A lot of us make circuit boards at home. I find it a useful skill to have in my bag of tricks for intermediate steps along the way to a finished project, even if the finished version is going to be sent out to a PCB fab. When I need a breakout board that meshes with other development tools, for instance, there’s nothing like being able to whip something up that plugs right in. Doing it quickly, and getting on with the rest of the project instead of placing an order and waiting for delivery, helps keep me in the flow.
Toner transfer is by far the fastest way to make a circuit board at home — simply print the circuit out on a laser printer, iron it onto the copper, and etch. When it works, it’s awesome. When it doesn’t, it can be a hair-pulling exercise in figuring out which of myriad factors are misaligned.
For a long time now, I’ve been using a method that’s very reliable and repeatable. Recently, I’ve been tweaking a bit on the performance of the system, and I thought I’d share what I’ve got. At the moment, I’m able to very reliably produce boards with 6 mil (0.15 mm) traces and 8 mil (0.20 mm) spacing. With a little care in post-production, 4 mil / 6 mil is entirely plausible.
That’s good enough for most of my prototyping needs, covering TSSOP parts down to 0.65 mm pin pitch and allowing me to feed two traces through an 0805 surface-mount resistor or capacitor. It’s on par with the cheapest of professional manufacturing houses, but I can turn around a board in about fifteen minutes. Beats the heck out of ordering and waiting, in my book. When I need luxuries like silkscreen and through-hole plating there are workarounds, but mostly I’ll skimp on those until the final version is made.
The secret? Science! Or at least taking what can be an overwhelming number of variables and secret techniques, reducing them to experiments that change one variable at a time, and then optimizing along that one dimension. There are a dizzying array of techniques out there, and a good number of them only work when you have exactly the right brand of toner, the right paper, or the right touch in wielding a hot clothes iron. In short, they’re not reproducible. When your setup doesn’t exactly match someone else’s, all bets are off. Here, I’ll lay out a reproducible method and show you how to calibrate it.
A lot of Internet toner-transfer guides focus on details that don’t really matter, like the type of paper used for the transfer medium or the method of cleaning the copper before ironing. That’s not to say that you don’t need to clean the copper first — you absolutely do — but just that it doesn’t matter really how you do it. I’ve used fine-grit sandpaper, green pot scrubbies, and these days I use some sponges that I got in the hardware store for shining up pipe joints before brazing. I follow with a wipe-down of acetone. The point is that you remove the oxidation and grease from the surface. I don’t care how.
Similarly, the choice of transfer paper is pretty open. I alternate between glossy magazine pages (the Economist is my favorite at the moment) and the plastic-coated backing papers that peel-off stickers come on. These two surfaces work entirely differently; the sticker backing paper peels off directly, while the magazine paper dissolves off in cold water when rubbed with a thumb. There is fancier stuff out there. The point of the transfer paper is that it’s glossy so that it doesn’t deform the image, and that releases easily leaving the toner on the copper. The rest is convenience.
Finally, the choice of etchant doesn’t matter. I use oxygen-refreshed copper chloride in acid because it’s essentially infinitely recyclable and I hate the hassle of disposing of toxic chemicals. People use ferric chloride or ammonium persulfate. Other folks use vinegar and salt, or dragon spit and accountants’ tears. Some people agitate liquids in a tank, others spray, and still others sponge. Whatever works.
These things are all absolutely vital to getting a good toner-transfer PCB made, of course, but none of them are uniquely irreplaceable.
On the other hand, there are three fundamental factors that matter for the adhesion of the toner to copper, and it’s down to physics: time, temperature, and pressure. It’s a huge step towards reliability and repeatability to reduce these three down to one, and that’s our first step.
As far as pressure goes, more is almost always better. In fact, there are some laser copiers that work without heated rollers at all. They squish the plastic toner particles so hard that they adhere to the paper at room temperature or thereabouts. We’re not going to be able to apply that much pressure, but we’re going to fix the pressure variable at “as much as possible”.
I put the PCB on my inverted iron and press down with all of my weight on a rolling pin. The resulting pressure is fairly high because the rolling pin has a small contact patch with the board, and it’s fairly constant because I only weigh so much.
I took my inspiration from this method which does essentially the same thing upside-down. My experience is that the board never rolls evenly across the dowel using this procedure, but I can understand if you don’t want the exposed electrical horror that is my setup in your house. Any strong hot surface should work, and an improved setup would have better temperature control. If you have a modified laminator that puts out enough pressure, that’s probably even better. I have a scrap clothes iron.
Unlike pressure where more is better, the effect of longer dwell times drops off rapidly. In laser printers, where the number of pages per minute is a critical selling point, they try to get this dwell time down to the minimum. If you roll too quickly over the board, adhesion can be uneven, so the solution is to simply go slowly and roll back and forth. You can remove time from the equation by simply rolling so slowly that any decrease in rolling speed doesn’t substantively change anything. A minute or two should do.
All of this is relative, of course. You don’t use the same toner as I do, weigh the same as I do, or even have the same idea of “slow” that I do. But as long as you keep your own practice consistent, these variables won’t affect the outcome very much, and we’re free to tackle the variable that will: temperature is the secret sauce. Here comes science!
When toner is heated up, it goes through a few distinct phases. At first it’s a hard plastic, then it gets sticky and slightly malleable as it passes through the glass transition, and then as the temperature increases even further it melts and becomes a liquid.
I have no idea how the Internet rumor that you should turn up your iron “as hot as possible” came about. But I do know what the result of doing so is — smeary transfers that are extremely sensitive to the amount of pressure applied. That’s the exact opposite of what we’re looking for here. Instead, the goal is to keep the toner in the glass transition, with the temperature as low as possible so that it will fuse with the copper at our roughly constant maximum pressure.
The first step of calibrating the procedure, then, is to run a series of transfers at increasing temperature. Your temperatures won’t be the same as mine, but that’s fine because you’re not using my toner or making PCBs over at my house. The trick is being consistent.Lower temp on top, higher on bottom 98 degrees 125 degrees 142 degrees 166 degrees
The four examples here were transferred at 98 °C, 125 °C, 142 °C, and 166 °C. The first clearly doesn’t adhere well at all, and in fact none of the traces except the 16 mil trace are actually continuous after etching. The highest temperature board has short-circuits all over, despite using 20 mil clearance on the ground pour. Surprisingly, the TSSOP footprint worked out basically OK, even though the two traces at 6 mil spacing are shorted.
The problem with the extreme temperatures are also obvious from looking at the transfer sheets. In the case of the low-temperature board, there is a lot of toner left on the paper. It simply hasn’t stuck well enough to the copper. The high temperature sheet ripped when pulling it off, probably because its plastic layer melted into the toner.
The two middle boards look pretty good, but looks can be deceiving so it’s probably worth etching them. There’s actually a minor defect in the 125 °C board — the 6 mil trace sample is broken just at the bottom where it connects to the pad. On the 142 °C board, the two close traces passing through the 0805 part are shorted although all of the other traces are continuous. A temperature somewhere between these two is probably going to be the sweet spot.
I re-did a run of this test board at 132 °C, printed onto magazine paper, and it came out perfectly. Except for the minor hassle of cutting out the magazine stock and taping it to a sheet of regular paper so that it can feed correctly through the printer, magazine paper is probably the ideal substrate. When it dissolves in water, there’s very little stress on the toner and a much lower risk of pulling up a spot here or there.
Not coincidentally, I’ve been using a temperature around 130 °C since I last did this calibration myself after buying my current laser printer. Getting a consistent procedure with a dialed-in temperature is much more than half the battle.
At this point, with the toner transfer process itself pretty much dialed in, the limiting factor in how reliably and at what resolution you can print is going to be the printer itself and the software that drives it. I found surprisingly large differences between different printer drivers (PCL and Postscript) as well as between different file formats used to save the artwork, and the programs that read them.
In particular, both of the drivers on my system seem to be interpreting the Postscript and PDF output of KiCad as being color files, and applying dithering to the result. With the Postscript driver, this results in jagged edges, an effect that other people have noticed before. With the transfer working well, these jaggies end up in the PCB’s copper.
With the PCL driver that I had been using, it seems to be applying a much higher-resolution dithering algorithm, with the result that the lines appear thinner to the point that they have discontinuities. It turns out that the problems I was having with thin lines printing was caused by the driver.
Finally, when I save the file as an SVG graphic and print it out from within Inkscape, all of the color-dithering artifacts go away, but the result is nice smooth dark traces that are a tiny bit too thick once transferred. Perhaps with these improved traces I can lower the temperature a little bit? There’s also some dust in the transfer on the sticky-backed paper that doesn’t occur with magazine paper, so maybe paper type does matter a little bit after all.
It’s important to note that you can’t really diagnose fine details like these until you’ve gotten the main variables of temperature and pressure under control. But once you do, and the smallest glitches that arise from the software stage of the chain are visible and reproducible, you’ve got a new limiting factor, and it’s time to tune up the software.
Unfortunately, everyone has a different software and printer setup, and not all the tweaks that work for me will work for you. At this point, you’re on your own. Just know that the printer’s resolution, printer configuration options, and even the drivers and file types can matter. That’s a lot to experiment with, so take them one variable at a time.
So everything interacts: temperature, pressure, paper type, and even the file format and printer drivers. I gave up experimenting for today because I’m content with reliable 6/8 traces, and when I need to push it even further for some project I’ll probably start by decreasing the temperature further to see if that solves the thickening lines. If I really need 6 mil clearances, I’ll either double-check those locations before etching and clean them up with a scalpel blade, or I’ll experiment with 4 mil trace widths. After all, with the SVG output and Inkscape doing the printing, I’ve not had a single discontinuous trace at 6 mil.
One thing I absolutely won’t change, however, is the basic technique. I first started applying a consistent maximum pressure through a roller and tweaking the temperature about ten years ago, and it’s been a very reliable method of making PCBs since. This recent quest for maximum resolution has been a fun diversion and I’m both happy to know that I can do small features when I need to, and saddened that the paths to further improvement seem to lead through tweaking software and drivers because there are just so many options.
At some point, it does become easier to let the pros do the work for you. But for me, toner transfer works for almost all of the initial stages of prototyping, and with turnaround times that absolutely can’t be beat.