The implementation changes and first live tests of BeiDou and Galileo on Teseo-3 GNSS chips developed in 2013 are covered, bringing it to a four-constellation machine. By 2020, we expect to have four global constellations all on the same band, giving us more than 100 satellites — under clear sky, as many as 30 or 40 simultaneously. By Philip G. Mattos and Fabio Pisoni Multi-constellation GNSS first became widely available in 2010/2011, but only as two constellations, GPS+GLONASS. Although receivers at that time may have supported Galileo, there were no usable satellites. BeiDou was a name only, as without a spec (an interface control document, or ICD), no receivers could be built. However, the hardware development time of receivers had been effectively shortened: the Galileo ICD had been available for years, BeiDou codes had been reverse-engineered by Grace Gao and colleagues at Stanford, and at the end of 2011 they were confirmed by the so-called test ICD, which allowed signal testing without yet releasing message characteristics or content. The last weeks of 2012 saw two great leaps forward for GNSS. Galileo IOV3 and 4 started transmitting at the beginning of December, bringing the constellation to four and making positioning possible for about two hours a day. At the end of December, the Chinese issued the BeiDou ICD, allowing the final steps of message decode and ephemeris calculation to be added to systems that had been tracking BeiDou for many months, and thus supporting positioning. The Teseo-2 receiver from STMicroelectronics has been available for some years, so apart from software development, it was just waiting for Galileo satellites; however, for BeiDou it needed hardware support in the form of an additional RF front end. Additionally, while it could support all four constellations, it could not support BeiDou and GPS/Galileo at the same time, as without the BeiDou ICD the spreading codes had to be software-generated and used from a memory-based code generator, thus blocking the GPS/Galileo part of the machine. The Teseo-3 receiver appeared late in 2013, returning to the optimum single-chip form factor: RF integrated with digital silicon and flash memory in the same package, enabling simultaneous use of BeiDou and GPS/Galileo signals. Multi-constellation in 2012 was GPS+GLONASS, which brought huge benefits in urban canyons with up to 20 visible satellites in an open sky. Now, for two hours a day in Europe while the Galileo IOVs are visible, we can run three constellations, and in the China region, GPS/BeiDou/Galileo is the preferred choice. This article covers the first tracking of four Galileo satellites on December 4, 2012, first positioning with Galileo, and first positioning with BeiDou in January 2013. It will cover static and road tests of each constellation individually and together as a single positioning solution. Road tests in the United States/Europe will combine GPS/GLONASS/Galileo, while tests in the China region will combine GPS/Galileo/BeiDou. Results will be discussed from a technical point of view, while the market future of multi-constellation hardware will also be considered. In the 2010–2020 timeframe, GLONASS and BeiDou (1602 MHz FDMA and 1561 MHz respectively) cost extra silicon in both RF and digital hardware, and cause marginal extra jamming vulnerability due to the 50 MHz bandwidth of the front end. The extra silicon also causes extra power consumption. After 2020, GLONASS is expected to have the L1OC signal operational, CDMA on the GPS/Galileo frequency, and BeiDou is expected both to have expanded worldwide, and also to have the B3 signal fully operational, again on 1575 MHz. At that point we will have four global constellations all on the same band, giving us more than 100 satellites. With a clear sky, the user might expect to see more than 30, sometimes 40, satellites simultaneously. Besides the performance benefits in terms of urban canyon availability and accuracy, this allows the receiver to be greatly simplified. While code generators will require great flexibility to generate any of the code families at will, the actual signal path will be greatly simplified: just one path in both RF (analog) and baseband (digital) processing, including all the notch filters, derotation, and so on. And this will greatly reduce the power consumption. Will the market want to take the benefit in power consumption and silicon area, or will it prefer to reuse those resources by becoming dual-frequency, adding also the lower-L-band signals, initially L5/E5, but possibly also L2/L3/L6 ? The current view is that the consumer receiver will go no further than L5/E5, but that the hooks will be built-in to allow the same silicon to be used in professional receivers also, or in L2C implementations to take advantage of the earlier availability of a full constellation of GPS-L2C rather than GPS-L5. This article presents both technical results of field trials of the quad-constellation receiver, and also the forward looking view of how receivers will grow through multi-frequency and shrink through the growing signal commonalities over this decade. History Galileo was put into the ST GPS/GNSS receiver hardware from 2006 to 2008, with a new RF and an FPGA-based baseband under the EU-funded GR-PosTer project. While a production baseband (Cartesio-plus) followed in high volume from 2009, in real life it was still plain GPS due to the absence of Galileo satellites. The changed characteristics in Galileo that drove hardware upgrades are shown in Figure 1. The binary offset carrier BOC(1,1) modulation stretches the bandwidth, affecting the RF, while both the BOC and the memory codes affect the baseband silicon in the code-generator area. Figure 1. Changes for Galileo. Next was the return to strength of the GLONASS constellation, meaning receivers were actually needed before Galileo. However the different center frequency (1602 MHz), and the multi-channel nature of the FDMA meant more major changes to the hardware. As shown in Figure 2 in orange, a second mixer was added, with second IF path and A/D converter. Figure 2. Teseo-2 RF hardware changes for GLONASS. Figure 3. Teseo-2 and Teseo-3 baseband changes for GLONASS. The baseband changes added a second pre-processing chain and configured all the acquisition channels and tracking channels to flexibly select either input chain. Less visible, the code-generators were modified to support 511 chip codes and 511kchips/sec rates. Teseo-2 appeared with GPS/GLONASS support in 2010, and demonstrated the benefit of GNSS in urban canyons, as shown by the dilution of precision (DOP) plot for central London in Figure 4. The GPS-only receiver (in red) has frequent DOP excursions beyond limits, resulting either in bad accuracy or even interrupted fix availability. In contrast, the GNSS version (in blue) has a DOP generally below 1, with a single maximum of 1.4, and thus 100 percent availability. Tracking 16 satellites, even if many are via non-line-of-sight (NLOS) reflected paths, allows sophisticated elimination of distorted measurements but still continuous, and hence accurate, positioning. Figure 4. DOP/accuracy benefits of GNSS. BeiDou Like Galileo, BeiDou is a story of chapters. Chapter 1 was no ICD, and running on a demo dual-RF architecture as per the schematic shown in Figure 5. Chapter 2 was the same hardware with the test ICD, so all satellites, but still no positioning. Chapter 3 was the full ICD giving positioning in January 2013 (Figure 6), then running on the real Teseo-3 silicon in September 2013, shown in Figure 7. Figure 5. Demo Teseo-2 dual RF implementation of BeiDou. Figure 6. Beidou positioning results. Figure 7. Teseo 3 development board. The Teseo-3 has an on-chip RF section capable of GPS, Galileo, GLONASS and BeiDou, so no external RF is needed. The clear green space around the Teseo-3 chip in the photo and the four mounting holes are for the bolt-down socket used to hold chips during testing, while the chip shown is soldered directly to the board. Figure 8A shows the development board tracking eight BeiDou satellites visible from Taiwan. However, the silicon is not designed to be single-constellation; it is designed to use all the satellites in the sky. Figure 8b shows another test using GPS and BeiDou satellites simultaneously. Figure 8A. Beidou. Figure 8b. GPS+Beidou. A mobile demo on the Teseo-3 model is shown running GPS plus BeiDou in Figure 9, a road test in Taipei. Satellites (SV) up to 32 are GPS, those over 140 are BeiDou, in the status window shown: total 13 satellites in a high-rise city area, though many are non-LOS. Figure 9. GPS + Beidou roadtrack in Taipei. Extending the hardware to add BeiDou, which is on 1561 MHz and thus a third center frequency, meant adding another path through the IF stages of the on-chip radio. After the first mixer, GPS is at 4 MHz, and GLONASS at about 30 MHz, but BeiDou is at minus 10 MHz. While the IF strip in general is real, rather than complex (IQ), the output of the mixer and input to the first filter stage is complex, and thus can discriminate between positive frequencies (from the upper sideband) and negative ones (from the lower sideband), and this is normally used to give good image rejection. In the case of BeiDou, the filter input is modified to take the lower sideband, that is, negative frequencies, and a second mixer is not required; the IF filter is tuned to 10 MHz. The new blocks for BeiDou are shown in green in Figure 10. The baseband has no new blocks, but the code generator has been modified to generate the BeiDou codes (and, in fact, made flexible to generate many other code types and lengths). Two forms of Teseo-3 baseband are envisaged, the first being for low-cost, low-current continues to have two input paths, so must choose between GLONASS and BeiDou as required. A future high-end model may have an extra input processing path to allow use of BeiDou and GLONASS simultaneously. Figure 10. Teseo-3 RF changes for Beidou shown in green. Galileo Again Maintaining the chronological sequence, Galileo gets a second chapter in three steps. In December 2012, it was possible for the first time to track four IOV satellites simultaneously, though not to position due to the absence of valid orbit data. In March 2012, it was possible for the first time to demonstrate live positioning, and this was done using Teseo-2 simultaneously by ESA at ESTEC and STMicro in Naples and Milan, our software development centres. The demos were repeated in public for the press on July 24, 2013, at Fucino, Italy’s satellite earth station, with ESA/EC using the test user receiver (TUR) from Septentrio, and ST running simultaneous tests at its Italian labs. Figure 11 and Figure 12 show the position results for the data and pilot channels respectively, with independent LMS fixes. In real life, the fixes would be from a Kalman filter, and would be from a combined E1-B/E1-C channel, to take advantage of the better tracking on the pilot. Figure 11. Galileo positioning, E1-B. Figure 12. Galileo positioning, E1-C. Good accuracy is not expected from Galileo at this stage. The four satellites, while orbited to give good common visibility, do not also give a good DOP; the full set of ground monitoring stations is not yet implemented and cannot be well calibrated with such a small constellation. Finally, the ionospheric correction data is not yet available. Despite these problems, the residuals on the solutions, against a known fixed position for the rooftop antenna, are very respectable, shown in Figure 13. Figure 13. Galileo residuals, L1-B. The common mode value is unimportant, representing only an offset in the receiver clock, and 10 meters is about 30 nanoseconds. The accuracy indicator is the spread between satellites, which is very respectable for a code-only receiver without full iono correction, especially around 640 on the TOW scale, where it is less than 2 meters. The rapid and major variation on the green data around t=400 is considered to be multipath, as the roof antenna is not ideally positioned with respect to other machinery and equipment also installed on the roof. QZSS and GPS-III/L1C Teseo-2 has supported the legacy (C/A code) signal on QZSS for some time, but Teseo-3 has been upgraded to handle the GPS-III/L1-C signal, waiting for modernized GPS. This signal is already available on the QZSS satellite, allowing tests with real signals. Significant changes were required in the baseband hardware, as the spreading code is a Weill code, whose generation complexity is such that it is generated once when the satellite is selected, then replayed real time from memory. Additionally it is long, in two domains. It is 10230 chips — that is, long to store but also long in time, with a 10-millisecond epoch. On Teseo-3, the legacy C/A code is used to determine code-phase and frequency before handing over to the Weill code for tracking. Using a long-range crystal ball and looking far into the future, a model of the future Teseo-4 DSP hardware is available, with 64 correlation taps per satellite. Running this on the captured QZSS L1-C signal gives the correlation response shown in Figure 14. Having multiple taps removes all ambiguity from the BOC signal, simultaneously removing data transitions, which can alternatively be pre-stripped using the known pilot secondary code (which on GPS III is 5 dB stronger than the data signal). The resultant plot represents 2,000 epochs, each of 10 milliseconds, plotted in blue, with integrated result for the full 20 seconds shown in the black dashed line. Assuming vehicle dynamics is taken out using carrier Doppler, this allows extremely precise measurement of the code phase, or analysis of any multipath in order to remove it. This RF data was captured on a benign site with a static antenna, so it shows little distortion. Figure 14. L1-C tracking on QZSS satellite. Figure 15. Dual RF implementation of dual-band front end. The Future Having already built in extreme flexibility to the code generators to support all known signals and generalized likely future ones, the main step for the future is to support multiple frequencies, starting with adding L5 and/or L2, but as before, ensuring that enough flexibility is built in to allow any rational user/customer choice. It is not viable for us to make silicon for low-volume combinations, nor to divide the overall market over different chips. Thus our mainstream chip must also support the lower volume options. We cannot, however, impose silicon area or power consumption penalties on the high-volume customer, or he will not buy our product. Thus, our solution to multi-frequency is to make an RF that can support either band switchably, with the high band integrated on the volume single-chip GNSS. Customers who also need the low band can then add a second RF of identical design externally, connected to the expansion port on the baseband, which has always existed for diagnostic purposes, and was how BeiDou was demonstrated on T2. By being an RF of identical design to the internal one, it incurs no extra design effort, and would probably be produced anyway as a test chip during the development of the integrated single-chip version. Without this approach, the low volume of sales of a dual-band radio, or a low-band radio, would never repay its development costs. Conclusions All four constellations have been demonstrated with live satellite signals on Teseo-2, a high-volume production chip for several years, and on Teseo-3 including use in combinations as a single multi-constellation positioning solution. With the advent of Teseo-3, with optimized BeiDou processing and hardware support for GPS-3/L1C, a long-term single-chip solution is offered. For the future, dual-frequency solutions are in the pipeline, allowing full advantage of carrier phase, and research into moving precise point positioning and real-time kinematic into the automotive market for fields such as advanced driver-assistance systems. Acknowledgments Teseo III design and development is supported by the  European Commission HIMALAYA FP-7 project. This article is based on a technical paper first presented at ION-GNSS+ 2013 in Nashville, Tennessee. ST GPS products, chipsets and software, baseband and RF are developed by a distributed team in: Bristol, UK (system R&D, software R&D; Milan, Italy (Silicon implementation, algorithm modelling and verification); Naples, Italy (software implementation and validation); Catania, Sicily, Italy (Galileo software, RF design and production); Noida, India (verification and FPGA). The contribution of all these teams is gratefully acknowledged. Philip G. Mattos received an external Ph.D. on his GPS work from Bristol University. Since 1989 he has worked exclusively on GNSS implementations, RF, baseband and applications. He is consulting on the next-generation GNSS chips, including one-chip GPS (RF+digital), and high-sensitivity GPS and Galileo for indoor applications, and combined GPS/Galileo/GLONASS chipsets. In 2008-2009, he re-implemented LORAN on the GPS CPU, and in 2009-2010 led the GLONASS implementation team. He is leading the team on L1C and BeiDou implementation, and the creation of totally generic hardware that can handle even future unknown systems. Fabio Pisoni has been with the GNSS System Team at STMicroelectronics since 2009. He received a master’s degree in electronics from Politecnico di Milano, Italy, in 1994. He was previously with the GNSS DSP and System Team in Nemerix SA and has earlier working experience in communications (multi-carrier receivers).