Apparently the OSHPark order is already shipped but there was no notification. Apologies to those waiting.
OK I've more or less done Rev3 PCB layout, which has a few changes from Rev1 and 2:
- Surface-mount-hack mounting of edge connector, allowing board to lie flat and be used with 3278s etc.
- Move to SPI-protocol DAC for Vref (trigger voltage); this in theory means that Vref can be changed per column or even key if necessary
- Removal of pushbuttons for Program and Reset. There are test pads instead.
- Larger HC49-style crystal instead of the special tiny thing that was used before. Saves some cost.
- Solder-mounted chassis ground connection instead of massive through-hole hole
- Last but not least---15% reduction in board size. Makes it a bit cheaper.
I really hope this will fit inside a 3278 (remember the edge connector will be soldered at 90 degrees to the board):
- 2013-07-30-170158_1679x1030_scrot.png (25.48 KiB) Viewed 11090 times
The DAC for Vref is interesting. I2C (that was used with the last DAC) is very slow and thus precludes changing Vref on the fly (if there was a dodgy column etc.). SPI is way faster (with 16MHz AVR the multipliers work out so we can communicate at 8MHz, instead of 400KHz max for I2C, although it was probably about 100-200KHz as we were bit-banging).
Unfortunately SPI DACs are more expensive, so instead of using a 12-bit DAC I went with a 10-bit. In theory this means we have a quarter of the resolution with Vref, but I'm using a three-resistor window divider to output between 0.8V and 1.6V (with 1024 possible values, so that's 0.78mV resolution) instead of before where it was outputting the full range between 0V and 5V (with 4096 possible values, so 1.22mV resolution).
The pushbuttons have gone as they are relatively expensive, take up a lot of board space (and we only really have one side now that the board is mounting `on-flat'), and are mostly useless as the controller can be reset into the bootloader for updating the firmware using the GUI util. I've left test pads on there so it can be manually reset by shorting out the pads with a screwdriver or whatever.
After I order this and make a few the next project is a Displaywriter converter---those are quite different from the other beamsprings, having 8 rows. They seem similar to a Model F controller in some respects.
(I seem to recall it needs some odd voltage, but one can dream...)
