Laser cut flexure based microscope


I am currently looking into the possibility to create a laser cut microscope. To be totally honest this kinda a solution looking for a problem. I am the author of a box generator for laser cutting and want to get into flexures a bit more. While there is some hype around flexures there are not that much usable examples out there in the Open Source space other than the OFM.

I am currently toying with double parallelogram flexures which look like a good building block for an xy or z stage. I at first played with decoupled x and y axis but I decided to go with putting the axis in series for now - which seems to the the case for many metal microscopes, too.

So I will start with a simple z axis and probably add something like the flat top microscope around it. Probably with an option for mounting a cheap xy slide holder from Aliexpress. I am also thinking about making this into a z axis for the PUMA microscope to get a purely optical variant.

After that a xy stage is the next step - then a combination of both. The idea is to not reinvent the wheel more than necessary and re-use optic modules and software from the Openflexure Microscope and may be PUMA.

Another thing that I want to look into in the future is backlash-free actuation. I am still missing a good idea how to attach the 28BYJ-48 without producing backlash. Directly driving the lead screws with NEMA steppers should work but is much more expensive and bulky.
For manually actuation I want to try a double lever system with a coarse and a file lever who’s movement gets added together. But I have no idea yet if this is even feasible. For now I am happy if the stages move at all.

Has anyone looked into laser cut stages already? Have double parallelogram flexures been considered for the OFM and if so what where the reasons to not use them? Or are the discussions on the OFM layout available online anywhere? This forum seems to be too young for that. The published papers do describe the final result very well but don’t give much information on alternative solutions.


1 Like

Hi @ffesti. I don’t have any answer to most of your questions, but the reasons for not using double parallelogram flexures is I think because they are more complicated and have not been found to be essential. With the single parallelogram flexures there is some parasitic movement in the orthogonal direction, but this is very small in the case of Openflexure systems. That is partly because the 3D printed flexure hinges that we use have a practical limit of around 6 degrees of deflection. I have used double parallelogram 3D printed flexures in other systems requiring more motion, but not very successfully.

This is a non-engineer talking here. :wink: I think it could be possible to break all the individual components (legs, stage, optics, illumination, etc) and laser cut them. The challenge is how to you make a hinge to connect them without loosing the structural rigidity of the monolithic build.
Do I make any sense at all? :stuck_out_tongue_closed_eyes:

That was not quite what I had in mind but that should work pretty well actually. I am more interested in cutting the flexures from the same sheet as the rigid parts. But let’s set that aside for a moment:

Looking at figure 5 b in A one-piece 3D printed flexure translation stage for open-source microscopy it is very apparent that all four moving parts at each “corner” can be joint together by a single sheet of bendable material. The horizontal pieces could be kept together with tabs for assembly that are broken off afterwards. Alternatively or additionally holes for locating pins could be cut. This won’t work for the “legs” though as they need to be cut “from the side” so we can’t drill a hole at the top or bottom with a 2D manufacturing process. But we could just add a finger that fits in a hole of flexible sheet. Yes, this cuts into the hinge in the middle but that should not be a big deal.

For assembly one could either just glue things together or add a second layer to clamp the flexure material in place.

This would allow to make the stage quite a bit bigger. May be big enough to fit the actuator to the inside which would prevent the overall size growing, too much. Otoh I am not sure if that is really practical from an assembly and maintenance POV.

I really like your idea but I’m having trouble seeing your concept. I wondered what would it take to build the “OFM max pro plus”. In any event, wouldn’t bending metal be more likely to break than bending plastic? If I understand this correctly, it will be highly unstable.

Well, the flexures could still be made from a polymer sheet. POM aka Delerin comes to mind. It is also very easy to laser cut but is said to be close to impossible to be glued. So it needs to be clamped. I may still have some PET or PC sheets around. May be super glue sticks to those.

As the stage itself is only made of a few, very simple parts I may be able to come up with a working prototype this weekend.

1 Like

Here the first prototype. Flexures are 0.5mm PC the rest is 3mm birch ply. Lessons so far:

  • Stiffness difference between the birch ply and the PC is not big enough
  • Those flexures are too wide and thus too stiff. Two small bridges at one either end is enough
  • Glueing this kinda stuff with super glue is a pain I’ll try a variant with clamping next
  • This is just 5 minutes of machine time. This makes creating microscopes in the dozens a realistic goal

Those dots are 1mm alignment holes. I used small nails to keep things in place during the glue-up.


That looks nice. Polycarbonate I would think is too stiff. 0.5mm polypropylene that we use for the diffuser in the microscope is nice and flexible, although even then it might be better to have the hinge as a couple of narrower strips as you suggest. 8 hinges need to bend for the microscope to translate along one axis.

For me this is an interesting investigation because one of the underlying assumptions of the Openflexure development is that minimising assembly time is an overriding issue. This leads to having single pieces as much as possible, which is enabled by 3D printing. You are challenging that assumption with a piece by piece design that has many features to make it easy to assemble.

1 Like

Yes, 3D printing and laser cutting are very different processes requiring different approaches with very different pros and cons. From an OFM POV it would make sense to replace the stand / base and electronics drawer by laser cut boxes first rather than messing with the delicate stages. Those are huge but don’t offer much functionality. Yes, they wouldn’t look as good but they could be made with much thicker wall. It would also allow adding a cover / lid. Which would just be prohibitively long to print.

But that is another project for another time. I am already far of my original course.

1 Like

Also I am not sure if PC being stiff is actually an issue. Using thinner material, narrower or longer blades is always an option. While I have not completely gone thorough the maths I think the important thing is that elongation at yield is not reached with the geometry. So a stiffer material with better elongation may work just fine.

1 Like

The 0.5mm polypropylene does not laser cut very well (at least in my experience cutting the diffusers), I would avoid it for anything requiring precise cuts wherever possible.

Very impressive. That is kind of what I pictured in my head. Perhaps a combination of laser cut and 3D printed parts? TPU is almost indestructible and very flexible.

OK, I tried making the clamping thing work. But I am not happy with the result. May be the flexure blades need to go into a slot instead of being glued or held on top of a piece. I will try to turn them vertical and basically stick them into top / bottom of the pieces. The vertical pieces can be replaced by I beams if I need a flat surface at the top or bottom.

Turns out my flexure material isn’t PC but PET-G. Unfortunately it is pretty difficult to find reliable data on the maximum elongation data. Values found vary by a large factor. Young’s modulus otoh is in the same ball park for all the polymers - 2000 to 2800, most being around 2200 to 2400 MPa. So I don’t expect the material making too much of an impact on the stiffness. So PETG should not be a bad choice.

The thinner flexure are ofc much less stiff. Reducing the thickness from 0.72 to 0.5mm should already reduce the stiffness by a factor of ~2.75.

If all fails I at least have now a better idea how this all works and may be able to cut the flexures from the same piece as the stiff parts.

1 Like

Ok, here prototype 2:

All flexures now vertical. Note that it is assembled wrong. The stage and base plate need to be on the same level as the top / bottom of the vertical legs. Otherwise the mechanism binds up as the different hinges try to move the stage different amounts. I knew this in advance but this version lacks clearance to put together properly.

Lessons learned:

  • These single leaves are very small, may be I need to make them bigger without widening the actual flexing parts
  • The thin super glue I use does not mix well with the wood as it gets wicked away inside instead of filling the gaps
  • You can’t be too clever with these tiny parts as the heat from the cut will warp them otherwise
  • Being too stupid makes for a fiddly assembly, though

These small flexures bend pretty nicely when the mechanism is not binding. My impression is that 20 to 30° is possible. Things will have a long of cosine error by that. Will try with one that actually works…


It just occurred to me that this kind of construction is even better suited for a delta. Here we need hinges with one DOF to separate the different motions. A delta can use ball joints with 2 or 3 DOF. So instead of a blade flexure I could just use a wire flexure or a notched blade. Those would allow bending in two directions. So the delta arms can directly be connected to the stage and the driving arms. That would reduce the amounts of flexures by quite a lot and make assembly a lot easier. The delta also doesn’t need a separate Z axis saving even more pieces.

1 Like

Oh yes, delta would be just three legs not 4+z-axis. 2D wire flexures are used in some of the commercial metal translators that I use. It will be interesting to see how they work here.

Edit: The metal ones with wires are the Thorlabs ManoMax series. I had to take one apart and was surprised to see wires not the plate flexures that are in their other multi-axis stages. I don’t know whether the wires are actually used for 2-axis bending, I did not examine the mechanism in much detail.

Can you post a picture of those here or a link to the company page?

Here my first try of a flexure based xy stage I did quite a while ago already:

Issue here is that the material is too thin. The blade flexures need to be much higher than wide. That’s why these are typically cut out of thick metal with an water jet or EDM. With a ration of just 1:2 width to height the stiffnness in plane is totally insufficient while the axis are still pretty stiff. This might be able to be fixed using acrylic which can be cut even in pretty thick sheets (20mm+). But that pushes the limits of the CO2 laser cutter. But the whole assembly is pretty bulky in my eye anyway even though is is seen as a rather compact flexure in the field.

Thinking about supporting the intermediate stage from below I realized I can make a compact one DOF stage by putting 3 flexures on 3 sides of a cube and that I might be able to nest those inside each other.

Here is the first try for a simple one DOF stage:

Displacement is 10mm to each side 20mm in total:

Kerf is a bit too large here. I am too used to put my boxes together with a hammer. These are way too fragile for that and I instantly broke some of the flexure blades sigh

Otherwise the flexure perform very nicely. The stage is really stiff in all directions except x. At the end of travel some force is required but nothing excessive. The nominal size is 100x100x100mm but that’s on the inside of the cube so the size is more like 11.2cm + 1cm left and right for the table extending. The later ofc is optional. The table can easily be made smaller to stay inside the overall outline.

Now I need to figure out how to fit the second stage. There are basically two options:

  1. Flip the x stage and the horizontal flexure and add the y stage upside down and 90° rotated. That would keep the x stage at the bottom and the x flexures at the same size.

  2. Make the horizontal flexure shorter so there is a gap left and right that allows the y stage flexures to poke through while traveling left and right. So we need more than 2 time the x travel. This can either be done by shortening the horizontal flexure or by elongation the stage in x direction. Ofc the sizes are parametric so this isn’t a choice in the design we need to make now.

I guess I will try option 2 first as it will use less material and I like having a large opening at the bottom to fit optics or condensers. Especially as we still need a z stage somewhere.

It turns out that for that two flexures in a 90° arrangement may be good enough. This is pretty stiff already. Nominally it lacks stabilization for rotation around the axis of movement. This is an issue for a microscope stage but an optics mount should be fine with that. There is nothing to really put stress in that direction.

I currently don’t have a actuator fitted but I guess just having a screw and nut should work fine in conjunction with a spring or O-ring for pretension. I am actually puzzled by the large reduction on the stage of the OFM given that most metal microscopes have a rack and pinion drive on the stage without any reduction at all.

On a different note I also quickly put together a delta stage. It uses 1.75mm PET-G 3D printer filament as flexures with 4mm length (of the gap):

The flexures are not ideal but allow something similar to the 6° of the original. Just sticking the filament into holes is also less than ideal. If this goes anywhere I need a smarter clamping mechanism. May be an arm that can be bend and has a latch/hook at the end. It’s also worth looking at other flexure materials. Thin POM wire seems to be hard to come by. May be 1mm nylon line works.


I enjoy how you’re recording your results and appreciate that you’re sharing them with us. I built something similar, first with PLA, then with wood, using a 40W laser cutter I received for Father’s Day. It just didn’t work for me. Most likely because I lack an engineering expertise and my designs are quite basic.
I found several publications throughout my investigation. This is the one that inspired me the most. I used Google Sketchup to create the model, which took some time. It did not move the claimed 50mm. It’s only 5mm before it snaps. Perhaps you can understand it better.

Design and characterization of a two-axis, flexure-based nanopositioning stage with 50 mm travel and reduced higher order modes - ScienceDirect


This looks like it is basically the same flexure as I showed above. Only the segments at the corners are mirrored to the other side to make things more symmetric. So I don’t see a reason why this wouldn’t work.

Ofc there are quite a few possible pitfalls. The paper uses 0.5mm thin aluminium blades. In polymers they have to be a bit thicker to achieve similar characteristics. Cutting this kinda of thing from plywood has it’s own challenges. Wood has basically zero strength across the grain and cheaper plywood has knots and defects hidden in the inner layers. Hitting just one of those in a section that has cross grain on the outside will often doom delicate pieces like flexures. My flexure cum has the benefit that I can align all the blades with the outside wood layers. Also I use pretty good plywood. Several plastics also don’t take the heat from the laser too well. Normally that won’t make a huge difference. But if the part is very thin there may be no unaltered material left. I have seen that with my PET-G blades which are also borderline too tiny - even if they still worked.

I might cut a final model out of POM which has a pretty high elongation. With that 50mm travel might actually not be that far out of reach. it probably still needs a cube with 20 or 25cm of size.

1 Like