RPN-duino

[Chasfield]

Attached are a couple of pictures of my latest calculator build. I had been looking around for a better platform for these projects after taking the regular 8 bit Arduino as far as I wanted. The ARM-based Due model used in my last lunch-box sized calculator was great to program and it handles double precision floating point maths effortlessly but the PCB is physically quite large and that project's power consumption is relatively high at 50+ mA. After convincing myself that The PJRC Teensy 3.2 would do 64 bit maths out of the box, I laid out 20 GBP to acquire one. The specifications are just right: Cortex M4, 64K RAM, 256K flash ROM and it is powerable from a single 3.7V LiPo cell. The unit is smaller still than an Arduino Nano. The resulting calculator is almost Casio sized!

The picture of the underside shows the layout of the components and this time the construction is boxless, with the siamesed 16 button keypads forming the top deck. Side rails are 8x8mm acrylic bar and the lower case cover is cut from 3mm acrylic sheet. The Teensy 3.2 is soldered directly onto the lower keyboard's header pins, which takes care of most of the wiring. Keyboard column pins are commoned via ribbon cable. A second ribbon cable connects the row pins of the top keyboard to the Teensy. Charge control of the LiPo cell is taken care of by one of those ubiquitous little controller PCBs that seem to cost only their postage on eBay. The 1.3 inch OLED display connects to the designatd I2C pins on the Teensy (these displays are getting cheaper too - 6 GBP).

The display is driven with the u8g2 display library and the display code from my Due project only required minimal tweaking to work on the Teensy. The standard Arduino IDE accepts Teensy extensions that are downloadable from the seller's web site. For minimal power consumption, the device is throttled down to 48 MHz clock speed and the calculator draws around 35mA from the 800mAHr LiPo. It took me a while to shake out the full 64 bits from the math.h library. By default, a compiler flag (-fsingle-precision-constant) is set by the IDE that restricts declared constants to 32 bit float precision even when they are defined as 'double'. Once this was edited out of the Teensy 3.1/3.2 section of the IDE's boards.txt file, the math.h library worked properly and the calculator delivered full precision.