The new schematic:
I have made a few changes: I now use a Li-Ion button cell instead of the lithium cell CR2032, which simplifies the circuit. Since >= 3 V are available, one step-up converter can be omitted. LT3494, a low power boost converter, provides 23 V, which is increased to 46 V (without load) by the downstream voltage doubler. for driving the droplets. U5, MAX6920, is a serial interfaced VFD tube driver. Six of its twelve CMOS push-pull switches are connected to the six electrode connections of the display. U5 can be easily controlled by a SPI interface. U3 is the Attiny1614 microcontroller, that is most of the time in deep sleep and only woken up by user demand or by U1, DS3231, a highly accurate real time clock (once every hour, minute, (second)). U1 communicates with the microcontroller over its I2C data bus. Before going to sleep, the microcontroller shuts down U4, U5 and partly U1 (I2C bus). In this mode, U1 will draw ~ 2 µA and the Attiny in sleep 0.1 µA.
Moving a droplet:
every push-pull switch of U5 can be considered as a half bridge. Two of them build a H-bridge, that can be driven either in phase or phase inverted. As the outputs of U5 are pulsed with 70 Hz, in the first case the voltage between the two outputs is 0, in the second case it is a square wave AC voltage. The droplet moves to the electrodes with the electric alternating field between them.
To inhibit the movement of a droplet, its corresponding counter electrode has to be grounded.
The (previous version) PCB:
The PCB shown above, is the first, I have made – ever. My requirements were:
- avoid acquisition of new devices (etcher, exposure machine...)
- avoid making exposure drafts
- avoid chemical processes if possible
my first approach was isolation routing with the laser. Although possible, I soon discarded this idea. The isolation trenches are to narrow (only ~ 5 µm) because the laser has to be focused. And even more important, the vaporised copper contaminates the optics.
My second approach was surprisingly successful:
The layout is made with KiCad. It was a bit of a hurdle at the beginning, to get along with the libraries. After finishing the design, the next step is to enlarge all pads by 0.2 mm. This doesn't take too much time because KiCad allows to change pads of same shape at once.
For the solder mask clearance I chose -0.01 mm.
Then the copper layer and solder mask are plotted as .dxf files. The copper layer .dxf file is opened in LibreCAD, where some minor corrections can be made and the solder mask .dxf file is imported as a block. Both drawings are aligned exactly about each other and saved as a new .dxf file. This file is finally imported into LaserCad, which allows to process each layer separately by the laser. (Picture below)
To make things easier, I decided to abandon through-connections and make a few connections by wires. On the bottom side of the PCB, there is only the CR2032 cell.
The double sided base material is now hot laminated on both sides with the same foil as used for the display's dielectric and thermally cured at 140 °C. After removing the thermoplastic layers with acetone, each coating has a thickness of 4 µm. The laser is defocused, to increase the spot size on the working surface from ~5 µm to ~70 µm. The laser only ablates the coating around the traces to create an etching mask. After etching for 20 minutes with iron trichloride, the PCB is processed a second time on the laser. This time, the clearances over the pads are ablated in engraving mode, processing the solder mask layer. The TSSOP has a pitch of 0.65 mm.
The coating has two tasks: it serves as an etching mask and as the solder mask.
The PCB now is ready for hot air levelling as a surface finish. I use a zinc paste used for pipe fittings from the construction market and my hot air gun at 250 °C.
A variation of the process is possible: After etching and before lasering the clearances, the PCB could be coated a second time with the foil, to cover the isolation trenches and increase the thickness to 8 µm.