ESP32 FTMS Indoor Bike Sensor

Arduino IDE

Arduino® Nano ESP32 with headers

1 Potentiometer 10K

<html><body>Introduction <figure><iframe src="https://www.youtube.com/watch?v=H2Jkb1zzhYA"></iframe></figure> This project converts a conventional magnetic resistance exercise bike into a smart FTMS-compatible indoor bike using an ESP32. The system measures real-time pedaling cadence through a magnetic reed switch sensor, applies an adaptive EMA filtering algorithm to stabilize RPM readings, estimates rider power output using a quadratic mathematical model, and transmits cadence, speed, and power data over Bluetooth Low Energy (BLE) using the official FTMS standard. The firmware was specifically designed to improve compatibility with indoor cycling applications such as MyWhoosh, Zwift, Kinomap, and other FTMS-compatible platforms while maintaining smooth and realistic sensor behavior. Main Features <ol><li>Real-time cadence acquisition using a magnetic sensor </li><li>Adaptive EMA filtering with dynamic α adjustment based on RPM delta </li><li>Quadratic power estimation model </li><li>BLE FTMS Indoor Bike compatibility </li><li>Simulated speed generation for app compatibility </li><li>Resistance level detection using a potentiometer </li><li>Automatic braking and stop detection </li><li>BLE data transmission every 300 ms </li></ol>Instantaneous Cadence Acquisition A magnetic reed switch sensor is mounted on the bicycle frame. The sensor detects the passage of a magnet attached to the crank pulley. Most magnetic resistance exercise bikes are already equipped with a reed switch cadence sensor connected to the factory console. Therefore, the original sensor can be reused without any mechanical modifications. The only required step is to unplug the two sensor wires from the original console and connect them to the Arduino Nano ESP32, which will then read the cadence pulses directly.Each time the crank completes one full revolution, the magnet activates the sensor and generates a discrete electrical pulse. This creates a sequence of time events: <code class="code">t₀, t₁, t₂, ..., tₙ</code>To calculate instantaneous pedaling cadence, the firmware measures the elapsed time between the current pulse (tₙ) and the previous pulse (tₙ₋₁): <code class="code">Δt = tₙ - tₙ₋₁</code>Since time is measured in microseconds for maximum precision, cadence in revolutions per minute (RPM) is calculated as: <code class="code">Cadence (RPM) = 60,000,000 / Δt</code><figure><img src="https://projects.arduinocontent.cc/07095519-6516-4f00-b495-a97b0c47dff0.jpg"/><figcaption></figcaption></figure> Cadence Filtering Directly transmitting raw RPM values produces abrupt fluctuations and unstable visual behavior inside cycling applications. To smooth the signal, the firmware implements an EMA (Exponential Moving Average) filter. EMA Filter Unlike a simple average, the EMA filter gives greater weight to recent samples while progressively reducing the influence of older values. <code class="code">RPMfiltered(n) = α · RPMinstant(n) + (1 - α) · RPMfiltered(n - 1)</code>Where: α is the smoothing factor (0 &lt; α &lt; 1) High α → fast response but more noise Low α → smoother signal but slower response Adaptive Improvement (Delta-Based α) To avoid choosing between smoothness and responsiveness, the firmware dynamically adjusts the value of α according to RPM variation intensity. First, the algorithm calculates the absolute RPM difference between the current and previous pedal stroke: <code class="code">δₙ = | RPMinstant(n) - RPMinstant(n - 1) |</code>Then, α is automatically adjusted: RPM Delta	Behavior δₙ &lt; 3	Maximum smoothing → α = 0.12 3 ≤ δₙ &lt; 10	Progressive transition 10 ≤ δₙ &lt; 20	Fast response δₙ ≥ 20	Instant response → α = 0.95 Interpolation formulas: α = 0.12 + ((δₙ - 3) / 7) · 0.23 α = 0.35 + ((δₙ - 10) / 10) · 0.45 Internal Alpha Smoothing To prevent abrupt jumps in the smoothing factor itself, α is also stabilized using a second internal EMA filter: <code class="code">αsmooth(n) = 0.70 · α(δₙ) + 0.30 · αsmooth(n - 1)</code>This significantly improves visual stability while preserving fast responsiveness during accelerations and sprint efforts. Power Estimation The exercise bike does not include physical torque sensors or strain gauges. Therefore, rider power output is mathematically estimated using: <ol><li>Filtered cadence (RPM) </li><li>User-selected magnetic resistance level </li></ol>Magnetic Resistance Principle The braking system operates using Eddy Currents. As the metallic flywheel rotates, it cuts through the magnetic field generated by permanent magnets. This induces microscopic electrical currents inside the flywheel, creating an opposing magnetic field that generates resistance without physical contact. The user manually changes braking intensity by modifying the distance between magnets and flywheel. Mathematical Power Model Magnetic braking force increases approximately linearly with flywheel speed. However, since power equals force multiplied by velocity, rider effort grows quadratically. The firmware implements the following model: <code class="code">Power (W) = (K · RPM²) + (C · RPM)</code>Where: K → quadratic coefficient associated with magnetic braking C → linear coefficient associated with mechanical friction losses Resistance Level Detection To indicate the resistance level, a potentiometer is installed on the bike's console. The cyclist manually sets the resistance level they are riding at, and this value is used by the ESP32 to estimate braking force and power output. As with any manually entered parameter, accurate results depend on the resistance setting matching the actual riding conditions.The ESP32 reads this value through its 12-bit ADC: ADC Range: 0 → 4095 To reduce electrical noise, the firmware averages 8 consecutive samples before converting them into 8 discrete resistance levels. <code class="code">Level = constrain( ((ADC - 120) / (3980 - 120) * 8) + 1, 1, 8 )</code>Dynamic Coefficient Table Each resistance level uses specific K and C coefficients: Level	K	C 1	0.006	0.05 2	0.009	0.07 3	0.012	0.10 4	0.017	0.14 5	0.022	0.18 6	0.030	0.24 7	0.039	0.30 8	0.050	0.38 Calculation Example Example: Resistance level: 4 Cadence: 80 RPM K = 0.017 C = 0.14 Quadratic component: 0.017 × 80² = 108.8 W Linear component: 0.14 × 80 = 11.2 W Estimated total power: 120 W BLE FTMS Implementation <figure><img src="https://projects.arduinocontent.cc/dba85e58-3393-43d9-b1e7-095b3cb9af7c.jpg"/><figcaption>MyWhoos Free APP</figcaption></figure> The ESP32 implements the official FTMS profile defined by Bluetooth SIG. Every 300 ms, an Indoor Bike Data packet is automatically transmitted using BLE notifications. FTMS Packet Structure The packet contains three main fields declared through the 0x0044 flag mask: Bytes	Field 0–1	Flags 2–3	Simulated speed 4–5	Cadence 6–7	Power All values are transmitted using little-endian format. Cadence Encoding The FTMS standard uses a 0.5 RPM resolution. Therefore, cadence is transmitted as: cadenceValue = cadence × 2 The application internally divides the received value by two. Simulated Speed Some cycling applications reject FTMS packets if the speed field is absent. Since the bike does not include a physical speed sensor, speed is artificially derived from cadence: speedValue = cadence × 30 This improves compatibility with third-party applications. BLE Notify vs Read The ESP32 uses the BLE notify mechanism. Once the application subscribes to the Indoor Bike Data characteristic, the ESP32 automatically transmits new packets every 300 ms without requiring manual requests. If no device is connected, transmission is automatically skipped. </body></html>