The Garden Sensor Noise Filter

JSS 2 MakeCode micro:bit Simulation first | Rolling Sensor Filter

📑 Project sections ▾

Simulation lab Build guide Learn & Story Test model Try a challenge Quick check & Quiz

Simulation lab

Board: BBC micro:bit

⛶ Expand Open MakeCode ↗

Simulator preview

☀️ 80

Change the slider level and observe the 5x5 LED display.

📁 Tools ▴

• Web browser (Chrome, Edge, Firefox)
• Internet connection
• MakeCode (micro:bit) online editor

📖 Learn 🔧 Build </> Code ▷ Test ⚑ Challenge ❓ Quiz

🔧 Build Your Rolling Sensor Filter

1 Connect an analog soil moisture probe or potentiometer to analog input A0.

2 Declare a 3-element integer array 'readings' to hold recent sample history.

3 On each loop cycle, subtract the oldest reading, insert the newest reading, and compute the average.

4 Plot raw noisy readings against smoothed filtered values in the Serial Plotter to verify spike rejection!

🧩 Block plan / Starter code ▴

const int SOIL_PIN = A0;
const int NUM_READINGS = 3;

int readings[NUM_READINGS] = {500, 500, 500};
int readIndex = 0;
long total = 1500;
int filteredAverage = 500;

void setup() {
  Serial.begin(9600);
  Serial.println("Raw,Filtered");
}

void loop() {
  int rawVal = analogRead(SOIL_PIN);

  // Rolling Window Moving Average
  total = total - readings[readIndex];
  readings[readIndex] = rawVal;
  total = total + readings[readIndex];
  readIndex = (readIndex + 1) % NUM_READINGS;

  filteredAverage = total / NUM_READINGS;

  Serial.print(rawVal);
  Serial.print(",");
  Serial.println(filteredAverage);
  delay(300);
}

The rolling average algorithm attenuates sudden transient spikes while preserving true analog trends.

⚗️ Test your model

Test Condition / Action	Expected Result
Transient Electrical Spike (900 raw)	Filtered output only rises mildly to 630; prevents false alarm
Sustained Dry Soil Transition	Filtered output smoothly matches true dry condition over 3 samples
Noise-free Steady Input	Filtered output tracks raw input perfectly without offset

⚑ Try a challenge

Increase filter size from 3 samples to 7 samples and compare noise reduction versus response lag!

❓ Quick check

What causes 'Sensor Noise' in electronic soil moisture and agricultural sensors?

• Electromagnetic interference, loose soil contact, wind vibrations, and power supply voltage ripples
• The soil being too dark in color

Check answer

📖 Our Story & Concepts

Analog sensors in agricultural fields, irrigation pumps, and factories suffer from environmental electrical noise: motor start-ups induce inductive voltage spikes, and loose wires generate transient artifacts. If an irrigation system shut off whenever a 10ms spike occurred, crops would wither.

Digital signal processing (DSP) filters out noise mathematically. A Simple Moving Average (SMA) acts as a low-pass filter, smoothing out rapid high-frequency noise while letting the genuine, slow physical process shine through cleanly.

</> Full Annotated Code Walkthrough

// Complete Circular Buffer Filter with Threshold Pump Control
const int SOIL_PIN = A0;
const int PUMP_PIN = 7;
const int WINDOW_SIZE = 5;

int buffer[WINDOW_SIZE];
int bufIndex = 0;
long runningSum = 0;

void setup() {
  pinMode(PUMP_PIN, OUTPUT);
  Serial.begin(115200);
  for (int i = 0; i < WINDOW_SIZE; i++) {
    buffer[i] = analogRead(SOIL_PIN);
    runningSum += buffer[i];
  }
}

void loop() {
  int rawVal = analogRead(SOIL_PIN);

  runningSum -= buffer[bufIndex];
  buffer[bufIndex] = rawVal;
  runningSum += buffer[bufIndex];
  bufIndex = (bufIndex + 1) % WINDOW_SIZE;

  int smoothed = runningSum / WINDOW_SIZE;

  // Actuator triggers strictly on filtered data
  if (smoothed < 350) {
    digitalWrite(PUMP_PIN, HIGH);
  } else {
    digitalWrite(PUMP_PIN, LOW);
  }

  Serial.print("Raw:"); Serial.print(rawVal);
  Serial.print(",Filtered:"); Serial.println(smoothed);
  delay(200);
}

The rolling average algorithm attenuates sudden transient spikes while preserving true analog trends.

▷ Expected Results & Troubleshooting

Test Condition / Action	Expected Result
Transient Electrical Spike (900 raw)	Filtered output only rises mildly to 630; prevents false alarm
Sustained Dry Soil Transition	Filtered output smoothly matches true dry condition over 3 samples
Noise-free Steady Input	Filtered output tracks raw input perfectly without offset

⚑ Extension Challenges & Blueprints

⚑ Extension Challenges

• Level 1: Add a median filter that rejects the absolute highest and lowest outlier in every sample window.
• Level 2: Display both raw and smoothed signals on an I2C OLED display screen.
• Level 3: Implement an exponential moving average (EMA) using floating-point alpha weighting.

❓ Check Your Understanding

Question 1: What causes 'Sensor Noise' in electronic soil moisture and agricultural sensors?

• Electromagnetic interference, loose soil contact, wind vibrations, and power supply voltage ripples
• The soil being too dark in color
• The micro:bit battery having too much charge
• The plants drinking water too quietly

Question 2: If three consecutive raw soil readings are 30, 90 (noise spike), and 30, what is their 3-point Moving Average?

• 50 ((30 + 90 + 30) / 3 = 150 / 3 = 50)
• 90
• 30
• 150

Question 3: Why is acting directly on raw unfiltered sensor data dangerous for an automated farm irrigation pump?

• A brief 1-second noise spike could trigger the pump unnecessarily, flooding crops and wasting precious water
• The pump will run in reverse
• The soil will turn to concrete
• The computer will run out of storage space

Question 4: What is the primary advantage of a rolling FIFO (First-In, First-Out) buffer when filtering stream data?

• It always retains the newest samples while discarding the oldest sample, providing smooth real-time filtering without lag
• It doubles the memory capacity of the chip
• It encrypts the sensor data with RSA
• It eliminates the need for battery power

Check answer Try again Feedback appears after you check your answer.

← Back to project overview Next project →