T. Pany, N. Falk, B. Riedl, T. Hartmann, J. Winkel. IFEN GmbH, Alte Gruber Straße 6, Poing, Germany, G.

Transcription

1 Wide-band signal processing features for reference station use of a PC-based software receiver: cross-correlation tracking on GPS LP, AltBOC and the inter-frontend link for up to eight frequency bands T Pany, N Falk, B Riedl, T Hartmann, J Winkel IFEN GmbH, Alte Gruber Straße 6, Poing, Germany, TPany@ifencom G Stangl Space Research Institute of the Austrian Academy of Science, Schmiedlstraße 6, 84 Graz, Austria guenterstangl@oeawacat BIOGRAPHIES Dr Thomas Pany works for IFEN GmbH as a senior research engineer in the GNSS receiver department In particular, he is concerned with algorithm development and C/C++/assembler coding He also works as a lecturer (Priv-Doz) at the University FAF Munich His research interests include GNSS receivers, GNSS/INS integration, signal processing and GNSS science Nico Falk received his diploma in Electrical Engineering from the University of Applied Sciences in Offenburg Since then he works for the IFEN GmbH in the receiver technology department focusing on signal processing, hardware and FPGA development Dr Bernhard Riedl received his diploma in Electrical Engineering and Information Technology from the Technical University of Munich Since 1994 he has been concerned with research in the field of real-time GNSS applications at the University FAF Munich, where he also received his PhD In 6 he joined IFEN GmbH, where he is currently working as product manager SX-NSR Tobias Hartmann received his diploma in Telecommunication Engineering from the University of Applied Sciences in Ulm Since then he works for the IFEN GmbH in the receiver technology department focusing on RF Front End development Dr Jón Ó Winkel is head of receiver technology at IFEN GmbH since 1 He studied physics at the universities in Hamburg and Regensburg He received a PhD (Dr-Ing) from the University of the FAF in Munich in 3 on GNSS modeling and simulations G Stangl is an officer of the Federal Office for Metrology and Surveying of Austria and is working by contract to about 5% at the Space Research Institute of the Austrian Academy of Sciences He is working since 1985 in the filed Applications of GNSS for Reference Frames, Crustal Dynamics and Atmosphere Research Main topics are the maintenance of EUREF, contributions to the Global velocity fields and the development of regional ionosphere models ABSTRACT This paper describes a recent update of the PC-based realtime software receiver SX-NSR to process L-P signals via cross-correlation tracking and the Galileo AltBOC signal via split-band processing The NavPort-4 interfrontend link allows synchronizing two frontends for up to eight RF chains or for dual antenna usage The SX- NSR is now able to acquire and track all civil GPS, Galileo and GLONASS signals Around highbandwidth channels can be tracked in real-time per x86 CPU core (ie 8 channels on a laptop, 4 channels on a high-performance PC) The receiver is running semipermanently on the observatory Lustbühel/Austria, collocated to the IGS point GRAZ The data is analyzed with the Bernese Software 5 within a local and a regional GPS network confirming its geodetic data quality INTRODUCTION PC-based software receivers have found nowadays broad use as R&D tools to test new signal processing algorithms, to analyze received GNSS signals, or to integrate various sensors with GNSS So far, software receivers are typically not used as operational receivers, neither in the mass-market, nor in the professional sector, nor as a reference station where a PC would already be available The last point can be explained by the fact that most software receivers can only process a limited number of frequency bands (sometimes just L1) and are often limited to small bandwidth signals (eg L1C/A or LC) Recent improvements achieved at the end of 1 and early 11 of the PC-based software receiver SX- NSR try to overcome this limitations They include: the first real-time implementation of P code processing on L, a unique method to process the ultra-wide Galileo AltBOC signals on E5 and to synchronize two NavPort-4 frontends (each supporting four frequency bands of 15 MHz bandwidth) via a hardware link The SX-NSR is developed in cooperation with the University FAF Munich, runs under the Windows operating system (XP or 7) and allows to process GNSS signals plus sensor data (eg IMU) in real-time and post-

2 processing It supports all the civil GPS, Galileo and GLONASS signals as listed in Figure 4 User defined signals can be included by providing the PRN codes and the tracking parameters All-in-view tracking of the complete GPS, GLONASS and Galileo constellation on L1, L, G1 and G (or L1, L, E5a, E5b) can be done with a standard laptop, as shown in Figure 1 The SX- NSR typically connects to the NavPort-4 frontend via a single USB connector One frontend supports four RF paths with 15 MHz bandwidth in the L-band One band is sampled with 496 MHz and bit Small batches of samples are captures with 1 bit in regular intervals for increased spectral analysis possibilities Decimation and/or bit reduction are options to limit the data transfer bandwidth on the USB bus The NavPort also includes configurable notch and FIR filters working with 1-bit and 496 MHz The SX-NSR further supports standard output formats (RINEX, RTCM, ), has a graphical user interface and an application programming interface in C Figure NavPort-4 synchronization Each device generates its own 496 MHz sample rate out of this reference The phase difference of the 496 MHz sample rate is measured in the master and slave with a phase detector The first input of the detector is the local 496 MHz clock The second input is the 496 MHz clock from the other NavPort-4 with different phase alignment due to ambiguities in its generation and cable delay The phase detector measures the phase difference between both clocks The low pass filtered output of this measurement is digitized with an Analog-to-Digital Converter If this measurement is in-between a phase range of ±7 (±475ps), the coarse synchronization is finished If the value is not in-between this range the synchronization algorithm repeats Figure 1 Skyplot for GPS (G) + GLONASS (R) + GIOVE (E) at Munich, Sept 6 th, 11, 1: UTC INTER-FRONTEND LINK The inter-frontend link enables to synchronize two NavPort-4 devices It generates a synchronous reference clock for a proper phase relationship Moreover, a trigger is used to adjust the digital data stream of both devices Thus, the amount of available GNSS frequencies could be doubled Another possible application could be to build a dual antenna solution For this purpose, each NavPort-4 device handles the same GNSS frequencies, but from different antennas For the inter-frontend link both devices have to use the same 1 MHz clock reference for a synchronous setup This could be reached by using the reference clock output of the master device as reference clock input of the slave device as depicted in Figure It is also possible to connect both NavPort-4 devices to an external clock reference Figure 3 Phase measurements for bias correction After starting the data processing for both devices simultaneously with an implemented digital trigger, the phase difference between master and slave clock could be measured continuously for later fine-tuning in SX-NSR Figure 3 shows the results of such a phase measurement in the master device (upper plot) and in the slave device (lower plot) To compare the internal NavPort-4 measurements, an oscilloscope was used to measure the phase difference between master and slave Each blue cross indicates an independent measurement after powercycle and synchronization The x-axis represents the phase difference measured with the oscilloscope, and the y-axis is the internal NavPort-4 measurement value From these measurements, a straight line could be calculated as a best fit which could be used as a model for calibration in the SX-NSR One can see that the accuracy of the measurement is in a range much below 1 which results in a measured bias accuracy of much below 7 ps An exemplary dual frontend spectrum captured with the antenna on the roof top of the IFEN office in Poing near Munich is shown in Figure 5 The L1 band is sampled by both frontends showing nearly identical spectra

3 Signal Acq & Track Ranging Range in PVT Nav Extracted Nav Decoded Nav in PVT L1 C/A X X X X X X L1 P NA NA NA L P X (cross-corr X X NA NA NA with L1) LC X X X X X X L5-I X X X X X X L5-Q X X X NA NA NA E1-A E1-B X X X X X X E1-C X X X X X X E6-A E6-B X HO X E6-C X X X NA NA NA E5a-I X HO X E5a-Q X X X NA NA NA E5b-I X HO X E5b-Q X X X NA NA NA G1 C/A X X X X X X G1 P G C/A X X X X X X G P E5ab-Q Trk X X NA NA NA (AltBOC) QZSS (L1 C/A) X X X X (not fully supported) IMES (L1 C/A) X X PL (L1 C/A) X Legend: 1 Acquisition and Tracking = C/N is generated, code phase is tracked Ranging = full pseudo range is generated AND available at an output interface 3 Range in PVT = Pseudo range is used in PVT, but may rely on extrinsic navigation data 4 Nav Data Extracted = raw Navigation data frames are extracted, de-fed'd, de-interlaced, etc and available at an output interface 5 Nav Data Decoded = Navigation data frames are decoded to readable integer / float values and available at an output interface 6 Nav in PVT = This navigation data is or can optionally be used in PVT solution Remarks: X: Yes NA: not applicable HO: Code ambiguity is resolved by an absolute time information either from the PVT solution or from another frequency band Trk: only tracking via handover from another frequency, no acquisition L5-Q, E1-C, E6-C, E5a-Q, E5-b-Q, E5ab-Q: Typically the NSR tracks pilot signals as signal pairs with the data component Standalone tracking of pilot signals is possible but requires a precise receiver clock to resolve the code ambiguity LCL: possible with a limited number of channels in real-time due to large memory requirements Upgrade to x64 platform in future E1B: Iono parameters not decoded QZSS: Nav decoder available, not all subframes decoded PL: proprietary pseudolite signal For sensitivity and computational performance reasons it is recommend to perform the acquisition on L1C/A or E1B and to perform a handover to the other frequency bands Galileo and GIOVE signals supported GLONASS upgrade to P code in near future Figure 4 GNSS signal tracking capability of the SX-NSR (Sept 11)

4 fsc + B'/ fsc + B'/ G f S f df + f S f df ( ) ( ) fsc B'/ fsc B'/ () f sc BOC subcarrier [Hz] B Considered (dual-sided) bandwidth of one BOC lobe [Hz] Figure 5 Spectrum analyzer plot for 8 frequency bands, L1(twice), L, L5, E5b, E6, G1, G Interference from the Munich airport is visible on L5 and E6 GALILEO ALTBOC The AltBOC processing inside the SX-NSR relies on the fact that both frequency bands E5a and E5b are sampled coherently and this coherency can be exploited to reconstruct the full AltBOC signal This section gives an overview on the processing scheme together with the core block diagram The later sub-sections give details on the internal signal generation, implementation aspects and on the inevitable biases A conventional view on the AltBOC signal processing would require a sample rate of at least two times the total signal bandwidth [1,] Depending on how many outer AltBOC sidelobes are considered, this results in a sampling rate of 1 Msamples/s or more A more refined view considers the fact, that most of the accuracy of the AltBOC navigation signal is concentrated in the main BOC sidelobes itself More specifically, the thermal noise and multipath performance are dependent on the Gabor bandwidth, which represents the curvature of the correlation function at the tracking point [3]: ( ) R '' + B/ G = = f S ( f ) df 4π B/ G Gabor bandwidth [Hz ] (1) R Second derivative of the signal autocorrelation function [1/s ] B f S Dual-sided navigation signal bandwidth [Hz] Frequency component in baseband representation Power spectral density in baseband representation [Hz -1 ]; normalized to + S ( f ) df = 1 For BOC signals it is possible to neglect parts of the power spectral density, because the main BOC lobes span the available bandwidth more or less completely; thus By choosing eg B = 1 MHz, we approximate the AltBOC signal by two (QPSK) signals of 1 MHz bandwidth Both signals can be sampled with Msamples/s to fulfill the Nyquist criterion individually, yielding a total effective sampling rate of 4 Msamples/s This is more than a 6 % reduction In the following we call a full bandwidth sampling of the AltBOC signal like (1) as single band method and a method based on () as split band Using the following normalized expression for the AltBOC power spectral density from [7] S ( f ) = π f π f π f cos cos cos f f fsc f c c sc π f π f cos π f π f cos cos f + sc fsc 4 f sc f c Code rate [chip/s] allows to compute the Gabor bandwidth for various bandwidth values The result is shown in Figure 6 The split band bandwidth of () is B and the single band bandwidth B is from (1) The exactly same Gabor bandwidth is obtained for B=4f sc~6 MHz and B = f sc ~3 MHz G [(MHz) ] Single band Split band B, B' [MHz] (3) Figure 6 Gabor bandwidth for single band AltBOC processing and split band processing

5 The Gabor bandwidth can used to compute the Cramér- Rao lower bound for the code noise using [3] B DLL CRLB = R '' C / N CRLB ( ) Code noise CRLB in [s] The result is shown in the figure unterhalb and demonstrates the high code tracking accuracy with the split band and the single band processing of the AltBOC signal Code noise [mm], CRLB Single band Split band B, B' [MHz] (4) The block diagram of the AltBOC split band processing is shown in Figure 8 Inside the NavPort-4 frontend, the E5a and E5b signal are processed through two separate RF chains This is basically a design constraint to allow sharing the same RF channel design for all four NavPort- 4 RF channels and because the NavPort-4 supports a maximum sample rate of 496 Msamples/s per RF channel On the positive side, this design is more robust than a wide band E5 channel against interference as it provides fallback options if one of the two frequency bands E5a/b is jammed The local oscillator of E5a and E5b are derived from the same clock and remain in a constant phase relationship throughout one power cycle However, the difference of the initial E5a, E5b phases are randomly distributed within 36 All local oscillator frequencies are derived from a common reference frequency of 51 MHz Each RF path uses a separate fractional PLL to phase-lock the LO frequency to the reference frequency The PLLs are housed in separate chips and are initialized independently The PLL chip experience the problem that any propagation delay through the counters is not defined ( communications from RFMD) As a consequence the initial LO phases are unrelated to each other A later hardware revision of the NavPort-4 might foresee to use exactly the same LO for both E5a and E5b Then the E5a/b phase difference will be constant (apart from thermal variations) over power cycles Figure 7 Code noise Cramér-Rao Lower Bound for C/N = 45 dbhz and B DLL = 1 Hz NavPort-4 E5a filter x Lowpass ADC GNSS RF signal USB E5b filter x Lowpass ADC E5a LO E5b LO Clock E5a samples x x Sum Ringbuffer NCO transform SX-NSR Sin/ cos E5a NCO E5a-Q USB E5a loop E5b loop E5 loop E5 NCO + Code/carrier/ freq discrimintors Sin/ cos E5b NCO E5b-Q E5 code carrier pseudoranges E5b samples x x Sum Ringbuffer NCO transform Figure 8 AltBOC tracking inside the SX-NSR (colored elements run within the same execution thread of the SX-NSR)

6 The E5a and E5b signal samples are generated synchronously inside the same ADC chip and are transferred via the USB bus to the PC running the SX- NSR The SX-NSR first acquires and tracks the signal separately on E5a and E5b For performance reasons this is done in separate threads The threads are synchronized to each other with a fixed rate; currently every 1 ms In between the synchronization points the threads (and thus the E5a/b tracking channels) run independently The integrate-and-dump rate is related to the chosen coherent integration time Here a value of ms provides a good compromise between squaring loss and frequency/phase tracking stability As it is quite efficient to run the E5a and E5b on separate threads (and on separate CPU cores), the combination of E5a and E5b correlation values to E5 correlation values is done at the post-correlation level There is no feedback from the E5 channel to the E5a/b channels The E5a/b correlation values are stored in ring buffers together with the E5a/b NCO values that were used for the correlation Once the E5a and the E5b signals of a certain transmitter are tracked, an E5 channel is started This channel does not have own correlators, but retrieves E5a and E5b correlation values from the ring buffer The channel maintains its own NCO A dedicated transformation is used to account for NCO differences between the E5a/b NCO values and the E5 NCO values It is basically a sinc-interpolation in the code-phase direction and accounts for Doppler and carrier phase differences The transformed correlation values are added and are used to compute the discriminator values, as will be outlined below The E5 NCO values are used to compute the code and carrier pseudoranges, the Doppler and the C/N values, which are the primary output of the E5 receiver Although the E5 receiver is a somehow virtual receiver (ie without correlators) it has the same user interface including most of the configuration parameters, output (eg multi-correlator) and API In the current implementation the E5 channel considers only the pilot component The inclusion of the data components is considered to be of less importance due to the following reasons First, the data component contributes only a high C/N values, as at lower C/N values the short integration time (4 ms) and the data bit cause a significant squaring loss At high C/N values, the only benefit of the data channel (+3 db) is not that important as the error budget is dominated not by noise, but rather by multipath Furthermore, E5a and E5b data bits are generally different and don t allow to form the AltBOC Post-Correlation Correlator Merge The AltBOC tracking method correlates the E5a and E5b AltBOC pilot component separately and adds the correlation values to obtain full AltBOC correlation values This is described mathematically in this subsection The pilot component of the AltBOC can be approximated as: ( ) exp( π ) + ( ) exp( π ) ee 5a Q tµ if sc t µ r ( tµ ) a e E5b Q t µ if sc t µ e E5x-y a (5) Spreading code on E5x, y component Signal amplitude The AltBOC pilot is approximated as two BPSK(1) pilot signals, separated by twice the subcarrier The approximation of the rectangular subcarrier by a sinusoidal signal causes a correlation loss of around 9 db To track the two BPSK signals as a single AltBOC signals requires that the E5a and E5b NCOs are derived from a single E5 NCO In other words, the code and carrier generators of E5a and E5b need to be steered in a way, to precisely reproduce the respective AltBOC components A model for an AltBOC signal (with the sine approximation for the subcarrier) at RF level can be derived by separating the E5a and E5b signal component into two separate signal paths This is a bijective transformation as E5a and E5b are well separated in frequency domain r µ, AltBOC, RF = ( α τ ) iπ fsc ( αd tµ τ G ) ( α τ ) + iπ fsc ( αd tµ τ G ) e t e a ee5b Q D tµ G e τ G τ P α D f ab e ( ) E5a Q D µ G iπ fab αd tµ τ P r µ,altboc,rf φ µ ϕ µ Code delay [s] at E5 Phase delay [s] at E5 Dimensionless Doppler E5 center frequency [Hz] (6) x1 vector for two complex valued signal samples at RF (upper E5a, lower E5b) NCO code phase for sample µ [s] NCO carrier phase for sample µ [cyc] The NCO code phase φ µ is related to the code pseudorange (or code delay) τ G via φ = α t τ µ D µ G and NCO carrier phase ϕ µ to the carrier pseudorange (or carrier delay) τ P via ( t ) ϕ = π f α τ µ ab D µ P The model can be translated to the intermediate frequency level Two different local oscillator frequencies can be chosen for E5a and E5b In general their numerical values (7) (8)

7 do not have any influence on the processing result, as the values cancels in the correlation ( φ ) ( φ ) iπ fsc φµ iπ f LO, a t µ ee 5a Q µ e e iϕµ µ, AltBOC, IF = a e + iπ fsc φµ iπ f LO, b tµ ee5b Q µ e e r f LO,a f LO,b r µ,altboc,if (9) Local oscillator frequency for E5a [Hz] Local oscillator frequency for E5b [Hz] x1 vector for a two complex valued signal samples at IF (upper E5a, lower E5b) Comparing this expression to a signal model for two E5a,b BPSK signals allows us to establish a relationship between the E5 NCO values and the E5a,b NCO values The NCO carrier phase relationship reads as ( ) ( ) ϕ = ϕ φ f + t f f a, µ µ µ sc µ LO, ab LO, a ϕb, µ = ϕ µ + φ µ f sc + t µ f LO, ab f LO, b (1) ϕ a,µ E5a NCO carrier phase for sample µ [cyc] ϕ b,µ E5b NCO carrier phase for sample µ [cyc] f LO,ab and the code phase relationship as φ φ = φ = φ a, µ µ b, µ µ φ a,µ φ b,µ Local oscillator frequency for E5 [Hz] (11) E5a NCO code phase for sample µ [s] E5b NCO code phase for sample µ [s] The corresponding relationships for the NCO carrier rates (which is basically the Doppler) and the NCO code rates can be obtained by computing the first derivatives The received signal is modeled as s = r + n s µ µ µ, AltBOC, IF µ n µ (1) x1 vector for a two complex valued received signal samples (upper E5a, lower E5b) x1 vector for a two complex valued received noise samples (upper E5a, lower E5b); each component of n (real and imaginary parts included) has unit variance Noise is white Gaussian and E5 correlator values are obtained by computing L µ µ, AltBOC, IF µ = 1 C = s r C L (13) Complex valued E5 correlator value Number of samples in integration interval Assuming that code rate errors and Doppler errors are much smaller than the inverse coherent integration time, one ends up with a model for the correlation values [3] In the following formula for the correlation model, code/phase delay values with an index denote replica parameters; values without this index denote received signal parameters C ( τ G,, τ P, ) = alr ( τ G τ G, ) { i π fab ( τ P, τ P )} π fsc ( τ G, τ G ) exp cos R { } (14) Cross-correlation function of the E5a or E5b component of the received signal at baseband with the baseband representation of the replica The correlation values depend on the code delay error and this dependency is the product of the BPSK correlation function (a triangle) with the subcarrier This proves, that the split band method approximates the AltBOC correlation function Correlator Transformation The block NCO transform is used to transform multicorrelator values obtained by independent E5a,b tracking The goal is to get multi-correlator values for E5a,b NCO values obtained by transforming E5 NCO values via (1) and (11) The independent-tracking E5a,b NCO values and the transformed E5a,b NCO values will be slightly different One should note that multi-correlator values represent a sufficient statistic of the received signal provided that the code phase offsets of the multicorrelator values are small enough (eg smaller than the inverse of the bandwidth) Thus all information of the navigation signal is contained in the multi-correlator values and this transformation can be in theory lossless In practice, a larger multi-correlator spacing s or numerical errors might introduce errors To transform multi-correlator values from one set of NCO values to another set of NCO values, one has to apply a phase rotation to account for the NCO carrier phase difference and to interpolate the multi-correlator values corresponding to the NCO code phase difference The phase rotation is easily realized by multiplication with the complex exponential of the phase difference The code interpolation is done via a sinc-interpolation, which is exact provided that the Nyquist criterion (for the E5a or E5b BSPK signals) is fulfilled Doppler differences are accounted for intrinsically via the phase correction We

8 assume that Doppler difference are much smaller than the coherent integration time Code rate difference can be neglected Biases Every GNSS signal is influenced by several receiver, transmitter and propagation biases, which degrade the measurement accuracy Whereas all commonly known biases (eg hardware delays, atmospheric delays, ) also affect the AltBOC signal, the AltBOC signal is specifically influenced by two more biases due to its wide bandwidth: filter and LO biases and ionospheric dispersion effects RF Filter and Local Oscillator Biases RF filters, LNAs or mixers generally introduce a frequency dependent group and phase delay These delays can not be expected to be exactly identical on the centre frequencies of E5a and E5b Delay variations of the 9 MHz ceramic AltBOC filter used in the NavX-NTR (not SX-NSR) on the centre frequencies of E5a and E5b are on the order of a few nanoseconds The phase delay difference is 6 As mentioned above, the NavPort-4 frontend uses two independent local oscillators for E5a and E5b, which causes a phase delay between E5a and E5b of maximally +/- 5 cycles It is important to note, that the difference changes with each power cycle of the NavPort-4 (current hardware revision) Mathematically the signal model changes due to hardware delays to ( φ + ) ( φ ) = iπ fsc φµ + iϕ µ + iπη ee5a Q µ D e µ, AltBOC, RF a + iπ fsc φµ + iϕ µ iπη e E5b Q µ D e r D η and the correlation model to C (15) E5a, E5b group delay difference [s] E5a, E5b phase delay difference [s] ( τ G,, τ P, ) alr( τ G τ G, ) { i fab ( P, P )} fsc ( G, G ) { } exp π τ τ cos π τ τ η (16) Assuming that the group delay difference is much smaller than the inverse BPSK signal bandwidth, we may approximate the BPSK correlation function by the average of the E5a and E5b correlation function The phase delay shifts the subcarrier away from the BPSK correlation function, which significantly distorts the overall AltBOC correlation function (this has been explicitly verified by visual inspection of the AltBOC correlation function) The effect should be identical for all channels As a net effect an E5a/E5b phase delay difference is absorbed into the total AltBOC service delay An estimate of the phase delay can be obtained via introducing an artificial phase delay and choosing it in a way to maximize the total received signal power at the prompt correlator from all AltBOC signals: ˆ η = arg max k η k R ( τ G, k τ G,, k ) fsc ( G,, k G, k ) { π τ τ η } cos (17) Index to count the different transmitters Such an estimation can obviously not differentiate between a frontend caused phase delay difference and phase delay differences caused by the propagation channel Most importantly multipath effects, will have a different phase on E5a and E5b Those effects are not constant in time The estimation procedure has been tested with the GATE test signals, known to show strong multipath effects The estimation is based on 1 correlation values (each with a coherent integration time of ms) Correlation values from all 8 GATE transmitters are considered (overall the calibration phase lasts 1 * ms / 8 = 5 s) After an estimation has been performed, the receiver (but not the frontend) is reset, and the procedure is repeated (for the purpose of obtaining this plot) The antenna was static The figure unterhalb shows the resulting total received power (maximum normalized to 1) for all 1 test runs The estimated η values are in the range of 14 to 144 Total normalized received power E5a/b phase delay difference [deg] Figure 9 Total received power as a function of assumed E5a/b phase difference η for various runs in the GATE test area Ionospheric Dispersion The AltBOC ionospheric dispersion can be modeled by applying the group and phase delay on the E5a/E5b signal component separately:

9 ( φ + ) ( φ + ) = iπ fsc φµ + iϕ µ iπ f ai a ee5a Q µ I a e µ, AltBOC, RF a + iπ fsc φµ + iϕ µ iπ f bi e b E5b Q µ I b e r I a I b Ionospheric delay on E5a [s] Ionospheric delay on E5b [s] Then the correlation model changes to ( τ G,, τ P, ) ( τ G τ G, G, ab ) i π fab ( τ P, τ P ) fabi ab C = al R + I { } { π fsc ( τ G, τ G + I ) } exp cos (18) (19) With centimetre accuracy, the conventional delay model can thus be applied also to the AltBOC signal Most notable, the dispersion causes a shift of the subcarrier compared to the BPSK correlation function This finally causes a slight (probably nonlinear) distortion of the AltBOC correlation function Verification The AltBOC tracking module is verified by the NavX- NCS RF signal simulator and by tracking the GIOVE-B satellite The computational load to combine the E5a and E5b correlator values is very low and is a few percent for a standard quadcore laptop NavX-NCS where the following delay parameters have been introduced I ab f I f I f I f I I = = f f f I G, ab faia + fbib = f ab a a b b a a b b Ia + Ib = sc b a () For the laboratory tests, the NSR has been connected to a NCS signal generator, which was configured to output the Galileo E1 and the Galileo E5 signal To better asses the achievable accuracy, no atmospheric errors have been simulated The core NSR parameters are listed in Table 1 The reader might note that a sampling rate below the Nyquist rate is used Although the NSR can be configured to work with twice the sampling rate (eg 496 MHz), we chose to keep the lower one, as the processing results show already an excellent precision Sub sampling losses are ~16 db To derive (19) we approximated both correlation functions R by their average, which is obtained by replacing I a and I b by their average I G,ab This is reasonable, as I a and I b differ by maximally 15 m assuming a maximum ionospheric delay on the E5 centre frequency of 3 m Assuming a 1/f law for the ionospheric delays, one can plot the parameters () as a function of the delay on the E5 centre frequency As all three parameters are numerically very near to the delay on the centre frequency, we plot the difference to the delay on the centre frequency in Figure 1 Parameter RF Bandwidth E5a/E5b Sampling rate (real valued) Sampling rate (complex valued) AltBOC correlator spacing E5a/E5b correlator spacing E1 correlator spacing Value 15 MHz 48 MHz 14 MHz 5 chip chip chip 15 I ab -I E5, I-I E5 DLL bandwidth 1 Hz I G,ab -I E5 PLL bandwidth 15 Hz I ab, I, I G,ab [m] I E5 [m] Figure 1 Difference of the AltBOC ionospheric delay parameters wrt the delay on the E5 centre frequency RF Bandwidth E5a/E5b Table 1 AltBOC SX-NSR parameters Correlation Functions 15 MHz The I/Q correlation functions (plotted as a function of code delay and Doppler offset) for PRN are shown in Figure 11 The E5a/b phase difference has been fixed and all channels are in phase lock As expected, virtually all power is in the I channel and the correlation function has the expected shape

10 The performance of the bump jumper has been verified by intentionally shifting the code phase by 1 m The bump jumper detects this offsets and corrects this intentional shift Figure 1 E5a (upper) or E5b (lower) correlation function I while in phase lock, PRN Figure 11 AltBOC correlation function I (upper)/q (lower) while in phase lock, PRN Looking more carefully on the correlation function, one realizes that it does not approach zero for code delays > 1 chip Instead, a small residuum remains This can be explained by looking at the respective E5a or E5b correlation function shown in Figure 1 Probably, the low bandwidth of the RF filter (15 MHz) tends to smear out the correlation function, or the approximation of the rectangular subcarrier by a complex exponential has a similar effect Since neither of the two issues can be changed and because the code tracking is near the theoretical value, no further investigations on this issue have been performed Code Tracking Accuracy The code tracking accuracy is determined from code minus carrier plots for a single PRN and all four tracked frequency bands (E1, E5a, E5b and the AltBOC) in Figure 13 This plot is mainly used to assess the code noise as the carrier noise is at least one magnitude smaller than the code noise The E1 and E5a/b code noise are quite similar due to the different correlator spacing used (cf Table 1) For PRN the reported C/N value is 4834 db for E5a(I+Q), E5b(I+Q) and the AltBOC (E5aQ+Eb5Q) Taking into account that our Nyquist bandwidth is 14 MHz and the 3 db RF bandwidth is 15 MHz, the effective C/N is reduced by 16 db due to aliasing of noise into the Nyquist bandwidth; this results into an effective C/N of 4674 db For the AltBOC signal we measure a standard deviation of the code noise of 184 mm Recomputing the AltBOC CRLB for a C/N = 4674 db and a split band bandwidth of 15 MHz yields code CRLB of 1645 mm, which is in excellent agreement with the measured result

11 5 Galileo_E1_dump_PRNdumplog 5 Galileo_E5a_dump_PRNdumplog Code minus carrier [m] Code minus carrier [m] GPS time - t = s [s] GPS time - t = s [s] 5 Galileo_E5b_dump_PRNdumplog 5 Galileo_E5ab_dump_PRNdumplog Code minus carrier [m] Code minus carrier [m] GPS time - t = s [s] GPS time - t = s [s] Figure 13 Code pseudorange minus carrier pseudorange (from upper left: E1, E5a, E5b, E5), PRN Positioning Results To asses the accuracy of the AltBOC signal on the positioning level, a single point positioning solution was calculated on an epoch-per-epoch basis based only on code pseudoranges using the settings of Table 1 All 6 AltBOC signals were used as shown in Figure 14 GIOVE-B The SX-NSR can be configured to track the GIOVE test satellites and GIOVE-B is known to broadcast currently the AltBOC signal on E5 The figure unterhalb shows the code noise for two hours of GIOVE-B tracking It is a code minus carrier plot where the ionospheric codecarrier divergence was removed via phase observations from E1 and E5a Since the phase noise and multipath is usually much smaller than the code contributions, we assess the code accuracy for GIOVE-B on E1 6 m, on E5a 34 m and on E5b 33 m The AltBOC on E5 delivers 57 ranging accuracy (DLL bandwidth = 5 Hz) On E1, a correlator spacing of chip was used, and on E5a/b a spacing of chip The AltBOC spacing was 5 chip Figure 14 NCS AltBOC Sky plot The North error is 88 mm +/- 14 mm, and the East error is -4 mm +/- 16 mm This is obviously far beyond of what can be reached in a real-world single point code positioning applications due to inevitable multipath, orbital errors or atmospheric errors

12 Code error [m] Code error [m] Code error [m] C/N [dbhz] E1B+C, E, code Tracking Performance, Iono Removed E5aI+Q, E, code Tracking Performance, Iono Removed E5 AltBOC, E, code Tracking Performance, Iono Removed E5bI+Q, E, code Tracking Performance, Iono Removed 1 Code error [m] Figure 15 Code noise (plus multipath and residual iono delay) from station Poing near Munich, DOY 54, 11, black=e1b+c, red=e5ai+q, green=e5 AltBOC, blue=e5bi+q, GIOVE-B CROSS-CORRELATION TRACKING ON L Although the GPS modernization process is ongoing and ten LC capable satellites are in orbit, tracking of the encrypted P code signal on L is still the key element for any receiver to be considered as a reference station receiver, because dual-frequency observations need to be available for the full GPS constellation Cross-correlation tracking of the encrypted P code signal on GPS L is described in [6] The receiver computes the cross-correlation function between the raw L1 and L samples over a long coherent interval as shown in Figure 16 A receiver internal complex carrier is generated (Freq compensation), whose frequency equals the Doppler + IF difference between L1 and L This frequency is generally different for each satellite The L signal is delayed to compute the cross-correlation function for several code phase taps A performance analysis of crosscorrelation tracking can be found in [3] Figure 16 Cross-correlation block diagram from [3] Cross-correlation tracking shows a relative high squaring loss, but provides full wavelength carrier phase estimates It can be implemented in a numerically efficient way, as no P-code needs to be generated and the Doppler compensation frequency is low If six code phase bins are selected, then the computational load of one crosscorrelation channel equals approximately the load of one GPS C/A code channel More sophisticated tracking methods require the generation of the P code in real-time Such options are currently under investigation and might be realized with the FPGA inside the NavPort-4 frontend, the graphics card or directly with the CPU To compute the cross-correlation function, the received signals are converted from a real-valued signal representation at the IF to a complex-valued baseband representation Then the sample stream is decimated by a factor of like in a Hilbert transformation The cross-correlation function is computed using the predicted Doppler difference based on the Doppler estimated from L1 A number of batches are collected and a post-correlation FFT transformation is applied [4] The parameters of Table 1 result in a total coherent integration time of 56 s Shorter batch values of eg 1 have also been tested successfully The result is the crosscorrelation function as a function of code phase and Doppler as shown in Figure 17 Using interpolation techniques, the position of the peak is determined, which then gives the delay and Doppler shift of the L signal wrt the L1 signal The complex argument of the peak value gives the L-L1 carrier phase differences Those differences are filtered and are then added to the L1 parameters to give the LP code estimates A formula of [3] is used to determine the received signal power of the L signal It implicitly assumes that the L1-L P-code transmitted power difference is 3 db Of special importance is a zero latency link between the L1 and L channel in order that the cross-correlation process precisely follows the L1 carrier phase tracking We use two first order Kalman filters (on for the code, one for the phase) to smooth the cross-correlation estimates The code filter is updated with the estimated delay and the Doppler; the phase filter is updated with the estimated Doppler and phase Cycle-slips are detected if the L1-L phase changes are too high Loss-of-lock is detected by comparing the estimated L C/N value against a threshold We found that the algorithm works

13 reliable if the GPS C/A C/N is at least in the upper thirties depending on the tracked satellite LP(Y) G5, code Tracking Performance, Iono Removed Parameter RF Bandwidth L1/L Sampling rate (real valued) Value 15 MHz 48 MHz Code error [m] Code error [m] LCM+L G5, code Tracking Performance, Iono Removed - Sampling rate (complex valued) 14 MHz L1C/A G5, code Tracking Performance, Iono Removed LP correlator spacing Correlator positions in (complex) samples 1 chip, 1,, 1,, -3 Code error [m] L5I+Q G5, code Tracking Performance, Iono Removed Batch length 4 s Number of batches 64 Table 1 SX-NSR cross-correlation parameters parameters Code error [m] C/N [dbhz] Figure 18 Code noise (plus multipath and residual iono delay) from station GRAB, DOY 49, 11, black=lp, red=lcm+l, green=l1c/a, blue=l5i+q, GPS PRN 5 GPS REFERENCE STATION OPERATION Figure 17 GPS L1/L cross-correlation function for PRN3 The resulting performance is best analyzed by running the SX-NSR as permanent GPS reference station as will be described in the next section Results at single channel level are shown in Figure 18 The Block IIF satellite PRN5 is tracked over a whole pass and the C/N values plus the code-minus-carrier pseudorange is shown A 6 th order polynomial has been subtracted from the difference to remove the ionospheric code-carrier divergence prior to plotting Small stars indicate a [rarely occurring, only on LP] cycle slip and disrupt the code-minus-carrier computation The code noise and C/N values conform to our expectations A code tracking loop bandwidth of 5 Hz was used C/N variations due to multipath are visible in the LCM+L signal as well as in the LP code signals, but are more expressed in LCM+L One of the key applications of a professional GNSS receiver, is its usage as a GNSS reference station Using a software receiver for this purpose would also provide increased monitoring capabilities to detect (un)intentional inference via a RF spectral analysis or to detect signal anomalies due to satellite failures or multipath [5] Furthermore, it is well useable for a number of scientific experiments, like scintillation monitoring or atmospheric occultation The software receiver was attached to the Austrian site GRAB which is an eccenter to the IGS-station GRAZ (Graz Lustbuehel, see Figure 19) Using a Topcon antenna the site was included into two of the presently six OLG (Observatory Lustbuehel Graz) networks which are routinely processed in post-processing mode according to EPN ( lines_analysis_centrespdf, EUREF Permanent Network) standards The analysis is done by the Bernese Software version 5 ( Before the network analysis the RINEX files were checked by the quality check of teqc (UNAVCO, org/software/teqc/teqchtml) As an example the multipath estimations of the original Topcon receiver at DOY346 1 and the new software receiver at DOY3 11 (first day) are for L1/L 39/36 to 9/359 m In the meantime the noise could be reduced to about 5 m It was detected that the main part of the noise is coming from low elevations, at higher elevations >45 the value is

14 around 5-6 m 1 Combined with a mapping function these noisy data lead to major distortions which decrease the coordinate values to decimeter accuracy only Therefore it was decided to choose a cut-off angle of degrees (standard is 3 degrees) for GRAB in one network As can be seen from Figure the precision increases to better than 1 mm with some instabilities in the East and Up component Additionally there is a bias of about 1- cm in the Up component resulting from the different elevation cut-off angles caused by the GPS geometry Running the SX-NSR as a reference station has been partly successful, as the GPS data fitted into established processing schemes without problems On the other hand, multipath at lower elevations is higher than for hardware receivers; this will be corrected during future parameter and software updates We plan to contribute to the IGS M-GEX experiment ( not only providing RINEX data but also additional observables (spectra, multi-correlator values, ) to demonstrate the benefits of a software receiver solution for this kind of application ACKNOWLEDGEMENTS The research leading to the AltBOC results has received funding from the European Community's Seventh Framework Programme (FP7/7-13) under grant agreement n REFERENCES [1] Dötterböck, D, Stöber, C, Kneissl, F, and Eissfeller, B, "Tracking AltBOC with the ipexsr Software Receiver," Proc ION- GNSS 1, Portland, pp , 1 Figure 19 IGS-Site GRAZ and eccenters, GRAB is the location of the SX-NSR, J Weingrill [] Lück, T, Winkel, J, and Bodenbach, M, "A Complex Channel Structure for Generic GNSS Signal Tracking," Proc ION-GNSS 9, Savannah, 9 [3] T Pany Navigation Signal Processing for GNSS Software Receivers, Norwood: Artech House, 1 [4] Stoeber, C, Kneissl, F, Eissfeller, B, and Pany, T, "Analysis and Verification of Synthetic Multicorrelators," Proc ION- GNSS 11, Portland, 11 Figure GRAB coordinate residuals in [mm] of the new receiver compared to the a-priori from a 1-year-time series in North, East, Up CONCLUSIONS The increased number of available signals adds value to the SX-NSR as an R&D tool Applications like GNSS/INS integration, vector tracking, reflectometry and many more benefit from this increase drastically Also the broadband AltBOC signal will offer unique applications 1 For this test, the C/A signal was tracked with a correlator spacing of chip an no dedicated multipath mitigation was performed This will be done in future work [5] Stöber, C, Kneißl, F, Krämer, I, Pany, T, and Hein, G, "Implementing Real-time Signal Monitoring within a GNSS Software Receiver," Proc ENC-GNSS 8, Toulouse, 8 [6] Wu, J Y T F, Single-Epoch Weighting Adjustment of GPS Phase Observables NAVIGATION, Journal of The Institute of Navigation, vol 5, pp 39-48, 5 [7] Ávila Rodríguez, J Á, On Generalized Signal Waveforms for Satellite Navigation 8 University of Federal Armed Forces MunichWerner-Heisenberg-Weg 39, D Neubiberg