The new generation of Android smartphones is equipped with high performance Global Navigation Satellite System (GNSS) chips capable of tracking dual frequency multi-constellation data. Moreover, starting from version 9 of Android users can disable the duty cycle power saving option thus good quality pseudorange and carrier phase raw data are available thus the application of Precise Point Positioning (PPP) algorithm becomes more and more interesting. The main aim of this work is to assess the PPP performance of the first dual-frequency GNSS smartphone produced by Xiaomi equipped with a Broadcom BCM47755. The advantage of acquire dual frequency data is highlighted by comparing the performance obtained by Xiaomi with that of a single frequency smartphone the Samsung S8. The horizontal and vertical accuracy achieved by Xiaomi are of 0.51 m and 6 m respectively while those achieved by Samsung are 5.64 m for 15 m for horizontal and vertical.
The ability to access raw data started in May 2016 when Google announced that raw GNSS
measurements would be available to apps in the Android Nougat operating system was a
revolution. From that day on, it became possible to use D-GNSS algorithms with mobile
phone, try to reduce errors in urban areas, to fuse GNSS data with data from the other
phone sensors such as the Inertial Measurement Unit (IMU). Since that date many steps
forward have been made thanks also to the appearance of new chips designed exclusively for
mobile devices. Having access to smartphone raw data together with all the GNSS products
made available free of charge by organizations such as International GNSS Service (IGS) [
Precise Point Positioning algorithms are well known and ample literature is available (see
[10] for an exhaustive description). The data collection has been carried out by using two
Android smartphones and a geodetic receiver installed on the roof of a two-storey building
of the Engineering Department at the University of Perugia in an open-sky area. Geo ++
RINEX Logger app was used to acquire 1 Hz data in RINEX 3.03 format over a time span
of about 1 hour (13:20 - 14:20 UTC) on March 4, 2019. Figure 1 shows the setup of the two
smartphones and the geodetic GNSS receiver/antenna Topcon Hiper HR. This configuration
was chosen in order to simultaneously acquire measurements from the three receivers. The
ifrst smartphone used was a Xiaomi Mi 8 running Android 9 operating system, embedded with
a Broadcom BCM47755 chip. This chip is the first dual frequency chip expressly developed
for a smartphone [
The second smartphone employed was a Samsung Galaxy S8 equipped with a Exynos 8895 Octa – EMEA, running Android 7 operating system. It support only L1/E1 frequency for GPS, GLONASS, Galileo, BeiDou, QZSS, SBAS constellations.
The geodetic receiver used was a Topcon HiPer HR with integrated antenna capable of receiving GPS, GLONASS, Galileo, BeiDou, QZSS, SBAS signals on all frequencies.
The data acquired by the geodetic receiver were post processed in the static relative positioning mode. In detail, a Continuously Operating Reference Station (CORS) of the Umbria GNSS Network ([11, 12]) was used as base station. The results obtained were used as reference in order to evaluate the accuracy achieved by the two smartphones.
Starting from version 9 of Android users have the possibility to disable the duty cycle via software. This can be done by enabling the “Force full GNSS measurements” item in the developers menu. Thus for the Xiaomi Mi 8 duty cycle has been disabled while for Samsung S8 this was not possible.
As discussed in 2 section the smartphones and geodetic receiver have been installed on the roof of a building in an open-sky area. Figure 2 shows the sky-plots of Xiaomi Mi 8 and Samsung S8 in the left and right subplot respectively. The arcs depicted in red refer to satellites whose signals are acquired both on E1/L1 and E5a/L5 while arcs in orange represent satellites whose signals are acquired only on L1/E1 frequencies. From the figure it is clear that the Samsung S8 only receives the E1/L1 signals.
In figure 3 are shown the number of visible satellites with cut-of angle of 15 degrees and the evolution of GDOP versus time in top subplot and in bottom subplot respectively. Data relative to Xiaomi Mi 8 are depicted in blue while those relative to Samsung S8 are depicted in red. By observing the figure it can be noticed that the number of visible satellites acquired by Xiaomi Mi 8 is always higher than those acquired by Samsung S8. This is expected due to the higher tracking performance of Broadcom chip. Clearly the higher number of acquired satellites also afects the GDOP values which are always better for the Xiaomi Mi 8. Finally we want to point out the presence of some data gaps present in the data acquired by Samsung S8 (for example around 134600, 136700, 137600 seconds) caused by unwanted hardware stand-by.
Despite these diferences, however, we can state that the number of acquired satellites and their geometry are good for both smartphones having an average number of visible satellites equal to 22 for the Xiaomi and 15 for the Samsung and an average of GDOP of 1.3 and 1.7 respectively.
In order to assess the quality of received signals we compared the average of signal to noise ratio of each satellites received by Xiaomi Mi 8 and Samsung S8 with those of the same satellites received by geodetic receiver on L1/E1 frequency. This comparison is showed in figure 4: green, blue and red bars represent Topcon, Xiaomi Mi 8 and Samsung S8 respectively. Due the high number of satellites the comparison figure has been divided in four subplots representing the four constellations used. The figure clearly highlights the better performance of Xiaomi Mi 8 with respect to Samsung S8. Moreover observing the subplot relative to BeiDou (the bottom one) it can be noticed that only the geodetic receiver tracks C11, C26, C32 and C34 satellites.
Results are obtained using the Demo5 b33a (downloadable from http://rtkexplorer.com/ download/demo5-b33a-binaries modified version of the RTKLIB software originally developed by Takasu ( [13, 14]), using the PPP static algorithm, setting forward and backward option for iflter solution, a cutof angle of 15 degrees. Precise ephemeris and clocks correction used are the final CODE (Center for Orbit Determination in Europe) products. We also considered the earth rotation parameters (ERP) and satellite antenna corrections provided by IGS for PCV (phase center variations). For the ambiguity resolution we used the “continuous” method with a threshold equal to 30 to guarantee a better and more feasible estimation (See Dabove [15, 16]).
We want to emphasize that currently there are a large number of online applications that allow raw data PPP processing. In this study we also processed data by using the CSRS-PPP (Canadian Spatial Reference System) service [17]. However we will show the results obtained using RTKLIB.
Many processing cases using diferent combinations of GNSS systems have been investigated. However in order to facilitate the readability of the article, in this section are shown only the results obtained using signals from all the tracked constellations namely: GPS, GLONASS, Galileo and BeiDou constellation with the exception of Milena and Doresa Galileo satellites due to their elliptical orbits ([18, 19, 20, 21]). All the achieved results are summarized in tables 1 and 2. Firsts test are conducted using L1/E1 single frequency PPP algorithm in order to compare results achieved by the two diferent smartphones. Figure 5 shows the scatter plot of achieved results, two diferent scales are used in order to get better readability. Blue markers represent errors obtained by using Xiaomi Mi 8 measurements, red circles represent error for Samsung S8. The horizontal accuracy achieved is of 3.15 m and 5.75 m for Xiaomi Mi 8 and Samsung S8 respectively.
In figure 6 are depicted east, north and up solution components in subplot a, b and c respectively for Xiaomi Mi 8 (in blue) and Samsung S8 (in red). Two things are worth noting. The first is that in this test the vertical accuracy of the Xiaomi (about 20 m) is worse than that of the Samsung of about 5 m. The second is that the number of positions obtained by Samsung S8 are 684 while those obtained by Xiaomi Mi 8 are 3607. This is very evident by observing in the figure numerous gaps in the red line probably due to the activation of the duty cycle power saving mechanism.
In order to take full advantage of the potential of the Broadcom chipset embedded in the Xiaomi, further processing has been carried out. By exploiting the dual frequency measurements, it was possible to use the ionospheric-free combination in order to eliminate the first-order ionospheric error.
The scatter plot of this analysis is shown in figure 7. Blue markers represent errors obtained by using Xiaomi Mi 8 measurements. The horizontal accuracy achieved expressed in terms of DRMS is 0.52 m. East, north and up solution components are depicted in the subplot a, b and c of figure 8 respectively. The vertical RMS achieved is 6.0 m and the number of positions is equal to 1756.
The statistics related to all the processing carried out with diferent combinations of GNSS systems are shown in table 1, and 2.
We analyzed and assessed Xiaomi Mi 8 and Samsung S8 Android smartphone performance in open sky environment using PPP algorithms. More than twenty diferent elaborations have been carried out.
E1/L1 single frequency scatter plots reveals a better horizontal accuracy for Xiaomi Mi 8 with respect to Samsung S8. Vice versa the vertical accuracy achieved is better for Samsung than Xiaomi. However analyzing the east, north, up coordinate error plots in time domain we found that the number of solution achieved by Samsung S8 is about 1/6 than those achieved by Xiaomi Mi 8. This could be addressed to the activation of the duty cycle power saving mechanism; in particular, Samsung S8 carrier phase measurements are afected by cycle slips.
Xiaomi Mi 8 dual frequency L1/E1 & L5/E5a scatter plot shows an horizontal accuracy of 0.52 m while the vertical RMS is 6.0 m. By observing east, north, up coordinate error plots in time domain it can be noticed that the dual frequency PPP algorithm takes about 350 seconds before providing solutions with a fairly good continuity unlike the single frequency PPP algorithm.
To facilitate the readability of the article, in the previous section we have chosen to show only some graphics of the elaborations carried out. By comparing results achieved by Xiaomi in both ionospheric-free and dual frequency strategy it can be noticed that the exclusion of GLONASS and BeiDou constellations (table 2) worsens the accuracy obtained when all constellations data are used (table 1).
Results obtained show that the Xiaomi Mi 8 dual-frequency multi-constellation smartphone can be used in application with low precision requirements such as map updating and cadastral survey. (2019). URL: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85066060689&doi= 10.3390%2fs19092189&partnerID=40&md5=accffa636601d6efe7503aa2a8e0c0. doi: 10. 3390/s19092189. [9] Q. Wen, J. Geng, G. Li, J. Guo, Precise point positioning with ambiguity resolution using an external survey-grade antenna enhanced dual-frequency android gnss data, Measurement (2020) 107634. [10] P. Teunissen, O. Montenbruck, Springer handbook of global navigation satellite systems,
Springer, 2017. [11] F. Radicioni, A. Stoppini, La nuova rete gnss multicostellazione dell’umbria, GEOmedia 23 (2019). [12] G. Dardanelli, M. Lo Brutto, C. Pipitone, Gnss cors network of the university of palermo:
Design and first analysis of data, Geographia Technica 15 (2020). [13] T. Takasu, Ppp ambiguity resolution implementation, in: RTKLIB v 2.4.2, PP-RTK and open standards symposium, March 12-13, Frankfurt am Main, Germany, 2012. [14] T. Takasu, Rtklib: an open source program package for gnss positioning, in: RTKLIB: an open source program package for GNSS positioning, 2016. Http://www.rtklib.com. [15] P. Dabove, N. Linty, F. Dovis, Analysis of multi-constellation gnss ppp solutions under phase scintillations at high latitudes, Applied Geomatics (2019) 1–8. [16] P. Dabove, V. Di Pietra, Single-baseline rtk positioning using dual-frequency gnss receivers inside smartphones, Sensors 19 (2019) 4302. [17] Canadian spatial reference system, https://www.nrcan.gc.ca/, ???? Accessed: 2020-02-20. [18] U. Robustelli, G. Pugliano, Assessment of pseudorange measurements of galileo foc satellites with incorrect highly eccentric orbits, volume 2116, 2019. doi:10.1063/1.5114288. [19] U. Robustelli, G. Pugliano, Galileo single point positioning assessment including foc satellites in eccentric orbits, Remote Sensing 11 (2019). doi:10.3390/rs11131555. [20] U. Robustelli, G. Benassai, G. Pugliano, Signal in space error and ephemeris validity time evaluation of milena and doresa galileo satellites, Sensors (Switzerland) 19 (2019). doi:10.3390/s19081786. [21] U. Robustelli, G. Benassai, G. Pugliano, Accuracy evaluation of doresa and milena galileo satellites broadcast ephemerides, 2019, pp. 217–221. doi:10.1109/MetroSea.2018. 8657859.