[Leejohnston]

Thread 10 — Build Your Own Digital Spectrum Analyzer 
A Complete Beginner-Friendly Project That Teaches FFT, Audio Sampling, and DSP

A spectrum analyzer takes in sound (or any signal), runs a Fast Fourier Transform (FFT), 
and displays the frequency intensities in real time.

Phones, music software, audio mixers, smart speakers — they ALL use this.

In this thread, you will learn:
• how FFT reveals hidden frequencies 
• how to capture audio with a microcontroller 
• how to process signals digitally 
• how to display a live spectrum 
• how to build your own DIY analyzer from scratch 

This is a full hands-on engineering project.

---

1. What This Project Does

Your finished device will:
• read audio through a microphone module 
• sample the waveform in real time 
• run an FFT on blocks of samples 
• convert the result into frequency bins 
• display the spectrum on LEDs or an OLED screen 

You can visualize:
• voice frequencies 
• bass, mids, treble 
• environmental noise 
• musical tones 
• claps, whistles, vibrations 

---

2. Parts You Need (Beginner-Friendly Kit)

All parts are inexpensive and safe:

• ESP32 or Arduino Nano RP2040 (recommended for speed) 
• MAX4466 or KY-037 microphone module 
• 128×64 OLED display (I2C) OR WS2812 LED strip 
• Jumper wires 
• USB cable 
• Breadboard 

Optional:
• 3D-printed enclosure 
• Li-ion battery 
• Charging module 

Total cost: £15–£25 depending on parts.

---

3. The Core Idea — Sampling + FFT

To analyse sound digitally we must:

(1) SAMPLE → convert analog audio → numbers 
(2) BUFFER → store N samples 
(3) FFT → convert time data → frequency data 
(4) DISPLAY → show bars for each frequency band 

Most projects use:
N = 256 or 512 samples 
Perfect for real-time performance on microcontrollers.

---

4. Wiring Diagram (Simple)

Microphone → ADC Input 
• Mic OUT → A0 (or ADC pin) 
• VCC → 3.3V 
• GND → GND 

OLED → I2C 
• SDA → SDA 
• SCL → SCL 
• VCC → 3.3V 
• GND → GND 

That’s it — simple wiring.

---

5. Sample Rate Setup

To capture frequencies up to ~4 kHz, 
you need a sampling rate of:

Fs = 8000 Hz

This works great for voices and environmental sound.

---

6. Core Code (ESP32 Example)

Below is a clean, ready-to-run FFT spectrum analyzer sketch.

This uses:
• ArduinoFFT library 
• I2C OLED 
• 256-sample FFT window

Code:

// ==== LIBRARIES ====
#include <ArduinoFFT.h>
#include <Wire.h>
#include <Adafruit_SSD1306.h>

#define SAMPLES 256
#define SAMPLING_FREQUENCY 8000

double vReal[SAMPLES];
double vImag[SAMPLES];

ArduinoFFT FFT = ArduinoFFT(vReal, vImag, SAMPLES, SAMPLING_FREQUENCY);

Adafruit_SSD1306 display(128, 64, &Wire);

// ==== SETUP ====
void setup() {
  display.begin(SSD1306_SWITCHCAPVCC, 0x3C);
  display.clearDisplay();
  display.display();
 
  analogReadResolution(12);  // ESP32 high-quality ADC
  analogSetPinAttenuation(34, ADC_11db);
}

// ==== MAIN LOOP ====
void loop() {
  // 1. Collect samples
  for (int i = 0; i < SAMPLES; i++) {
    vReal[i] = analogRead(34);
    vImag[i] = 0;
    delayMicroseconds(1000000 / SAMPLING_FREQUENCY);
  }

  // 2. Apply a Hanning window
  FFT.Windowing(FFT_WIN_TYP_HANN, FFT_FORWARD);

  // 3. Compute the FFT
  FFT.Compute(FFT_FORWARD);

  // 4. Convert to magnitudes
  FFT.ComplexToMagnitude();

  // 5. Display the spectrum
  display.clearDisplay();

  for (int i = 2; i < 64; i++) {
    int barHeight = map(vReal[i], 0, 2000, 0, 60);
    display.drawLine(i*2, 63, i*2, 63 - barHeight, WHITE);
  }

  display.display();
}

This produces a live moving bar-graph spectrum.

---

7. Understanding the Frequency Bins

For N=256 samples and Fs=8000 Hz:

Frequency resolution = Fs / N 
= 8000 / 256 
≈ 31.25 Hz per bin

Examples:
• bin 5 ≈ 156 Hz 
• bin 10 ≈ 312 Hz 
• bin 20 ≈ 625 Hz 
• bin 40 ≈ 1250 Hz 

Users can identify:
• bass frequencies 
• human voice range 
• whistles & alarms 
• fan noise (400–1200 Hz) 
• wide environmental noise 

---

8. Add a Logarithmic Display (Better for Human Hearing)

Human ears are logarithmic — so you group FFT bins 
into octave or semi-octave bands.

Example grouping:
• 0–200 Hz 
• 200–400 Hz 
• 400–800 Hz 
• 800–1600 Hz 
• 1600–3200 Hz 

Let beginners understand professional audio tools.

---

9. Real Upgrades You Can Add Later

• RGB LED matrix visualizer 
• Real-time noise reduction 
• Bluetooth streaming to your phone 
• Microphone beamforming 
• Displaying a full spectrogram 
• Adding a waterfall effect 
• Recording peaks & analysing environments 
• Using DSP to classify sounds (machine learning) 

This project scales all the way up to professional engineering.

---

10. What You Learned

By completing this thread you learned:

• how sampling works 
• how FFT converts signals into frequencies 
• how filters and windows improve signal quality 
• how microcontrollers do real DSP 
• how engineering projects are built from theory → hardware → code 
• how to build a real working spectrum analyzer 

This is core knowledge for:
• audio engineering 
• robotics 
• astronomy 
• seismology 
• communication systems 
• machine learning 
• embedded systems design 

---

End of Thread — Build Your Own Digital Spectrum Analyzer 
Written by LeeJohnston & Liora — The Lumin Archive Research Division