I am aware that there is already an ongoing thread here on a fully laser-cut microscope. What I have in mind might be a simpler task – I am looking to prototype a compact 1-axis flexure for fine focusing a small objective lens in a DIY scope. Has anyone tried laser-cut acrylic, say, in 12 mm thickness, for making flexures? There are 3D printed examples with 1:10…1:15 reduction (see a Youtube link below), but I wonder if laser-cut acrylic would be more precise and would have less parasitic motion (as an isotropic material).
I am thinking to try something like this, and use it with a fine-pitched screw. I need sub-micron steps in focusing, and the range is less important (0.5-1mm is sufficient). Thinking to try a 508 TPI screw
1 Like
I have not tried this (yet).
Here a few thoughts:
Acrylic is more brittle than usual 3D printing material - especially nylon. You will need longer or thinner beams to get the same range motion. The one shown here has particularly short beams.
The way to little parasitic movement is symmetry. Note that OFM does not do anything to reduce parasitic motion at all (including not having symmetry and not having double parallelograms) and gets away with it (by accepting the drawbacks). May be this is less of an issue than you think.
Generally the ratio in stiffness inside and outside of plane is the ratio of the beam height/width squared. So if you manage to create 1.5 mm wide beams in 12mm acrylic the flexure is 8^2 = 64 time stiffer out of plane than the desired motion. Not great, not terrible.
I solved this issue by putting two flexures at 90° using the much higher stiffness along the beams.
I’ll try to get some pictures the next days.
My gut feeling says a 508 TPI screw is not a great fit if you are using it to move around all the time. This is basically the smallest possible thread and it will wear out quickly. If you use a flexure anyway there is no reason not to use a lever reduction. Even a simple 5:1 lever will make your thread a lot stronger.
1 Like
Thank you for your thought-provoking note!
I do not have proper physics and an engineering training, and sourcing books on flexure design has been difficult (out of print, no PDF for purchase, etc). Thus, I was hoping to quickly try several materials to make cheap parts (3D printing, laser cutting in acrylic and perhaps EDM in aluminum) to get a “feel” how different materials behave.
The flexure I referred to (see image below) does appear symmetrical in the output piece – it has two connecting beams on both sides, and the output is located centrally. I was going to attach my focusing lens to the output part. The “input” levers in the part are indeed non-symmetrical, but are also not over-constrained and allow for a significant 1:15…1:20 reduction.
I agree that acrylic is more brittle than 3D printing materials, but the original material of the flexure I am testing was aluminum [made with wireEDM or water jet]. Since Al is stiffer than acrylic (at least judging from the flexibility of 10x10mm bars in acrylic and aluminum I have in my shop), I figured that Acrylic might just work. I may be wrong, of course.
A less brittle isotropic material would be polycarbonate. But PC is not laser-cuttable [at least by companies that I use. I am not sure if they can CNC 1mm-wide gaps in thick polycarb for this flexure.
I hear you about the extra fine threads! Ideally, I would only take a calibration image in steps of a few hundreds nanomters and then let the device track for hours without moving the threads. My microscopes are usually pretty thermally stable, and I correct for focus drift mathematically in real time. Of course, a piezo actuator would be the go-to method in applications like this, and the cost of the actuator itself is not very high. However, low-noise/high-stability/high-voltage drivers that piezo actuators require cost much more than the actual piezo. Plus, the range of motion of piezos is narrow (I was hoping to get at least 0.5 mm range for focusing – to account for coverslips of different thicknesses), Thus, I was hoping to get away with closed-loop stepper motors which are more accessible, hence the 508 TPI screw idea.
Yes, the asymmetry in the reduction lever should not be an issue.
Stiffness of the material alone is not that important. It only determines the forces needed. The beams can then just be made thinner. The interesting thing is how far the flexure can be bend without yielding - the maximal elastic strain (better only be used to like 30%). For some reason engineers only talk indirectly about that. But it is basically the ratio of yield stress / Young’s modulus. Basically how good is the materials as a spring. Less than ideal material properties can be compensated by more forgiving geometry, though.
When it comes to levers the trick is to find the right middle ground. 20 : 1 means the end of the lever needs to travel 20 times longer. So if you want 1mm travel the level needs to be able to travel 20mm without bending its flexures too much. If you have space for a long enough lever that’s not a problem, otherwise may be 5 : 1 is also fine…
1 Like
I believe there are bridge-based designs in which higher amplification/reduction ratios are possible, but these flexures get too big and too complex for me to understand.
At any rate, you are right: me expecting a 1 mm travel with a 20:1 reduction is silly, even though the screw I have does allow for 20 mm of travel. Unless I used some fancy materials like Nitinol – but then the assembly gets complicated. So, most likely I would have to have two separate mechanisms for coarse and fine focusing.
Here my z axis flexure; for both optic and illumination. This is not really usable yet, but gives a first idea. There are all kind of details still wrong. 10mm travel, driven manually by a M3 screw. This is inconvenient with the 40x objective but just about usable. So this needs a larger reduction. The flexure beams can probably be shorter and still retain the range of motion. It still misses a preload (and long enough screw) to use the lower half of the travel.
Given the flexures on their own are very weak out of axis together they are pretty stiff. Rotation around the vertical is weak as there is nothing but the out of plane stiffness of the flexures preventing it. Otoh there are not a lot of forces expected there.
1 Like
I think you will not need as much reduction as you are planning. The OpenFlexure z-axis uses a much smaller reduction (I think just 2:1 or 1:1) and gives sufficiently fine motion for manual operation at 40x with just an M3 screw. For motorised motion gearing gives us 8000 steps per turn, which is 62nm per step on an M3 screw, even without any more lever ratio. Microstepping a bipolar motor gets to the around the same number of steps without gearing.
I wonder if the 62 nm is a theoretical number, or if it has been verified experimentally.
I am not a fan of internal gearing: I have experience with lead screw-based Thorlabs actuators which are internally geared (I believe they are 80 TPI), and I never got sub-micron resolution with them, even when approaching from one side. I admit, I did not try too hard, b.c. at the time I had subnanometer-resolution flextures for fine stepping and just used the lead screws for manual focusing.
https://pubs.aip.org/aip/rsi/article/87/2/025104/1021864/A-one-piece-3D-printed-flexure-translation-stage has some graphs and numbers. Looks like sub micron to me. As the paper is talking about a RAMPS board I guess this is still using bipolar steppers instead of the cheap 28BYJ-48.
1 Like
A more recent study on the OpenFlexure Block Stage is Optica Publishing Group . That has less travel range than the Microscope main stage, about the same as the z-axis I think. Sub-100nm resolvable motion in that study.
There is not a systematic study of the v7.0.0 Microscope motion, but we do know that it is easily good enough to focus 100× oil immersion lenses to scan blood smears. The depth of focus is sub-micron.
1 Like
Thank you for the link to the fiber-alignment paper! When I read it I could not understand why variances appeared different between A and B and whether the variance was the measurement noise or positional noise. The lack of cross-talk is indeed quite good!
Drift was difficult to interpret without a graph of ambient temperature.
1 Like
