GPS World, January 2018

Even though we do not know the precise location of the phase center of the antenna because the difference was performed between two devices that are the same model and have the same orientation the error in the phase center location is common and will cancel Various pairs of devices were tested by being mounted on a wooden board on a tripod at approximately waist height The test configuration is shown in FIGURE 4 FIGURE 5 provides the double difference GPS L1 C A pseudorange residuals between two Android devices We see errors beyond 100 meters and a standard deviation across all data of 144 meters A precision system RTK or RTX PPP would use a standard survey quality base or network of bases and not an Android device for the correction data Consequently in a typical operating mode where a precision data stream provides corrections the contribution in a double difference from the pseudorange on the Android devices would be roughly half the Android to Android residual seen in this test or approximately 72 meters for this example For comparison the same metric was generated between two precision GNSS units connected to antennas on the same roof While the data was not from the same time period we observe very consistent performance over time FIGURE 6 shows the same pseudorange double difference across a short baseline over 24 hours When comparing Figures 5 and 6 note the difference in the scale on the pseudorange residual axis The standard deviation from a pair of precision devices is 53 centimeters cm or 27 times lower noise than an example pair of Android devices All phones that provide GNSS measurements via the Android API publish the phase data in the accumulated delta range field An accumulated delta range is not necessarily a full phase measurement it can have an arbitrary starting phase For example in a precision GNSS receiver if the receiver locks to a satellite and some time later locks a second channel to the same satellite the phase measurement from the two channels may have a different integer cycle component but the subcycle component would be the same except for millimetric tracking noise If the two channels are providing 28 GPS WORLD WWW GPSWORLD COM JANUARY 2018 FIGURE 8 Cycle slips Android device FIGURE 9 Cycle slips precision device accumulated delta range the initial phase offset may differ by up to one cycle From the population of Android devices that publish phase that we have tested we have not observed any devices that deliver true full phase They all deliver an accumulated delta range with an arbitrary phase offset This limits a phase engine to float processing and ambiguity fixing is not possible The Android phase data collected from the previously described experiment was processed to provide the double difference carrier residuals This is shown in FIGURE 7 The y axis is in millicycles 1000 millicycles 1 cycle or approximately 19 cm FIGURE 4 Android device test configuration FIGURE 5 Short baseline double difference pseudorange Android devices FIGURE 7 Short baseline double difference phase residuals Android devices FIGURE 6 Short baseline double difference pseudorange precision devices MOBILE PRECISION