Transcript
PAU-GNSS/R, a Real-Time GPS-Reflectometer for Earth Observation Applications: Architecture Insights and Preliminary Results J. F. Marchan-Hernandez, I. Ramos-Perez, X. Bosch-Lluis, A. Camps, N. Rodríguez-Álvarez, D. Albiol Remote Sensing Lab, Dept. Teoria del Senyal i Comunicacions D3, Universitat Politècnica de Catalunya. Barcelona, Spain
[email protected] Abstract—The assumed potential of Global Navigation Satellite System Reflections (GNSS-R) to retrieve geophysical parameters has yet to be fully shown. The implementation of a GNSS-R receiver to compute real-time Delay-Doppler Maps (DDM) will be a significant step forward. In the PAU instrument these DDMs will be obtained along with co-located data of brightness temperature. To relate both measures a parameterization of the DDM is performed. Preliminary results have been obtained. Keywords- GNSSR; DDM; FPGA; GPS; Delay;Doppler;
I.
INTRODUCTION
The retrieval of ocean geophysical parameters has many applications from both a scientific and a commercial point of view. The Passive Advanced Unit for ocean monitoring (PAU) is described in [1]. It is a new hybrid instrument that combines in a single receiver an L-band radiometer and a GNSSreflectometer and also includes a thermal infrared radiometer. One of the main goals of PAU is to obtain the first collocated measurements of L-band brightness temperature, GNSS reflected signals, and sea surface temperature (SST), which will significantly broaden the synergic capabilities of the instrument. Reflectometry using geosynchronous navigation satellite system signals (GNSS/R) is becoming increasingly mature [25]. The correlation of the received GPS signal with several replicas with different delays of the locally generated C/A code, and also for several carrier frequencies that account for the Doppler shift of the signal is known as a Delay-Doppler Map (DDM). The present paper will briefly describe the system architecture. Then it will deal with the use of data from the direct link GPS signal, and the DDM parameterization will be introduced. Finally the integration and test within the PAUOR demonstrator will be explained. II.
SYSTEM ARCHITECTURE
Most of the existing GNSS-R receiver architectures just measure the peak or the cut in the delay axis of the DDMs once the Doppler is compensated, which is called a
1-4244-1212-9/07/$25.00 ©2007 IEEE.
5113
‘waveform’. The architecture of an FPGA-based real-time GPS reflectometer that computes the full two-dimensional DDM is described in [6]. The key components of that system are the DDM generator, the buffer for the incoming 1-bit lefthand polarized band-base I/Q signal for the 4 different beams to be synthesized, and the RS-232 serial interface that receives data from the commercial GPS receiver used to provide apriori information regarding the satellites on the scenario, their geometries and their Doppler offsets. The buffer is composed of two storage units that alternatively switch their input and output ports: Whenever one is receiving and storing the I-Q data the other is being unloaded (the four beams, one at a time) to compute a new DDM waveform. The clock frequency to receive the data bits is four times the sampling frequency of the data to allow simultaneous recording for four different beams. On the other hand, the unloading of raw data and the generation of DDM operations are performed at the clock frequency of the reflectometer, which is significantly higher than that of the incoming data. Thus hardware reuse is possible and four satellites can be processed at a time. The switching between the storage units takes place whenever the one being written reaches its capacity limit (i.e, 5745 samples for each of the four beams), and an interrupt is issued by the buffer. This interrupt has top priority, since no data loss is allowed, and it triggers the generation of the DDMs associated to the data of the 1ms under consideration. To do so it is necessary to translate the delay and Doppler values for signals received from different satellites to a set of parameters used by the DDM generator, and send these equivalent parameters to the DDM generator before it starts computing the new waveform values. The RS-232 interface also interrupts the system, but with a lower priority. Its interrupt service routine (ISR) reads the received data bytes, parses them and extracts from the several data packets (using a TSIP protocol) the parameters of interest, updating the chart where, for each of the 32 possible satellites, they are stored. The elevation, azimuth and power level fields of the chart are used to select which are the most
powerful satellites within a given spatial mask that are going to be selected to work with. The transfer of up to four complex-valued DDMs every 1 ms (period of the civilian GPS C/A codes at the L1 frequency band) was achieved using the USB 2.0 protocol. The resulting throughput is 2 (real and imaginary parts) x 32 bits per point x 4 beams x (16 x 16) points = 8 Kbytes every 1 ms, that is 7.8 MB/s.
Figure 1. Interface to receive, process and store the generated DDMs. In the figure a calibration test is being performed, using the same satellite ID code for the four channels
In fact, together with the DDM additional information travels, such as the time tag of the raw data and the delay and Doppler center coordinates used for its computation, because not all of the 32 bytes are used to code the DDM point values. At the other end of the USB link a program receives and performs the desired averaging of the DDM (Fig. 1). Whenever the first of the 4 DDMs has been successfully received by the host computer, an interrupt is issued in the reflectometer system by the USB FPGA controller and the parameters for the generation of the next DDM are transferred to the DDM generator, which starts the computation using the data for the corresponding beam. After 5745 clock cycles (1 ms of data sampled at 5.745 MHz) the DDM generator issues the ‘data_ready’ signal, and the microprocessor program proceeds to ask the USB controller to read the DDM values from the output interface of the DDM generator, and to send these values to the host PC, thus closing the cycle of processing and sending data. The integration time can be freely configured, since every 1ms (the shortest integration time for L1 GPS signals) one DDM is obtained. It is up to this receiving program to accumulate as many of them as configured, either coherently (in amplitude and phase) or incoherently (in absolute values). An exhaustive analysis regarding the coherent and incoherent integration times as a function of the sea state shall be conducted, and hopefully it
1-4244-1212-9/07/$25.00 ©2007 IEEE.
5114
will contribute to the already existent knowledge regarding this issue in the literature. III.
USE OF THE GPS DIRECT SIGNAL
In order to have the waveform maximum in the center of the generated DDM it is necessary to have a good estimate of its delay and Doppler coordinates. At low and moderate altitudes the Doppler maximum value for both direct and reflected signals is roughly the same, and the difference lies only in the magnitude of the Doppler spread. Taking this into account, and also considering that the temporal derivative of these Doppler shifts (~ 1Hz/s according to [7]) is much smaller than the 1-second update rate of the parameters provided by the GPS commercial receiver, its value can be used straightforwardly as the Doppler center value. Unfortunately the situation is quite the opposite when it comes to the delay value: it is necessary to estimate geometrically the range difference between the direct and reflected signals. This value can be added to the received value for the direct delay to obtain the reflected delay. Now, however, the temporal derivative cannot be considered zero over the 1-s update interval. Therefore, should the delay value obtained through the serial interface be used, the maximum of the DDM will move along the delay domain (at a speed depending on the position of the satellite) until eventually getting out of view. This drift is very inconvenient, since integration is needed to improve the low SNR of the reflected signal. If the peak moves, the integration will only further degrade the waveform. And even assuming a zero rate of change of the delay value, still an estimation of the code offset between the received signal and that provided by the commercial receiver should be somehow obtained, since both devices are independent and thus have different time references (i.e., they are not synchronized). Therefore it is of capital importance to have readily available an estimate for the delay updated the more frequently the better. To do so a whole new block is required: It performs the circular correlation by means of Fast Fourier Transform operations to find the maximum of the correlation of the direct signal for four different satellites simultaneously.
I Q
FFT
IFFT Complex Product
CA
MOD IFFT MAX
Figure 2. Block diagram of the FPGA core that computes the delay offset (1 chain of the 4 simultaneous channels)
IV. The applied equation is:
R x ,ca = IFFT ( FFT ( x) ⋅ IFFT (ca))
(1)
where x is the base band signal I + j*Q. Before computing such a correlation it is necessary to compensate their respective Doppler offsets by using the data provided by the commercial receiver and to generate for each of the satellites a local replica with zero delay offset of their C/A codes. This block has been implemented with standard FFT and complex multiplier cores, ensuring its performance and preventing an excessive use of the FPGA resources (Fig. 2). Since the number of samples (5745 samples for 1 ms of sampled data) is not a power of 2, as needed to perform an FFT, it is necessary to fill the data values until reaching 8192 (2^13) samples. This results in a correlation having several maxima with amplitudes that depend on their positions, instead of a single maximum with position-independent amplitude (Figure 3). Additional logic performs the translation of this information into the desired 0 to 5744 value for the delay offset.
PARAMETRIZATION OF THE DDM
In order to extract the geophysical parameters of interest from the shapes of the DDMs, and to reduce the high data throughput, a parameterization of these waveforms is implemented using an analytic function with 5 parameters. Each DDM is then represented by those 5 parameters that minimize the minimum squared error (MSE) between the measured and the analytically generated function. This parameterization will make easier to link the DDM and the geophysical parameters of interest, such as the sea brightness temperature.
Figure 4. Experimental setup at the top of the D3 University building to capture GPS reflections from the Collserola range (left of the image). In the the picture a test adquiring direct signal was in progress.
Figure 5. Captured DDM (SV 25 ) from a satellite signal reflected on the Collserola range
V.
Figure 3. Autocorration function of a GPS C/A code sampled at Fs = 5.745 MHz. In the upper figure 5745 samples are used, whereas in the lower one 8192 samples are considered. The added samples have been set to zero
1-4244-1212-9/07/$25.00 ©2007 IEEE.
5115
INTEGRATION IN PAU-OR
PAU-OR (one receiver) is a small-scale version of the PAU instrument [8 and 9]: a compact instrument to be placed at a fixed point to observe sea and terrain surfaces. It has been developed to test the concept before assembling the whole instrument, and also to allow more flexibility when trying to gather field data, because of its compactness. The reflectometer core, common to both instruments, has been
assembled in PAU-OR to debug the reflectometer by means of gathering the first real GPS data. The instrument was deployed in the roof of the department building, aiming to the Collserola range, to collect reflections of GPS signal from the mountain slope (Fig. 4). Taking into account the position of the satellites at the time of the measurements, it was determined which one should be captured (Fig. 5). It was determined that further precision with the delay reference should be obtained in order not to lose the lock on the peaks of the DDMs. The required adjustments are being implemented, and as soon as the first measurements over sea are available, they will be validated by comparing them with the outcome of a GNSS_R simulator [10] that generates DDM as a function of location, time, wind speed and other parameters (figure 6). It is known that the shape of the DDMs associated to the signals reflected over a certain surface depends on the nature of the surface: its roughness, spatial orientation, and dielectric permittivity. We believe that using the whole DDM instead of just the peak of the DDM or the “waveform” shall allow a better insight in the retrieval of the geophysical parameters associated with the surface. Also, depending on the surface under observation (land or sea, for example), the maximum coherent integration time must be selected according to the surface’s correlation time, so that the averaging of uncorrelated waveforms is not performed.
synchronization of the various parts that compose the system is achieved by means of interrupts and semaphore-like registers. The processing of the generated DDM (integration, parameterization, etc.) is fully configurable, and performed in the receiver terminal computer. An extra effort on the retrieval of the maximum of the delay for the reflected signal is needed not to lose track of the DDM. The first measurements show the potential of the instrument, even though further adjustments are needed. Measurement results will be presented at the conference. ACKNOWLEDGEMENTS This work, conducted as part of the award “Passive Advanced Unit (PAU): A Hybrid L-band Radiometer, GNSSReflectometer and IR-Radiometer for Passive Remote Sensing of the Ocean” made under the European Heads of Research Councils and European Science Foundation EURYI (European Young Investigator) Awards scheme in 2004, was supported by funds from the Participating Organisations of EURYI and the EC Sixth Framework Programme. It was also written with the support of the Department of Universities of the Catalan Autonomous Government and the European Social Fund. REFERENCES [1]
Measurements from the Catalan coast will be conducted under different sea state conditions. A first processing approach shall yield qualitative relationships between the sea state (mainly driven by the surface’s wind and the swell) and the overall DDM shape. The ultimate goal is to use the DDM shape to correct the sea state influence on the L-band brightness temperature to improve the retrieval of the sea surface salinity (SSS). d d
Figure 6. Simulated Reflected DDM for a surface wind speed of 5 m/s and a incoherent integration time of 10 ms.
CONCLUSION The implementation of a real-time GNSS-R receiver that outputs delay-Doppler maps is in its final stage. The
1-4244-1212-9/07/$25.00 ©2007 IEEE.
5116
A. Camps, J.F. Marchan-Hernandez, I. Ramos-Perez, X. Bosch-Lluis, R. Prehn, “New Radiometer Concepts for Ocean Remote Sensing: Description of the PAU (Passive Advanced Unit)”, Proceedings of the IGARSS 2006, July 31- August 4, Denver, Colorado. [2] M. Martín-Neira, “A Passive Reflectometry and Interferometry System (PARIS): Application to Ocean Altimetry”, ESA Journal 1993, vol. 17. pp 331-355. [3] G. Ruffini, M. Caparrini, B. Chapron, F. Soulat, O. Germain, L. Ruffini, “Oceanpal: an instrument for remote sensing of the ocean and other water surfaces using GNSS reflections”, Proceeding of EuroGOOS, 3-6 December 2002, Athens, ISBN 0-444-51550-X (Elsevier). [4] E. Cardellach, G. Ruffini, D. Pino, A. Rius, A. Komjathy, J.L. Garrison, “2003 Mediterranean Balloon Experiment: ocean wind speed sensing from the stratosphere, using GPS reflections”, Remote Sensing of Environment Volume 88, Issue 3 , 15 December 2003, pp 351-362. [5] S. T. Lowe, C. Zuffada, J.L. LaBrecque, M. Lough, J. Lerma, L. E. Young, “An ocean-altimetry measurement using reflected GPS signals observed from a low-altitude aircraft,” Proceedings of the IGARSS 2000, July 24-28, Honolulu, HI. [6] J. F. Marchan-Hernandez, I. Ramos-Perez, X. Bosch-Lluis, A. Camps, R. Prehn, “FPGA-based implementation of a DDM-generator for GPSreflectometry”, Proceedings of the IGARSS 2006, July 31- August 4, Denver, Colorado. [7] J. B. Y. Tsui, “Fundamentals of Global Positioning System Receivers. A Software Approach”, Ed. Wiley Interscience, New York, 2000 [8] A.Camps, “Passive Advanced Unit (PAU): A hybrid L-band radiometer, GNSS-reflectometer and IR-radiometer for passive remote sensing of the ocean”, European Science Foundation EURYI 2004 Awards proposal. www.esf.org/activities/euryi/awards_results/2004.htlm. [9] A. Camps, A. Aguasca, X. Bosch-Lluis, J. F. Marchan-Hernandez, I. Ramos-Perez, N. Rodríguez-Álvarez, F. Bou, C. Ibáñez, X. Banqué, R. Prehn, “PAU One-Receiver Ground Based and Airborne Instruments”, Proceedings of the IGARSS 2007, July 23-27, Barcelona, Spain. [10] D. Albiol, “Contributions to the PAU-GNSS/R receiver design and performance analysis”, UPC Master’s Thesis, 2007.
