• Current schematic: VOUT → R2 100kΩ → Net 24 → REED1 → GND, and Net 24 → R6 → U1 IO7.
• Because the door/window is closed most of the time, the magnet closes REED1, so R2 continuously conducts.
• With VOUT = 3.3 V and R2 = 100kΩ, idle current is:

Text
I = 3.3 V / 100kΩ = 33 µA

Text
3V3 / VOUT
  |
 R2 = 100kΩ
  |
  +---- R6 ---- U1 IO7
  |
 REED1
  |
 GND

1. Bad default-power state: closed door/window is the normal state, but that is the state that burns pull-up current.
2. IO7 deep-sleep limitation: IO7 is not an RTC GPIO, so it should not be used as the deep-sleep wake input.
3. 100kΩ is already relatively weak, but still too much compared with a 5 µA MCU sleep target.

1. Draw near-zero current in both steady states if possible.
2. Wake the MCU from both transitions: closed→open and open→closed.
3. Be robust against noise and contact bounce.
4. Use a GPIO that supports deep-sleep wakeup.
5. Avoid boot-strapping conflicts on ESP8684.

Text
3V3 / VOUT
  |
 R2 = 1MΩ to 4.7MΩ
  |
  +---- series resistor ---- U1 RTC GPIO, GPIO0–GPIO5
  |
 REED1
  |
 GND

• Minimal schematic change.
• Can reduce idle reed current by 10× to 47×.
• Works with a normally-open reed switch used in the usual magnet-closed state.
• No active parts required.

• Still consumes current in the closed state.
• Higher resistor values are more sensitive to PCB leakage, humidity, contamination, and EMI.
• Needs firmware to set the opposite wake level after every event.
• Contact bounce must be filtered in firmware and/or with a small capacitor.

• Keep R6 as a GPIO protection/RC isolation resistor if already present.
• Add 10nF–100nF from the GPIO side of the reed signal to GND only if wake latency and edge speed are acceptable.
• Configure wake on the opposite level each time:
  • If current state is closed/low, sleep waiting for high.
  • If current state is open/high, sleep waiting for low.

Text
3V3 / VOUT
  |
 REED contact, open when magnet present
  |
  +---- GPIO input
  |
 Rpull-down = 1MΩ–4.7MΩ
  |
 GND

• Optimizes the most common state.
• Can make normal idle current close to zero.
• Simple passive circuit.

• Requires a reed switch/contact orientation that behaves correctly with the magnet.
• Open-door state consumes current until the door/window is closed again.
• Not always possible with the chosen reed switch package.

Text
I = 3.3 V / 45kΩ ≈ 73 µA

• Fewer external parts.
• Firmware-configurable.

• Higher current than your 100kΩ pull-up in the conducting state.
• Pull value tolerance may be broad.
• Espressif notes that internal and external pull resistors can interfere if both are enabled.

Text
GPIO_A ---- REED1 ---- GPIO_B

1. Drive GPIO_A high or low briefly.
2. Read GPIO_B.
3. Return both pins to high impedance before sleep.

• Near-zero steady-state current when both pins are high-Z.
• No pull-up burning continuously.

• By itself, it does not provide a passive deep-sleep wake signal unless one side is biased during sleep.
• If both sides are high-Z during deep sleep, the MCU may not wake when the reed changes.
• More firmware complexity.

• Nanopower comparator with internal/reference bias.
• Low-power latch / flip-flop triggered by reed transitions.
• Dedicated magnetic/reed wake interface if available.

• Can achieve very low sleep current.
• Can generate clean wake pulses for both transitions.
• Can add debounce/latching outside the ESP8684.

• Adds cost, BOM complexity, PCB space, sourcing risk.
• Needs careful part selection from datasheets.
• The external IC must itself be lower current than the resistor current you are trying to save.

• Lets hardware optimize the normal state.
• Can improve tamper detection or detect missing magnet/contact faults.

• Larger or more expensive sensor.
• More pins and routing.
• May not be available in the same footprint/form factor.

• IO7 is used for REED1 through R6 — this should be changed.
• IO4, IO5, IO6 are currently used for RGB LED control.
• GPIO0–GPIO5 are RTC-capable, but some also have caveats.

1. Use IO4 for REED1 and move the LED channel currently on IO4 to IO7/IO10/IO18 if available.
2. Avoid using strapping pins for reed if the external state could affect boot.
3. Avoid pins with power-up glitches for visible LEDs if that matters.

1. Move REED1 sense from U1 IO7 to an RTC-capable GPIO, preferably one of IO0–IO5 after checking boot/LED conflicts.
2. Change R2 from 100kΩ to 1MΩ as the first practical low-power value.
3. Configure firmware so each sleep cycle wakes on the opposite level:
   • Door closed / GPIO low → wake on high.
   • Door open / GPIO high → wake on low.
4. Keep LED fully off except installation/setup.
5. If false wakes occur, add small RC filtering and/or debounce in firmware.

1. Read the reed GPIO.
2. Determine current state.
3. Send the new state to the gateway/hub.
4. Configure deep-sleep wake for the opposite level.
5. Enter deep sleep.

Text
state = read_reed_gpio()
if state == CLOSED_LOW:
    configure_wake_on_high()
else:
    configure_wake_on_low()
enter_deep_sleep()

• On wake, wait 10–50 ms.
• Re-read the input.
• Optionally require two or three consistent reads.
• Then transmit only the stable state.

• Keep the reed input trace short.
• Avoid routing near noisy SW node of TPS613221A.
• Keep flux residue and contamination low.
• Consider guard/spacing if going above 2.2MΩ.
• Add ESD protection if the reed wire leaves the PCB or runs near the enclosure edge.

Text
Move REED1 sense from IO7 to an RTC GPIO, and change R2 from 100kΩ to 1MΩ.