🎵 Audio Spectrum Analyzer

Real-Time FFT-Based Audio Visualization on STM32F411

A real-time audio spectrum analyzer that visualizes sound frequencies on an OLED display. Built with STM32F411, INMP441 MEMS microphone, and SSD1306 display.

✨ Features

Real-Time FFT Processing using ARM CMSIS-DSP (512-point)
8-Band Frequency Visualization with logarithmic scaling
Peak Hold Effect for smooth, professional-looking bars
DMA-Driven Audio Capture for zero CPU overhead
30+ FPS Display Update rate
Low Latency (~40ms total pipeline)

📸 Demo

┌────────────────────────────────────┐
│  OLED Display (128x64)             │
│                                    │
│  ████                              │
│  ████  ████  ████                  │
│  ████  ████  ████  ████            │
│  ████  ████  ████  ████  ████      │
│  ████  ████  ████  ████  ████  ████│
│  ────  ────  ────  ────  ────  ────│
│   0     1     2     3     4     5  │
│ Bass   Mid   Treble                │
└────────────────────────────────────┘

(Add actual photo/GIF of your working project here)

🛠️ Hardware Requirements

Component	Model	Quantity	Price (₹)
Microcontroller	STM32F411 (Black Pill)	1	450
MEMS Microphone	INMP441	1	180
OLED Display	SSD1306 (128x64, SPI)	1	350
Breadboard	Standard	1	50
Jumper Wires	M-M, M-F	20	30
Total			₹1,060

Optional (Recommended):

Logic Analyzer (~₹800) - for I2S debugging
Oscilloscope access (university lab)

📐 Wiring Diagram

INMP441 → STM32F411

INMP441          STM32F411
VDD     ────────► 3.3V
GND     ────────► GND
WS      ────────► PA4  (I2S2_WS)
SCK     ────────► PB13 (I2S2_CK)  ← NOT PB10 (conflict with I2C2)
SD      ────────► PC3  (I2S2_SD)
L/R     ────────► GND  (Left channel)

SSD1306 → STM32F411 (I2C2)

SSD1306          STM32F411
VCC     ────────► 3.3V
GND     ────────► GND
SCL     ────────► PB10 (I2C2_SCL)
SDA     ────────► PB3  (I2C2_SDA)

🚀 Quick Start

1. Clone or Download

git clone https://github.com/yourusername/audio-spectrum-analyzer.git
cd audio-spectrum-analyzer

2. Open in STM32CubeIDE

File → Open Projects from File System
Select the project folder
Click Finish

3. Configure STM32CubeMX

Open the .ioc file and verify:

I2S2 Settings:

Mode: Master Receive
Standard: Philips I2S
Data Format: 24 bits
Sample Rate: 44.1 kHz
DMA: Circular mode, High priority

I2C2 Settings:

Speed Mode: Fast Mode (400 kHz)
Clock Speed: 400000
Addressing Mode: 7-bit

4. Enable CMSIS-DSP

Critical step - don't skip!

Right-click project → Properties
C/C++ Build → Settings → MCU Settings
Check "Use float with printf"
Project → Manage Middleware → CMSIS → DSP ✅
Add defines: ARM_MATH_CM4, __FPU_PRESENT=1

5. Build & Flash

# In STM32CubeIDE
Project → Build All (Ctrl+B)
Run → Debug (F11)

6. Test!

Power on STM32
Play music near the microphone
Watch the spectrum bars dance! 🎶

📂 Project Structure

Sound_analyzer/
├── README.md                ← You are here
├── README_SETUP.md          ← Detailed setup guide
├── LICENSE                  ← MIT License
├── .gitignore               ← Git ignore rules
│
├── MIC/                     ← I2S microphone driver
│   ├── i2s_mic.h
│   └── i2s_mic.c
│
├── OLED/                    ← SSD1306 display driver (I2C)
│   ├── ssd1306.h / .c
│   ├── ssd1306_fonts.h / .c
│   ├── ssd1306_conf.h       ← Active config (I2C2)
│   └── ssd1306_conf_i2c.h   ← I2C config backup
│
├── FFT/                     ← FFT processing (CMSIS-DSP)
│   ├── fft_processor.h / .c ← CMSIS-DSP version (~1.5 ms)
│   └── fft_processor_simple.h / .c  ← Fallback (~15 ms)
│
├── SPECTRUM/                ← Frequency band analyzer
│   ├── spectrum.h
│   └── spectrum.c
│
├── DISPLAY/                 ← Spectrum visualization
│   ├── spectrum_display.h
│   └── spectrum_display.c
│
└── MAIN_EXAMPLE/            ← Integration examples
    ├── main_integration_example.c  ← Primary (I2C2 OLED)
    └── main_i2c_version.c          ← Full self-contained version

🧠 How It Works

Signal Flow

┌─────────────┐
│  INMP441    │  I2S Digital Microphone
│  24-bit     │  44.1 kHz sampling
└──────┬──────┘
       │ I2S DMA (512 samples)
       ▼
┌─────────────┐
│ DMA Buffer  │  Circular buffer, double-buffering
│ 512 samples │
└──────┬──────┘
       │ Half/Full transfer interrupts
       ▼
┌─────────────┐
│ FFT (CMSIS) │  512-point FFT with Hanning window
│ ~1.5ms      │  Magnitude spectrum (256 bins)
└──────┬──────┘
       │ Frequency domain data
       ▼
┌─────────────┐
│  Frequency  │  Map 256 bins → 8 logarithmic bands
│  Mapping    │  Bass, Mid, Treble separation
└──────┬──────┘
       │ Bar heights[8]
       ▼
┌─────────────┐
│  Display    │  Draw bars on SSD1306 via SPI
│  30+ FPS    │  Peak hold + smoothing effects
└─────────────┘

Frequency Band Mapping

Bar	Freq Range	Description
0	86-172 Hz	Sub-bass (kick drum)
1	258-431 Hz	Bass guitar, bass notes
2	517-861 Hz	Low midrange (male vocals)
3	947-1550 Hz	Midrange (vocals, guitar)
4	1636-2584 Hz	Upper midrange (female vocals)
5	2670-4307 Hz	Presence (clarity, definition)
6	4393-6890 Hz	Brilliance (cymbals, hi-hats)
7	6977-10336 Hz	Air (sparkle, ambience)

⚙️ Configuration Options

Adjust Sensitivity

Edit SPECTRUM/spectrum.c:

#define SCALE_FACTOR    3.5f    // Increase for taller bars (try 2.0-6.0)
#define MIN_THRESHOLD   0.01f   // Lower for more sensitivity

Change Smoothing

#define SMOOTHING_FACTOR    0.7f    // 0.0 = no smooth, 0.9 = very smooth
#define PEAK_DECAY_RATE     0.95f   // How fast peaks fall (0.9-0.99)

Customize Display Style

In main loop:

// Simple bars
DISPLAY_DrawBars(bar_heights);

// Bars with peak hold (recommended)
DISPLAY_DrawBarsWithPeaks(bar_heights, peaks);

// Custom style
DISPLAY_DrawSpectrum(bars, peaks, STYLE_OUTLINED_BARS);

📊 Performance Metrics

Metric	Value
Frame Rate	30+ FPS
FFT Processing Time	~1.5 ms
Display Update Time	~15 ms
Total Latency	~40 ms
CPU Usage	~60% @ 100 MHz
Memory (RAM)	~12 KB
Flash Usage	~45 KB

🐛 Troubleshooting

Display shows nothing

Check SPI wiring (SCK, MOSI, CS, DC, RST)
Verify ssd1306_conf.h pin definitions match your wiring
Test with ssd1306_TestAll() function

"Undefined reference to arm_cfft_f32"

Enable CMSIS-DSP library in project settings
Add ARM_MATH_CM4 and __FPU_PRESENT=1 defines
Rebuild project (clean first)

All bars at zero

Check I2S wiring (WS, SCK, SD)
Verify microphone VDD = 3.3V (multimeter)
Check DMA is in circular mode
Print mic_data_ready flag to verify data arrival

Bars jumping erratically

Increase SMOOTHING_FACTOR (try 0.8)
Adjust SCALE_FACTOR (try 2.0-5.0)
Check for EMI/noise near microphone

See README_SETUP.md for detailed troubleshooting.

🎓 Technical Deep Dive

Why Hanning Window?

Without windowing, rectangular truncation of the signal causes spectral leakage - energy from one frequency bin spills into adjacent bins, creating false peaks.

The Hanning window smoothly tapers the signal to zero at edges:

w[n] = 0.5 * (1 - cos(2πn/N))

This reduces sidelobes by ~31 dB, giving cleaner frequency separation.

Why Logarithmic Band Grouping?

Human hearing is logarithmic - we perceive octaves (doubling of frequency) as equal intervals. Mapping FFT bins logarithmically creates a more musically-relevant display.

Linear spacing would waste resolution on high frequencies we barely hear.

DMA Double Buffering

The 512-sample buffer is split into two halves:

Half-transfer interrupt: Process samples 0-255 while DMA fills 256-511
Full-transfer interrupt: Process samples 256-511 while DMA wraps to 0-255

This zero-copy design prevents audio dropouts and maximizes throughput.

📚 Resources & References

Documentation

Learning Resources

🔮 Future Enhancements

Potential features to add:

Beat Detection - Flash on bass hits
Bluetooth Audio Input - Analyze phone music
RGB LED Strip - WS2812B color visualization
BPM Counter - Detect tempo from bass frequencies
Multiple Display Modes - Waveform, waterfall, VU meter
Auto-Gain Control - Adapt to loud/quiet environments
PC Companion App - Python GUI for tuning parameters

🤝 Contributing

Pull requests welcome! Areas for improvement:

Optimize FFT for lower latency
Add alternative windowing functions (Hamming, Blackman)
Implement adaptive frequency ranges based on content
Create PCB design for permanent installation
Port to other STM32 families (F0, F7, H7)

📄 License

MIT License - see LICENSE file for details.

👤 Author

Kushal Pitaliya

GitHub: @kushalpitaliya

🙏 Acknowledgments

ARM CMSIS-DSP team for optimized FFT algorithms
afiskon for SSD1306 display library
STMicroelectronics for HAL drivers
InvenSense for INMP441 microphone design
3Blue1Brown for Fourier transform visualization

📈 Project Stats

Lines of Code: ~1,500
Development Time: 8 weeks
Components: 3 major ICs
Total Cost: ₹1,060
Frame Rate: 30+ FPS
Latency: <50 ms

Built with ❤️ using STM32, CMSIS-DSP, and lots of ☕

📌 Quick Links

Setup Guide - Detailed installation instructions
Implementation Plan - Development roadmap
Example Code - Complete integration
Troubleshooting - Common issues & fixes

⭐ Star this repo if you found it useful!

Name		Name	Last commit message	Last commit date
.github/workflows		.github/workflows
DISPLAY		DISPLAY
FFT		FFT
MAIN_EXAMPLE		MAIN_EXAMPLE
MIC		MIC
OLED		OLED
SPECTRUM		SPECTRUM
docs		docs
.gitignore		.gitignore
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
SECURITY.md		SECURITY.md
audio_config.h		audio_config.h

Latest commit

History

Folders and files

Repository files navigation