iot hardware design

    IoT Hardware Design: Low-Power Architecture and Wireless Selection That Survives the Field

    IoT hardware design guide — sub-100uA sleep architecture, BLE vs WiFi vs LoRa vs 433MHz vs cellular selection, antenna design, and battery math. Lessons from Apollo2 terminals, ESP32 sensors, smart locks, and a 7-day smartwatch.

    ·XTELL Engineering Team

    IoT Hardware Lives or Dies on Two Numbers

    Strip away the dashboards and the cloud platforms, and every IoT hardware project is judged on two numbers: how long the battery lasts, and whether the wireless link works where the device actually gets installed. Get both right and the software team gets to be heroes; get either wrong and no amount of firmware cleverness saves the product. A sensor that dies in three months or drops off the network behind a concrete wall is not a product — it is a support ticket generator.

    We have designed IoT hardware from Apollo2-based ultra-low-power terminals to ESP32 sensor nodes, smart locks, and tracking modules. This guide covers the two numbers: the low-power architecture that gets you to years of battery life, and the wireless selection framework that keeps devices connected.

    Low-Power Architecture: Think in Duty Cycles, Not Datasheets

    The most common low-power mistake is reading the MCU's sleep current from the datasheet and declaring victory. The datasheet number is real, but your product's battery life is determined by the duty cycle — what fraction of time the device spends awake, and how much everything else on the board draws while the MCU sleeps.

    The Apollo2 low-power IoT terminal (Wuhan_Apollo2) shows the full discipline: Deep Sleep modes with driver code for GPRS, GPS, RTC, ADC, Flash, and watchdog, supporting multiple toolchains (Keil/IAR/GCC). The Apollo2 MCU is famous for its sub-threshold operation — sleep currents measured in microamps — but the project still had to manage every peripheral: the GPS that draws milliamps if left on, the GPRS module with its amp-scale transmit bursts, the RTC that must keep time through it all.

    The low-power design method we apply to every IoT board:

    • Build the power budget as a spreadsheet first. Every state (sleep, measure, transmit, receive) gets a current and a duration. Multiply, sum, divide into the battery capacity. If the math does not close on paper, it will not close on the bench.
    • Attack the biggest consumer first. It is almost never the MCU — it is the radio transmit burst, the GPS fix, or the sensor warm-up time. A gas sensor that needs 60 seconds of heater warm-up dominates the budget of a gas monitoring node far more than the ESP8266's sleep current.
    • Sleep the peripherals, not just the MCU. Every sensor, regulator, and level shifter needs a power-gating strategy. Quiescent current is death by a thousand cuts.
    • Measure, don't estimate. A $30 current-measurement tool (or a shunt resistor and a scope) beats a week of datasheet arithmetic. Measure each state transition — the wake-up spike is where surprises live.

    A 7-Day Smartwatch: What Aggressive Power Design Looks Like

    The low-power smartwatch project (smartwatch) set brutal targets and hit them: 100µA standby current, 7-day battery life, heart-rate accuracy ±2bpm, 50+ sport modes, 5ATM waterproofing. Hitting 100µA standby with an always-on display-capable device required every trick in the book — the Apollo2-class MCU philosophy applied to a consumer product: aggressive clock gating, display memory that holds the frame without the MCU, sensor sampling scheduled in bursts, and a radio that spends 99.9% of its life asleep.

    The lesson generalizes: the difference between a 1-day and a 7-day device is rarely one big win. It is twenty small wins — each peripheral's sleep mode, each wake-up source, each millisecond shaved off the transmit window — compounded.

    Wireless Selection: A Framework, Not a Favorite

    Engineers tend to have a favorite radio. Resist it. The right wireless technology falls out of five questions:

    • Range? Meters (BLE), tens of meters through walls (WiFi/433MHz), kilometers (LoRa), anywhere with coverage (cellular).
    • Data rate? A few bytes per hour (gas sensor readings) vs. streaming (video doorbell).
    • Power budget? Coin cell for years rules out WiFi and cellular; mains power removes the constraint entirely.
    • Infrastructure? Who provides the network — the customer's WiFi, your LoRa gateways, or a carrier's towers?
    • Cost per unit? A cellular module plus data plan vs. a $1 433MHz transceiver changes the BOM math completely.

    Our projects span the whole map. The gas sensor IoT terminal (HH_GasSensor) uses ESP32 with WiFi — the right call for a mains-or-large-battery device that needs OTA firmware upgrades and cloud device management. The ESP32 air quality collector (env_monitor) similarly leans on WiFi for straightforward data upload. The multi-sensor tracking module (BLE_GSENSOR_GPS_GSM) combines BLE, G-sensor, GPS, and GSM on the QMS7926 platform — short-range phone connectivity plus wide-area cellular tracking, because a tracker that only works near a phone is not a tracker.

    And sometimes the unfashionable choice wins: the marine data transmission unit (BOAT_DTU) uses 433MHz wireless with an STM32F1 for shipboard data aggregation — sub-GHz penetrates the steel environment better than 2.4GHz, the modules cost almost nothing, and the data rate fits the application. The 433MHz security linkage system (433_IPC) pairs STC32G-based 433MHz terminals with cameras and a Flutter mobile app — proving sub-GHz still earns its place in 2026.

    Antennas: The Part Everyone Underestimates

    A perfect radio with a bad antenna is a bad radio. Antenna mistakes we see repeatedly:

    • PCB trace antennas need keep-out zones. Copper pours, batteries, and metal enclosures detune them. The antenna section of the layout is designed first, not squeezed in last.
    • Certification is part of the design. An intentional radiator needs FCC/CE/SRRC certification. Modular-approved radio modules shortcut this enormously — one reason we often recommend pre-certified modules for the first product generation.
    • Test in the real enclosure. Antenna performance measured on a bare PCB means nothing. Measure it in the final housing, held the way users hold it, in the environment where it installs.

    Battery Math and the Last 10%

    Battery capacity is not what the label says. Derate for temperature (cold kills capacity), for discharge rate (pulse loads reduce usable capacity), for self-discharge over the product's lifetime, and for the cutoff voltage of your regulators. Then add margin. The smart lock (smart_lock) — with fingerprint, face, password, IC card, WiFi remote control, and temporary passwords — has to budget for the motor-driven bolt (the pulse load), the always-listening touch wake, and the WiFi bursts, all from batteries the user replaces. Getting the battery math right is what separates a lock that warns "low battery" gracefully from one that strands someone outside.

    Certification: Budget It Early

    Wireless products need radio certification (FCC, CE, SRRC/MIC/TELEC depending on market), and most markets also require EMC and safety testing. Certification is slow — lab slots book weeks out, failures require redesign and retest — so it belongs in the project plan from the start, not as a surprise at the end. Using pre-certified radio modules dramatically simplifies radio certification (in many regimes the module grant transfers), though the final product still needs EMC testing as a complete system. We schedule pre-scan EMC testing on EVT boards: catching a radiated-emissions failure on a $500 pre-scan beats discovering it during the $15,000 formal test.

    Enclosure and Environment: The Forgotten Half of IoT Hardware

    Electrical engineers obsess over schematics and forget that the enclosure is part of the circuit. A sealed outdoor enclosure turns into a greenhouse — internal temperatures 20°C above ambient are normal in direct sun, which derates battery capacity, shifts crystal frequencies, and pushes regulators toward thermal shutdown. Venting helps, but every vent is a path for water and dust; Gore-style breathable membranes balance the two for genuinely outdoor products.

    Material choice matters for the radio too. A metal enclosure around a 2.4GHz antenna is a Faraday cage with extra steps — either the antenna goes outside the metal, or the enclosure gets a plastic RF window. We specify the enclosure material alongside the PCB stackup, not after the mechanical design is "done," because moving an antenna late means re-tuning, re-testing, and sometimes re-certifying. The smart doorbell (smart_doorbell), with its camera, PIR sensor, and WiFi living behind a weather-exposed faceplate, is a good example of a product where the enclosure, the optics, and the RF design had to be solved as one problem.

    Conclusion

    IoT hardware design is applied physics: duty-cycle the power budget until the battery math closes, pick the radio from requirements rather than habit, design the antenna as a first-class citizen, and derate the battery like a pessimist. The devices that survive the field — the 7-day watch, the terminal that sleeps at microamps, the tracker that roams on cellular — all share the same trait: their two numbers were computed before the schematic was drawn.

    If you are designing a connected device — sensor node, tracker, smart home product — our IoT development service covers hardware, firmware, and cloud together. For the system-level view, see our IoT system architecture guide; for the AI-on-device angle, see AI meets IoT.

    Need Professional Services?

    Contact us for customized solutions and quotes