used in the monitoring stations of the first Galileo satellite, the Galileo In- Orbit Validation Element–A (GIOVE- A). The articles by W. De Wilde et al and Andrew Simsky et alia listed in the Additional Resources section near the end of this article provide additional details about our receiver technology mentioned here and later in this discus- sion. Figure 1 shows an overview of the receiver. Its front-end receives all Gali- leo and GPS navigation bands (L1/L2/ E6/E5). After splitting and conversion to an intermediate frequency, the L1 and L2 signals are routed to a CA/P-code dual-frequency GPS baseband proces- sor ASIC (application specific integrated circuit). The intermediate frequency (IF) signals of the Galileo bands (L1/E6/E5) More Compass Points enter a field programmable gate array (FPGA) after digitization for baseband FIGURE 1 GPS/Galileo Receiver Block Diagram Tracking China’s MEO Satellite processing. This FPGA contains eight generic We logged the L1 Band tracking channels. These channels con- IF signal of the front L1 E6 L2 E5 on a Hardware Receiver sist of a mixer for removal of the carrier, end on four random Centre Frequency (MHz) 1575.42 1278.75 1227.6 1191.795 two multicorrelator banks for concur- moments because Wim De Wilde, On April 14 China launched its first medium-Earth-orbit (MEO) satellite code- rent despreading of data and pilot com- we d id n’t have Bandwidth (MHz) 40 40 25 55 Frank Boon, named Compass-M1, the first in a planned constellation with global coverage ponents, and two code generators that access to the two- Front-end Type Dual conversion heterodyning J-M Sleewaegen, provide associated spreading codes. line element set of Base-band Clock 56 MHz and Frank Wilms based on the same principles as GPS and Galileo. In order to investigate the The channels were designed to sup- the Compass satel- Number of GPS CA/P channels 9 Septentrio N.V. space vehicle’s characteristics and data structure, researchers modified a port all GIOVE-A signals (including lite. After post-pro- Number of Flexible Channels 8 GPS/Galileo hardware receiver so as to track the Compass MEO satellite. This AltBOC), all open Galileo-signals, GPS cessing in a Matlab TABLE 1. GNSS Receiver Parameters article tells how they did it and presents the first results of their efforts. L1 CA-code, GPS L5, GPS L2C, and all software receiver, GLONASS signals. two out of the four Because most Galileo and modern- log-files clearly showed correlation peaks code is an 11-bit shift register truncated n 2000 China deployed the Beidou- This explains the burst of research a normal hemispherical antenna. ized GPS signals have a significantly after correlation with the codes available Gold code. The overlaying secondary 1 navigation system. Originally this activity after launch of the first Compass Rather than building a receiver from longer spreading code than GPS CA, the from the SU website. This encouraged code is 20 bits long, resulting in a 50 S-band system provided ranging satellite (Compass-M1) in April this scratch, we found we could reuse a search space grows. To support acqui- us to go ahead with the development of Hz data rate. Both frequencies have the Iinformation via geostationary sat- year, particularly because China didn’t generic GNSS platform developed by sition in a reasonable amount of time, real-time tracking software. same primary and secondary code. ellites that operate as transponders. This disclose any details about its operation. our company. we implemented a parallel acquisition To conclude the description of the These concepts have much in com- system design required bulky two-way In an article published the May/June After a software-modification, this unit (PAU) in the FPGA. This unit uses GNSS receiver, Table 1 shows its main mon with GPS and Galileo: radios, had a limited capacity, and cover- issue of Inside GNSS, researchers at GPS/Galileo receiver supported track- a combination of a matched filter and an functional parameters. • The chipping rate is a multiple of the age was restricted to East Asia. CNES, the French space agency, pre- ing of the E2 and E5B BPSK(2) signals FFT (fast Fourier transform) algorithm 0.5115 Mcps, just like the rates of all Currently, China is developing the sented dish-antenna measurements of of Compass-M1. This provided a lot of supporting the same signals as the track- Enabling Compass-M1 GPS and Galileo signals. successor to this system, called Beidou- the Compass-M1 signal and unveiled information about signal strength and ing channels. Support • The concept of a truncated Gold code 2 or Compass. The Compass Navigation the main code properties. Subsequently, ranging quality. Because the receiver is The FPGA connects to a large Our modification of the receiver for has already been introduced in GPS Satellite System (CNSS) will consist of a Stanford University (SU) team under- capable of logging navigation symbols, SDRAM (synchronous dynamic random tracking Compass-M1 focused on the L5 and Galileo, though with other 30 medium-Earth-orbit (MEO) satel- took complementary measurements and we were able to figure out the main data access memory) intended for logging of E2 and E5B BPSK(2) signals that are parameters. lites broadcasting code division multiple worked out the spreading code param- structure properties of Compass-M1 as digitized IF samples. The logged samples believed to be the open service signals. • Secondary codes are defined for GPS access (CDMA) signals in the L-band. eters, which are outlined in an article well. can be downloaded over Ethernet and As discussed in the CNES research men- L5 and several Galileo signals. Unlike the relatively large user equip- elsewhere in this issue. used for post-processing. This func- tioned earlier, these codes have a 2.046 It turned out that all Compass-M1 ment of Beidou-1, Compass will sup- The information published by CNES Receiver Fundamentals tionality was very useful to find out the Mcps chipping rate and a primary code signal characteristics were already sup- port global navigation by means of small and SU is sufficient to build a hardware The receiver we used to track Compass Compass-M1 signal was actually avail- length of 2,046 chips. ported in the generic channel design of handheld receivers. receiver able to track the signal through is the commercial version of the receiver able at the antenna. Stanford University figured out this our receiver.

44 InsideGNSS july/august 2007 www.insidegnss.com www.insidegnss.com july/august 2007 InsideGNSS 45 more compass points

150 I Q code can be sourced via a memory, as of power. Conversion into received iso- 100 required for the non-Gold-code L1BC tropic power results in -153 dBW for E2 and E6BC Galileo signals. and -150 dBW for E5B, compared to -157 50 We chose to use the Gold-code dBW for GPS L1CA. This enables acqui- generator option because this uses the sition at very low elevations and offers 0 hardware more efficiently, envisaging new perspectives for indoor navigation. -50 later implementation in an optimized The second observation is that the product currently in development. Compass ranging quality is similar to -100 The E5B and E2 frequency bands GPS. In the 2 hour to 5 hour time frame Normalized Correlation Amplitude (%) are both within the bandwidth of the in Figure 4 we found a ranging error -150 receiver’s analog front end. Here we need standard deviation as indicated in Table -1.5 -1 -0.5 0 0.5 1 1.5 to remark that the E2 center frequency 2. The contribution due to thermal noise Lab (Chips) happens to be at 1561.098 MHz, rather is displayed in the table as well. FIGURE 5 Compass-M1 Correlation Peak E5B than 1561.1 MHz as stated in earlier These data clearly show that the per- papers. This is exactly equal to the GPS formance is limited by multipath, and center frequency minus 14 times 1.023 we see a similar behavior in this area 30 second periodicity Large peak at 720 seconds MHz and perfectly symmetrical to the for both systems. This indicates that the × 104 E2 ACF E1 band. This may suggest CNSS will higher chipping rate doesn’t have much FIGURE 2 Generic Code Generator 10 support AltBOC(14,2) modulation. influence on the ranging performance. The E2 offset with respect to L1 was This is as could be expected, because 6 E5b RH CP Gain vs. theta L1 RH CP Gain vs. theta well within the pulling-range of the most multipath has a delay much shorter 2 20 20 oscillators in the tracking channel. We than the 500-nanosecond Compass chip -2 theta 0 200 400 600 800 1000 1200 0 0 concluded the the receiver’s hardware length. 4 E5b ACF Seconds × 104 Label didn’t need to be modified in order to Our receiver has a built-in correla- × 10 5 -20 -20 10

Gain (dB) Gain (dB) be Compass-M1–compliant. Enabling tion peak monitoring function. Figure 5 4.5 -40 -40 Compass-M1 reception only required an displays the E5B correlation peak. The 6 4 -60 -60 -180 -120 -60 0 60 120 180 -180 -120 -60 0 60 120 180 upgrade of the embedded software and E2 peak looks very similar. 2 Close3.5 Up theta (*) theta (*) of the graphical user interface. -2 3 0 200 400 600 FIGURE 3 Antenna Gain at E5B and L1 Compass-M1 Data Structure 2.5 4 E2-E5b CCF Seconds Tracking Compass-M1 The GNSS receiver can output the data × 10 2 10 C/Nc We hooked the receiver up to a GPS/ symbols received from the satellite. We 1.5 65 6 Compass-M1 E2 Galileo/GLONASS choke ring antenna. logged the Compass E2 and E5B data 1 60 Compass-M1 E5B Figure 3 shows the gain profile of this during the satellite pass. To find out the GPS-BIIR-02 L1CA 2 0.5 55 antenna. The receiver acquired Com- data structure we calculated the auto- -2 0 50 pass-M1 straightaway via the PAU, prov- correlation functions of the logged data 0 200 400 600 0 5 10 15 20 25 30 35 45 Seconds X axis (dB-Hz) ing the versatility of the platform. (Figure 6). This clearly shows a periodic- 40 We logged the signal strength and ity of 30 seconds. FIGURE 6 Navigation Data Correlation 35 pseudoranges during a complete high- A close-up view of the 30-second 30 0 1 2 3 4 5 6 7 8 elevation pass on July 5. The ranging interval reveals small peaks at a spacing Time (h) accuracy was investigated by subtracting of 6 seconds. This suggests 30-second are exactly identi- Satellite Compass-M1 E2 Compass-M1 E5B GPS PRN13 L1 Ranging Error (Ionosphere Removed) the precise but ambiguous carrier-based frames and 6-second subframes. Note cal to that of GPS C/No 55.5 dB-Hz 60 dB-Hz 52 dB-Hz Compass-M1 E2 4 Compass-M1 E5B pseudorange from the code-based pseu- that a very high correlation exists for a CA-code. An obvi- Antenna Gain 7.2 dBi 7.5 dBi 7.2 dBi GPS-BIIR-02 L1CA 2 dorange. The ionospheric effect (around 12-minute shift. This may indicate that ous thing to do was System NF 2.4 dB 1.4 dB 2.2 dB 10 meters) was removed via a curve-fit. the Compass navigation message has a to try to route the 0 Isotropic Power -153 dBW -150 dBW -157 dBW (m) We did this for both frequencies and 24-frame period. received symbols to -2 compared the results to similar data Because the same format is detected the GPS CA naviga- Ranging Error 13.5 cm 13.4 cm 11.9 cm Thermal Noise Contribution 2.4 cm 1.7 cm 3.7 cm -4 obtained from the GPS satellite Block at both Compass frequencies, we found tion data decoder. IIR-02 (PRN13) through the same set- it interesting to calculate the cross-cor- Unfortunately, the TABLE 2. Comparisons of Selected Compass and GPS Signals 0 1 2 3 4 5 6 7 8 up. This data was recorded during a pass relation function between E2 and E5B decoder didn’t syn- Time (h) through zenith. The resulting graphs are data, which shows perfect correlation. chronize. Apparently the navigation the data, we came to the structure pre- FIGURE 4 Compass-M1 C/No and Ranging Error Compared to GPS shown in Figure 4. The ranging errors Further investigation indicates that the message preamble is different from the sented in Figure 7. This figure shows the have been shifted apart for clarity. Actual satellite is sending the same data mes- GPS preamble. data received during three succeeding The receiver’s channel and PAU have generator. The feedback taps and initial- ranging errors are shown in Table 2. sage simultaneously at E2 and E5B. We detected the correct preamble frames. two code-generation options (see Figure ization vectors can be programmed by Our first observation is that the We need to underline the fact that by looking for a pattern that was repeat- The preamble of each subframe is 2). The first option is a generic Gold code software. Alternatively, the spreading Compass-M1 satellite is sending a lot the subframe and frame periodicity ing every six seconds. By reorganizing followed by a subframe counter that

46 InsideGNSS july/august 2007 www.insidegnss.com www.insidegnss.com july/august 2007 InsideGNSS 47 more compass points

and E5B. The fram- Authors ing structure is Wim De Wilde received similar to GPS CA- his master’s degree in code, but the field electrical engineering ordering is differ- from the University of Ghent. Upon graduation FIGURE 7 Compass-M1 Navigation Data ent. he joined the DSL research team of Alcatel counts from 1 to 5. For all subframes Acknowledgment Bell. Since 2002 he has been an R&D engineer at except frame 4, the bits after bit 60 Septentrio Satellite Navigation. His areas of inter- didn’t change over similar subframes We thank Orban Microwave Prod- est are analog and digital signal processing and in the three frames. Because we see ucts (www.orbanmicrowave.com) for system design. data switching over in repeated zero- providing their excellent GPS/Gali- Frank Boon holds a Ms. Sc. sequences after odd multiples of 30 bits, leo/GLONASS choke ring antenna and in aerospace engineer- this suggests 30-bit wording. This would associated gain-plots. Part of the equip- ing from the Delft Uni- be similar to GPS CA-code. However, ment used for this study is based on the versity of Technology. the parity mechanism is different from work performed under ESA contract He is responsible for the GPS CA. 18834/05/NL/MM. design and implementa- After support of preamble and sub- tion of positioning algo- frame detection in the embedded soft- Manufacturers rithms in the firmware of Septentrio’s GNSS receiv- ware, the Compass-M1 timing was com- The GeNeRx1 GPS/Galileo receiver ers. His research interests include error models of pared to GPS. The second subframe of used to track the Compass-M1 signals is GNSS measurements and optimization of posi- tioning algorithms for high dynamic applica- Compass-M1 lags the first GPS subframe manufactured by Septentrio N.V., Leu- tions. by exactly 14 seconds. This relationship ven, Belgium. The choke ring antenna suggests the Compass-M1 timing is is from Orban Microwave Products Jean-Marie Sleewaegen is currently responsible for UTC-based. (OMP), Leuven, Belgium. GNSS signal processing, data analysis, and tech- Conclusion References nology development at A software modification has been [1] De Wilde, W., and J.-M. Sleewaegen, “New Septentrio Satellite Nav- applied to an existing GPS/Galileo/ Fast Signal Acquisition Unit for GPS/Galileo igation. He received his GLONASS receiver to enable real-time Receivers,” ENC-GNSS 2006, Manchester, UK , M.Sc. and Ph.D. in electrical engineering from the Compass-M1 tracking at E2 and E5B May 8-10, 2006 University of Brussels in Belgium. through a hemispherical antenna. The [2] Gao, Grace Xingxin, and Alan Chen, Sherman Frank Wilms has a diplo- signal strength was clearly higher than Lo, David De Lorenzo, and Per Enge “The Compass ma in mechanical engi- the one of GPS and exceeded 60 dB-Hz MEO Satellite Codes,” Inside GNSS, July/August neering, with specializa- at E5B. No anomalies were observed in 2007 tion in spaceflight the ranging accuracy. [3] Grelier, T., and J. Dantepal, A. Delatour, A. technology, from the Identical data is transmitted at E2 Ghion, and L. Ries, “Initial Observations and Technical University of Analysis of Compass Braunschweig. At the MEO Satellite Signals,” university’s Institute of Flight Guidance and Con- Inside GNSS May/June trol he worked on real-time data-acquisition sys- 2007S tems and integrated navigation systems based on [4] Simsky, A., and J.-M. GNSS. In 1999 he was with the Navigation Section Sleewaegen, “Overview of Dornier Satellitensysteme/DaimlerChrysler of Septentrio’s Galileo Aerospace and involved in system architecture Receiver Development tasks within the ESA GNSS-2 Comparative System Strategy,” September Studies. In 2000 he joined Septentrio Satellite 13-16, 2005, ION GNSS Navigation in Leuven (Belgium) where he was 2005, Long Beach, Cali- responsible for GNSS receiver interface design as fornia USA well as application project management. Since 2003 he has served as manager of Septentrio’s [5] Steingass, A., and A. Galileo projects with ESA, CDE1 Test User Segment Lehner, “Measuring the (CDE1 TUS) and GSTB-V2 Experimental Test Navigation Multipath Receiver (GETR). Wilms is also involved in Galileo Channel … A Statistical and security-related business development. Analysis,” Proceedings of ION GNSS 2004, Long Modified GPS/Galileo/GLONASS receiver tracking Compass-M1 Beach, California USA

48 InsideGNSS july/august 2007 www.insidegnss.com