that allows one to select different parts on the building platform in pause mode and basically “delete” them so this/these part(s) is/are taken out of the laser path, it might be a good extra. I realize that this feature might not have to be used too often but for sure would help in critical situations. Some people might realize they have mixed older with newer parts half way through the building process and could “delete” the wrong parts while the good ones would continue building.
    This saves time and material.

  3. This one I believe some have mentioned before, a system that senses part failure. That actually would be the most important new feature. There is nothing worse than having a part failure and the machine keeps going, wasting everything in its path…the machine is not bullet-proof even though it’s promoted to be, a feature like a “part failure stop” would help tremendously. With this feature it might also be possible to safe the build, since the machine would be put on pause.

On a different note - the firmware update did fix the sensor error.
one image attached - simple idea of a resin gauge incorporated

Well, your times show about an 18-23% underestimate. That suggests that whatever the source of the error, it’s fairly deterministic and should be something that can be accounted for. In the meantime, if you just add 20% to the time estimate you’re going to be pretty close.

Item 3 - layer lines result from misregistration of the print from one layer to the next. When you pause the print and touch either the build platform or resin tank, you’re probably changing the alignment. Pausing to add some resin in Open mode is about all that you can do and not move anything. Prying some resin off the bottom of the tank, or popping one failed object off the build platform will definitely cause misregistration. This happens because there’s a little play in all the mechanical systems. if the mechanisms were all designed with no play (or “backlash”) they would wear out too quickly.

Item 4 - a good idea I think would be very challenging to implement. It would have to be done from the touch screen of the printer. I’m not sure it’ll be the best interface for doing something like this…

Item 5 - how would the printer do this? Optically? You’d need a camera and the software to figure it out. There’d be no way to sense this directly that I can think of.

Let me quickly respond to Item 3) first. Pulling out the building platform or the tank in Pause mode of course will mess with the registration…I am well aware of that fact, as well as prying off parts while the building platform is still in the machine - that’s not even a question. Or ripping off stuck resin…it’s a precision machine and needs to be handled that way, like any precision machine.
Whenever I have to peel off resin that’s stuck on the tank, I can say with absolute certainty that the tank is not moved whatsoever, I can only speak for myself here not for others.
Different scenario - I had 2 “missing cartridge” errors - without any apparent reason - the machine stayed in pause mode for about 3 hours. When I realized it had stopped - I never even opened the cover, I carefully lifted the cartridge then put it back into place, the error went away and the machine continued - 2 times that happened and both times the machine failed to adhere the continuing layers.
But did finish the parts, one ended up being to “messed” up the other I could salvage.
Whenever I pause the machine, open the cover for visually inspection, then close it again - it doesn’t matter I always have a slight layer shift, which the machine still does off and on, even though layer shifts are occurring less, compared to the Form 1+, but it’s always visible where the machine had paused. Could be my machine (but I had the same problem with all my form 1+ machines as well - 3 altogether).

The machine is capable of recovering parts when caught early (as you of course are aware of)- if stuck resin is caught - carefully - peeled off and has a thickness of 300-350 microns it’s safe to say the part is recoverable. This also depends on geometry and where the failure happened.

Item 2) accurate building time - how is the building time calculated? - it seems to be a straight forward process, we know how long different laser cycles (exposure schedules) are, we know the different sizes of cross sections of a model and how those change in measurement as the part is being built, we know the technical specs of the stepper motors, z-axis spindle, belt specs, therefore wiper speed and cycle, z-spindle speed etc, we know how long the machine takes to reset - therefore should be able to get a quite accurate time - it’s part of a professional machine, giving correct numbers.
No-one would want a car that stops (out of fuel) with the fuel gauge showing 20% of fuel left - checking the manual it states just deduct 20% of what your fuel gauge is showing and you’ll have a better idea of how much fuel is left in the tank.
Don’t quite understand why it can’t be an accurate number or at least far more accurate than it is right now.

ITEM 4) It seems the key here is to have all the parts on the building platform being separate parts, not one big stl file containing all parts. Each part is separately loaded into preform. This way each part can be manipulated separately and the software knows which part is which (preform). I don’t know if preform calculates the laser path already or if the Machine’s interface software does this calculation. If preform calculates the path, there could be an added “delete part mode”, which can be activated on the machine’s touch screen, pausing the machine. Preform then will open, either automatically or by hand and the model will reappear in preform showing the “delete part mode window”. There one can delete the specific part(s) and resend the file back to the machine, like in pause mode it remembers the last finished layer, this time the laser path has been recalculated eliminating the deleted part and continuing the other parts with the next layer.
If the laser path is calculated within the machine’s interface, a screen window would have to be added showing the parts, in a table type arrangement, let’s say we have 10 parts on the building platform, there could be 2 columns with each showing 5 parts, which then can be selected via touch-screen and deleted. when resuming the building process, the interface software recalculates the laser path and continues without the deleted part(s).

Item 5) Yes that is a tricky one. Making it in the way of the machine, simple but effective is going to be tough. Optically - we have different types of resins, the surface of the tank will have to be “scanned”. Using a camera that would utilize a software based on image recognition - or some type of scanner, would not only be a fairly complex and expensive set up but it also would extend building time, in-between every laser cycle this “scanning” process would have to be activated. Further we have clear and opaque resins - using clear resin one can see with the naked eye if there is some resin stuck on the PDMS layer. Opaque resins one can not tell until the wiper goes across the tank and exposes a potential problem.
Which brings me to the wiper itself. The wiper goes across the PDMS layer frequently and is in constant contact with the silicone surface. There might be a way of adding a sensing mechanism to the wiper’s connection arm. if the wiper experiences an obstacle it’s otherwise continuous motion is being altered, the sensor picks up on that and sends a signal to the machine to pause. It could be a mechanical sensor, maybe a magnetic sensor. The wiper arm is spring loaded (like the tank tray) and is dialed in for the surface friction and resin resistance. it will shift slightly whenever there is an obstacle on the tank surface, this shift can be picked up by possibly a tiny magnet sensor. I am not sure how fine this can be tuned and if that’s even a possibility, but It seems like it’s feasible. This system might not be sensitive enough, maybe the wiper has to be redesigned completely, not just partially.
The idea is to keep as much of the machine as it is right now…a complex second optical system doesn’t make sense.

  1. I believe this is mostly due to the resin sensing and filling operations (assuming this is a Form 2 in regular mode). That is what FormLabs has told me in the past when I brought up this very issue. It is rather significant. I believe they have enough data now from all the thousands of prints that have been done to where they should be able to correct their estimations. As it stands, i just do what Randy mentioned and add a buffer for planning purposes. It is pretty consistent.

  2. I have seen this too. It doesn’t seem like a x/y shift to me but more of a denser area which leads me to believe it could be material that has cured differently since it was not immediately exposed to the next layer.

  3. I would probably use this once in a blue moon.

Accurate build time.

If you print the same form file on different occasions you will get different build times. This I’m guessing has to do with environmental factors like, for instance, the ambient temperature affecting the heating cycle and the resin flow rate. I suspect there are a number of factors like this that are not allowed for in the build time calculations and it’s these that are causing the discrepancies.

Yes. But if the prints are always “long” by about 20%, FL could just add that as a constant to the calculation and the print time estimates would be better than they are now… with the current implementation, prints always go longer than estimated. With a 20% “adder” on the estimates, some prints will end up finishing sooner than predicted, but the overall sigma on the statistical variation from “reality” should be a lot less.

The guys from Formlabs did a nice job in, first, fixing the resin sensor error and second, speeding up the heating process (tremendously) in the latest firmware. If you have not downloaded it, I’d highly recommend it.

Once the operating temperature is reached resin flow is consistent - one of the advantages having controlled resin temperature.
The only true unknown is, as you mentioned, billb, the environmental temperature. Even though I suspect most people operate their machines in air conditioned environments, narrowing that temperature span, therefore also the difference in time until operating temperature is reached.
And yes, Randy, I agree 100% it should be added if this is a known 20% error, even though much more interesting would be why that is. A few minutes - I can see - but hours?
Essentially that is not too critical, as long as the machine keeps kicking out good parts consistently, which mine has stopped doing lately (after about 400 hours running time), I can live with the time differences as well, even though it would be nice having a reliable number there, after all the machine is being promoted to be the best in its field…

Interestingly, I pulled my F1+ out of mothballs to evaluate a 3rd party resin. It’s running a print now. I’m 63 minutes in to the print on layer 221 of 516, which calculates to a projected completion time of 516/221*63=147 minutes. Print time varies as a function of layer complexity, I know, so it’s possible layers yet to be printed will take much longer than the ones already printed. Except my parts are fairly uniform density in the Z axis, the complexity at any one layer is not very different than any other.

PreForm says this print will take 239 minutes. So in the case of my F1+ PreForm’s estimate appears to be overly pessimistic. The opposite of the F2 estimates.

wow you still got your F1+, I got rid of my machines as soon as the f2 came along, I had begun to really dislike the F1+, not to use the word hate.
But that’s great you’re doing the time test!
Quite interesting that the old machine would rather give one a longer building time - would make more sense from a Marketing point of view, either way an accurate number is what a professional machine should be able to put out.
By now your F1+ should be finished. Let me know know it went.

The F1+ finished the job about when I estimated it would. Around 1.5 hours less than predicted by PreForm.

Thanks for posting and testing!
Interesting how the 2 machines seem to have a different procedure on how to determine the building time, getting opposing results

I suspect the machines have very similar procedures. The difference is that the F2 has more variables like heating time and refilling time and wiping time. But I did not expect the F1’s estimate to come out higher, I expected it to be closer to correct.

even with more variables they are all still set within the operating cycles. So they are just added on. Which brings us back to the fundamental question why is the time off.
Interesting is also that none of the formlabs guys have chimed in. If it was my product I’d make sure that it’s understood why things are the way they are.

I’m sure FL is watching, and equally certain they have nothing to add or they would have added something by now… though I bet they’re looking at it and will say something when they have something to say. :slight_smile:

IMO, it really isn’t super critical that this be accurate. I put it in the “annoying” category. I just wait until the print is well past the base layers and do print_time/layers_printed and then multiply by remaining_layers and this results in a fairly accurate estimate of the remaining print time.

But I could see how someone running the printer for commercial purposes would care about this more than me. If you’re charging for your services, you want an accurate estimate before you’ve started printing the job, not partway through printing it.

yeah, I said the same thing before not really super critical but it’s the principle of it - I build a machine and claim all kinds of things but can’t even keep correct building time…not particularly good self-promotion. I have not checked the accuracy of resin use. I just assume it is correct. That’s why it would be good having a visual on the touch screen in form of a resin gauge, that would help keeping resin in stock as needed and not suddenly run out in the middle of a project.
On the other hand it’s much easier to calculate the volume of a part, which would be the use of resin + the volume of the supports, if you want to be really accurate you’d have to add the wasted resin that’s coming off when cleaning the parts and building platform, but now we’re getting tricky again - some might call this hair splitting, others might call it accuracy…
If I was to build machines I would want my machines to be right on - simple as that.
Anyone who aims to build the best product in its field should have that mind set.