1. Introduction

Environmental monitoring in off-grid areas needs nodes that survive for months on a small battery and still report over kilometres. Low-power wide-area networks, and LoRaWAN in particular, are well matched to the low duty cycle and small payloads involved [1], [2]. The engineering question is not whether the radio reaches, but whether the whole node's energy budget is dominated by sleep rather than by sensing and transmission.

We describe an open-hardware node designed around that budget, give a simple model that predicts its lifetime, and validate the model against a three-month field deployment on a smallholder farm.

2. Node design

The node pairs a low-power microcontroller with a sub-gigahertz LoRa transceiver and calibrated temperature, humidity, and soil-moisture sensors. The firmware keeps the microcontroller in deep sleep between readings, wakes to sample and transmit, then returns to sleep. The design targets three levers that dominate consumption:

• deep-sleep current, which sets the floor because the node sleeps for almost the whole period;

• the active duration per cycle (sample plus transmit), kept short by using a fast sensor warm-up and a fixed spreading factor;

• the reporting interval, traded against data resolution.

3. Energy model

Over one reporting period $T$ the node spends $t_a$ awake at current $I_a$ and the remainder asleep at $I_s$. The average current is

$$I_{\text{avg}} = \frac{I_a\, t_a + I_s\,(T - t_a)}{T},$$

and, for a usable battery capacity $C$, the expected lifetime is

$$L = \frac{C}{I_{\text{avg}}}.$$

Because $t_a \ll T$, the first term is small and $I_{\text{avg}} \approx I_s$ until the interval becomes short. This is why reducing sleep current, not transmit power, is the highest-leverage design choice [3].

4. Measurements and field results

We measured current with a precision shunt across a full cycle. Table 1 gives the components at a fifteen-minute interval. The measured average of 38 microamps implies, from equation (2) with a 1200 mAh usable capacity, a lifetime of about 14 months.

A three-month deployment of eight nodes reported reliably, with an end-to-end packet delivery ratio of 97.4 percent; the lost packets clustered during heavy rain, consistent with attenuation at the gateway rather than node faults.

5. Conclusion

A node whose firmware is written around the sleep-dominated energy budget of equations (1)-(2) reaches a multi-month lifetime without exotic parts. The schematic, firmware, and calibration procedure are open [4]; the main avenue for improvement is adaptive data-rate selection to cut the transmit share further.