It took some work to get the factory program included with the LilyGo T-Display S3 sample code to get running without the need for their specialized phone app, but I made it work. Imagine my disappointment when I found their flip clock did not actually "flip" - it just re-painted one digit on top of another.
So I decided to write my own version of a flip clock with some simple animation of the digits rolling over:
.......
I wrote a Python program (detailed below) to generate the images I needed. This consisted of each whole digit, plus a series of three additional images with the first digit tilted at 45, 90 and 135 degrees overlaid on the second digit being "flipped" to. These images get stored on the device's flash memory using the LittleFS file system.
At run-time, pressing the top/right-of-USB button will toggle the display between 24-hour format, 12-hour format, and 12-hour with a leading blank insted of '0'. Pressing the bottom/left-of-USB button will toggle "demo mode" on and off. When "on", the time is set to a few minutes before a full 4-digit cascade of digits will occur and runs at 6 seconds per minute instead of the normal 60 seconds per minute.
First, either rename or make a copy of the my-secrets.h file as secrets.h and insert your WiFi credentials in there.
The flipclock.h file contains definitions you may wish to change, including your host name and time zone. You can also choose the intial time display format with the INITIAL_DISP_FMT definition.
The partitions_16M_app6M_data3M.csv file may require modification if you wish to use a different device partition layout.
Below are brief descriptions of the code. Additional information is in the comments in each.
In flipclock.cpp, the setup() function initializes the WiFi and time service as well as the display. loop() checks for and then over-writes a single digit for the "cascading, rolling" effect if needed. Otherwise, it checks for a change to the time (hours and minutes) and will kick off the rolling animation for, at a minimum, the minute units digit. It also flashes the colon image on and off.
The module initiates the WiFi connection and NTP time service to fetch the initial time.
This module deals with reading image files from "disk" and displaying them using the TFT_eSPI library. The image for the flashing colon is cached to avoid re-reading it many times per minute. (When the colon is off, the area is simply filled with a black rectangle.)
Images are stored in a format optimzed for this display. Each pixel is stored as a 16-bit value with 5 bits for RED, 6 bits for GREEN and 5 bits for BLUE. They are provided in "big-endian" order so they can be copied directly into display buffers (even though the ESP32-S3 is a "little endian" machine).
This code holds descriptors for each screen digit image and the routines needed to initialize and change them. Each digit is given a string listing the possible characters for that digit position. (The units positions always use all 10 digits, but the tens positions have more restricted sets of digits to display.)
This Python program uses the Pillow (PIL) library to generate image files that are written to the data directory. It can generate the files in standard formats such as PNG, BMP, etc., but I use it here for a custom RGB565 format. This file format contains:
| offset | length | use |
|---|---|---|
| 0 | 4 | magic 'R565' flag |
| 4 | 4 | image width in pixels |
| 8 | 4 | image height in pixels |
| 12 | 4 | reserved |
| 16 | * | width * length 16-bit RGB565 values |
All values are in "big-endian" format (although ESP32's are natively "little-endian") because that is the most efficient way to load the display buffer.
Files all have 3-character names: The first char denotes the "main" digit being displayed, the second char is the digit being rolled to and the third is the step number 0-3. For example, the '4' digit file is named '440', the step files when rolling are named '451', '452', '453' and the final file in the sequence is '550' (as shown above).
Although this format has a certain advantage in loading efficiency and is 1/3 smaller than a .bmp file, the processing savings are not as big as I originally imagined - the bulk of image loading time is spent in reading from the LittleFS file system, which can vary by 100% simply based on where the file is located in the flash, and even the fastest reads are still 5 times longer than the time needed to actually process the data, whatever format it is in. Had I known this before, I would probably have stuck with the .bmp format as those files can be easily processed by many other programs.
Items you may wish to change in flip_digits.py:
- FONT_FACE: choose a font you have on your desktop.
- SCREEN_W and SCREEN_H: if using a different-sized display.
- *_FG and *_BG: colors for the digits.
Here is a picture of the final result:
