Write on paper.
See it digitally.

A smart pen that streams real-time sensor data to a host, where deep learning translates your physical writing into digital strokes on any canvas app.

3D Pen concept render showing the smart pen design

How It Works

No computation on the pen. All intelligence lives on the host.

1

Write

Use the 3D Pen like any normal gel pen. Write, draw, or sketch on real paper.

2

Sensors Stream

Pressure, dual IMUs, and capacitive touch sensors stream data at 8kHz wirelessly.

3

ML Inference

A deep learning model on the host translates raw sensor streams into 2D stroke coordinates in real time.

4

Digital Output

The pen registers as an OS-level input device. Your strokes appear in any canvas app.

4-panel lifestyle view showing the 3D Pen in different use scenarios

Hardware

All electronics fit in a 2.75mm annular gap between the refill and the outer shell.

Translucent X-ray view of the 3D Pen showing internal components
Zone N

Nib Section

0 – 15mm

Removable threaded nib (M10x0.75) in black POM. Unscrews to insert standard gel refills.

Zone A

Pressure + IMU

15 – 55mm

FSR 400 Short pressure sensor (40mm zone) plus IMU #1 (ICM-42688-P) for nib-end motion tracking.

Zone B

Touch + Wake

55 – 80mm

Azoteq IQS263 capacitive touch slider with 4 electrode pads. ADXL367 wake controller for power management.

Zone C

MCU + Radio

80 – 110mm

nRF52840 (ARM Cortex-M4F) with integrated 2.4GHz radio. PCB IFA antenna for wireless streaming.

Zone D

Power Management

110 – 135mm

10180 Li-ion battery (100mAh), BQ51003 Qi receiver, BQ25100 charger, MAX17048 fuel gauge, IMU #2.

Zone E

End Cap + Charging

135 – 150mm

Sealed end cap with 9mm Qi receiver coil and ferrite shield. Debug flex tail exits before seal.

Exploded isometric view of the 3D Pen showing all 12 components

Exploded view: 12 components from nib cone to end cap

Specifications

32kHz
Max IMU sample rate
8kHz
Wireless polling rate
<10ms
ML inference latency
$156
Prototype BOM cost
ParameterValue
Dimensions150mm L × 11.5mm D
Weight20 – 28g (polycarbonate shell)
RefillStandard gel pen (110mm × 6mm, Pilot G2 compatible)
MCUnRF52840 — ARM Cortex-M4F, 256KB RAM, 1MB flash
IMU2× TDK ICM-42688-P (6-axis, 32kHz max ODR)
Pressure SensorInterlink FSR 400 Short (0.3mm, 5.6mm active area)
TouchAzoteq IQS263 capacitive slider + proximity
WirelessNordic ESB at 2Mbps (BLE secondary for config)
Data throughput6ch × 8kHz × 16-bit = 768kbps
Battery10180 Li-ion, 100mAh, 8–14+ hours active
ChargingWireless (Qi-style) via cradle
PCB2-layer flex polyimide, helical wrap, 230mm × 10mm
FirmwareZephyr RTOS via nRF Connect SDK
Host inferenceONNX Runtime, <10ms per window
HIDOS-level digitizer pen (Usage Page 0x0D)
OS SupportWindows, macOS, Linux

Roadmap

Phase 1

Research

Complete
  • 48+ documents across hardware, embedded, ML, and software domains
  • Component selection and feasibility analysis
  • Architecture design and requirements specification
  • AI concept renders and hardware design spec
Phase 2

Design

In Progress
  • Flex PCB schematic and layout in KiCad
  • Mechanical CAD for shell, nib, and inner structure
  • Firmware skeleton with Zephyr RTOS
  • ML model architecture and training pipeline setup
Phase 3

Prototype

Upcoming
  • Flex PCB fabrication and component assembly
  • CNC machined shell and inner parts
  • Firmware bring-up and sensor validation
  • First sensor data capture sessions
Phase 4

Integration

Future
  • ML model training on real sensor data
  • Host software with virtual HID driver
  • End-to-end: write on paper, see strokes digitally
  • Iterative refinement and second hardware revision

Support the Project

This is an independent open-source hardware project. Your support funds components, PCB fabrication, and prototyping.

$3/mo

Follower

  • Project updates and progress posts
  • Behind-the-scenes design decisions
  • Discord community access
Join
$25/mo

Builder

  • Everything in Supporter
  • Access to raw CAD/KiCad files
  • Monthly video call updates
  • Priority input on hardware decisions
Join
$100/mo

Sponsor

  • Everything in Builder
  • Logo on project website and repo
  • First-run prototype unit (when available)
  • Direct collaboration channel
Join

Funding Goals

First PCB order $200
Full prototype build $500
Second revision + ML training hardware $2,000

Fully Open Source

Hardware designs, firmware, ML models, and host software — everything is open. Explore the 50+ document knowledge base, review the architecture, or contribute.

Explore the Repository