Why the EoRa-S3-900TB?
Choosing the right hardware for a battery-powered, always-listening receiver isn't
just about picking a board with LoRa on it. The wrong board kills your battery in
days. The right one survives for months — or longer. Here's why the Ebyte
EoRa-S3-900TB earned its place in this project.
Everything in One Package
The EoRa-S3-900TB packs an ESP32-S3FH4R2 (dual-core LX7 @ 240 MHz, 4 MB Flash,
2 MB PSRAM, Wi-Fi, BLE 5.0) and an Ebyte E22-900MM22S LoRa module (SX1262, 22 dBm,
~7 km range) onto a single 27 × 64 mm board, along with a 0.96-inch OLED, SD card
slot, Type-C interface, and — critically — a built-in lithium battery charging
circuit. No separate modules to wire up. No charge IC to source. One board, one
LiPo connector, done.
A Power Architecture Built for Battery Life
This is the detail most Ebyte reviews skip over. The EoRa-S3-900TB uses two
separate LDO regulators — LDOA and LDOB — feeding the ESP32 and the LoRa
peripherals on isolated 3.3 V rails:
| Rail |
Regulator |
Powers |
Max Continuous |
Peak Pulse |
| LDOA |
HT7533 |
ESP32-S3 |
250 mA → 3.3 V |
340 mA |
| LDOB |
ME6214 |
E22 LoRa + OLED + peripherals |
150 mA → 3.3 V |
100 mA |
Why does this matter? The ESP32's Wi-Fi and Bluetooth transceiver draws sudden
300+ mA spikes. On a single-LDO board, those transients collapse the supply rail
just long enough to corrupt a LoRa packet in flight, or worse, trigger a false
wakeup. The split-rail design keeps the SX1262 on clean, isolated power — exactly
what you need when a false wakeup means the camera switches on at 3 AM with nobody
watching.
The OEM Spec vs. What You Actually Achieve
Ebyte's datasheet quotes a deep sleep floor of ~25 µA — measured with USB
disconnected, OLED off, LoRa radio in full sleep, no peripherals active, running
on battery alone. That's the theoretical minimum: just the ESP32 RTC domain alive,
waiting for an external wakeup event.
Real-world with Wake-on-Radio running, deep sleep is ~174 µA. That is not a
separate mode — it is deep sleep for this project. The SX1262 autoDutyCycle
wakes for 9.2 ms every duty cycle interval to sniff for a preamble, then the
system returns to the ESP32 RTC baseline. Those two states time-average to 174 µA:
- 9.2 ms active sniff @ ~11 mA — SX1262 listening
- Remainder of duty cycle @ ~25.38 µA — ESP32 RTC baseline
The 25.38 µA baseline, measured on the Nordic PPK2, validates the OEM ~25 µA spec
exactly. It is the floor beneath the duty cycle, not a separate operating mode.
| Mode |
Current |
| OEM deep sleep floor spec (no USB, peripherals off, battery) |
~25 µA |
| Deep sleep + WOR duty cycle — measured (Nordic PPK2) |
~174 µA |
| Active sniff spike (9.2 ms, SX1262) |
~11 mA |
radio.sleep() — total system, 120 s viewing window (Nordic PPK2) |
18.39 µA avg / 19.55 µA max |
| High activity — TX / processing |
37–50 mA |
That 174 µA — achieved on a development board with an OLED, SD slot, and
battery charger on the PCB — is the headline. Bare modules can go lower. This
board, with everything integrated, holds it to 174 µA. That's the number that
makes 13–19 months of field deployment on a 3000 mAh LiPo realistic.
One Complete Cycle — Power States in Sequence
Once a wake packet arrives and the camera switches on, the power profile shifts
entirely:
- Preamble detected — EXT0 fires on GPIO_NUM_16, ESP32 wakes from deep sleep
- Receive + ACK — ESP32 and SX1262 active, 37–50 mA
- Camera ON — KY-002S latching switch pulsed, power bank energised
radio.sleep() called — total system current drops to a flat ~18.39 µA average (19.55 µA max, measured Nordic PPK2 ampere mode, in series with battery); SX1262 datasheet sleep spec is 600 nA–1.2 µA, so the ~18.39 µA is almost entirely the ESP32 RTC domain; ESP32 sets 120 s timer and holds in deep sleep for the full viewing window- Timer fires — ESP32 wakes, pulses switch to power camera OFF
- WOR resumes — SX1262 returns to autoDutyCycle, system back to ~174 µA average
radio.sleep() only appears in step 4 — the quiet interval between camera ON and
camera OFF. It plays no part in the idle WOR listening phase.
The One Hardware Gotcha
The SX1262's DIO1 interrupt line — the pin that fires when a preamble is detected
— is wired to GPIO33 on the EoRa-S3-900TB. GPIO33 is not RTC-capable on the
ESP32-S3, so it cannot serve as an EXT0 deep sleep wakeup source.
The fix: a single jumper wire from the DIO1 pad to GPIO_NUM_16, which is
RTC-capable. No extra ICs, no inverter buffer circuit. One wire, and the SX1262
can wake the ESP32 directly from deep sleep the moment a valid preamble arrives.
Why Not a Bare Module?
A bare ESP32-S3 + separate SX1262 module would get you to a lower idle floor, but
you'd be designing your own LDO split, your own charge circuit, your own PCB
layout. The EoRa-S3-900TB ships with all of that already solved, at ~$18–20 USD,
in a form factor that fits in a weatherproof enclosure with room to spare. For a
field-deployed, battery-powered receiver with no mains power at the site, that
tradeoff is an easy one.
