Project Caretaker. Part 3. Photoresist

Since we've already covered (and planned even more) a lot of technologies - why not include another one?

I decided on this spontaneously after coming across an article mentioning photoresist, solder masks, and photopolymer printers.

I had minimal previous experience in designing and manufacturing PCBs - I designed in Sprint Layout and implemented using the Toner Transfer method, but since my technical capabilities now allow it - let's try something new.

TL;DR; Instructions for manufacturing a PCB with photoresist on a photopolymer printer

If you're not interested in my personal experience, here's the workflow I ultimately developed:

1. File preparation
   1. Export SVG of all layers from KiCAD: you need the negative of the copper layer, negative mask, and regular silkscreen. It's important that the outer contour of all SVGs matches and each layer has a common center, otherwise it won't be possible to align the mask.
   2. Convert SVG to STL using my developed converter - https://github.com/positron48/svg2stl
   3. Prepare print files on the photopolymer printer - 10s exposure of the bottom layer, 6 bottom layers. Center the parts in the slicer.
2. Photoresist
   1. Sand the copper layer with 1000 grit sandpaper
   2. Degrease the board
   3. Apply the photoresist
   4. Iron a couple of times on "synthetic" mode for 1-2 seconds
   5. On the printer, create positioning guides in 2 directions to accurately position the board
   6. Expose the photoresist on the printer for 20 seconds
   7. Develop in a solution of soda ash for ~12 minutes (250 ml water, 1/2 teaspoon of soda ash)
   8. Post-expose the photoresist for a couple of minutes under ultraviolet light
   9. Etch the board in solution (100ml of 3% hydrogen peroxide, 3 heaping teaspoons of citric acid, 1 teaspoon of salt, heated to 40℃)
3. Solder mask
   1. Degrease the board
   2. Apply a small amount of mask to the board (if you have a silkscreen mesh - you can skip a couple of steps)
   3. Distribute the mask evenly across the board through a transparent film (by hand, with a spatula, or glass)
   4. Wrap the board with film and secure it with tape
   5. Expose on the photopolymer printer for 1 minute
   6. Peel off the film and wash away unexposed mask with alcohol (you can wipe with a cotton swab or cloth)
   7. Post-expose the mask for several more minutes under UV lamps
4. Silkscreen - completely identical to the case with the solder mask.

What we need for manufacturing a PCB:

• routed PCB design, this time I used KiCAD (I also looked at Eagle, but it's in its final months);
• copper-clad board;
• photoresist - exists as a liquid or film, but I took a board already covered with photopolymer film - we'll try to skip the step of applying photopolymer to the board;
• negative film photoresisit;
• soda ash - for developing the photopolymer;
• citric acid, hydrogen peroxide, and salt - for etching the board;
• caustic soda (drain cleaner) - for washing off exposed photopolymer;
• solder mask - protects the main part of the board and gives it color;
• silkscreen mesh - to apply the mask in a perfect thin layer;
• UV lamp / LED strip / exposure box for photopolymer parts - for final exposure of the mask after development and washing off excess.

I also use a hot air gun to heat the board so the mask dries after application, and of course, a photopolymer printer for exposing the board and mask.

KiCAD

Schematic

KiCAD can transfer elements from schematics to boards with auto-routing, but this didn't work for me - some parts were missing from the library.

And I drew the schematic for the entire device as a whole, while the board serves as an esp32-cam shield for connecting wires to connectors.

The schematic turned out like this:

I included everything except the power adapter connected to the BMS charging board (that's a USB-C board with a 12.6V output).

Board

For the reasons mentioned above, I had to draw the entire board manually. There was also a strange bug (or unclear feature) - if you copy a component and paste it - traces don't attach to the copies.

The board itself:

And its 3D render. If all elements were from the library - KiCAD would have rendered them in 3D as well, but for now it looks like this:

Photoresist

Now to the most challenging part for first-timers, let's break it down into steps.

1. Preparing the file for the printer

In brief - we have a schematic in KiCAD, from which we need essentially 2 layers - copper and solder mask. We could also take the silkscreen layer and apply it to the front side using LUT for correct visibility of labels, but that seems unnecessary for now.

Drawing the negative - SVG - we export the copper layer in negative, and the mask and silkscreen layers in normal mode.

Next, we need to somehow convert these SVGs into a format the printer understands. My Photon Mono X6Ks doesn't understand anything except its own format, so we need to make an STL model. Online there are many different methods with lots of software (Photoshop, various CADs, UVTools, etc.) - even the instructions look quite cumbersome.

But we're in the AI boom era, and the task of extruding SVG to STL seems quite trivial. It took several attempts to make it all work, ultimately resulting in the svg2stl utility written by Sonnet 3.7 Thinking in Cursor.

The only thing I couldn't overcome was building for Windows and eliminating false virus detection. It turns out Windows Defender recognizes executables built with either pyInstaller (which packages source code, Python interpreter, and all dependencies) or nuitka (which translates source to C and compiles) as trojans:

If anyone wants to overcome this - welcome to the repository. I also haven't checked running on Mac due to not having one, but there shouldn't be problems due to its Linux-like nature.

To briefly describe what happens - SVG is converted to PNG. The parameter specifies the physical pixel size, for my printer it's 0.034, there's not much point in making it smaller - this size essentially affects the DPI of vector rasterization and the physical dimensions of the model in STL. Then, not in the most optimal way, rectangles are collected based on black pixels, and "extrusion" is performed at a specified thickness - i.e., all corresponding polygons are generated.

STL with a pixel size of 0.034 looks like this (0.5mm traces):

2. Photoresist exposure

There are positive and negative photoresists, mine on the ready-made board turned out to be positive. The board was enough for 3 attempts, the first failed due to positivity, the second because I forgot to update the file on the flash drive after inversion, and the third I exposed correctly, but in the soda ash solution it developed so slowly that it came off even in areas where it shouldn't have. Whether it was the board quality, photoresist, its application, an unsuitable solution, or exposure parameters - will remain a mystery, as the ready-made board was used up.

I didn't want to order a new one and experiment with parameter selection, so I decided to go back to the canon and manually apply film photoresist - I bought film from Chip and Dip, which should be more than enough for all experiments. The photoresist is negative, with instructions on the box, so everything should be fine.

Just to clarify:

• negative photoresist - exposed areas solidify, so you need to expose the parts of the board that should be covered with copper;
• positive photoresist - exposed areas dissolve, so you need to expose the parts of the board that should remain without copper.

I exposed it on a photopolymer printer (Photon Mono X6Ks) - for positioning the board, I constructed something from document file plastic and rectangular plates left over from printer testing.

I placed it roughly, eyeballing the image from the printer (this is ultraviolet, look only through glasses or a camera):

By the way, if you notice discrepancies with the schematic and masks above - kudos for your attentiveness, initially I incorrectly routed the voltage divider.

I also held the board with my hands, as 20 seconds isn't too difficult:

After exposure, we get some vague pattern that should become more contrasted during development:

2.1 Applying film photoresist

My photoresist comes as a film in a roll - with protective films on both sides. It's easy to determine which is the adhesive layer, but harder to peel off these protective films. In my case, the adhesive layer was only slightly sticky, and the upper one completely smooth and not sticky.

Due to the weak adhesiveness, I managed to stick the film without bubbles on the first try, smoothing it just with my hands, as I don't have a laminator (yet?). Then I heated the resulting board, clamped to another with clothespins, to ~100 degrees with a heat gun (as required by the instructions). This was enough for the photoresist to develop normally even with a little brush action.

But it wasn't enough for etching - the traces came off the board during the process.

I had covered a piece of board large enough for 4 PCBs with photoresist, and only a quarter was used for this unsuccessful attempt. For the remaining 3, I ironed them through paper on synthetic mode; the photoresist became less uniform but clearly adhered better.

And still, the subsequent development was unsuccessful - helping the process with a brush, I also brushed off traces that should have been fixed.

Then with one board I repeated the same exposure process for 10s and put it in developer, and decided to expose the last board longer - 20s. The more intensely exposed board got a sufficiently high contrast, and in comparison with it, it became clear that the photoresist had previously been washed away precisely due to insufficient fixation.

It was with this board that I managed to reach the final stage.

3. Development

We'll develop in a solution of soda ash 1-2% (in household stores you can find a kilogram or so package, usually intended for laundry).

Solution recipe - half a teaspoon of soda for 250ml of water.

I tried heating it, but with or without heating - development took about 12 minutes. Here the board is halfway in the development solution:

As you can see - everything developed excellently, the copper is clean. If you doubt whether all the film has dissolved - keep the board in the solution for another 10% of the time already spent.

4. Etching

Some people dry the board with an iron after development, some wait several hours; I decided to post-expose the board in my improvised solarium for 10 minutes - ultimately the contrast significantly increased.

We'll etch in a solution of peroxide and citric acid: 100ml of 3% hydrogen peroxide, 3 heaping teaspoons of citric acid, 1 teaspoon of salt. It's best to heat the solution to 40℃.

Just place the board and periodically stir/remove it so that the forming bubbles don't interfere with the process:

5. Washing off the photoresist

It can be washed off in the same soda ash solution, but I had already disposed of it for some reason, so I mixed drain cleaner (caustic soda, sodium hydroxide, "Krot") in a small amount of water. The solution is very dangerous, work strictly with gloves - it's concentrated alkali.

In this solution, the photoresist washes off in just a minute; I helped it with a cotton swab, but you could just leave the board to soak.

6. Mask

You could stop here - tin the board and solder the components, but we're dealing with high technology after all, not schoolchildren - time to try the solder mask too.

I have this small tube of black mask:

Unfortunately, by the time of the experiments, I had not yet received the silkscreen mesh through which this mask could be applied in an even thin layer - so we'll spread it under the film ourselves.

I don't have photos of the process, I'll describe it in text - degrease the board if you've managed to smudge it or touch it with your hands. Drop some amount of paste and cover with film (this can be a Mylar film, something thicker, or, as in my case - part of a document "file").

You can press down with flat glass for even distribution. I didn't have such a thing, so I distributed literally with my hands - fingers. It turned out not very even, but I think acceptable for the first time.

I couldn't wait for the mask to dry to the point of not being tacky, so I wrapped the film and secured it with tape.

Then we expose on the printer, positioning the board in the same way as when exposing the photoresist. But the exposure time should be longer - I held it for 50 seconds.

Since the mask hasn't dried yet - when removing the film, unexposed areas come off easily, you can just wipe with something; I washed with a cotton swab in isopropyl alcohol.

In some places the mask chipped off, in some there are small glimpses of copper - this can be fixed with a toothpick or left as is.

After washing, I sent it for post-exposure in the solarium for about 10 more minutes.

7. Silkscreen

I initially planned either not to apply inscriptions to the back side or to apply them using LUT, but since the experiment with the mask was successful, I decided to repeat it here.

The process is the same, we'll just expose only the inscriptions.

After washing off the mask with alcohol, I sent it for post-exposure for some minutes in the solarium.

There's one small flaw with "5v" - the tail of the digit came off, but this can also be fixed by hand or left as is.

Behind the scenes, I also drilled holes with a 0.8 drill on a homemade drill - it probably collected dust for about 8 years in the closet:

And that's it, just need to solder the components and test it (I checked with a multimeter - everything is fine). Another technology can be considered mastered, and it didn't take that much time - managed it over the weekend.

The board was manufactured for my Caretaker project - you can learn about other technologies and the project as a whole in the series of articles.