Sensor Integration Best Practices¶

Sensors lie. They drift, they glitch, they pick up interference, and they have specs that only hold at 25 °C with a stable 3.3 V supply. These practices help you get useful data anyway.

Interface Selection¶

Prefer I2C or SPI for new designs. Analog is messy.

I2C Wiring Best Practices¶

ESP32          Sensor
─────          ──────
GPIO21 (SDA) ──┬── SDA
GPIO22 (SCL) ──┼── SCL
               │
            [4.7 kΩ to 3.3 V]   ← required external pull-ups
               │
             3.3 V rail

• Always use external 4.7 kΩ pull-ups to 3.3 V (not 5 V). Internal ~45 kΩ pull-ups are too weak for reliable I2C.
• Bus capacitance limit: keep total trace/wire length < 1 m for 400 kHz operation. Longer runs → lower speed.
• Address conflicts: many cheap sensors ship with fixed I2C addresses. When two sensors share an address, use an I2C multiplexer (TCA9548A) or separate I2C buses.
• Scan at startup to verify all sensors are present:

void i2cScan() {
    for (uint8_t addr = 1; addr < 127; addr++) {
        Wire.beginTransmission(addr);
        if (Wire.endTransmission() == 0) {
            Serial.printf("Found device at 0x%02X\n", addr);
        }
    }
}

ESP32 ADC Caveats¶

The ESP32's onboard ADC is notoriously inaccurate: - Non-linear — especially below 150 mV and above 3.0 V (on a 3.3 V supply) - Reference drift — varies with temperature and supply voltage - ADC2 conflicts with WiFi — ADC2 cannot be used while WiFi is active; use ADC1 pins (GPIO32–GPIO39) only - Effective resolution: ~11 bits usable despite 12-bit register

For precision analog measurements: - Use an external ADC (ADS1115 over I2C — 16-bit, differential, true reference) - Or use the internal ADC with calibration compensation:

#include <esp_adc_cal.h>
esp_adc_cal_characteristics_t adcCal;
esp_adc_cal_characterize(ADC_UNIT_1, ADC_ATTEN_DB_11,
                         ADC_WIDTH_BIT_12, 1100, &adcCal);
uint32_t voltage_mV = esp_adc_cal_raw_to_voltage(analogRead(A0), &adcCal);

Calibration Strategy¶

Every sensor needs calibration. "Factory calibration" from a $2 module is optimistic.

Two-point calibration (most common):

// Map raw reading to physical units using two known reference points
float calibrate(int raw, int rawLow, int rawHigh, float physLow, float physHigh) {
    return physLow + (float)(raw - rawLow) / (rawHigh - rawLow) * (physHigh - physLow);
}

// Example: soil moisture sensor
// rawDry=3200, rawWet=1200, physDry=0%, physWet=100%
float moisture = calibrate(analogRead(SOIL_PIN), 3200, 1200, 0.0f, 100.0f);

Store calibration constants in NVS so they survive firmware updates.

Debouncing Digital Inputs¶

Mechanical switches bounce — they register multiple edges for a single press.

Software debounce (simple):

const uint32_t DEBOUNCE_MS = 50;
bool lastState = HIGH;
uint32_t lastChangeMs = 0;

void checkButton() {
    bool state = digitalRead(BTN_PIN);
    if (state != lastState && (millis() - lastChangeMs) > DEBOUNCE_MS) {
        lastState = state;
        lastChangeMs = millis();
        if (state == LOW) onButtonPressed();
    }
}

For interrupt-driven debounce: set a timer in the ISR; process the stable state when the timer fires.

Filtering Noisy Data¶

Raw sensor readings fluctuate. Filtering smooths the signal without adding significant latency.

Simple moving average (fixed window, low lag):

template<typename T, size_t N>
class MovingAverage {
    T buf[N] = {};
    size_t idx = 0;
    T sum = 0;
public:
    T update(T val) {
        sum -= buf[idx];
        buf[idx] = val;
        sum += val;
        idx = (idx + 1) % N;
        return sum / N;
    }
};

MovingAverage<float, 8> tempFilter;
float smoothTemp = tempFilter.update(rawTemp);

Exponential moving average (lower memory, adjustable decay):

class EMA {
    float alpha;  // 0 < alpha < 1; higher = faster response
    float value;
    bool first = true;
public:
    explicit EMA(float alpha) : alpha(alpha), value(0) {}
    float update(float x) {
        if (first) { value = x; first = false; }
        else value = alpha * x + (1 - alpha) * value;
        return value;
    }
};

EMA humidityFilter(0.2f);  // alpha=0.2 → smooth, slow to react
float smoothHum = humidityFilter.update(rawHum);

Use EMA for slow-changing values (temperature, humidity, pressure). Use a sliding window for sensors with occasional spikes (light, soil moisture).

Shielded Wiring and Noise¶

For long sensor cable runs or electrically noisy environments (near motors, switching supplies):

• Use shielded twisted pair (STP) cable; connect shield to GND at one end only
• Add series resistors (100–300 Ω) on I2C/SPI lines to dampen reflections
• Keep analog signal wires away from power wires and PWM lines
• Add decoupling capacitors (100 nF) on sensor VCC pins, close to the sensor
• Ferrite beads on power lines if switching noise is an issue

See Also¶