OboeTester: Add chirp and multi Data Paths tests by robertwu1 · Pull Request #2326 · google/oboe

robertwu1 · 2025-10-31T02:47:06Z

This PR enhances the Data Path test in OboeTester by adding two new analysis signals: a swept-sine (chirp) and a multi-tone signal. This provides a more comprehensive and robust way to test the audio data path, especially for input presets where OEMs might apply audio processing.

Problem

The existing sine wave test is not always sufficient. For some processed input presets (e.g., Camcorder), OEMs can apply effects like noise reduction that might filter out a simple, steady sine wave. This can lead to test failures that don't accurately reflect the device's ability to record real-world audio like music.

Solution

This PR introduces two new test signals that are more representative of complex audio and less likely to be misinterpreted as noise by audio processing algorithms:

Swept-sine (Chirp): Tests the device's frequency response over a wide range of frequencies.
Multi-tone: Tests for distortion and other non-linear effects by playing multiple frequencies at once.

Key Changes

UI: Added a new UI element in the Data Path test to allow selecting the desired test signal (Sine, Chirp, or Multi-tone).
Signal Generation: Implemented the generation of the chirp and multi-tone signals in the native code.
FFT-based Analysis: Implemented a new analysis path based on FFT to analyze the recorded chirp and multi-tone signals.
- For multi-tone, the analysis calculates the Signal-to-Noise and Distortion Ratio (SINAD).
- For chirp, the analysis computes a spectrogram and verifies that the frequency sweep is recorded correctly.
Pass/Fail Criteria: Added pass/fail criteria for the new analysis modes to allow for automated testing.
Code Quality: Refactored the new code to improve readability, maintainability, and remove magic numbers.

flamme · 2025-11-07T23:32:15Z

+                        if (maxFreq > nyquist) {
+                            maxFreq = nyquist;
+                        }
+                        double freq = kChirpStartFrequency + (maxFreq - kChirpStartFrequency) * mFrameCounter / (getSampleRate() * kChirpDurationSeconds);


Q: This looks like will go out of the range of nyquist freq when framecounter is greater than two times of kChirpDurantionSeconds * sampleRate.

Leaving as is because the increment code should not let this happen as it's caught by the while loop. If this happens, then we have a bug with our test code

OboeTester: Add chirp and multi Data Paths tests

64db51f

robertwu1 requested a review from flamme October 31, 2025 02:47