Sometimes (frequently…) I get sidetracked from photography because I need some kind of equipment or trinket. In such a case I could just go online and order whatever it is that I believe I need. But most of the time I end up DIY-ing something, especially if it’s an electrical device. And that involves making PCB’s. Out of necessity, I have become reasonably good at making these at home. Let me share my current way of working, as it involves a few tricks that make this a little easier.
What this is not going to be about, is how to layout a PCB. I might do a write-up on this at a later point. On the other hand, many people are far better at this than I am. I particularly recommend the videos by Phil’s Lab. Personally I use DesignSpark from RS Components, which does pretty much the same as the more popular KiCad. It just so happens I once got onto the DesignSpark boat and when I recently gave KiCad a spin, I was just annoyed that I had to relearn everything. Admittedly, KiCad’s 3D previewing abilities are way nicer than DesignSpark’s, which look like they escaped from an early 1990s PC racing game.
Introduction
Anyway, we start with a layout, which in this case is a LED driver board. I’ll do a write-up on the project it’s part of, as it’s photography related and sort of worthy of a story on its own. If I get it to work, of course, so let’s start with that!
Pay no heed to the silk screen design – I mostly ignore it, since I very rarely print silk screens. Very crucial bits of information I usually just write on the PCB with indelible marker.
The process I’m going to use to make these PCB’s is based on how it’s done commercially, but with a few twists to make it more feasible to do this at home with relatively simple materials. In essence, the process boils down to making both sides of a double-sided copper laminated FR4 board light sensitive, exposing the artwork onto it and then developing out the resist. After this, the board can be etched. Next, a solder mask is applied, and finally the board is drilled and populated.
The whole thing typically takes me a couple of hours for a double-sided board. If you put a price tag on labor, place just scoot over to JLCPCB or PCBWAY and have your board made for you, because evidently, doing this at home makes no sense at all from an economic viewpoint. Besides, most of the stuff used stinks, it’s probably not all that good for you and things get either hot or wet or they stain like mad, etc. All considered, it’s a nuisance, but it’s a nuisance that puts a working device in my hand within a day, and for me, that’s why it’s worth the hassle.
Working with liquid / paste photo resist
To make this real, let’s start by printing the design on some transparency sheets, since it’s going to essentially be an optical process. I use generic screen printing film – the same I use for digital negatives for photographic ‘alt process’ prints. The Epson 3880 I use for generic photo use also works very well indeed for making PCB’s. How convenient.
Now, how to get the printed design as copper traces on a board? The raw material comes as a board that’s laminated on one or two sides with a thin copper film, typically 35um or 70um copper thickness. The base material is usually a stuff called FR4, which apparently is a mat of fiberglass glued together with epoxy. Thus, we actually need a way to remove the bits of copper we don’t want. We do this by etching the copper with an appropriate chemical, and to do this selectively, we use a so-called ‘resist’ to selecitvely block out the access of the etchant to the copper.
As a brief interlude: some people have good success with the so-called ‘toner transfer’ method. It involves printing the artwork with a laser printer and then transferring the actual laser print onto the PCB as a resist. This sounds great in principle, but I’ve never had much luck with it, so I gave up on it. It’s also not very convenient for manufacturing (small) series in my opinion.
There are several types of resist, but the cheapest and most pervasive stuff (e.g. via AliExpress) comes as either a film or a solvent-carried paste, both of a blue color. Apparently it’s the same stuff in a different form; the film is just the paste without (most of) the solvent. I’ve used both, but usually go with the paste, which is what I will use in this example as well. If anyone is interested in the film stuff: I used to use it for photopolymer intaglio printing in the past. I even made a couple of videos on that process, and this particular video shows how I laminated the film to PVC plates. I’ve used exactly the same approach for PCB’s from time to time, and this works perfectly as well. Here’s another tutorial for working with the dry film.
The ‘liquid’ resist is a very smelly, blue, thick paste that dries very slowly at room temperature. It’s in fact not so nice to work with – until you figure out how to deal with its particularities.
In online tutorials on using this paste, I saw someone using a spatula to apply this paste as a thick layer onto a board and then drying it for a day or so. Yeah…but no. I’m not going to sit around for an entire day waiting for my resist to dry; I might as well send out an order and have my PCB’s shipped to me next week. There must be a quicker way, right?
Well, there is: the paste works best if it’s applied in a thin layer. I’ve read about people diluting the paste with acetone or similar solvents and then brushing it on. The problem here is that the acetone evaporates quickly and it’s very difficult to get an even coating on a PCB this way, at least using a brush. I experimented with some other methods, e.g. flowing diluted paste across a PCB just like you’d pour collodion plates, but with not much luck.
What did work for me, is a simple rubber roller, or ‘brayer’ in printmaker’s vocabulary. I use the small red one in the photo above for PCB’s. It helps to first roughen the PCB with some fine grit sandpaper. Then simply dab a little paste on the PCB and roll it out into a thin, even film with the roller. Don’t work it too long, as the paste will start to dry, clump together and just stick to the roller instead of the plate. With a little practice it’s fairly easy to get a usable coating. A very sticky one, at that.
A film of good evenness helps later on in the development process. Too thin a layer tends to have gaps, and too thick a layer tends to give problems with drying and developing as well, so it’s a matter of practice to get it just right.
The fresh layer is super sticky, so it needs to dry before we can expose it. I use a soldering hot plate for this. I have this around anyway, since I do a lot of projects involving SMD components. But I keep finding more and more uses for it – it’s probably the best €75 or so I’ve spent! So it also doubles as a resist-baking oven. I set it to 150C or so and put the freshly coated PCB onto it.
Drying the PCB actually doesn’t take very long – a minute or so, maybe two. If it takes longer, the coating is usually too thick. A properly dried layer is easy to spot by looking it at an angle. Here’s what the sticky/wet and dry PCB looks like:
When properly dried, the coating has a dull matte appearance without any shininess to it. As soon as it’s cooled down sufficient, it won’t be sticky anymore and the artwork can be laid on top of it without everything fusing together.
For the exposure, I use a simple contact printing frame and a UV LED light source. The contact printing frame could be replaced with a heavy sheet of glass, or a glass sheet that’s weighted down. Good contact between the printed transparency and the PCB is crucial for imaging fine details.
Exposure with this particular light source is about 40 seconds. Sunlight should work too, or basically any other kind of UV light source. The best exposure time is a matter of determining it with trial & error. It turns out to not be very critical. After exposure, nothing is visible on the PCB, but this changes when we ‘develop’ the image.
Development is a fancy word here for just washing out the unexposed resist with a sodium carbonate solution. I use regular washing soda from the supermarket. The concentration is…ah, perhaps 5% or so? I don’t know really, and it’s not critical either. Anything from 2% to 20% will probably work just fine. This is what the PCB looks like within a few seconds when it goes into the developer bath:
The exposed pattern now becomes visible, with the exposed part of the image remaining unchanged. The unexposed parts become darker and the material actually becomes brittle and unstable. It’s almost possible to wash it away just by agitating the tray.
To help the development process along, I usually gently brush away the unexposed material while keeping the plate submerged. It should take no effort at all to remove the unexposed resist. If this isn’t the case and the unexposed resist only comes away with difficult, or not at all, any of the following causes can play a role:
- Overexposure, so the resist that was under the dark parts of the printed transparency are also exposed.
- Insufficient ink coverage of the transparency – especially due to poor transparency materials. I’ve got good experiences with screen printing film and recommend it for this application. Banding issues (inkjet) or poor coverage (laser) can also be part of the problem and should be visible on the transparencies using a loupe.
- Poor contact between the transparency and the coated PCB. This causes light to leak underneath the parts that shouldn’t be exposed, hardening those areas.
- Resist coating is too thick. Too thick layers tend to be of a deeper blue color. Uneven thickness can also be a problem, with thicker areas requiring more time to develop properly, and too thick layers not developing entirely at all.
- Resist was dried at too high temperature and/or for too long. Excessive heating of the freshly coated resist tends to harden it in my experience, and it will then not develop fully, or with much difficulty.
It’s also possible that even the gentlest of brushing removes resist that should have remained on the plate. I find this is much less likely than the opposite (see above). The major reason for this happening is usually underexposure. Try again and expose a little longer.
After developing the resist, I inspect the PCB carefully for any minor defects. If there are a lot, I will strip the resist off with a sodium hydroxide solution (see later on) and try again. 9 times out of 10, the problem is due to coating issues, with too little or too much paste being applied to the plate. Practice, practice…
Minor faults I usually fix ‘on the fly’. Little bits of resist that shouldn’t be there can be scraped away with an X-ACTO knife. Areas where resist is gone, but where it should have stayed, can be filled in with purpose-made indelible pens (see e.g. AliExpress). You could even hand draw your artwork with these, but I only use them for retouching.
Catch you on the flipside: alignment
Most PCB’s I make are double-sided. Having done the resist on the top layer, now comes the tricky bit of doing the bottom layer as well. This of course has to be in perfect alignment, otherwise there will be massive problems with any components and vias that involve drilled holes. In my experience, any alignment error beyond ca. 100um (0.1mm) is problematic. Fortunately, it’s not insanely difficult to get good alignment, at least on the small-ish PCB’s (up to 160 x 100mm) I tend to make.
I always include at least three alignment holes as far apart as possible on my PCB’s. I usually drill these out in the end to become mounting holes, so they are double-duty. In my PCB software, I use a ‘bullseye’ shape for these holes, which makes the alignment process just a little easier. You can see these at the corners of the PCB in the image above.
After applying the resist to the first layer and developing the artwork on it, I drill out the alignment holes using a small drill. I use the generic ‘PCB drills’, which are apparently tungsten carbide, from AliExpress. They work great, but you MUST use a drill press. Any lateral movement WILL break the drill, as they’re very brittle. However, using a drill press that the dremel tool bolts into, I very rarely break one. Really, the drill press is another pro-tip. It makes all the difference.
Now, it’s a matter of repeating the resist coating and drying procedure for the bottom side. This is done exactly the same as the top layer. Since the top layer is already exposed and developed, we don’t have to worry too much about it anymore. Once exposed, the resist is actually pretty darn tough and doesn’t get damaged very easily.
I use a simple LED tracing pad (poor-man’s light table) to do the alignment as it makes it easy to overlay the alignment marks over the drilled holes. Some masking tape holds the artwork in place as it goes into the contact printing frame.
From here, it’s the same process that was already done on the top layer, i.e. develop with the aid of some gentle brushing:
Etching using ferric chloride
The go-to etching agent for PCB’s is ferric chloride. It has been used for decades and for good reason: it’s fairly fast, doesn’t release harmful vapors and relatively non-toxic. The main drawbacks are that if you drop it onto your pristine stainless steel sink or cutlery, they WILL rust. Trust me. Also, it tends to leave nasty stains. And I personally don’t really like its smell, although that’s more of a minor inconvenience.
Etching the PCB is straightforward: just dunk the PCB into a tray with some ferric chloride. Nothing here is particularly critical. I dilute the ferric chloride solution 2:1 or 1:1 with water – but it’s not critical. I reuse the etchant until it gets soo slow I’m getting too impatient when using it – again, it’s not critical. Time of etching, is…you guessed it: not critical. Sometimes I just leave the PCB in the etching tray and go have dinner or something. Come back in an hour and it’s usually almost done, except for a few spots that touched the bottom of the tray or so.
But usually I like to speed things up by brushing. That’s how that brush in the previous photos got that horribly rusted ferrule! I now use a brush with a plastic ferrule. With active brushing across the surface and a fresh etchant, it takes only a few minutes to complete etch a double sided PCB. Easy!
There’s not all that much to say about the actual etching, really. The challenging bit is to get a good resist image on both sides in proper alignment. Once that’s done, the etching is straightforward and generally doesn’t present any problems.
Stripping the resist
Now that the resist has done its job, it needs to be removed. Exposed resist is pretty tough stuff. Heating it afterwards toughens it up even more, I find, so especially the resist on the first layer of a double sided PCB tends to be very resilient. Removing the resist is done easily (but not necessarily quickly) by soaking the PCB in a solution of sodium hydroxide. I’ve tried potassium hydroxide as well, but I have a feeling the sodium hydroxide works better in this application for some reason.
I use something like a 5% or 10% solution – pretty strong. The stronger, the quicker it works. Of course, a strong hydroxide solution wreaks havoc on your skin, so don’t touch it and/or wear gloves. Just leave the PCB in the solution for an hour or so to allow the resist to soften. Eventually it will and most of it can then be brushed off.
When I’m in a hurry – or simply impatient – I soak the PCB for maybe 10 minutes and then take some fine grit sandpaper to it and rub off the resist. This takes quite a bit of elbow grease, but works OK. It does take off a little bit of the copper in the process, although I’ve never found this to be a problem.
This leaves us with the etched PCB. In principle, this is ready to be used. There is no strict necessity to process it further; components can be soldered onto it and the PCB will work in this state for many years.
However, ultimately, the copper will corrode and more importantly, it’s very tricky to solder (small) components onto a PCB like this without getting globs of tin to stick literally all over the place. Therefore, I prefer to apply a solder mask as well. But first, let’s have a look at the technically finished PCB:
The photos are evidently crap (pardon my Gaulle), so I tried getting a scan of a small part to show the image quality. Not much luck with that either, given how shiny the copper is! Anyway, here’s an attempt:
The square box is a 6-pin small outline IC (a Monolithic MP24894 LED driver) with a 1.0mm pin pitch. The 0603 box is the outline of a capacitor with a 300um trace running between its pads. Clearance between this trace and the cap’s pads is 190um / 7.5 mil. A 0805 decoupling capacitor is shown vertically to the right; its upper pad is connected to the ground fill with 190um / 7.5 mil spokes. The other pad connects with a 400um trace to one of the IC’s pins.
For me, this kind of resolution is more than good enough. It allows me to make PCB’s that use leadless parts with fine pitches of e.g. 0.5mm, like the Texas Instruments TCA8418 keypad controller with 24 pads on a 4x4mm QFN package or the STM8S003F3U6 8-bit microcontroller (3x3mm QFN, 20 pads).
The process is generally not entirely defect-free – there are often little blemishes, but much of the time they’re in non-critical areas where they don’t hurt a thing. The occasional functionally harmful errors can usually be corrected afterwards. But to be honest, the major source of problems are not in the manufacturing part, but plain old & stupid bugs resulting from momentary lapses of reason on my behalf.
Solder mask
A solder mask is a coating that leaves only the solder pads bare and otherwise seals off the copper traces. This has two functions: it prevents the copper from oxidizing – although it generally takes a rather long time before this becomes a threat to the functioning of the circuit. More acutely, a solder mask is helpful when soldering components onto the board, since the tin will only be able to adhere to the bare copper. Especially pads that are close to each other will easily bridge without a solder mask and it’s difficult to clean the mess up once that happens. What’s in a name, indeed.
In terms of making a PCB, the solder mask behaves very similarly to the etch resist. It’s more smelly though, and slightly more tricky to work with. When I started using this stuff, I went online and found how-to guides very much like this one by Adafruit. They’re not the only one to suggest the ‘soggy sandwich method’ where the wet solder mask paste is sandwiched between the PCB and a transparency and exposed that way. I found it an annoying way of working, for two reasons:
1: Using a clear sheet means that there’s generally more distance between the printed artwork and the solder mask paste, and that is detrimental to image resolution. Small features, like tiny QFN chip pads, tend to not image well. They end up being exposed and thus impossible to solder without manual retouching by scraping the excess solder mask away. You can actually see this happening in the Adafruit video linked to above. But since they’re using gigantic through-hole pads, it’s not all that much of an issue for them. I found it counterproductive, though.
2: It’s a mess! Sure, it looks clean enough in the video, but in reality, you end up with little snippets of foil that have a horribly smelly and sticky paste on them. These bits of foil then stick to whatever comes into contact with them, and to cut a long story short: you’ll be cleaning off little solder mask stains all over the place. Better do it quickly, too, because as soon as they expose to UV, they harden and are virtually impossible to remove without damaging the surface they’re on.
Moreover, I just don’t see the point in wasting 75% of the material. If you look at the video linked to earlier, I’d be surprised if 25% of the paste actually made it onto the PCB, with the rest of it being squished to the side, making a mess. What’s up with that?
There must be a better way, right? And there is – just do exactly the same as with the etch resist. Just put some dabs of the mask paste onto the PCB and roll it out with the small rubber roller. A thin layer is all it takes. Easy peasy! This part is actually easier than rolling out the etching resist, because that one dries out much quicker. The solder mask virtually doesn’t dry out unaided. But we’ll fix that.
How about drying, then? Turns out that again, we can drive off the nasty solvents with the hot plate. I set it at around 160C or 170C and pop the PCB onto it. It’ll start to fume quite badly – no doubt the solvent vapors aren’t exactly healthy, so better ventilate well. Boy, this stuff stinks, I can tell you that. Anyway, drying the plate out takes only a minute – remove it immediately once the entire surface has transitioned to a dull matte sheen like so:
Take care not to overheat the solder mask. I find that if I leave it to bake longer than strictly necessary, it won’t harden properly anymore when exposed. It seems to do the opposite of the etching resist, which hardens fully if overheated. The solder mask seems to disintegrate chemically, resulting in exposed areas dissolving during ‘development’.
After drying it, it’s still a bit tacky, especially when it’s still hot. Therefore, I don’t apply much force at all when sandwiching the PCB together with the artwork to expose the mask. In fact, I just lay the transparency with the artwork on top, followed by a piece of sheet glass and that’s it. No pressure frame or anything. For very tiny pads, some pressure may be needed to get really good contact between the artwork and the PCB surface, but some of the unexposed mask tends to then stick to the sheet with the artwork. It’s usually not a big problem when that happens, though.
The solder mask requires far more expose than the etching resist. The latter I exposed for 40 seconds under my light source, but the solder mask requires something like 240 seconds at least. Err on the long side with this stuff. It’s annoying to have to do it all over if too much of it brushes off in the next step.
On the other hand, it is possible to add more layers on top of each other. And in fact I’ve exploited this once or twice to make a sort of silkscreen, using one solder mask layer as the base and another one on top of it for the ‘silkscreen’ print. Works quite well, but it’s an additional step that I mostly find unnecessary.
After exposure comes a developer step. Again, it’s not so much development, but actually dissolving away the unexposed mask. The solder mask is a lot more chemical-resistant than the etching resist, so the sodium carbonate solution doesn’t work for this one.
Experimentally I found that a mixture of equal parts of acetone and ethanol (no additional water) works quite well to soften the unexposed solder mask. It can then be gently brushed off, just like with the etching resist development. The downside is that this step smells quite badly, again.
Brushing off the unexposed mask is the proof of the pudding, so to speak. If the mask is unexposed or overheated, parts of the exposed mask tend to also brush away. It’s best to brush very gently, and of course don’t overheat and underexpose the mask to begin with. In any case, the freshly exposed mask is very fragile and only becomes tough when it’s heated some more (once the unexposed stuff has been removed) and/or additional UV exposure. Treat it with care when it’s still fragile.
The process is then repeated for the bottom side:
Drilling and vias
Drilling is just what it is – there’s not all that much to be said about it. In a DIY workflow, drilling is one of the final steps, in contrast to factory PCB manufacturing, where it’s usually one of the first. As said earlier, I like to use a drill press because I don’t waste nearly as many tungsten carbide drills that way. It also makes aligning the drill far easier.
I use an assortment of tungsten carbid drills; they can be had fairly cheaply from AliExpress. Diameters I use the most, are:
- 0.3 mm for vias (see below)
- 0.9 mm for through-hole components like connectors
- 2 mm, 2.5 mm and 3 mm for mounting holes
I usually drill from the side where the tolerance is the most critical. For instance, on through-hole connectors I always route the tracks to the connector on the bottom side of the PCB. When drilling, I drill from this side, because a slight alignment error to the other side isn’t as problematic as on the side where the tracks connect to the pins. But with good alignment between both PCB sides and some care during drilling, it really doesn’t matter much.
Concerning holes and connectors: there’s a notable difference between a home-made PCB and a factory-made one. In factory-made PCB’s, the inside of the holes are generally copper-plated, so each hole is effectively a via from one side to the other. On home-made PCB’s it’s not very feasible to copper plate the holes. As a result, for connectors, you have to make sure to route the tracks to the side where you solder the pins – i.e. the bottom side. This requires some care and planning during routing.
The lack of plated-through holes also has mechanical implications. A connector sits a little less firmly on the board if it’s only held in place by the soldered contacts on the bottom side. On connectors that see some force in regular use, I make sure to glue them to the PCB to compensate for this a bit. Still, the kind of PCB’s I make don’t withstand the kind of abuse a factory-made one may survive. It’s not a problem in practice, but during testing and assembly something to keep in mind.
Then there’s the issue of vias. Yes, there’s this guy who has managed to copper-plate some holes, so theoretically, vias could be done this way. It would be neat, but boy, it looks like a lot of work for sure. More people have tried this, but I still don’t see this as a viable solution given the complexity and time-consuming nature of the process. It does look neat though!
Another approach is the use of rivets. This seems easy enough, but I haven’t tried it. Mostly because I feel it’s still quite some work and it seems easy enough to me to damage the PCB in the process. More importantly, these rivets are pretty big by my standards, with typical hole diameters of something like 0.9mm and an annulus that’s significantly larger still. My vias are generally 1.3 mm diameter pads with 0.4 mm holes, and even that makes for some crammed areas around high density components. Larger vias would be too much of a compromise for me.
So I do it the old-fashioned, low-tech way. And that actually works quite well. It’s simple: take a thin piece of copper wire, stitch it through all the holes that need to become vias, solder the wire on both sides of the PCB and cut off the wires that stick out.
I’ve found some tinned copper wire on AliExpress that’s apparently intended for jewelry. I’ve got no need for that, but it does solder really nice, so it fits my use. And it comes in small diameters – I got some 0.2 mm and 0.3 mm wire. The thinner kind is a little fragile and breaks easily, but 0.3 mm turns out to be a very useful compromise and goes through 0.4 mm holes fairly easily.
Well, that’s about it. At this point, I usually poke around the PCB a bit with a continuity tester to see if everything checks out. This one presented no problems, but let’s see how it performs when it has some parts stacked on top of it. But that’s perhaps for another day!
