Sleight of hand

Project: ESP32-driven analog wall clock - replace the timing IC and drive the coil directly from GPIO, enabling custom movement modes.

Repo: https://github.com/skorokithakis/sleight-of-hand

Details

ESP32-C3 SuperMini (esp32-c3-devkitm-1 board definition)
Cheap no-name sweeping quartz wall clock with timing IC removed
GPIO 4 → Coil pin A, GPIO 5 → Coil pin B
Coil resistance: 600 ohm
820 ohm series resistor on one coil pin (delivers 2.32 mA at 3.3V)
960 pulses per revolution (measured: 5761 pulses over 6 revolutions = 960.17)
Original IC drives at ~62.5ms per cycle (31.25ms pulse + 31.25ms pause)
WiFi AP name: SleightOfHand
Three movements purchased total: two sweeping, one ticking

This is a sweeping-motion clock, not a ticking clock. The original IC drives continuous sweeping pulses (~30ms pulse, ~30ms pause), not one-pulse-per-second.
All initial debugging failures were because the firmware was driving one pulse per second (ticking mode), which is the wrong drive mode entirely.
Confirmed by buying a second identical clock, testing with battery first, and analyzing the original PCB with multimeter and oscilloscope.

ESP32-C3 GPIO default drive strength is 40mA, much more than needed. Can reduce with gpio_set_drive_capability() (5/10/20/40mA). Using GPIO_DRIVE_CAP_0.
Tested timing limits: pulses down to 15ms (pulse and pause) work reliably. 10ms and below fail.
Firmware uses 960-pulse model anchored to NTP minute boundaries. Two modes: steady (62ms cycle) and rush_wait (50ms cycle, ~12s idle at end of minute).

Original battery: 1.5V across 600 ohm = 2.5 mA, 3.75 mW
At 3.3V with no series resistor: only ~5.5 mA
To match original 2.5 mA at 3.3V: need 1320 ohm total → 680-750 ohm series resistor
Final: 820 ohm series resistor → 2.32 mA, slightly under original but works perfectly

MQTT control commands: stop, start, start_at_minute (waits for NTP minute boundary)
MQTT reconnect only attempted during idle gap between revolutions to avoid stalling pulses
UDP broadcast logging on port 37243 for wireless debugging (nc -kul 37243)

Serial output on SuperMini requires USB CDC flags: -DARDUINO_USB_MODE=1 -DARDUINO_USB_CDC_ON_BOOT=1
Original driver PCB was desoldered and removed
Idle state between pulses: high-impedance (pinMode(..., INPUT))

2026-02-10: Project initialized with PlatformIO (ESP32 Arduino framework). Basic coil driver written, repo created on GitHub. Switched board to ESP32-C3 SuperMini.
2026-02-11: Switched coil pins to GPIO 4 and 5. Removed series resistors. Extensive debugging - rotor flails/oscillates instead of making clean steps. Tested pulse widths 5-200ms, various drive strengths, both polarities. Tried short-brake after drive pulse. Measured coil at 600 ohm.
2026-02-12: Bought a second identical clock and tested with battery first - discovered it’s a sweeping clock, not ticking. Retrieved original PCB from trash, confirmed sweeping drive on oscilloscope.
2026-02-12: Updated firmware to 30ms pulse / 30ms pause continuous drive. Installed 820 ohm series resistor. Clock works perfectly.
2026-02-12: Measured 960 pulses/revolution. Rewrote firmware to NTP-anchored 960-pulse model with steady and rush_wait modes.
2026-02-12: Added MQTT control and UDP broadcast logging. Done and working.
2026-02-23: Working on adding ticking movement support. Debugging an off-by-one error in the ticking timing code - hard to pin down.
2026-02-24: Experimenting with a 150 Ω series resistor to reduce tick strength on the ticking movement. Still tuning.

Last updated on February 24, 2026. For any questions/feedback, email me at hi@stavros.io.