The most significant risk to this team’s success is getting the firmware done in time for the demo. This risk is managed by the fact that we are planning to spend the bulk of our time this week on Capstone. Contingency plan is to get as much work done by the end of Thursday, and decide what to include in the demo on Friday after that.
Progress is behind schedule, but since there are no classes this week, we have extra time to spend on integrating all of the firmware and testing it before the demo.
There are a couple of small design changes we decided to make this week, such as only including battery statuses instead of fine percentage levels. This is due to the variability, and therefore unreliability, in voltage as Remigotchi is being used.
Below is a state diagram of our firmware.
Unit Tests
- GNSS test: Ran firmware on Remigotchi to interface with its GNSS module. Measured how long it typically takes to get a verified time. We found that we could only reliably get a verified time when outside, or sometimes by a window.
- RTC test: Synchronized the RTC with the GNSS time once, put the ESP32 to sleep, woke it up several hours later and measured the time difference. We found that over the course of roughly 8 hours, the ESP32’s internal RTC was roughly 2 min, 15 sec ahead. Given this, we are considering synchronizing to GNSS time more frequently than once a week.
- Power tests: Enabled/disabled various combinations of peripherals connected to the ESP32 and measure how much power the device draws overall. We found that having the GNSS module in use vs not in use didn’t have much of an impact. However, the module itself is embedded onto the board, and its power lines are hard-wired to the board’s main power. This means that even if the GNSS module is not in use, it still does have some amount of power going to it. Should this project be worked on in the future, we would likely have some sort of way to dynamically control access to power for various devices.
- ADC battery monitor: Read out millivolts measured on the ESP32’s ADC pin, and compared it to what I measured with a multimeter. It was accurate to about 0.01 volts. However, we concluded that voltage is not typically a good indicator of overall battery charge level due to the fact that voltage varies a bit during usage, and the lipo discharge curve is very non-linear. Therefore, we decided to not include a precise battery percentage readout visible to users, but instead, opt for battery indicators when the device needs to be charged or when it is fully charged.
- Bluetooth connection time: We measured how long it took for the desktop app to connect to Remigotchi at varying distances between the two devices. Average is around 2 seconds.
- Bluetooth data upload time: We measured how long it took to send a user’s schedule from the desktop app to Remigotchi, once it’s connected. Typical schedule sizes (under 10 reminders/week) result in send times of less than 1 second.