Skip navigation

Tag Archives: cnc

Ah, the bliss of populating THT circuits on manufactured PCBs… just follow the silk-screen and you’ll be fine. in contrast, the fun-factor of soldering perfboard is way lower. your eyes move from your plan (if you happen to have one) to the board and back, your brain is trying to mirror the layout while you flip the board back and forth all the time. while this might be OK for a single board, what if you have to assemble two? or three? or even more?

for my latest project, i have to populate 10 identical boards. why not have a PCB manufactured, you may ask. well for different reasons: first, it’s too expensive, because the boards are rather large. secondly, i have to get rid of a larger amount of perfboard. third, the layout is not complicated enough to justify a manufactured board.

anyway, in a former similar project, i used a CNC mill to plot the traces onto the board:

plotted pcbs

this was extremely helpful, but setting up the files and the mill took some time. now in the recent project, i experimented with toner transfer to not only put the layout on a perfboard but also a silkscreen on the top. the result is a board that is as easy and flawlessly to assemble as your first beginners electronics kit back in the day and it is neither complicated nor time comsuming:


although this is supposed to be some kind of tutorial, i’m not going into details about the toner transfer process, because this is described in other tutorials.
the base of the process is to have a board layout. i normally create layouts even for my perfboard circuits in Cadsoft Eagle. if you set the grid to 0.1 inch and stick to 90 degree angles, you’ll be fine. after that, you usually print out the layout and use it as a template while soldering.
but soldering is still pain in the rear. so why not spend a few minutes with a laser printer, an iron and a sip of water to create a professional perfboard™?

first, i CNC’ed out the boards, because they have a rather odd shape.


if you have no access to a CNC, just use a hacksaw, jigsaw, scroll saw, chainsaw, dynamite, whatever. if you cut your boards manually, you maybe want to cut it after the silkscreening to have an outline for cutting.

for the toner transfer, laserprint the layout onto a piece of catalog or anything else that does not absorb the toner. you may want to stick to one of the toner transfer tutorials i linked earlier for details.
now it gets a little counter-intuitive: keep in mind that during ironing, the layout gets mirrored. that’s why you should only mirror the TOP layer before printing, not the bottom layer. it’s a good idea to print out of Eagle directly. printing into a PDF first is likely to result in unwanted scaling of the layout.




this is a little tricky: you have to carefully align the perfboard grid to the layout grid. i made good experiences with aligning towards a window or a ceiling light, then holding both layout and board with two fingers, carefully putting it on a table upside down and carefully taping it into place. did it mention that you should be very careful? before ironing, check the alignment again.





after ironing, let the board cool down and put in into a bowl of water. peeling/rubbing off the paper is easy if done underwater (only the board and your hands need to be under water). if you decided to do layout AND silkscreen, just repeat the whole process, but be sure not to iron the second layer too long, because you will re-liquify the toner on the other side and the board will stick to the surface.
satisfied? then go solder your professional perfboard™.




summary of things to keep in mind:

* print directly out of Eagle/yourLayoutSoftware
* mirror the top layer, print bottom layer normal
* carefully align layout and perfboard grid
* check alignment before ironing
* iron second layer only half a minute or so

if you have any questions or ideas to improve the process, please do comment!

UPDATE: here’s timelapse of soldering the board. it took about one hour and the video is speed up 10x:

So i had these three servos and thought about how to use them in a project.  i saw flexpicker robots on the internets which were really impressive. had to have one. found six ball joints in my drawer and started to do some cad. i needed twelve ball joints or at least something comparable, so i had to make six from scratch. a few days of cnc-milling and drilling later i had my delta-robot. now it’s time to find a purpose for this thing. i thought about attaching some kind of head to it with an integrated display. i plan to control robot and display via rs232 and a microcontroller. so there’s more to come.


delta robot in uncentered position


delta robot closeup

Some time ago i got into rubik’s cubes by accident. after a while i thought it would be nice to solve the cube using a different perception than the visual. i decided that i wanted something haptic.  so i went and designed some patterns that were

  • easily distinguishable
  • systematic
  • plain
  • easy to make
  • rotation invariant


    the result were three different basic shapes: cross, box, hollow box

    Cube Patterns

    “why no circle?”. well i decided to stick to the concept of straight lines. but i added some roundings just for a more comfortable feeling. anyway, it’s three basic patterns in a normal and diagonal version each since the colors on a rubik’s cube are also paired. we have [red/orange], warm colors on opposite sides of the cube, [white/yellow], bright colors and [blue/green], cold colors.  i wanted to keep that concept in order to have a better orientation when solving the cube. so each [normal/diagonal]-pair is located on opposite sides of the cube.

    finally the plates were cnc-milled out of  pcv and it was a really painful amount of work to clean them with steel wool.  it took a while till i got my fingerprints back. i peeled off the original labels of the rubik’s cube and mounted the plates using double-sided tape.

    54 plates later the cube was finished. the first days the cube felt a little uncomfortable, but it became better because the edges wore off. solving the cube takes me about five times longer than solving a normal cube, but it’s an interesting challenge every time.

    here’s a picture of the finished cube:

    finished cube

    %d bloggers like this: