Feature Request: Numerical positioning


Hey there! Quick feature request because I am seeing a lot of people taking about the camera positioning being off depending on where on the bed it is.

It would great to see some numerical x/y positioning implemented. Some of us are planning on doing precise work using jigs and a manual drag and drop style align really messes with that flow. Especially if that preview isnt exactly where it should be. As example. I have a bunch of dice I would like to engrave. Right now it would be impossible to line everything up correctly without numerical x/y input, and just say I was engraving the numbers on a d20 if even one engrave is off I’d need to toss the piece and start again.


It’s a feature many have asked for.

Apparently Glowforge went with a “Relative” positioning scheme based on the head/camera relationship.
They did not include a provision for limit switches, which almost all basic CNC machines use to home on.

I’m pretty sure the reason they didn’t think they needed limit switches, is because it would make camera alignment difficult, you would need to use one system or the other.

The option would have been nice at least.


FYI, here is an early thread on this topic: Feature request: Absolute positioning - CNC Mill style

It would be great to have. There’s one workaround while waiting, which is to make the file 20x12 inches, which will cause it to be positioned in the same place when loaded.


These are separate issues: I think the camera is supposed to “zero” at the same place every time, and I don’t doubt that it’s technologically credible to use cameras and image recognition to get the head to a very precise position. My objection is mainly that it’s slow and complicated, and prone to failure from lights and reflections, smudges, dirt, etc. And the lack of limit switches also means many of us have experienced the machine moving the wrong way and grinding against the end of the rail. It’s only a matter of time before that damages the motor or knocks something out of alignment.

Apart from that, it’s frustrating not to be able to move and position objects in the UI by typing in numbers. Even if the machine has a repeatable zero, unless you use the 20x12 trick (which isn’t always practical), you can’t position any given object in a repeatable place.


I’m sure they saved $5 on limit switches, by not installing them.


I’d be just as okay with the center of the bed being 0 and positioning based on that. Just say I wanted to position something in the top left corner I’d just input -x -y values


This one has been beat to death. They aren’t going to do it. Glowforge team thinks jigs and the 20x12 trick will get you what you need. Cut yourself a (0,0) corner using 20x12, hot glue it in place, and then make all your dice jigs off that position.

Good luck.


I wanted to be that blunt as well. But I’m glad I didn’t have to. :wink: I do believe those to be factual statements at this time. I know of no plans to ever implement numerical positioning. I’d classify it as “anything’s possible.”

I also wanted to say how I hope Support sees and closes this quickly before there’s yet another 2,183-post thread about it. :slight_smile:


Hey, the more people that chime in the more they’ll realize it is a good idea. Especially with all of the threads about misalignment that are currently open.


I think there are two separate issues.

The camera homing and lack of a fixed 0 in the bed is one. The other is that the GUI doesn’t show position, size and rotation of the object you are dragging. It should be easy to display these in fields that you can type into like virtually every 2D manipulation program. It isn’t hard to do, I have even written 2D UML diagram editors myself many years ago.

If it did that you would be able to position and scale items accurately relative to each other without laying them out in an external program.


That’s pretty obnoxious, Tom and Mark. Numerical positioning would be trivial to implement and make it easier for people to do precision work. Perhaps the reason there are multiple feature requests for it is because a lot of people could use the feature. If you’re not interested, why don’t you mute the thread and go away, instead of telling people off for asking?

Besides, Dan specifically described the 12x20 trick as a temporary workaround, and said they “have a lot of work in the hopper to help with positioning”.


I know it’s a kluge, but couldn’t a user semi-permanently mount a block with a target on it and have GF team add a software selectable homing to that fixed point using the camera? The user would calibrate to the target when needed.

Yes. $5 limit switches would have been smart even as a backup.


What’s that smell? OH! It’s fresh irony.

That doesn’t change the likelihood that it’s never going to happen.

LOL! Nah.


Hey, I know you have your own opinions on the matter and I respect that, but if you’re going to respond with outward hostility towards the subject can you please keep it to yourself? I started this thread to bring a feature request to the GF staff and didn’t really want to see the same conflict I see in other areas of the forum. By all means read and laugh but it seems like you and @chris1 can keep the argument to PMs


So you can’t simultaneously say you “respect my opinion” and then tell me to “keep it to myself.”

That said… It’s just that… well… it’s DONE. So I wasn’t sure if you knew that. So I agreed (and continue to agree) with @markwkruse. There’s been hundreds and hundreds of messages about the topic. Including with Glowforge staff. Hell, I even started one of those topics because I want it too. But it’s never been met with any hope that it will ever exist ever ever ever.


Glowforge is commercialized, the design of hardware is complete. You can wish there were switches, but it doesn’t change the facts on the ground.

  1. It’s 100% clear that they are not shipping a physical (0,0) point with each machine. No physical edges will be shipped. No mention of them working on it in any way, in any updates. It is clearly not a priority for them now. Factory fabrication optimization is clearly (and should be) the priority. If you need accuracy, you will need a repeatable physical location in the machine that can be achieved by 20x12 trick. Software numeric positioning won’t change the FACT that you will need to place YOUR OWN physical stop in your machine. You will know where that is by cutting a corner in some cardboard or material of your choosing utilizing the 20x12. Accurate positioning of your design can be placed utilizing the program of your choosing inside the 20x12 template. Software update may occur in the future, but without a physical zero in the machine it really doesn’t do you a lot of good.
  2. There are no limit switches in the hardware. There has been extensive discussion from multiple Glowforge posts on the accuracy and repeatability of the homing process using the “G” image. It does fail sometimes, but when it does work, the location is correct within sub-millimeter accuracy. The speed of homing, errors from dirt, errors from lights, errors in running into the end stops is unfortunately a software upgradable fix that they will continue to develop.
  3. When they eventually deliver on the promise of machines for all, they may have time to work on the software some more. The latest update described where their direction for locational accuracy is headed. They have extensively mapped every machine delivered, and will push out some kind of update that will magically make every machine accurate. Maybe at that time they will implement edge scales, rotation, edge snap to grid, etc. All of this visual positioning magic, edge scaling, snapping and the like, but you will STILL be required you to place the object in the machine. When you place it in there, the height will distort the visual position, and you will still need to utilize the 20x12 trick.

I hope they implement numeric positioning. It was discussed in 20 threads for over 2 years now with thousands of replies. Keep posting it? Sure why not? How many “hopper” items are there and how many have been implemented? Where is the hopper list? How is it prioritized? We are left with this forum, closing our eyes and wishing really hard. Unfortunately the track record for communication clarity on such things is limited to periodic updates, and only what they choose to reveal.

This one feels like a dead horse.


Unfortunately, we’ve had this discussion many times before, and usually it results in a lot of wasted time and a lot of bad feeling. I believe Tom and Mark are just trying to forestall that happening again.

I’m going to preface all of this by stating that this is my understanding of what’s going on. I might be off base somewhere, but I have no desire to get deeper into the weeds on anything or continue discussions about it, or argue whether it’s a good idea to add something or not. Already been through that.

The time to have this discussion was back while they were creating the software for the machines, about a year and a half ago, and we did. Several times. Also unfortunately, it’s not a trivial thing to implement, because their rendering software, (not the interface), is using a relative location method to create the code for the head movement.

They made the decision not to go with limit switches on this machine long before they ever built it. I know that because of an explanation Dan gave to a potential customer at a talk he gave once. (And I can’t even begin to remember which one, so i can’t find it again now.) In it, he described the potential for limit switches to fail rapidly with repeated banging of the head into them, requiring replacement and downtime that were unacceptable, expensive and requiring people have to take the machines apart and reassemble them correctly. (Which might have also caused the need for realigning the beam path on the laser, which is apparently a very precise thing.) In other words, they wanted to make the machine easy for people to use without having to be able to take the machine apart and put it back together. Or ship it back for repair.

So…no limit switch. There’s a reason for it, and now everyone knows what it is. I’m pretty sure they aren’t going to add one, no matter how many people think it’s a good idea.

Without a limit switch, the camera uses the Glowforge bug on the laser head to orient itself. It takes pictures, squares itself up, and then takes a few pictures of the bed and positions the design in relation to that. The problems with misplacement of designs on objects on the bed has to do with the curvature of the lens necessary to see a bed that large…the farther out you get from the camera in the center of the lid, the more fisheye distortion is introduced. They have an algorithm that compensates for that based on the metrics that they have captured for each machine. These early units might not have completely captured metrics yet so the compensation for the fisheye effect might not yet be implemented. When they ask you if you understood that the placement on objects on the bed could be off by up to 1/4", this is what they meant.

They can go back and adjust that remotely to a certain extent using their software, if they have the metrics for your machine. Dan has said in a recent status update that improvements to that process will start going out with machines being built going forward soon, and then applied retroactively to those machines that are currently already out there. So the 1/4" offset is going to improve for everyone, including the machines already out. (For those of you who haven’t received your units yet, if you don’t want to have to mess with it, delay receipt of your unit until they start rolling out with the improved metrics tracking. They give us a choice about receiving it now or waiting, although we don’t know when that will be and they’re probably not going to prognosticate.)

My pre-release unit has been in operation for seven months and it’s visual object placement is accurate to within a fraction of a mm. It didn’t start out that way, it started out being off by about a quarter of an inch, and during the times that they make updates to their firmware, it can get out of whack again. Unfortunately other mechanical issues can apparently also cause that to be off. The only ones who can determine what is causing the problem is Support. They have access to the metrics for individual machines and get feedback on what is causing problems that we don’t see here. If trying to reboot the machine doesn’t fix an alignment problem for you, which is about all that we as customers can recommend that you try, then you need to wait to hear back from Support.

Back to the original issue at hand - that of how to align objects accurately…you can do that very easily in outside design software (Illustrator, Inkscape, AD, CD) and save that as an SVG file that captures all of the information in perfectly aligned, perfectly scaled, perfect engraves and cuts. The secret to using it is relative addressing. The rendering software looks for one object, either a cut or an engrave in the file and then it moves the head relative to that object in order to perfectly place the other design elements. The advantage to doing it this way is that you can drag the artwork anywhere on your material to use up the leftovers without having to cut it up into a specific size to place the piece of wood up against an endstop where the artwork needs to happen. In addition, they have added a workaround for those who want to do numerical placement. If you make your artboard 12" x 20" in your design software, then where you place your artwork on the artboard is exactly where the item appears on your screen when you save the SVG file. Scaling is correct, and relationships between elements are maintained.

The use of jigs is a technical version of just adding a weeding box around your artwork in the SVG file. You can engrave one side of a double sided design, cut the weeding box, flip it over, engrave the other side of the object, and finally cut out the shape inside the weeding box. It’s a perfectly precise method. The placement of the weeding box around the artwork acts as the anchor here, you can place the artwork anywhere on the artboard, or move it anywhere and get perfectly accurate results with the build. I prefer to use jigs (ie: just add a weeding box to the design) but I only need to make one or two at a time. If you are engraving on thousands of precut items and you need repeat ability, you can use the artboard sized correctly for numerical placement.

A lot of people want answers to why Glowforge doesn’t jump on their wonderful suggestions and just make it happen. This is why. If you managed to make it through all of that, you now have a pretty good idea of why I think it ain’t gonna happen any time soon. You also might have a better idea of how to proceed to get excellent results using the tool you have at hand. Just use outside design software to work with the tool’s parameters. There are reasons it was created this way. Once you get a little comfortable with it, you’ll get an idea of the true power of this machine. It’s astounding.

Changes down the road are likely a long way out. They will capture the interest now, but I don’t expect to see anything major implemented any time soon. We should still give the suggestions so they can gauge interest, but keep in mind that it’s for the future.

TLDR : Quote “Thanks for the suggestion, we’ll let the team know.”


Might be a good idea not to make it sound like you work for GF.


ROFL! I’m quoting them. Does everything have to be made so difficult?

I’ll put quote marks on it or something… :smile:


Jules, I’m sorry, but you’re simply wrong. If the machine is capable of cutting in the same spot when you load a 20x12 SVG, it is capable of allowing the user to enter “1.0,2.175” in the UI and cut in the same spot.

Even besides that, just aligning things relative to each other would be a lot easier if we could type in their X and Y positions. I have little doubt that these features will be added eventually, because it’s not nearly as complicated as you imply, and it has nothing to do with whether or not there are limit switches. That’s why I tried to make it clear in my first response that these are two separate things. Of course my machine is not going to grow limit switches, that’s hardware that isn’t present and will never be in this model. Software changes, on the other hand… it’s a lot simpler to add what’s been suggested to the UI than to do all of images processing and convolution mapping to compensate for individual machines’ camera calibration.

I stopped responding here, because certain people need to have the last word, but it’s important to correct this misinformation that comes with a false air of authority.