After building the Nixie Tube it got a bit of attention from family and friends, particularly my sister. Fortunately her birthday was coming up and it made this year easy for shopping, but I wanted to try something different. Numitrons.
The numitron is the short lived successor to the seven segment display, quickly replaced by LEDs. Having 7 segments gave it advantages over the nixie tube as there were a wider range of patterns the tube could display, and where the nixie tubes needed one cathode per filament making the back most digits harder to read, the numitron was a true seven segment display. Nixie tubes relied on gas discharge and required very high voltages and low current, whereas numitrons worked via filament heating, typically running as low as 5VDC but with several mA per filament making them quite power hungry when displaying the number 8.
One of the more common numitrons, the IV-9 is also about half the size of the common IN-8 nixie tube.
The drafted schematic was unfortunately lost in a computer crash. All that remains is a hand sketch based on a PCB which was mid-production during the crash. The schematic was recreated from the negative image etched into the PCB. The scan is attached below.
The schematic is far simpler than that of the nixie clock due to the lack of high voltage powersupplies and switching circuits. The circuit takes a 5VDC regulated input so no PSU is needed on board, the ATMEGA88 is the same microcontroller and the clock source is also a DS1344 and 32.768kHz crystal hooked up via the two wire interface. What’s new is the method of driving the display.
The TLC59025 is a 16 channel constant current sink LED driver. The constant current sinks allow each filament to be driven independently. It takes a serial input into overflowing buffer. Two are used to drive the 21 signals needed for all clock digits. The drivers have their enable and clock lines hooked up in parallel while their serial data lines are in series.
The case is connected to a pin on the microcontroller allowing for a clever software based touch system.
The software is almost identical to that of the nixie clock so for details on the interface to the DS1344, state machine, or the fancy touch sensor check out the nixie clock page. The only difference for this clock is in the way the output is calculated.
The TLC59025 are essentially serial to parallel converters utilising 16bit shift registers. Any overflow of data coming in via the SDI pin is pushed out via SDO. This allows us to drive both TLC59025s and thus all 4 numitrons at once simply by writing 32bits to the serial line before pulsing the enable line of both chips. The output is bit-banged via PortB via a simple process of shifting the correct bit to PB0 and then cycling PB1 on and off. Once 32bits have been written PC0 is pulsed to latch the new data from the shift registers and the numitrons light up accordingly.
To simplify the routing of the numitron PCB which was a perfboard, the software makes use of a lookup table. Before writing the output to the TLC59025 the software compares the current digit against a table of hex values based on the current numitron it is trying to drive. The 8 bit number from the lookup table is then written out using logical shifts via PortB and the next numitron and lookup table is then consulted for new data. This ensured that regardless of which segment of each numitron was connected to which pin, providing all the pins of the numitron were connected to the same side of the TLC59025 (bits 0-7 or 8-15) the correct digit could be derived in software.
The numitron clock works wonderfully as expected. The IV-9 was very difficult to solder onto the PCB due to its tiny size but as the main clock body is smaller it looks proportionally quite good. One problem that did arise is when the clock is set in calibration mode for the touch sensor the current touch output is displayed on the left two digits and these change quite quickly. Numitrons turned out to be quite slow at updating their value so the rapidly changing numbers are completely unreadable. This is not a problem for normal clock operation.