Skip to content

FFT Microphone Spectrum Analyzer

Run the FFT Microphone Spectrum Analyzer MicroSim Fullscreen

You can include this MicroSim on your website using the following iframe:

1
2
<iframe src="https://dmccreary.github.io/fft-benchmarking/sims/fft-mic/main.html"
        height="492px" width="100%" scrolling="no"></iframe>

About This MicroSim

Every other spectrum in this book is computed from numbers someone typed in. This one is computed from you.

Click Start microphone, grant the browser permission, and whistle. The tall bar that appears is your voice's fundamental frequency, and it moves the instant your pitch does. That is the whistle test from Chapter 14, running in a browser tab instead of on a Pico.

No microphone, or a locked-down lab machine? Click Demo sweep for a synthetic whistle that glides up and down. Everything below works the same way.

What You Are Looking At

  • The bars are the magnitudes of a 512-bin FFT, colored blue at the low end of the displayed range and red at the high end.
  • The dashed line marks the strongest bin — the peak.
  • The peak trail across the top records where the peak has been over the last second or so. Glide your pitch upward and the trail becomes a comet tail climbing to the right. That trail is the whistle test's evidence.
  • The readout gives the peak in hertz, its bin index, and the nearest musical note with a cents offset.

Things to Try

  1. Whistle a slow glide from low to high. Watch the trail. The peak should track your pitch with no perceptible lag — that is a working real-time pipeline.
  2. Hum instead of whistling. A hum is far richer: you should see evenly spaced harmonics at 2x, 3x, 4x the fundamental. A whistle is nearly a pure tone and shows almost nothing above the fundamental.
  3. Say "ssss" and then "shhhh." Both are noise, not tones, so no single peak dominates. "Ssss" pushes energy much higher in frequency than "shhhh."
  4. Sing a steady note and read the cents offset. Can you hold a note within 10 cents? Most people cannot, and the display is unforgiving about it.
  5. Tick the decibel-scale box. The room's noise floor jumps up from nothing to clearly visible, and the peak stops towering over everything else. Nothing about the sound changed — only the mapping from magnitude to bar height.

Why the Readout Jumps

Whistle a very slow glide and watch the hertz value. It does not slide smoothly; it steps. At a 44.1 kHz sample rate with 512 bins the bin width is about 43 Hz, so the peak can only ever report a multiple of 43 Hz. Everything between two bins gets rounded to one of them.

That step size is exactly the limitation Chapter 15 fixes with parabolic interpolation, which reads the two bins on either side of the peak and estimates where the true frequency falls between them. This MicroSim deliberately does not do that, so you can see the raw resolution the FFT hands you.

Lesson Plan

Grade Level

Undergraduate (college junior/senior)

Duration

15 minutes

Prerequisites

  • An FFT turns a block of samples into a magnitude per frequency bin
  • Bin width equals the sample rate divided by the FFT size
  • Harmonics are integer multiples of a fundamental

Learning Objective

Students will be able to apply their understanding of the magnitude spectrum to real audio by predicting, then verifying, how whistles, hums, and noise differ in the frequency domain.

Activities

  1. Predict then test (5 min): Before touching the microphone, students write down what they expect the spectrum of a whistle, a hum, and "ssss" to look like. They then produce each sound and compare.
  2. The whistle test (4 min): Students glide their pitch from low to high and describe, in their own words, what the peak trail shows and why it counts as evidence that the whole pipeline works.
  3. Find the resolution limit (4 min): Students glide slowly and record the smallest change in reported frequency they can produce, then check that value against the bin width printed in the readout.
  4. Linear versus decibel (2 min): Students toggle the scale and state which view they would choose for spotting a quiet harmonic, and which for showing off a loud fundamental.

Assessment

Ask: "Your whistle reads 1077 Hz, and the bin width is 43.1 Hz. What is the widest range of true frequencies that could have produced that reading? What would you have to change to narrow it?"

Privacy Note

The microphone stream is analyzed entirely in your browser. Nothing is recorded, stored, or transmitted, and the audio is never routed to your speakers, so there is no feedback risk. Closing the page ends microphone access.

References

  1. Spectral density — what the bars are actually estimating.
  2. Web Audio API — the browser machinery p5.sound builds on to reach the microphone.
  3. Cent (music) — the logarithmic pitch unit used in the readout.