Writing · How-to
How to build an ESP32 e-paper meeting-room sign
A working DIY meeting-room sign needs four things: an ESP32 board, an e-paper panel, a battery, and a small server that turns the room’s calendar into an image. The ESP32 wakes on a schedule, downloads that image if it changed, draws it, and goes back to deep sleep. The hard part is not drawing the agenda. It is choosing when the sign wakes up.
Published 2026-11-11. We ship an ESP32-S3 + E Ink Spectra 6 stack in 8frames and adapt it for team projects. The firmware details below come from that codebase; the open-source projects we link to are other people’s work, described from their READMEs.
What you need
- A board with deep sleep that actually sleeps. Any ESP32 works; an ESP32-S3 with PSRAM gives you room for a full color framebuffer (800 × 480 at 4 bits per pixel is 192 kB). What matters more than the chip is that the board does not leak milliamps through an always-on LED or regulator.
- An e-paper panel. 4.2″ to 7.5″ is the usual room-sign range. Black-and-white is cheapest and fastest. Six-color Spectra 6 (for example the 7.3″ GDEP073E01) lets status be a color instead of a word.
- A battery and a way to charge it. A Li-ion cell around 3,000 mAh with USB-C charging is a sensible start. Size it after you pick a wake schedule, not before.
- A cut-off for panel power. A load switch or PMOS on the panel’s supply, so the display draws nothing between refreshes.
- Optionally, a button. Wired to an RTC-capable pin so it can wake the board from deep sleep for an immediate check.
- A server (a Raspberry Pi, a small VM, or a serverless function) to read the calendar and render images, unless you go fully on-device.
Two ways to get the calendar onto the sign
The open-source projects split cleanly into two camps, and the choice decides how painful the sign is to live with.
| Server renders, ESP32 displays | ESP32 reads the calendar itself | |
|---|---|---|
| Firmware | Small: wake, fetch image, draw, sleep | Larger: TLS, iCal or API parsing, layout, fonts |
| Change the layout | Edit the server, every sign updates | Reflash every sign |
| Calendar credentials | Live on the server only | Live on every device |
| Extra infrastructure | An always-on host | None |
| Time awake per wake | Short: one image download | Longer: parsing and rendering on device |
| Best for | More than one room | A single desk or home-office sign |
On the server-rendered side, jamct/DoorsignEPD ships a conference_room example: a PHP server draws the image for a 2.9″, 4.2″ or 7.5″ Waveshare panel and the ESP32 fetches it out of deep sleep. Piszmog/esp32-calendar does the same with a Go server reading a private iCal link. On the device side, VoIPshare/ESP32_ePaper_GoogleCalendar parses iCal links on the ESP32 with no server at all.
Our recommendation: For a meeting room, render on the server. Room signs get redesigned, calendars move between Google and Microsoft, and you do not want an OAuth token sitting in flash on a device screwed to a public corridor wall.
Build it in seven steps
1. Get the panel drawing from USB power
Run the vendor’s example until a test image draws cleanly. Then add the habit that keeps panels healthy: send the panel’s own deep-sleep command after every refresh, then cut its power. Our ESP32-S3 + Spectra 6 field guide has the pin map and init sequence we ship.
2. Write the render endpoint
One URL per room that returns a ready-to-draw image at the panel’s exact resolution: room name, current state, the next two or three bookings, and the time it was rendered. Do the dithering and color mapping on the server, where you have a real CPU. Return an identifier for the image (a hash works) so the sign can ask “has this changed?” cheaply.
3. Connect the calendar
The simplest source is the room calendar’s private iCal link. For a whole office, use the calendar provider’s API with a service account that can read room resources, so you are not handing out per-room secrets.
4. Make the device ask before it downloads
On each wake, send the identifier of the image already on screen. If the server says nothing changed, go straight back to sleep: no download, no refresh. Our firmware does this with per-slot content tags, which is why an unchanged check is a short exchange rather than a full image transfer.
5. Pick a wake schedule
This is the decision that sets battery life (next section). Our firmware accepts either up to 24 wall-clock wake times a day or a fixed interval in minutes, and falls back to hourly when no schedule is set.
6. Add a wake button
A button on an RTC-capable GPIO lets someone force a refresh (“I just booked this room”) without the sign polling constantly. On the ESP32 this is the ext0 wake source; arm it only when the pin is at its idle level, or the board wakes the instant it sleeps.
7. Provision, mount, and watch it for a week
Give the sign a way to join Wi-Fi on site (a captive portal is the usual trick), log battery voltage on every check-in, and leave it on the wall for a week before building a second one.
The wake schedule is the whole design
A battery e-paper sign does not hear about a new booking until it wakes and asks. Every wake costs charge, mostly for Wi-Fi. Using the same budget as our battery-life guide (about 3 mAh per Wi-Fi wake, 0.11 mAh per full refresh, 10 µA deep sleep, 3,000 mAh cell, every wake redrawing):
| Schedule | Wakes / day | Runtime at 3 mAh/wake | Runtime at 1 mAh/wake |
|---|---|---|---|
| Every 5 minutes, all day | 288 | ~3 days | ~9 days |
| Every 15 minutes, all day | 96 | ~10 days | ~28 days |
| :00 and :30, 08:00 to 19:30 | 24 | ~40 days | ~3.7 months |
| Hourly, 08:00 to 19:00 | 12 | ~2.6 months | ~7 months |
The third row is the useful one. Meetings mostly start on the hour or half hour, so waking just before those boundaries during the working day gets the sign right when it matters, and a button covers the rest. Polling every five minutes feels responsive and turns the sign into something you charge twice a week.
Pre-wall checklist
- Measured deep-sleep current with a meter, not taken from a datasheet
- Panel deep-sleep command sent and panel power cut after every refresh
- Unchanged-content check returns without downloading or redrawing
- Wake schedule set in local time, with a sane fallback if the clock is wrong
- Wake button armed only at its idle level
- Last good image cached, so a Wi-Fi outage leaves the sign readable
- Battery voltage reported on every check-in, with an alert threshold
- Wi-Fi signal checked at the mounting spot, not at your desk
- A way to re-provision Wi-Fi without opening the enclosure
When to stop building
A single sign is a weekend project. A fleet is a product: on-site provisioning, signed OTA updates, low-battery alerts, enclosures that survive cleaning crews, and calendar permissions across a whole tenant. For a small office, buying is cheaper; our Joan alternatives comparison says so plainly. For hundreds of rooms, or a workplace-software company that wants its own branded sign, see our workplace signage work and the dev-kit-to-product checklist.
FAQ
Can an ESP32 read Google Calendar directly?
Yes. Open-source projects do it two ways: the ESP32 parses a calendar’s private iCal (.ics) link itself, or it calls the Google Calendar API with OAuth credentials. Both work for one room. For more than a handful of rooms, a small server that reads the calendars and renders one image per sign is easier to maintain, because layout changes no longer need a firmware flash.
How long does a battery e-paper room sign last?
It depends on how often it wakes. On a 3,000 mAh cell with our conservative budget of about 3.1 mAh per wake, a sign that wakes 24 times a day lasts roughly 40 days, and one that wakes every five minutes lasts about 3 days. Cutting each Wi-Fi wake to about 1 mAh stretches the 24-wakes-a-day case to roughly 3.7 months.
How does the sign notice a meeting booked at the last minute?
A battery sign only knows what it fetched at its last wake. Common fixes are waking at the half-hour and hour boundaries when meetings start, adding a button that forces an immediate check, or accepting a short lag. Keeping the radio on to receive pushes defeats the point of a battery e-paper sign.
Which e-paper panel should I use for a room sign?
Black-and-white or black-white-red panels refresh fastest and cost least. A six-color Spectra 6 panel such as the 7.3-inch GDEP073E01 lets status read as green, yellow or red from down the corridor, at the cost of a roughly 20 to 30 second full refresh, which is fine for a sign that changes at most every half hour.
When should I stop building and buy or commission room signs?
When you need more than a pilot’s worth of rooms. Provisioning Wi-Fi on site, OTA updates, low-battery alerts, enclosures and calendar permissions across a whole tenant are where DIY projects stall. Below about 15 rooms an off-the-shelf product is usually cheaper; above a few hundred, or if you want your own branded product, a custom build starts to pay.