Smartphones for Science: Accuracy and Frequency Fidelity of Built-In Accelerometers

Since smartphones are widely available and relatively inexpensive, their built-in MEMS accelerometers are appealing for scientific sensing. In this study, we collected acceleration data from eight smartphones under controlled sinusoidal excitations (2-8 Hz, 0.1-0.7 g) using a hydraulic shake table, and benchmarked the results against a laboratory-grade reference sensor sampled at 100 Hz. Time-domain comparisons showed low error (MAE < 0.04 g) and consistently good phase alignment across devices. In the frequency domain, FFT analyses confirmed that the dominant frequencies were preserved, with only mild amplitude attenuation at the higher excitation levels. Overall, the smartphones provided frequencyaccurate, low-error measurements in this low-frequency range, supporting use cases such as seismology and Human Activity Recognition, and simple post-processing further improved similarity to the reference signal. All data from this study are publicly available for reuse and further analysis.