Pseudorange Generation in Android Smartphones: Converter Inconsistencies, Root-Cause Analysis, and Positioning Effects

Accurate pseudorange generation on Android smartphones depends on how device-related clock fields—especially FullBiasNanos (FBN) and BiasNanos (BN)—are handled. Although several converters exist, including GNSSLogger, Geo++ RINEX Logger, CSV2RINEX, and raw android to rinex, differences in how these fields are processed have led to inconsistent pseudorange outputs and uncertainty when comparing results across studies. This paper performs a detailed investigation of these four converters using controlled open-sky datasets collected on multiple Google Pixel devices. Reconstructed pseudoranges show that each converter applies a distinct and often undocumented FBN strategy—(1) fixed vs epoch-varying FBN, (2) truncation of nanosecond-level digits, and (3) floating-point precision loss when storing the bias terms—leading to systematic pseudorange offsets of up to 6–150 m. At the positioning level, Single Point Positioning (SPP) results demonstrate that these offsets map almost entirely into the estimated receiver clock term, while position components remain stable at the millimeter-to-centimeter level under consistent geometry. Additional temperature-controlled experiments reveal a clear thermal dependence of the smartphone clock, with the FBN drift rate decreasing as device temperature increases. Together, these findings identify the root causes of converter-dependent discrepancies, quantify their impact on pseudoranges, and show how these biases propagate into the positioning solution through the receiver clock estimate.