Warning: This fork is a work in progress to replace the deprecated MAG3110 chip with the ST-LIS3MDL. Do not use, it is not ready.
Replacement software will be available on Raphaël's repo
Pluto is a programmable digital watch that re-uses case and LCD panel of the Casio® F-91W. This is the hardware repo, for the software side of things, see pluto-fw. Looking for pictures? There you go. You can discuss the project and ask questions on the EEVblog forum
- Displays time in decimal/binary/hexadecimal base
- Multiple alarms
- Multiple countdown timers
- Uses RTTTL ringtones for alarm sound
- Stopwatch
- Compass
- Speedometer (Enter distance, watch measures time and displays speed in km/h)
- Generation of time-based one-time passwords according to RFC 6238 (WIP)
- Menu-driven interface
- Infrared receiver for software updates and TOTP secret transfer (WIP)
- Useless customisation (Key beep frequency, etc.)
- greater than 1 year battery life (estimate based on current consumption)
Pluto replaces the PCB of the Casio F-91W with a custom one containing a MSP430FR6972 MCU, a digital magnetometer (compass) and a 38 kHz infrared receiver for software updates and communication. Due to careful planning and luck, the Pluto PCB perfectly fits inside of the F-91W's mounting frame, although it houses many more components. Schematic as PDF BOM BOM with Mouser part numbers
The MSP430FR6972 is a low-power 16bit MCU with integrated segment LCD controller. It can be programmed and debugged using the two-wire SBW interface accessible through holes in the mounting frame.
The MAG3110LIS3MDL compass is connected to the MCU via I²C. To reduce its power
consumption to zero when not in use, it's powered from GPIO pin. Due to
layout constraints, the MCU's I²C interface can't be used, so the
MCU's firmware has to implement a bit-banged I²C master controller.
The TSOP57338 38kHz IR receiver is powered from GPIO pin as well since it doesn't have any power saving modes. Transmitter and firmware support are WIP.
The backlight LED is connected to two GPIO pins, so that LEDs of either orientation can be fitted. The driving circuit for the buzzer uses an inductor (measured approx. 10mH) to boost the battery voltage for increased volume.
We manufactured the PCB at elecrow. PCB thickness has to be 0.6mm for the PCB to fit between battery clip and plastic frame. Note that the PCB has "castellated vias" (for contacting the button springs) as PCB manufacturers usually surcharge for these. We chose to use ENIG (gold) finish since the original PCB is gold-plated as well and Casio wouldn't have done so if this wasn't strictly necessary. Unfortunately, not all vias are covered with solder mask due manufacturing details (in the gerber files, they are), so you'll have to insulate the bottom side of the PCB from the battery with sticky tape.
For details on the software, see the software repository at pluto-fw.
I used to own an ez430-Chronos watch, but after several years of use, the rubber wrist strap disintegrated, so I needed a new programmable watch. Pretty much all of the existing DIY watches are impractical for daily use because of bulk, being not waterproof or having miserable battery life.
Q: Why not just use a smartwatch?
A: Smartwatches have a miserable battery life of one week at best. Some
even require user interaction to display the time.
Q: Why isn't there some sort of radio interface? (BLE, nRF24L01)
A: Take look at this picture to get an idea of the space available
inside of the F-91W frame. The nRF24L01 is 4 mm×4 mm and barely fits, but
requires external components like balun and antenna. There's no space
for these. An IR receiver is the same size as the nRF24L01, but doesn't
require any external components. TI's CC430 MCUs seem like a perfect
match for pluto as since they combine MCU and radio, but since the radio operates at
sub-1GHz frequency, the antenna needs to be prohibitively large.
Q: Why is there no accelerometer?
A: Accelerometers are the most useful when they're permanently active
for applications like sleep trackers, pedometers or features like
"shake to do something". Unfortunately, even accelerometers marketed as
"ultralow power" draw approx 30 µA when running, draining the CR2016
coin cell in less than half a year. So we decided to dedicate the
available space to a digital compass.
Thanks go to sh-ow for adding the compass and helping to get the board ready for manufacturing.
Stuff loosely related to this project
For basics on bare glass LCD panels and how to drive them, see application notes from various MCU vendors.
For driving an unknown LCD panel, one needs to know which pins connect to the backplanes (COM) and which ones connect to segments (SEG). The easiest way to do so is to apply a square to all pins of the LCD panel and invert one pin at a time. When few segments light up, you've got a SEG pin, when many light up, you've got a COM pin. Sometimes apparently independent LCD segments are driven as one segment to save SEG pins.
The F91-W's original PCB has a rather complicated outline that needs to be duplicated with high precision, so that the Pluto PCB fits nicely within the existing mounting frame.
By desoldering the spring contacting the piezo buzzer, the PCB is completely flat on the bottom side, so it can be placed on a flat-bed scanner. This eliminates any kind of distortion you'd get by taking a photo. The scanned image can then be scaled to match the actual dimensions measured with a caliper. There may be unequal scaling in X and Y direction due to inaccuracies of the scanner. We've now got a pretty accurate digital representation of the PCB that can be used for tracing out the outline and registering photos from the top side of the PCB to get the position of pads on top side.
Inkscape and GIMP (for correcting camera distortion) work well for these tasks. The resulting outline has been exported as a dxf document for importing the outline in KiCad.
The project started off with the intent of re-purposing the the F-91W case and LCD. Unfortunately, the LCD panel of the F-91W is rather limited since some LCD segments are driven as one segment.
So we went out and took a look at the Casio W800, F-201 and DB-36 watch. After manufacturing a test jig for the LCD panels of the former two, it turned out that people at Casio /really/ like building LCDs with 5 backplanes. For some reason, the LCD panels need also to be driven with around 5 Vpp to achieve decent contrast. (Best guess: These watches are old designs, LCDs weren't that sensitive back then)
To my best knowledge, there's no (non-unobtainium) MCU on the market, that can boost a 3 V supply to 5 V for driving and LCDs /and/ can do 5-mux. The only MCUs meeting the first requirement are from the RL78 family from Renesas, but they only do 1,2,3,4 and 8-mux. Driving a 5-mux LCD with 8-mux waveforms leads to bad contrast. With 3 V and 5-mux, the contrast is much worse.
EPSON offers some obscure MCUs that can do both 5-mux and 5 V LCD drive, but these ones don't have enough segment pins the QFN64 package.
So we went back to F-91W.
For plotting current consumption over time, the EnergyTrace feature of the MSP430FR4133 LaunchPad has proven to be really useful: energytrace-util
Looking for more buttons and a radio? Take a look at goodwatch