First Signals of BeiDou Phase 3 Acquired at Ispra, Italy

August 21, 2015  - By

Editor’s Note: The article below has been greatly revised and expanded from the original version published Aug. 10.


By Michele Bavaro, James Curran and Joaquim Fortuny

On July 25, 2015, China launched two modernized BeiDou satellites. Although the nomenclature is still uncertain, the Joint Space Operations Centre / North American Aerospace Defense Command identifiers for the satellites are BEIDOU-3 M1 and BEIDOU-3 M2. The satellites have been placed in medium Earth orbit (MEO) and both satellites have reached their designated orbital slots.

On Aug. 9, some signals from these satellites were received with a software-defined radio sampler operated at the European Commission’s Joint Research Centre in Ispra, Italy. The sampler was driven by orbit-prediction software that triggers a synchronized acquisition on both 1575.42 MHz and 1278.75 MHz using 1-bit complex samples at 60 megasamples per second (about 60 MHz total bandwidth). The two-line element sets for the orbits were obtained from the CelesTrak website and predicted positions were computed using code developed following the Simplified General Perturbations Satellite Orbit Model 4 (SGP4) as documented in the U.S. Department of Defense Spacetrack Report No.3.

To confirm the identity of the satellite being tracked using codeless-tracking, the measured Doppler frequency shift measured by the codeless-tracking receiver was compared with the Doppler predicted using the SGP4. The local oscillator clock drift was modeled using GPS L1 C/A-code signals and taken into account when matching the Doppler shift.

According to a presentation given at Stanford University’s 2014 PNT Symposium by Mingquan Lu and Zheng Yao from Tsinghua University, modernized BeiDou satellites are expected to broadcast an MBOC(6,1,1/11) signal, being a multiplexing of BOC(6,1) and BOC(1,1) signals, and a BOC(14,2) signal on the L1 frequency. Neglecting the BOC(6,1) component, the two BPSK(1) lobes of the BOC signal were brought to baseband and cross-correlated by our equipment, in an effort to detect the presence of the broadcast signals. In Figure 1, the peak at 1756.41 kHz is believed to correspond to the MBOC(6,1,1/11) signal broadcast by the BEIDOU-3 M2 satellite.

Figure 1. BOC(1,1) cross-correlation.

Figure 1. BOC(1,1) cross-correlation.

Figure 2: Power spectral density of BeiDou-3 M2 on L1. Signal collected using a 1.8m steerable dish.

Figure 2: Power spectral density of BeiDou-3 M2 on L1. Signal collected using a 1.8m steerable dish.

On Aug. 10, a 1.8-meter dish was pointed at the satellite and a number of further datasets were collected. The power spectrum estimated from one of these datasets is shown in Figure 2. Upon first inspection, the spectrum shows very good agreement with the anticipated MBOC(6,1,1/11) signal, centred at 1575.42 MHz. Further testing revealed that the BeiDou pseudorandom noise code (PRN33) correlates with the low side lobe, indicating that the satellite is broadcasting a legacy B1I signal on 1561.098 MHz. A replica of the B1I PRN code was generated and correlated against the received signal, to confirm the presence of the legacy BPSK(2) signal. A trace of the cross-correlation function is shown in Figure 3.

Figure 3. Cross-correlation of a BPSK(2) BeiDou code PRN33 on a 1561.098-MHz carrier.

Figure 3. Cross-correlation of a BPSK(2) BeiDou code PRN33 on a 1561.098-MHz carrier.

Further examining the IF data, we were able to determine the exact modulation of the central MBOC -ike signal. The raw IQ samples corresponding to the center of the L1 band were filtered to approximately 10 MHz, keeping only the MBOC signal, and tracked using a phase-locked loop operating directly at the 60 MHz sample rate, to recover the phase of the signal, prior to correlation. These phase-coherent samples were examined to gain some insight into the modulation. A simple autocorrelation of these phase-coherent samples suggested that there was no power in the quadrature channel, but that the in-phase channel contained what appeared to be an MBOC signal. More interestingly was the observation that the autocorrelation was periodic with a period of 10 milliseconds, suggesting the presence of a 10,230-chip primary code. An example of the autocorrelation of these phase-coherent samples is shown in Figure 4.

Figure 4: Autocorrelation of the TMBOC(6,1,4/33) pilot signal centred at 1575.42 MHz.

Figure 4: Autocorrelation of the TMBOC(6,1,4/33) pilot signal centred at 1575.42 MHz.

As the signal appeared to have a repetition period of 10 milliseconds, it was possible to perform a coherent combination of many successive 10-millisecond periods to achieve a sufficiently high signal-to-noise ratio to examine the modulation in detail. This process revealed what appears to be a TMBOC(6,1,4/33) modulation. Interestingly, once we aligned the samples to a 20-millisecond system time boundary via tracking of the legacy signal, the allocation of the BOC(6,1) component and the BOC(1,1) component could be identified. This pattern appears to be identical to that of L1C signal, being four segments of one-chip duration in each successive group of 33 chips. Specifically, chips {1,5,7 and 30} are allocated a BOC(6,1) pulse, and the remainder of the 33 chips are allocated BOC(1,1) pulses, as shown in Figure 5. Having produced a replica for the BOC(1,1) and BOC(6,1) components, we were able to implement a matched filter to extract the spreading sequences.

Figure 5: TMBOC(6,1,4/33) modulation visible on the received signal. In magenta, the expected chip allocation of the time-division MBOC.

Figure 5: TMBOC(6,1,4/33) modulation visible on the received signal. In magenta, the expected chip allocation of the time-division MBOC.

Further analysis revealed that the pilot signal was modulated by an overlay sequence, having a repetition period of 18 seconds, again, similar to that of the L1C specification. Leveraging the alignment to the 6 seconds boundary of the legacy signal, the overlay code, having a length of 1800 symbols, was extracted. The code has been circularly rotated to match with the 18-second boundary of the legacy SOW (Seconds Of Week). Note, however, that while the relative values of the primary and secondary code are likely correct, there still exists a single uncertainty as to the overall sign of the two codes. The correct codes may indeed be the inverse of what was extracted.

The assumption that a power sharing favoring the pilot over the data as suggested by Lu and Yao was confirmed by the fact that the demodulated chips do not obviously appear as a three-level signal (as one would expect, for example, with Galileo E1B-E1C). Rather, the amplitude of the received signal was dominated by the pilot signal.

Figure 6: Autocorrelation of the BOC(1,1) data signal centred at 1575.42 MHz.

Figure 6: Autocorrelation of the BOC(1,1) data signal centred at 1575.42 MHz.

Figure 7: Coherent accumulation of modernised BeiDou OS data channel (pilot stripped with SIC) over 0.5 seconds. In magenta the expected subcarrier pattern is shown.

Figure 7: Coherent accumulation of modernised BeiDou OS data channel (pilot stripped with SIC) over 0.5 seconds. In magenta the expected subcarrier pattern is shown.

Having extracted the pilot code, we could perform successive interference cancellation (SIC) and strip its power contribution from the signal. An attenuation of 10 dB was sufficient to bring the spreading code of the data channel to the surface, which was readily achieved, even when using low-resolution samples. After we extracted the phase of the pilot signal, the samples were aligned to be phase coherent and the autocorrelation was examined. This appeared as a BOC(1,1) signal with a period of 10 milliseconds and aligned in phase with the pilot signal. The autocorrelation is shown in Figure 6. Using again the technique described above, the individual chips were examined, as shown in Figure 7, where it is clear that the TMBOC modulation is used only on the pilot signal, and that the data signals is, indeed, a simple BOC signal. Again, a PRN sequence of length 10,230 chips was extracted.

Knowing the PRN code for the data it was also possible to demodulate the navigation symbols, which do not show any particular repetition period, suggesting a weak or no preamble, similar to the approach taken by the GPS L1C design. Given both PRN code sequences, it was possible to track the signals and estimate the actual power-sharing ratio. Initial measurements suggest that there is a ratio of 2:1 favouring the TMBOC pilot signal.

A further effort was made to identify the modulation of the two lobes, spaced at ±14 MHz relative to the center frequency. Suggestions in the literature, had indicated that these were upper and lower lobes of a BOC(14,2) signal. Initial study, however, indicates that this may not be the entire picture.

Firstly, the upper lobe, centred at 1589.742 MHz, was examined. Placing a tight filter around the main lobe and examining the chips showed a number of interesting features. Firstly, it appears to be a simple BPSK(2) signal, having a 20,460 chip code, with a 10-millisecond repetition period. It also seems to be modulated by a secondary code. The secondary code is of 20-millisecond duration, with each symbol having a duration of 1 millisecond. Interestingly, this code was found to be the same Newman-Hoffman code used on the legacy B1I signal at 1561.098 MHz, and is given by:

00000100110101001110.

A peculiar feature of this signal is the appearance of a data modulation, having a symbol period of 10 milliseconds. In all, it appeared as though the signal has a primary code with a chip-rate of 2.046 megachips per second and a period of 10 milliseconds; a secondary code with a symbol period of 1 milliseconds, repeating every 20 milliseconds, and a further data modulation with a symbol period of 10 milliseconds.

When examining the received chips more closely, a number of gaps in the signal power were observed, corresponding to the same pattern as was observed on the pilot signal at L1. This suggested that it might also be a time-division signal, but that the narrow filtering that had been applied had rejected the second component.

Figure 8: Coherent accumulation of the time-division BPSK(2) and BOC(6,2) signal located at L1 + 14 MHz over 0.5 seconds. In magenta the expected time-division pattern between chips of the BPSK and the BOC components is shown.

Figure 8: Coherent accumulation of the time-division BPSK(2) and BOC(6,2) signal located at L1 + 14 MHz over 0.5 seconds. In magenta the expected time-division pattern between chips of the BPSK and the BOC components is shown.

To identify if there was another signal broadcast during these outages, we made a search for any signal in the vicinity that followed the same modulation of secondary and data symbols as the main BPSK(2) lobe. A BOC(6,2) component was found, with lobes spaced at 6 MHz on either side of the main BPSK lobe. Specifically, they were located at 1583.604 MHz and at 1595.88 MHz, representing a BOC(6,2) component for the BPSK(2) signal at 1589.742 MHz. Once this second signal component was identified, the received samples were re-processed with a wider bandwidth, such that the entire signal could be examined. A trace of the chips is presented in Figure 8.

Indeed, this seemed a very unique signal: being centred at 1589.742 MHz, being a time-division of a BPSK(2) signal and a BOC(6,2) signal, with a time-sharing of 8/66 and a pattern given by: {1, 2, 9, 10, 13, 14, 59, 60}. Having identified the signal period as 20 milliseconds, including the secondary code, we phase-aligned the received samples and measured their autocorrelation, as shown in Figure 9.

Figure 9: Autocorrelation of the time-division BPSK(2) and BOC(6,2) signal centred at  1589.742 MHz.

Figure 9: Autocorrelation of the time-division BPSK(2) and BOC(6,2) signal centred at 1589.742 MHz.

Although the work had identified three signals: the TMBOC pilot signal at L1; the BOC data signal at L1; the time-division BPSK and BOC signal at L1+14 MHz, and had confirmed the presence of the legacy B1I signal at L1-14 MHz, the question of the presence of the reported BOC(14,2) signal at L1 remained. Early experiments, showing the characteristic BPSK(2) cross-correlation between the signals at ±14 MHz had suggested its presence. A lack of periodicity in this cross-correlation had further suggested that it may have a non-repeating spreading sequence. However, as yet, no conclusive evidence of its presence has been found.

To further investigate the presence of the BOC(14,2) signal, the spectrum of the received signal was again examined. As it had been identified that there was a time-division modulation on both the TMBOC signal at L1, and on the BPSK(2)-BOC(6,2) signal L1 + 14 MHz, which used the same time-division pattern, a spectrum of the two different portions of time was estimated. The instantaneous power broadcast during each of these two periods is shown in Figure 10. This instantaneous power spectral density (PSD) represents the power broadcast at any instant by the satellite, and will be either one or other of the PSD shown, dictated by the time-division pattern (periods 1, 5, 7 and 30 out of every 33 periods). The average power of each spectrum will be, of course, scaled by 4/33 and 29/33, respectively. Note also that the relative power is distorted across frequency by the gain of the antenna element. In particular, the higher frequencies, above 1600 MHz, are attenuated by approximately 3 dB or more.

This confirmed a number of previous observations. Firstly, it was confirmed that the L1 signal contained a TMBOC signal, and a BOC signal having half of the power, as indicated by the 29/33 curve having twice the power of that of the 4/33 curve in the two BOC lobes at L1. Secondly, it confirmed the presence of the BOC(6,1) at L1 and of the BOC(6,2) at L1 + 14 MHz. Finally, it was noted that the two BPSK(2) lobes, at ±14 MHz both contain a non-time-division component. In the case of the -14 MHz lobe, there is no time division, as the power is constant for both the 4/33 and 29/33 periods. In the case of the 14 MHz lobe, it is clear that half of the power is time-multiplexed, as there is a 3 dB difference between the 29/33 and 4/33 periods. This indicates that there is a continuous BPSK(2) signal broadcast at +14 MHz. What is more, comparing the spectrum to that of the line-spectrum observed on the -14 MHz lobe, it is likely modulated with a very long, or non-repeating, spreading sequence. One explanation for this is the presence of a BOC(14,2) signal having a non-periodic spreading sequence.

Figure 10: The normalized Time-Division PSD measurement of the received signal, illustrating the instantaneous power broadcast during the 29/33 portion and the 4/33 portions.

Figure 10: The normalized Time-Division PSD measurement of the received signal, illustrating the instantaneous power broadcast during the 29/33 portion and the 4/33 portions.

An identical analysis of the second satellite, BEIDOU-3 M1, was conducted to determine whether or not the signal specifications of both satellites were similar. Indeed, they were found to be identical, again having a TMBOC and BOC data-pilot pair centered at L1, a Time-Division BPSK-BOC signal centered at L1+14 MHz and a legacy B1I signal at L1-14 MHz. Once again, there also appears to be a BOC(14,2) transmitting a non-repeating spreading sequence.

It appears, therefore, that the new BeiDou signal baseline will consist of a TMBOC civil signal at L1, containing a data TMBOC(6,1,4/33) pilot signal and a BOC(1,1) data signal very similar to that of the GPS L1C. This signal is complimented with a pair of open signals: one being the legacy B1I signal, located at L1 minus 14 MHz; and the second being a Time-Division of a BPSK(2) and a BOC(6,1). Beneath these two signals there also appears to be a BOC(14,2) signal having a non-repeating spreading sequence. This signal scheme is consistent between the two new satellites, BEIDOU-3 M1 and M2.
The authors invite any interested parties to contact them for more information.

References
http://scpnt.stanford.edu/pnt/PNT14/2014_Presentation_Files/10.LuMQ_GNSS_Signal_Design.pdf
http://www.javad.com/downloads/javadgnss/how-to/hardware/e6_decoding.doc

This article is tagged with , , and posted in Featured Stories, GNSS, Latest News

About the Author:

GPS World covers all aspects of the GPS and GNSS industry for our readers. To submit news, please send your release to gpsworld@gpsworld.com.

1 Comment on "First Signals of BeiDou Phase 3 Acquired at Ispra, Italy"

Trackback | Comments RSS Feed