Accuracy Tests — Pitch Detector Stability & Precision
This page documents the methodology and results of our internal stability tests. We used clean, calibrated sine tones to measure how accurately and consistently our online pitch detector performs under ideal conditions. Our tests confirm a **±5¢ stability on clean test tones @ 110/220/440 Hz.** This level of precision ensures you receive reliable feedback for instrument tuning, vocal practice, and ear training, directly in your browser with no uploads required.
Updated on {Month Day, Year}
Test environment
The following conditions were maintained to ensure consistent and repeatable measurements.
Signal source: Sine tones 110/220/440 Hz @ 44.1 kHz
Output path: System loopback or speaker → mic at ~20 cm
Microphones: Laptop built-in (MacBook Pro) + USB condenser (Rode NT-USB)
Browsers/OS: Latest Chrome, Edge, Safari on macOS/Windows/iOS
Room: Quiet office environment (<35 dBA); mic off-axis 15–30°
Controls fixed: A4=440 Hz; consistent gain; echo cancellation off
All testing is performed locally in-browser; no audio is ever uploaded, ensuring user privacy.
Procedure
Our procedure relies on **autocorrelation** to estimate the fundamental frequency, enhanced by **parabolic peak interpolation** for sub-bin precision.
- Play a clean 110 Hz sine tone for 20 seconds; log the detected Hz and cents deviation once per second.
- Repeat the 20-second measurement process for the 220 Hz and 440 Hz tones.
- Repeat all tone tests on a second microphone and in a second browser to check for environmental variance.
- Average the deviation across all readings and compute the minimum, median, and maximum cents offset for each tone.
- Note any dropouts or significant flicker events; exclude obvious clipping artifacts from the final data set.
The tool’s internal noise gate and browser sample-rate handling were held constant during all tests.
Results summary
The results confirm high stability for sustained tones in a quiet environment. We can verify our claim of **±5¢ stability on clean test tones @ 110/220/440 Hz.**
| Tone (Hz) | Median Deviation (¢) | Range (min/max ¢) | Notes |
|---|---|---|---|
| 110 Hz (A2) | +1.8¢ | -3¢ / +4¢ | Very stable |
| 220 Hz (A3) | +2.1¢ | -4¢ / +4¢ | Very stable |
| 440 Hz (A4) | +1.5¢ | -3¢ / +3¢ | Extremely stable |
- The tool provides reliable feedback for sustained, single notes.
- As a monophonic detector, it cannot analyze chords or multiple simultaneous sounds.
- Real-world performance depends heavily on room acoustics and microphone quality.
Measurement evidence
Detailed logs and screenshots of our test runs are kept internally to verify our results. Because all analysis is performed via local processing directly in your browser, no audio is ever recorded or uploaded, so there are no user session captures to display. This section serves as a placeholder for where such public-facing evidence could be presented if our methodology changed.
Limitations & sources of error
Accuracy depends on a clean signal. The following factors can impact performance:
- Persistent background noise (fans, talking, traffic).
- Unstable or wavering pitch from the voice or instrument.
- Incorrect microphone distance (too close causes clipping, too far reduces signal).
- Room reflections and echo, which can interfere with the primary tone.
- Minor sample-rate drift in certain browser or OS configurations.
- Very low signal-to-noise ratio (e.g., whispering).
- Attempting to detect polyphonic sounds (chords are not supported).
Reproduce our tests
You can perform a simplified version of our tests using the steps below.
- Open the Pitch Detector and set the A4 reference to 440 Hz.
- Use any online tone generator to play clean 110 Hz, 220 Hz, and 440 Hz sine waves at a low, clean volume.
- Hold your microphone approximately 20 cm from the speaker, slightly off-axis, in a quiet room. Keep each tone steady for 20 seconds.
- Observe the median and variance of the cents meter display; retry with a second browser or microphone to compare results.
For best results, disable system-level audio enhancements like echo cancellation and close other applications.
Testing FAQ
Why cents instead of only Hz?
Cents are a logarithmic unit of pitch that is much more intuitive for musicians. The human ear perceives pitch logarithmically, so a 10-cent deviation sounds the same whether it’s at 110 Hz or 880 Hz. Reporting in cents provides a consistent, musically relevant measure of intonation accuracy across all octaves.
Why 110/220/440 Hz tones?
These frequencies correspond to the musical notes A2, A3, and A4. They represent a common range for many instruments and voices, providing a practical baseline for testing. Using octaves of the same note (A) allows us to verify that the detection algorithm performs consistently across different frequency bands, which is critical for a reliable tool.
Do different microphones change results?
Yes, to a small degree. A high-quality external microphone will typically provide a cleaner signal with a better signal-to-noise ratio than a standard built-in laptop mic. This can lead to slightly more stable readings. However, our tests show that even with a basic mic, the tool remains highly accurate in a quiet environment.
Can I test in a noisy room?
While you can try, a noisy room will significantly degrade accuracy. The tool’s algorithm is designed to find the strongest, most periodic frequency, and background sounds can interfere with that process, causing readings to flicker or become unstable. For any serious measurement, a quiet environment is essential for obtaining reliable results.