Creating Lenses On Cheap CNC Machines

There are a lot of CNC machines sitting around in basements and garages, but we haven’t seen anything like this. It’s making lenses using a standard CNC machine and a lot of elbow grease.

The process of making a lens with a CNC machine begins by surfacing a waste board and taping an 8mm sheet of cast acrylic down with double-stick tape. The lens is then cut out with an 8mm endmill, removed from the stock material, and wet sanded to remove the tool marks. Wet sanding begins at 400 grit and progresses to 2000 grit, after which the lens is polished with a polishing compound meant for high-gloss car finishes. This was done by hand, but in this instance there’s no shame in using a real buffing wheel.

Several other lenses are demonstrated, including a cylindrical convex lens, but these are only planoconvex lenses, or lenses that are flat on one side. Biconvex lenses can be constructed by gluing two planoconvex lenses back to back, which is done with an acrylic glue, in this case Acrifix adhesive. The result is remarkable: with a lot — and we mean a lot — of sanding and polishing, you can make an acrylic lens on a cheap hobby CNC machine. The trick is just a very small stepover on your CNC path.

There are a few more videos planned in this series, including one on using Fusion 360 on defining the shape of the lens to have the right focal length. We can’t wait to see that.

48 thoughts on “Creating Lenses On Cheap CNC Machines

    1. most clear acrylics polish very fast in my experience. Unless they are intentionally hardened to prevent scratching.

      I made a number of acrylic cases about 20 years ago and spent alot of time hand polishing edges and other details.

    2. I have good results machining cast acrylic blanks, and polishing on a lathe by hand. For even better optical quality, I make a silicone mold and cast in a high index resin such as Polytek Poly-Optic 1420.

      1. Yeah. You can never tell when you’ll just pop an eyeball, or, suddenly realize after 10 days that a bad refraction is responsible for your crossed eyes and those blindingly headaches. (and, kids, always remember to never look at a laser with your good eye)

    1. Glasses lenses usually use glass, I believe because it’s got a better refractive index and fewer optical aberrations.
      It wouldn’t be a good way to make production glasses lenses, but might serve to prototype unusual prescriptions?

      1. Standard glasses have glass lenses, because they are standard, thus easier to manufacture in quantity. However non-standard ones are made from acrylic. For example my wife has a set with two different focal points in each lens. These can be done from glass too, but they are more expensive, and not everyone makes them…
        Also thicker lenses are made from acrylic, because are lighter. Imagine wearing two bottle bottoms on your nose…

        1. As someone who MAKES spectacle lenses, MOST glasses are made using polycarbonate lenses.

          Some of the newest machines use diamond bits, and produce finished lenses. Slightly older, but still common optical labs use ROUGH cnc milling to hog off most of the material from the lens blank, then machines with shaped sanding and polishing tools refine and finish the lenses.

          As to how long to go from block of plastic (cast resin #39, poly carbonate, mr-7, etc.) If you eliminate the necessary math, computer time, setup and physically moving the materials around.. maybe a half hour of actual machine running time.
          Glass lenses (almost an extinct material in the US lens industry) bump that up by several hours, and then there is almost another day of ‘tempering’ for the glass lens (though newest tech is bringing that number down too).

    2. You should definitely make your own glasses, naysayers have no clue. Go to endmyopia dot org and see why. Then make your own superglasses with tele, micro and reverse. I seriously doubt anyone ever got crosseyed by making their own glasses, except perhaps Steve Martin (The Jerk).

  1. Amateur telescope builders are able to hand grind very precise convex and concave spherical optics, but it’s naturally quite time consuming.

    This seems like a great way to reduce the time by “roughing out” the lens prior to hand lapping.

    I bet he could have achieved better end results by using the self-correcting features of spherical surfaces – employing the hand grinding techniques as a finishing method.

    In any case, the MPCNC is a cool machine.

  2. One can roughly rate lenses in three categories:
    1) Good enough for LED lamps
    2) Good enough as closeup / filter lenses on cameras
    3) Good enough as primary lens on camera

    I’ve found 1) to be very easy to achieve on hobby CNC machines, and 2) to be somewhat possible. 3) seems out of reach for now, though I guess one could continue with more traditional lens making methods after making the rough blank with CNC.

    1. Hmm, I’m sure it wouldn’t be too hard to automate the (grind surface finishing)/polishing on a simple low power even Meccano/Lego like machine which held the blank and positioned various polishing tools – much like a lathe but, with optical feedback paths through the lens being polished for acquisition to camera and control system and selection of polishers perhaps also spinning/wobulating at various ratios/trajectories.

      Eg Mount lens, pass “highest quality light” source through (of suitable/selected spectra) and spin up lens also with rate profile too, direct output light onto camera or suitable screen/surface such that digital camera can acquire image, select/process much like Newton-Raphsom method to incrementally/asymptotically apply best polishing tools (rotating too or not) appropriate from a selection and thus arrive at desired shape with minimal scattering selection criteria or even aim to a particular type of translucence for arty effects but, I digress. Calibration fairly simple and also at least semi automatic eg from so called “perfect blanks” etc. Would take a few passes but, once in place have a data set Ai like to more quickly converge on best approach for new blanks stock where most coarse shape in place first.
      You could make some interesting dichroic like effects too depending on materials and desired wavelength bands across the profile with variance – the mind boggles. Of course one need not be limited to rotationally symmetrical lens shapes either – that slightly more sophisticated approach would require more enhanced and complex rotating polishing tool approaches therefore would likely be in the master class at another time for the OCD grinding purists ;-)

        1. Easy, practical and standard practice as in possible, part of the control system process is to flush the paste and dust at appropriate intervals – which can be set knowing how much material is removed on each pass or asymptotically as the process incrementally approaches the desired outcome :-)

          1. Easier said than done.

            When you wash the grinding paste off the lens, you get water on it, and that messes up your optical measuring system, so you have to stop for an entire wash and dry cycle, which takes a lot of time, and you don’t get the continuous online feedback you described.

            The more you check, the slower you go. I don’t see improvement over standard lens grinding machines that already do this.

          2. As I have already written Luke, I mostly addressed the enthusiastic DIYer or amateur – if you like, that is not dissuaded easily from ‘having a go’, I never suggested a significant advance beyond existing tech would automatically be an imperative in a commercial context – though that’s easy to do once you position yourself in respect of analysing price vs effort that suits your interests. Ie lots of people value their effort without negativity in their assessment whatwver suits and fits their personality pattern where they think of price vs time limitation. Your earlier phrase “..far beyond usefulness..” tells heaps as it has no metric and appears as if you have more the pattern of unwillingness to progress by instead putting up obstacles and not aware how many can and do quite easily get around obstacles and vague issues of comparative value to prove a (positive) point for the drive to succeed.
            When you think about it though it’s not hard to arrange a relatively sophisticated wipe process it depends what you are trying to achieve ie prove something possible or agree with failure assuming it’s not. So, consider the wipe mechanism kept dry whilst grinding in progress that would save time as in parallel, eg when the control system assess a grind verification warranted it flushes the lens And wipes then takes its optical sample and processes which can either then progress the grind further and/or alter tool, rewipe with fresh material and repeat, suspend and alert and/or blast with clean dry air at suitable temperature followed by suitable cold air blast etc It’s all really simple stuff if that is your intent to progress, well other than criticise to quash paths in some sort of game. Instead of trying to claim how things can’t be done or find facile obstacles why not instead imagine how augmentation does allow progress opening new permutations and as I wrote under the positive drive of the enthusiastic DiYer/amateur – look at history many of those personality types do very well commercially whilst the naysayers cant maintain focus and fail. Which camp are you placing yourself in, is it a conscious decision or one by just mechanically reacting being clouded under the momentum for critique and negativity naively assuming something cannot be done ?

          3. Jeebus, Mike. If you want to do all of that work, knock yourself out. I don’t think Luke was saying “it can’t be done”; he was saying that this would be very difficult. And what you describe WOULD be difficult. I mean, if you want to develop an automated lens grinder to compete with the professional ones, it would be worth the effort. But then, you’d probably lose if you started with a cheap CNC engraver. If you’re just trying to grind a few lenses for yourself, the effort it would take to fully automate the process probably wouldn’t pay off. But again, if YOU want to do this, knock yourself out – I’m looking forward to the article.

      1. >simple low power even Meccano/Lego like machine which held the blank and positioned various polishing tools

        That’s how they are at the cheap end – simple linkage/cam-follower machines that shape glass blanks by reciprocating a grinding head along a path. They’re entirely mechanical systems.

        Problem is, they have mechanical tolerances, backlash and flex that change over time, with temperature, wear, contaminants, etc. which results in varying quality lenses, and they’re limited to few types of lenses, varying within the range of adjustment for the linkages.

        To make them CNC would increase the number of options, but you’re still running the system open loop and relying on the mechanical fidelity of the system. Implementing a feedback system would increase the complexity far beyond usefulness, because simple optical or contact measurement systems don’t work with the grinding methods. You could make a lens mount that tries to measure the lens optically from the backside, but that assumes one-sided lenses only.

        1. Hmm, well sure and hope you appreciate my post intended for the enthusiastic DIYer, of the general audience often associated with this forum, many of which I’ve met through the years too. All the complexities and issues raised are very simple for those focused individuals to address and these days without much difficulty at all. In any case if one wanted to go down a commercial path then net present costing will help you assess return on investment in terms of time spent on minutiae vs turnover and profit, ultimately each to his own :-)

  3. I’ve casted all type of lens (planoconvex, planoconcave, parabolic, cylindrical, Fresnel) out of clear silicone, the one used to encapsulate solar panels and LED components.
    Although it’s very expensive, it’s nearly eatable formula (not exactly, but reading the security data sheet is a pleasure for the lungs, skin and anything, not like epoxies) and very simple curing gives really impressive results.
    To have a spherical lens, just use acrylic half spheres, for the cylindrical, half pipe. It’s also very precise, so the mold must be perfectly clean and soft, but you can also cast it over an array of very small lines, like the ones found inside LCD screens (I can’t remember the name).
    And of course, casting it in a mat, sanded mold, will give a translucent effect, depending of the size of the imperfections.
    It’ easy to remove bubbles (if you want to, can be fun with) by putting the casts in the fridge for some hours before baking it at 100 ° C for 1 hour.
    And when cured, it’s silicone : INDESTRUCTIBLE… and also, soft, it bounces on the floor with no pain…
    I love that thing, but why is it so expensive ?!!
    And a fun fact : don’t eat or cook garlic or that kind of things, the sulphur will stop the cure (it’s in the data sheet ;o).

      1. I have used Dow Corning clear silicone, but very expensive, also UV clear so a mould containing photosensitive resin exposed to UV will cure the resin.
        I made an acrylic jenkins disk on a cnc some time ago, did not take very long to polish, but did not need to be grade A.

    1. This ^ .. Sylgard 84 is kind of the “go to” formulation for clear silicone. It’s used for solar cell encapsulation, so it’s rugged, and it’s used for soft lithography, so it captures extremely fine detail (10s of nm or better). For example, I have made diffraction gratings by taking impressions from diffraction gratings.

      If you have a polished surface that doesn’t want to stick to the silicone, you can cast a lens right into the mold. And, rather than polishing the surface of the mold, the trick might be to take something that is already polished, and then use some clever way to shape it .. like vacuum forming a plastic sheet to make a shell with an already-polished interior surface .. then pour the sylgard in.

      Really, the idea of CNCing a lens, and then polishing it it almost horrifying, since you can buy cast acrylic lenses dirt cheap. I’m nearly certain you can make better lenses with a lot less grief by casting silicone. Plus, you can make some crazy lenses for special applications.

        1. Any lightning uses, I made a table lamp that looks like a flower, using a cake cooking mold with concentric circles in the bottom (just like that one :âteau-de-vintage-273656827605.html), and adding a small sanded lens in the middle to diffuse the LED light.
          You can also make BIG lenses (if you can afford it).
          The parabolic one is interesting when lightening macrophotography if you want a spot light to have strong shadows.

      1. I use the ACC QSIL 216, which is about half the price of the Dow Corning here in Europe (and still very expensive), the 218 is a little harder, but more expensive, and I love the 24 hours pot life of the 216.
        Diffraction gratings, thanks !
        I vacuum formed parabolic lens molds, no polishing, very fast.

  4. One could use diamond ball-shaped endmills to cut lenses in glass. with proper fixture one could grind them on both sides without any problems. The same goes for acrylic lenses – just make a fixture to cut them on both sides…

    1. Machinist here. They do that, already.

      There are specialty machine shops that use CNC and diamond coated bits to machine glass directly. They have vacuum table systems and other various exotic fixturing to do so.

      What I find crazy is everyone is so surprised that you can do this, but noone here has sugguested using the same CNC to polish!

      If you get the right kind of special tooling, you can totally polish using the same CNC path, or at least polish a lot of the way there and get the hard part done. Commercially they do this on dedicated CNC polishing machines that are not mills, more like dedicated grinding and lapping machines, but I imagine for hobbyist purposes, you could get a steel gauge pin, machine a sleeve on a lathe or in the spindle of the mill (using it as a vertical lathe), fit the pin in till air pressure pushes it back with a very tight sliding fit, and glue a polishing pad of rubber charged with Linde A lapping compound on the end, and rerun the path.

      I’ve made holders like this to hold pens on commercial CNC Mills as I used to run vacuum table setups and I had to Precision scribe flat boards with assembly geometry in very fine tip pen 4 Transformer substations using this technique except putting a pen in the end in place of a lapping rubber. The idea is the air pressure self compensates for the out of flat condition of the item you’re tracing it over. If you aren’t a good Machinist I suppose a spring or gravity a weighted head works too.

    1. Fresnel lenses are a whole other problem. With spherical or even aspheric lenses, you have a smooth surface to make smoother and polish, with no abrupt edges. With a Fresnel, you would have to sand each ridge separately, without rounding off the corners. This is not something to be done without specialized tools.
      If you look at how lighthouse “Fresnels” are made, they usually use separate prisms polished separately, then assembled together. That, of course, you CAN do with DIY-grade equipment, but it’s still not a simple matter. If you’re looking for cheap, it might be easier to cut mirror tiles (glass or acrylic) into smaller squares, and mount them in a 3D printed frame with all of the angles calculated to direct each reflection to one point.
      Of course, if you would really like to see somebody do something, the best way is to do it yourself.

    1. Fun fact- Monocrystalline Diamond tooling (aka MCD) gives an incredible finish on many non-ferrous materials, and if you know how to make a brazed carbide lathe tool it should be possible to make your own MCD bit.

      I used to run MCD tooling near the beginning of my career machining green unsintered tungsten carbide parts. We used MCD tooling for anything that had complex geometry that needed to withstand wear more. I have some of the old bits from about 10 years ago laying around and they’re bonded to a carbide shank using a special brazing compound in the same way older carbide tools were just braised to a mild steel shank.

      There are suppliers online that sell grain oriented MCD polished wafers, while it sounds nuts it should be doable to buy your own and braze to a square carbide blank from MSC, at 1/4 the price of MCD tooling, for hobby work.

      Just remember MCD does not like ferrous materials. It breaks down at the molecular level due to a strange affinity under heat to give up carbon atoms to the iron and it immediately ruins the edge. As long as it doesn’t have iron you will get an incredible finish, but having an air bearing spindle or extremely well made normal spindle with no run out at all helps

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.