A printed circuit board
is a heart of modern electronics. There are many ways
to make one at home - photoresist, laser engraved resist, manual scribing, CNC milling,
and others... and all of them suck in one or another way, mostly by requiring costly
precursors, machines, or being finicky and prone to process variables.
A hopeful alternative method seems to be electrical discharge machining.
This method applies an electrical spark to the material surface, while immersed in a
dielectric liquid. The spark vaporizes some of the metal; the ejected particles are then
washed away by the liquid.
There are two main flavors of the method used in industry - die sink EDM, and wire EDM. The former uses a die that is gradually sunk into the workpiece; the latter pulls a cutting wire through the workpiece, yielding a linear cut. Even large chunks of hard materials can be cut this way. Five-axis cutting machines are commonly used.
A circuitboard is a sheet of thin metal (usually 35 µm) on a plastic backing. There are several flavors of the material, based on different polymers and their thickness, with different mechanical and electrical properties. For the purpose of EDM machining, only the copper layer plays a significant role.
Copper is usually not too suitable for EDM machining, due to its high thermal and electrical conductivity. When it is strongly planar, the bulk-material effects are somewhat attenuated, as the heat can be conducted away in only two directions.
First ad-hoc tests were done, using a small capacitor and a lab power supply.
The workpiece was connected to the positive end and the electrode/tool to the negative end. This ensures that the tool emits electrons, which cause higher heating of the workpiece and ensures less burning of the tool.
A 1000µF/35V aluminium electrolytic capacitor was used, fed from a current-limited lab power supply. The power supply was set to 20V for first tests, later dialed down to 14V. This equals spark energy of 200, resp. 100, millijoules.
For first tests, a thin (0.15mm?) springy brass wire was used. This wire is commonly used for wire EDM, here only its very end was employed. Alligator clips were used to hold both the board and the wire. For later tests, a yet thinner wire, a single strand from a stranded wire, was tested; this one worked well and left a thinner trace but was difficult to hold in the clip.
First sparks were made in air, at 200 mJ. They were large and noisy and sputtering particles around. Black "soot", probably minute metal oxide particles, was deposited around the spark sites. The edges of the sparked off sites were uneven and raised, showing molten copper.
Second phase used tap water as the working fluid. The sparks behaved way better; they were smaller and more focused towards the material, the ejected material was quenched faster, and the black goo was not depositing but was washed away. Electrolysis however occured at rather massive scale, quickly depositing a layer of insulating copper oxides and hydroxides all over the board and making it impossible to strike the spark. The deposit was removed by sandpaper. Distilled water may reduce the electrolysis, together with coating most of the electrode except its tip with insulating material. Contamination of the distilled water with dissolved copper salts may increase its conductivity and cause problems to reappear.
Third phase used mineral oil. A drop of machining oil was placed on the circuitboard, and the electrode was immersed in it. The spark behaved nicely, the electrode sticking was minimal though still present. The fine black goo was washed away immediately (and was hovering in the fluid due to lack of its circulation). In the thin layer the spark ejected oil mist to the air above, which caused contamination of the workspace and the stereomicroscope lenses (oops); thicker layer of oil may eliminate this effect.
In oil, a thinner wire was tested, together with lower voltage. The traces achieved were significantly thinner, with significantly less jagged edges. The wire however was only poorly held in the alligator clip tips, which hindered experimentation. The spark was also more difficult to ignite, possibly due to the lower voltage used.
The results are overall encouraging. Under oil, at lower spark energies, the cuts are even and with fairly straight edges. An individual spark can remove a circular area of about 100-200µm diameter; stacking such areas into lines and areas can create nonconductive traces in the copper-clad surface.
Some contamination was found to be sticking to the exposed insulator layer. This had a form of transparent gummy residue, sometimes black-discolored probably due to trapped copper oxide particles. It was easily scratched off with a tip of tweezers. Sandpaper or some solvent in ultrasound may be useful for its removal.
The spark with excessive energy may cause excessive damage to the surface of the underlying polymer. For applications where extremely low leakage currents or very high creep discharge resistance are required, such damage could cause problems; make sure this is not the case. The material behavior will be dependent on the resin used for making the composite; good or bad results with one won't directly translate to another material.
A SMD chip is used as size reference. Its pitch is 0.5mm (300µm between the pads, 200µm wide pads).
A stereomicroscope at 40x magnification was used to take the photos; a camera was hand-held to the eyepiece. A jig has to be 3d-printed.
More images have to be taken for subsequent tests.
 Test board, on 5x5mm raster |  Test board |  Detail |  Detail |
 Detail |  Detail, in air |  Detail, in air |  Detail |
 Detail, through-light, contamination of the groove visible |  Detail, contamination partly scratched off |  Detail, contamination partly scratched off |  Detail |
 Detail |  Detail, a 1mm scale on top |  Detail, too much energy used in air |  Detail |
 Detail |  Detail, just about right |  Detail, just about right |  Detail, just about right |
 Detail, just about right |
For circuitboard "etching", the head is a hybrid of a wire cutting and die sinking; the die is the end of the wire, sunk into the thin copper layer. Only a single spark is often needed for full-thickness removal of the copper; excessive energy actually creates a larger than needed crater, with ugly edges. This significantly simplifies the design.
As electrical connection of the layer is necessary for the operation, the circuitboard has to be processed in a raster fashion. Some aberrations are possible but the sheet of copper between the connector and the tool has to be intact; a closed curve would insulate a copper island and the sparking would then not be possible within.
A solenoid can be used for vibrating the electrode tip, to facilitate interrupted touch between it and the copper sheet without having to move the whole print head intermittently up and down, which could speed up the process. A slightly modified coil mechanism from a relay can be used with ease.
The solenoid can be also in all-down position, and its retraction triggered by the spark. At the same time it can act as a z-stop switch signal for the printer. The line-drawing sequence then would be "go down until z-stop triggers, then go slowly to the second end of the line, then go up by a millimeter". This will automatically compensate for both the uneven surface of the board and the tool tip burning away.
The spark-happened signal could be used for returning the solenoid to top position again. This can also be used as a zero-level reference for the 3d printer controller, so automated calibration for warped or otherwise uneven boards and their clamping mechanism would be possible and easy.
The jagged edges of the traces can be minimized by careful spark energy and distance settings. This may be important for high voltage and/or microwave frequencies, where the smoothness of the trace edges may play a role. For these, this technology may or may not be usable.
For bulk machining, where there's thicker metal layer to cope with that requires more than a single spark to get rid of, more sophisticated approach with fine z-axis head control is needed.
Some published designs of die-sink EDM machines use a capacitor-voltage feedback. A window is specified; when the voltage is below it, the die moves away from the workpiece, when the voltage is higher, the die moves in. This allows for sustained inward movement of the die; handy when e.g. removing a broken tap or a drill bit. If the reaction speed and the resolution of a 3d printer mechanics are sufficient, this can be implemented in software; otherwise, a hardware z-axis co-actuator (e.g. a solenoid or a voice coil) can be used that finely adjusts the die position in a realtime control loop in a small distance range, like the coarse-fine adjustments of the lens in a CDROM drive. The printer then can be moved in position coarsely, the coactuator lifts the die above the desired position, and slowly lets it move down. When the desired position is reached, the coarse actuators skip to next position and the process repeats.
It may be possible to spark complex shapes from thin metal foils; this could be of use as an alternative to photoetching of detailed parts for modelling of aircraft, ships, etc.
Holes may be possible to cut to thicker sheetmetal. This would be useful for machining instrument panels.
It may even be able to bulk-machine larger pieces of metal, with adequate power supply and patience. A pump may be needed for circulation of the fluid and forcing it under the tool to wash out the debris. Periodical lifting of the tool in peck-drill style may or may not be sufficient here, though.
Difficult-to-make internal structures, e.g. gun barrel rifling or grooves in wheels for groove-tongue assemblies, could be also achievable; the former would require a rotating plunger to achieve the twist needed. Ad-hoc tooling jigs with one or more stepper motors may have to be built for such advanced applications.