DIY Macro PCB Photography
niemczuk.techLabsmore is a commercial option with open source software, >10x cheaper than the existing microscope vendors. https://www.labsmore.com/
I want to point out that the Genmitsu 3018 desktop CNC that their $4400 microscope is based on is $330 on Amazon. It's hard not to recognize.
I think one could assemble the remaining optical path and camera for less than the remaining $4070.
Well, I know so, because, I did exactly that several years ago with a telecentric lens form Edmunds and a industrial C-mount camera back and an even cheaper 3018 CNC for automated inspection and archival scanning of PCBs. (telecentric lens due to the boards being populated with components, this helps capture them without focus stacking and without perspective).
Hi there! CTO of Labsmore here. Pricing is hard :) There is a lot that goes into the final price, especially once the other components, calibration, and overhead is factored in. Its our first time selling a product like this and would love input from others on how they arrive at sticker prices and PMF for commercial products.
Oh absolutely, I just couldn't help pointing out the little detail I noticed; given this is HN an all.
In my opinion, if you've shipped 2 or more units at the given prices... You've actually priced it right for your target market!
We have! And a few larger group buys / enterprise customers in the pipeline. We started out selling more expensive Mitutoyo systems (quality but expensive Japanese optics) and sold a few of those. But feedback from customers was that if we could do 80% of what that system does and get it on a credit card it would be very interesting to them (greatly simplifies purchasing flow). But at the same time distributors don't want to carry them because they loose margin over selling $30-200k comps which means we need to adapt more of a "DTC" approach than we were originally planning. So working on upping our sales / marketing game.
In my experience, for products that aren't medical devices, fighter jets, etc. The rule of thumb is 30-40%. 30-40% of the final cost to the customer is what you should be shooting to have your COGS leaving your factory.
Thanks! Yeah that's roughly inline with our calculations / how we priced things.
If we want to get the final price lower we need to switch to more basic components or tighter integration with suppliers. But we've been avoiding that under our premise having a system that can be purchased with a company credit card (say 5k or 10k limit) is the main requirement. Would love more feedback on how important that is for people's purchasing decisions. And even then if we lowered the cost of the chassis we might instead use that to improve the optics/camera following the "credit card purchase" philosophy.
I just picked up some Opto-E telecentric lenses with this exact idea in mind. (That plus possibly scanning microfiche so I don't have to mess with focus.)
If I can do it on a small scale with my long-suffering Ender 3 (which keeps getting weirder and weirder stuff stuck to it), then transplanting it to a larger motion platform is trivial.
What did you do for software?
I wrote a python script that sent G-code to step the machine in a grid via the USB-serial link, and then take photos that I later stitched with a script using OpenCV.
I think I spent more time trying to make a little tk-inter GUI for it ... that I never actually used because I all did was step in a regular grid pattern at a fixed speed... than everything else combined, and that included printing a combo ring light/lens holder-to-spindle-adapter to mount the lens in the CNC.
It was a fun single-day project. Seems like I should have started a company selling them.
Have you come across these type of things https://www.aliexpress.com/item/4001351566621.html that appear to behave like a metallurgical microscope.
I'm kind of tempted. I assume the objectives aren't infinity corrected?
Just found - https://www.youtube.com/watch?v=b_IWTq-TSiU which shows something similar. It looks like there's a ring that seems to adjust magnification on it, unless I'm mistaken.
I've got a DIY XY stage and would like to attach something like that too it, to image large objects (I've got a BHM microscope but the stage has only a small degree of movement).
Edit: Just noticed the objectives used are infinity corrected
> https://www.aliexpress.com/item/4001351566621.html
Ah yes, there, you found the other component to pair with the cheap 3018 cnc platform. :-)
Hi there! CTO of Labsmore here. We would love to chat with people involved at lab automation / digital microscopy to learn more about what we could do to help their workflows. Currently we are focusing on electronics / semiconductor but are exploring bio / life sciences given we've seen significant interest there.
Its also the first time we've started a company like this and are constantly exploring ways to validate PMF, iterate faster, and improve sales / marketing. Please reach out if you'd be interested to chat about the business side, grab coffee in silicon valley, etc.
Not quite the same thing, but I use the magnifier tool on iOS for closeup views of PCBs (it's under the accessibility settings).
With flash on, it can give quite a detailed look at your board. You could then strap a macro lens onto your phone if you need even more magnification.
I used to use this and also take still images, but I'm on an older iphone and apple introduced a bug which makes the still image blurry after you capture it.
https://www.youtube.com/watch?v=Mb659_4yIZ0
(I realize the youtube player really sucks for this, the effect is seen in the last second of the video but youtube just _has_ to shove its related videos screen in your face once you get near the end of the video)
What scale is this example?
If those are 5 mil traces with 5 mil spacing, then you might get better performance with a simple smartphone camera. The built in superresoloution algorithms seem to do a decent job on PCB's.
Yes, I've got the camera bodies, the macro lens, some historical experience stitching the pictures together to form panoramas, and I've just gotten into microscopy and am pretty sure I've got the microscope base he's using too. So I can see an alternative universe where I wrote this xD.
One of the first things I thought when I started was "how can I get the big camera onto the microscope" and the embarrassing conclusion I came to, is if you've got an expensive phone you can get cheap plastic adaptors and just attach it to the microscope body. Basic stereo microscopes are incredibly cheap these days, as are the phone attachments. I use droidcam obs to stream video so I don't even have to physically look into the microscopes any more. The main thing is going into the phone and getting some practice getting everything on manual so phone photo software doesn't do funky things.
My use case is slightly different, but the underlying principles are the same and I'm pretty sure the microscope hardware is identical to that in the post, so rather than rigging up a custom attachment to the expensive camera, I just rig up the phone camera to the already existent hardware of the microscope.
See crab spider for example of one of my earliest tests: https://imgur.com/xLS07WY
The Hugin output image at the end is pretty rough still :( With all that ado, I was hoping to see better final result
I take a different approach. I use a 2D stage driven by stepper motors and tag each image with the stage position. Then I use ImageJ's Grid Stitching plugin using my captured locations. This isn't really perfect, but you can also have the plugin improve the quality by doing various repairs.
Elsewhere on this thread, some company appears to be re-using the 3018 CNC form factor to build inspection microscopes. That's a remarkably cost-efficient way to get pretty good results in terms of motion. I've also purchased simple 1D stages and stacked them get get a full XYZ stage for very little money.
What's the magnification that you can get with such a setup?
If you just strap your phone to an actual microscope (see my other post), you're basically just limited to the magnification that the microscopes achieve. And very good performing microscopes are very affordable by hacker news standards these days.
Stereo microscope magnifications are typical at about 20x to 40x, but I'll commonly go to 640x on the biological one, though in practice you bug out at the physical limits of optical microscopy just above 1000x.
His specific setup is presumably limited by the magnification provided by whatever his objective lens achieves.