Skip to content

Spectrogram

Spectrogram

Overview

A tool that displays sound components in three dimensions: "Time (horizontal axis)," "Frequency (vertical axis)," and "Strength (color)". While a spectrum analyzer displays the "frequency distribution at the current moment," a spectrogram visualizes the ever-changing "audio signature (voiceprint)". It is ideal for observing the "transitions" of sound, such as voice intonation analysis, time-based changes in instrument overtones, bird call analysis, and discovering intermittent noise.

Operation

Starting and Stopping Measurements

  • Start / Stop Button: Toggles the measurement on and off.
    • When started, new data appears from the right edge of the graph, and old data flows to the left (waterfall display).

Understanding the Graph

  • Horizontal Axis (Time): Represents the passage of time. The current time is at the right edge, and going left represents past sounds.
  • Vertical Axis (Frequency): Represents frequency (pitch). Higher up means higher frequency.
  • Color: Represents the "strength" of the sound at that moment and frequency.
    • Bright colors (yellow, red, etc.): Strong sounds
    • Dark colors (blue, purple, black, etc.): Weak sounds or silence

Understanding the Distribution Map (Color Bar) on the Right

This vertical bar on the right side of the graph is a "correspondence table between color and volume (dB)" and also acts as a controller to adjust the brightness of the display.

  • Meaning of Colors: The top of the bar shows colors corresponding to strong sounds (near 0dB), and the bottom shows colors for weak sounds (near -120dB). Use it as a legend to understand "this color on the graph is roughly this volume."
  • Adjusting the Display Range (Contrast Adjustment):
    • Drag the white handles (triangle marks): You can adjust the range of colors displayed.
    • Changing the Overall Brightness: Drag the entire bar up or down, or scroll with the mouse wheel.
    • Changing the Contrast (Sharpness): Widen or narrow the width between the handles. Narrowing the width causes colors to change rapidly with small volume differences, making subtle changes easier to see.

Settings

Settings (Basic Settings)

  • Channel

    • Left / Right / Average: Select the audio channel to analyze.

  • FFT Size (Frequency Resolution)

    • Configures the granularity of the analysis.
    • 4096 / 8192 etc.: Finer frequency resolution, but time-domain response becomes slightly blurred.
    • 512 / 1024 etc.: Sharper time response (for rhythms, etc.), but frequency resolution becomes coarse.
    • 2048 is generally recommended for a good balance.
    • Note: When a faster "Speed" is selected, the maximum available FFT size is automatically limited to maintain real-time performance.

  • Window (Window Function)

    • hann (standard) or blackman are suitable for noise analysis.

  • Scale

    • Switches the display scale of the vertical axis (frequency).
    • Linear: Equal spacing. Suitable for observing high-frequency details.
    • Log: Logarithmic spacing. Expands the lower frequency range, matching human hearing characteristics. Suitable for musical pitch or low-frequency noise analysis.

  • Colormap

    • Changes the color scheme of the graph.
    • viridis / plasma / inferno / magma / cividis: Scientifically common color schemes where changes in brightness are uniform and easy to see.
    • turbo: A colorful rainbow-like scheme suitable for distinguishing fine level differences.

  • Speed (Flow Rate)

    • Adjusts the speed at which the graph scrolls.
    • Fast (Realtime): Flows in real-time. Suitable for viewing short-term changes.
    • Medium / Slow / Meteor: Flows slowly. Used for monitoring environmental sounds or observing changes over a long period (displaying several to 10 minutes of history on one screen). Note that these slow modes use a "Max Hold" mechanism to capture transient peaks within the time interval.

  • Min Freq / Max Freq

    • Narrows down the display range of the vertical axis (frequency).
    • For example, if you want to see low frequencies in detail, set Max Freq to 1000 Hz.

Usage Examples

Visualizing "Voice" (Voice Analysis)

Try analyzing your own voice or speech.

  1. Press the Start button.
  2. Talk into the microphone saying "ah" or "ee".
  3. You can see the positions of color bands (formants) change depending on the vowel.
  4. If you whistle, a very clear single line (close to a pure tone) appears.

Identifying the Source of Unusual Noises (Noise Identification)

Identify annoying sound components such as "whine" (high-frequency noise) or "hum" (low-frequency power line noise).

  1. Set FFT Size to a large value (4096 to 8192).
  2. Set Speed to about Medium.
  3. Observe the screen while the noise is occurring.
  4. A line extending horizontally: Constant noise such as a fan, motor, or power hum. You can identify the frequency by looking at the vertical axis.
  5. A line running vertically: Impact sounds or click noise.
  6. Hazy fog: Wind noise from air conditioning (white noise), etc.

Checking High-Res Audio Sources (High-Res Audio Check)

Confirm whether the music file being played is truly high-res (containing components up to high frequencies) or if it has simply been upsampled.

  1. Play music.
  2. Expand Max Freq to 48000 Hz (for 96kHz sampling), etc.
  3. Look at the region above 20kHz (20000Hz).
  4. If color is clearly present even above 20kHz, it contains high-res components. If it is cut off sharply at 20kHz and becomes pitch black, it is likely originally CD quality.