Building a Lightweight Curved E‑Ink Smartwatch

Most smartwatches behave like tiny phones. Bright screens, constant recharging, heavy cases, and interfaces that feel like they’re trying too hard. I wanted the opposite. Something with great battery life, something customizable, something light enough to forget you’re wearing it. Think more hospital band than gadget.

E‑ink made sense straight away. It’s gentle on the eyes, readable in full sun, and doesn’t wake your brain up before bed or during the night to see what time it is and if you have slept enough! That idea turned into a long experiment in flexible PCBs, e‑ink displays, backyard engineering, and a fair bit of trial and error.

This is the story of how I built a rugged, comfortable, very hackable e‑ink smartwatch around the Seeed Studio nRF52840 Sense and a 2.9 inch Waveshare display

The full code and hardware files are available on GitHub:

https://github.com/ephiras/EPD-FlexwatchV0.git

Brains

1. Seeed Studio nRF52840 Sense
2. Custom flex PCB (power + sensors)

Display

1. Waveshare 2.9 inch flexible e‑ink display

Sensors

1. LSM9DS1 9‑axis IMU
2. APDS‑9500 gesture sensor
3. MAX30102 heart rate + SpO₂ (early prototypes)
4. RV‑3028‑C7 ultra‑low‑power RTC
5. MAX‑M10S GPS module (early prototypes)

Power & Protection

1. AP2127K‑1.8 regulator
2. SI1304BDL MOSFET
3. TCA9406 I²C level shifter
4. Schottky + ESD diodes
5. LiPo battery 400mAH

Frame Materials

1. 50‑cent (cheap) plastic ruler
2. Pleather strip
3. Saucepan + boiling water

Before any flex PCBs, or broken e‑ink displays, the whole project started the classic way: a messy breadboard, jumper wires everywhere, and a very rough idea of what I wanted the watch to become.

This stage was all about proving the basics:

1. Could the nRF52840 Sense drive an e‑ink display reliably?
2. Would the IMU behave?
3. How slow is e‑ink refresh in real life?
4. How much power does everything actually draw?
5. Could I update the screen without the whole thing freezing?

I wired up the Seeed Studio nRF52840 Sense, the Waveshare display, and a handful of sensors. Nothing was pretty, but it worked well enough to convince me the idea wasn’t completely ridiculous.

What I Learned at This Stage

1. E‑ink looks fantastic in real life, even on a breadboard.
2. The nRF52840 Sense is a beast for prototyping — Bluetooth, IMU, mic, all in one.
3. The display refresh is slow, but perfect for a calm, low‑stress UI.
4. Power draw was low enough to make long battery life realistic.
5. The whole thing felt surprisingly doable.

This was the moment the project shifted from “fun idea” to “I’m actually building a smartwatch.”

Why This Step Matters

Breadboarding let me test:

1. display driving
2. SPI timing
3. sensor communication
4. power rails
5. UI experiments
6. sleep mode behaviour

…without committing to a PCB that I’d inevitably redesign three times.

It also gave me the confidence to move on to the flexible PCB concept. Which, as you you will read further, didn’t survive long, but was an important stepping stone.

After the breadboard stage I moved into proper design work. I modelled the watch in Fusion and designed the boards in Eagle. The plan was to create a segmented set of small rigid boards that would sit on top of a flexible PCB backing.Each section was its own sensor module so that if I was havining issues they could be removed or replaced individually. The idea was simple. Use cheaper rigid boards for the part placement and sensors, keep the rest flexible, and glue or solder or rivet the whole thing together. Black, white and gold (Love ENIG!) was the colour scheme.

Doing it this way kept the price down and let me experiment without committing to a full rigid flex stack up.

Where Things Started Going Wrong....

Rivets:

I tried brass rivets first. They looked great but they dug into the copper and I do not have the patience or the skill to make them behave. Glue and PLA filament rivets worked better, but the whole structure was already feeling fragile.

Flexible E Ink: Flexible in Name Only

I broke the first flexible display within minutes. The second one lasted a little longer, which I count as progress.

Flexible e ink sounds like magic until you actually work with it. Waveshare calls these panels flexible because the backing is a thin plastic sheet. The important part is what sits on top of that sheet. The active layer is full of tiny capsules and all of that sits on a very brittle thin film transistor backplane. The plastic can bend. The thin film transistor layer cannot. It behaves like a sheet of glass laminated onto cling wrap.

The weakest point was the ribbon connector. I lost several screens to that :( .

So yes, the display bends, but only in the gentle direction the factory intended. Bend it the wrong way, twist it slightly, or put any stress near the connector area and it cracks internally. You never see the crack. You only see the pixels in that region stop responding. You get streaks or dead zones that never recover.

The connector area is the worst place to put any bend because that is where the traces are densest, the lamination is stiffest, and the plastic backing transitions into the flex tail. It is a natural hinge point and a natural failure point. Forewarning! If your design puts the main bend there, the panel does not stand a chance!

Flexible e ink is flexible in the same way uncooked spaghetti is flexible. It will bend a little, but it is not thrilled about it.

The Heart Rate Bodge

The heart rate sensor refused to work. I spent far too long debugging it before realising I had forgotten to power the 3V3 to 1V8 I2C level shifter. A couple of bodge wires fixed it.

Broken Traces

After fixing the heart rate sensor I bent the band a few too many times and broke the traces in the flexible PCB. More layers or thicker copper might help, but that stops being cheap very quickly. Cheap was part of the charm and the wife approval factor.

The End of the Flexible Era

Between the fragile display, the fragile traces, and my fragile patience, it became clear that the fully flexible approach was not going to survive daily use. It was time to rethink the entire structure and move toward something more rigid and more realistic. I want to move back to this oneday and have some big plans....

After the flexible displays and the flexible PCB traces kept failing, I needed a new approach. The idea of a fully flexible watch was beautiful in theory, but the real world was not cooperating. I needed something that would protect the display, hold its shape, and survive daily use without snapping or tearing itself apart.

The answer came from an unexpected place. A Temu jewlery ad...

The First Rigid Frame

I found a cheap chrome bracelet on Temu that happened to be almost the right size. It was never meant to hold electronics, but it was rigid, curved, and easy to modify. It let me mount the display and the boards without stressing the fragile connector area. It also let me keep testing the electronics without destroying more screens.

It was not elegant, but it worked. Sometimes that is enough.

Why This Frame Helped

1. It stopped the display from bending in the wrong direction
2. It protected the ribbon connector
3. It gave the watch a stable shape
4. It let me keep experimenting without breaking everything

This was the first time the project felt like it might actually become a wearable device instead of a collection of broken parts.

The Turning Point

The rigid frame solved the biggest problem of the flexible era. The display finally survived long enough for me to test the rest of the system. I could walk around with it, move my wrist, and not worry about the screen cracking from a tiny twist. ( it did still break when I dropped it though :( )

This worked great for a while as a prototyping base. That was until I tried to order new flexible 2.13 inch epaper displays and waveshare announced they were discontinued.....

Once I had the rigid frame working, I realised the project needed a reset. The flexible era had been full of feature creep and fragile parts. The screensize was about to change( I even tried to use a flat rigid screen but this looked poo.).

Before adding anything fancy, I needed a version that simply worked every day without breaking. That meant stripping the design back to the essentials and removing anything that was causing trouble, taking up space, or draining power.

What I Removed and Why

The Gesture Sensor

The gesture sensor was fun, but the wider 2.9 inch display (the only one available) left no room for the backlight and button assembly that originally sat beside it. The sensor also added complexity to the layout and was not essential for a first working version. It had to go.

The GPS Module

The GPS was a challenge from the start. I soldered it using my wife’s old sandwich toaster as a reflow oven, which worked but was not ideal. It also drew a lot of power and the flexible antenna took up a surprising amount of space once the display grew. For a lightweight, low stress watch, GPS was not a priority. I removed it to simplify the design.

The Heart Rate Sensor

I had the heart rate sensor working after fixing the I2C level shifter, but placing it on a rigid frame was a new challenge. It needed skin contact, a window, and a stable mounting point. None of that fit well with the new structure. Rather than force it, I removed it until I could rethink the placement properly.

Why This Step Mattered

Removing these parts made the watch:

1. simpler
2. lighter
3. more reliable
4. easier to debug
5. easier to power
6. easier to fit inside the new frame

It also meant I could focus on the core experience: a comfortable, readable, long lasting e-ink watch.

After cutting the project back to a stable base version, the next big change came from something completely outside my control. Waveshare stopped making the smaller 2.13 inch display. That pushed me onto the larger 2.9 inch panel, and that one change reshaped the whole project.

The new display was wider and longer and needed a much gentler curve. The Temu bracelet frame that had carried the project through the early rigid era was suddenly too small. It squeezed the display, stressed the connector area, and made the whole thing feel cramped and awkward.

So I went back to backyard engineering.

The Plastic Ruler Solution

I found a cheap plastic ruler at kmart whilst buying my daughter her school supplies. It cost 50c and was never meant to be part of a smartwatch, but it was exactly what I needed.

I softened the ruler in a pot of boiling water, wrapped it around a jar to get the right curve, and held it in place until it cooled. The result was surprisingly good. It held its shape, it matched the curve of the larger display, and it was light enough to wear comfortably.

To make it wearable, I lined the inside with a strip of pleather so it would not feel like strapping a piece of hardware store plastic to my wrist.

The ruler frame allowed the project to move forward again. The display was safe. The electronics were stable. The watch could finally be worn without fear of snapping something. It was still rough, still experimental, and still very much a work in progress, but it worked.

It works!!!

A Stable and Wearable Version

The ruler frame protected the display and the connector area, which had been the main failure point in the flexible era. With the structure sorted, I could focus on the electronics and the user experience instead of constantly repairing damage.

This version survived daily use. I wore it on site in the rain and it held up better than expected. It was not waterproof, but the plastic frame kept the important parts safe. The battery lasted about six days once I tightened the sleep modes and removed the power hungry modules.

The Seeed Studio nRF52840 Sense handled the core features well. The IMU worked, the basic interface was stable, and the power system behaved. Bluetooth and the microphone are on the board but not implemented yet. They are simply waiting for the right chunk of coding time.

What Works Right Now

1. The 2.9 inch display is stable and readable in any light
2. The ruler frame is light, comfortable, and tougher than it looks
3. The watch lasts almost a week on a full charge
4. The IMU and core interface are solid
5. The electronics no longer break themselves during normal use
6. Bluetooth and the microphone are present but not coded yet
7. The whole device finally feels like something I can wear every day

This was the first time the project felt like a real watch. It simply showed the time and stayed out of the way.

What Still Needs Work

There is still plenty to improve. This is a project that grows every time I touch it.

1. The RTC needs a backup battery so it stops forgetting the time
2. Sleep modes can be improved to stretch the battery even further
3. The GPS and heart rate modules will return once I rethink the layout
4. Bluetooth and microphone support need to be coded
5. A rigid flex PCB is planned for a cleaner and more durable design
6. A UV sensor is on the list because it would be genuinely useful
7. A simple touch layer is still being explored, but the display is fragile
8. A custom flex PCB tied everything together. A proper rigid flex design is planned once the layout stops changing every few weeks.
9. A wristBand.
10. Titanium backplate!
11. There will probably be more feature creep because that is how these things go!

The current version is not the final one. The watch works. Now it is time to make it better!