ON-GROUND DIGITAL BEAMFORMING TECHNIQUES FOR SATELLITE SMART ANTENNAS

Transcription

1 ON-GROUND DIGITAL BEAMFORMING TECHNIQUES FOR SATELLITE SMART ANTENNAS ABSTRACT P. Angeletti (1), G. Gallinaro (), M. Lisi (1), A. Vernucci (). (1) Alenia Spazio SpA Via Saccomuro, Roma (Italy). p.angeletti@roma.alespazio.it () Space Engineering Via Dei Berio, 91, I Roma (Italy). gallinaro@space.it Flexibility boons offered by digital beamforming techniques are well known and regarded by the wireless telecommunications industry as a key elements of the upcoming mobile network of third generation. Regarding mobile satellite systems, the concept of "smart antennas" implementing adaptive beamforming to real-time counteract co-channel interference seems unreachable on-board the satellite without a drastic increase of the payloads complexity. In order to simplify the satellite architecture, the possibility to entrust to the gateway the exploitation of the more sophisticated adaptive techniques was envisaged. The paper presents results of the studies on the concept of on-ground digital beamforming applied to the return-link of a non-geostationary mobile satellite system. INTRODUCTION - ELEMENTS OF DIGITAL BEAMFORMING The potential advantages of adopting Digital Beam-Forming (DBF) techniques in the context of mobile communications systems are widely documented in the literature [1]. To provide more efficient and higher quality communications, DBF can offer to antennas a certain degree of "smartness" [] making them capable of various kinds of flexible and real time pattern control, for example: - Beams can be individually formed, steered and shaped. - Beamforming strategy can be software upgraded. - Interference can be minimised implementing Adaptive Beamforming. - Beams can be assigned to individual user. - DSP techniques (filtering, multiplexing, despreading, demodulation, signal information extraction, performance optimisation, etc.) can be implemented in a seamless digital environment. Similar advantages have been envisaged in satellite mobile communication [3] and DBF technologies are actually implemented in the payloads of the ICO MEO satellites and of the Thuraya GEO satellites. On-board the satellite, DBF allows the implementation of multibeam phased-array antennas either in a Direct Radiating Array or in an Array-Fed Reflector configuration with full control of their potential capabilities. 1

2 DBF for mobile satellite communications has often been proposed to generate a very large number of beams with the aim to recover the loss normally experienced, with a traditional coverage, by users located at the beam edge. The advantage of DBF networks over classical BFNs is represented by the simplicity in both operation and realisation. Indeed, conventional BFNs require a complex combination of physical splitters, phase shifters, attenuators and combiners. DBF network complexity is related to the number of feeds (both for the RF and Digital section), to the overall bandwidth required and the number of generated beams (principally for the Digital section). An advantage of paramount importance of DBF is the imposed simplification on the onboard routing arrangement (e.g., elimination of the space-switching stage). This can be achieved by the routing of an up-link carrier to the desired destination, by simply synthesising an additional adhoc beam which covers the area of interest, instead of a physical routing of the signal to the appropriate antenna port. Considering the receive DBF array architecture, the signal is down-converted and analog to digital converted at the element level and then processed in the digital domain. This approach permits to preserve each element information making feasible a large band of processing techniques. A typical configuration of a DBF array antenna is presented in Figure 1. Figure 1 - Receive DBF Array Antenna A static synthesis of beams (classical beamforming), known as spatial filtering in the signal processing community [4], is only one of the possible utilisations of the DBF features, but even in that application DBF networks result theoretically superior with respect to analog BFNs. Just to cite an example, the digital architecture permits to overcome the Stain limit for analog lossless networks.

3 EMERGING MOBILE COMMUNICATION MISSIONS New generation mobile satellite communications missions are required to realise the convergence of multimedia applications with mobility in the frame of a global communication village that is experiencing a radical revolution toward IMT-000 defined standards such as UMTS. For the satellite component of UMTS (S-UMTS) standardisation activities are ongoing and the ESA Satellite Wideband Code Division Multiple Access (SW-CDMA) seems a good reference access technique. It offers significant similarities to the terrestrial 3GPP (Third Generation Partnership Project) W-CDMA scheme (frequency division duplex mode - FDD), supporting a wide range of bearer services and permitting to exploit satellite path diversity and soft handover in the forward and reverse link. ON-GROUND DIGITAL BEAMFORMING In a similar scenario, the incorporation of a DBF processor in a the return link repeater (i.e. mobile terminal to gateway) would be quite a challenging task, due to the expectedly very high on-board digital-processing load, especially if in conjunction with a Code-Division Multiple Access (CDMA) scheme by which each up-link beam exploits the total bandwidth allocated to the system. An alternative solution to achieve similar performance is that of implementing DBF at the gateway (GW), in conjunction with a transparent return-link repeater, thus moving the processing burden from space to ground. A similar approach was adopted in NASA TDRSS, where multiple access service is provided by an on-board S-band antenna (30 helix element array) and a groundimplemented phased array beam forming [5]. The return-link of such an S-UMTS system can be modelled as three main sections (for our purposes, it is not necessary to also model the mobile terminals and the propagation S-band up-link channel): the transparent return repeater: from the S-band receive antenna feeds to the Ku or Ka band transmit antenna feed(s) (the choice of the K band is for bandwidth); the propagation medium relevant to the feeder-link band; the GW receive side: from receive antenna to the input of the Digital Signal Processing (DSP) subsystem in charge of performing all computations required to synthesise the desired beams. The return repeater is assumed to utilise a Direct Radiating Array (DRA) up-link antenna having a total of N receive feeds. Each feed is connected to a distinct receive chain operating on the same 10- MHz segment of the S-band (a 10-MHz total allocation was considered to be sufficient for an S- UMTS-compatible satellite, each such segment being supposed to host two adjacent 3.98-Mchip/s CDMA modules, about 5-MHz each). Each receive chain, which incorporates an LNA (Low-Noise Amplifier), frequency-converter and IF amplifier, produces a signal which feeds a Frequency-Division Multiplexer (FD-MUX). In reality we foreseen two of such FD-MUXs for polarisation reuse purposes, each one serving N/ receive chains, but Figure shows only one for simplicity. Irrespective of the preferred receive chain and FD-MUX internal architecture, the overall arrangement shall result in a FD-MUX output signal having a bandwidth directly corresponding to the sum of the bandwidths handled by each S-band feed and receive chain plus necessary guard 3

4 bands. Said frequency-multiplexed signals feed a power amplifier which in turn feeds a Ku or Ka band transmit antenna. Figure - Return Link On-Ground Digital Beamforming Scheme The propagation medium attenuates signals due to both free-space loss and absorption (gas and rain), and creates interference between the two orthogonal signals due to depolarisation effects. It should be noted that attenuation and depolarisation are time-variant effects as well as free space loss is for non-geostationary (NGSO) satellites. To minimise cross-polar interference, the frequency plan on the two polarisation is typically staggered by half bandwidth (in our case a.5-mhz shift). The propagation also causes a frequency shift due to the Doppler effect. At the gateway a classical RF chain supplies an analog to digital converter (ADC); in the case the down-link bandwidth exceeds the available ADC sampling rates, an RF sub-bands division could be adopted at the expense of an even multiplication of the ADCs. After the conversion, the frequency multiplexed feed signals can be digitally demultiplexed (e.g.: poliphase network + FFT). At this stage, to synthesise one beam, the samples of the various feeds can be managed with the normal DBF techniques (multiplication by a different amplitude / phase coefficient and sum of the modified samples). The same process is repeated for each beam to be generated, the set of amplitude / phase coefficients being different for each beam. ERRORS EFFECTS The process discussed in the previous section subtend the hypothesis that the mutual amplitude and phase relationships among the feed signals do not get randomly altered by the possibly different and non-constant behaviour of the various end-to-end paths (that is from each S-band antenna feed to 4

5 the DSP subsystem input). To achieve this, the gain and the phase response of such paths have to be kept under control. Precisely knowing the behaviour of the individual paths will allow to apply, at the gateway, proper amplitude and phase digital compensation to the signal samples so that, overall, all signals result to be multiplied by the very amplitude / phase coefficients which had been calculated for achieving the desired antenna performance in terms of beam shape, gain and sidelobes. With regard to amplitude, a static pre-compensation is likely to be insufficient to achieve the goal, both because deviations are not constant with time (e.g. due to temperature variations and ageing effects), and because one of the involved physical elements, namely the propagation medium, is inherently time-variant. A way to perform a dynamic compensation consists in the on-line variation of the amplitude DBF coefficients such that the effective amplitude weights always coincide with the calculated ones. To realise such dynamic compensation a control loop must be implemented. With regard to phase the situation is really serious for NGSO satellites, as the various feed signals will be subject to a random mutual phase shift just due to the fact that they are on different frequency slots across the feeder-link band and that the earth-satellite distance varies with time. The use of a dynamic phase compensation scheme therefore appears to be a must in such a situation. For what concerns the antenna performance sensitivity to amplitude and phase errors, let us assume that beamforming is only done by changing the feeds excitation phase, with constant amplitude (this assumption can be associated to the hypothesis of generation of co-phase beams steered by means of feed excitations phase variation). With reference to an array of N elements, in presence of independent, zero mean, random amplitude E ff * can be approximated by: and phase errors the expected beam power pattern, { } where: E { ff } = a ( 1 E{exp( jδφ )} ) * * E{exp( jδφ )} f0 f0 + N + σ a (1) * f f 0 0 and a respectively are the ideal beam power pattern and the ideal amplitude excitation of the feeds, σ a is the variance of the amplitude error; δφ is the random phase error per feed. When the number of feeds is significant, amplitude errors have a very small impact on beam gain. Viceversa phase errors produce a gain loss equal to E{exp( jδφ )} If the phase error is gaussian distributed, the gain loss becomes, exp(-σ φ ) 5

6 where σ φ is the variance of the phase error (gain loss of about 1 db corresponds to phase error standard deviation less than 0.5 radians). Also sidelobes peaks and, even more, the sidelobes nulls will change significantly. However, for CDMA system, what counts is the sidelobes average power rather than their peaks and nulls [6], so the issue will have a lower impact in determining the real acceptability of phase errors In addition to beam gain, the phase error will also yield an influence on beam pointing. In a D array the total pointing error standard deviation σ is proportional to: s σ s σ s φ () N where s = u + v, being u and v the beamwidths in the two dimensions. For σ φ less than 0.5 radians and N exceeding a few tens of elements the effect of beam pointing is negligible with respect to the induced gain loss. A DYNAMIC AMPLITUDE- AND PHASE-ERROR COMPENSATION APPROACH There is a clear need for a dynamic phase-error compensation strategy, especially in NGSO satellite systems, while such necessity is not mandatory with regard to amplitude errors, although it seems very likely. The adoption of a compensation scheme dealing with both phase and amplitude errors is proposed, which relies upon accurately calibrated reference signals to determine, in real time, the mutual amplitude and phase relationships among the end-to-end paths (from the S-band feed to the DSP subsystem input). Reference carriers are associated to each feed signal, each one travelling over the same end-to-end path, thus being subject to the same path response. The DSP subsystem at the gateway will derive the amplitude and phase relationships among the received reference carriers. This information will allow to calculate, in real time, the appropriately modified DBF coefficients to achieve the desired performance in presence of transmission paths characterised by randomly varying parameters. With respect to the reference carrier level, it has to be optimised to avoid power amplifier non linear effects, especially small signal suppression. Two variants of said approach were addressed: Reference signals generated on-ground, Reference signals generated on-board. The first approach (on-ground reference) does not require on-board hardware modifications or additions. Each gateway transmit one unmodulated S-band beacon in between the two CDMA spectra, which would be received by all the on-board feeds (beams have not yet been synthesised). A comb of closely-spaced carriers would so result (as many carriers as the number of gateways which can simultaneously get access to the satellite). The beacons transmitted by a gateway would so appear in each 10-MHz slot of the feeder link spectrum for each polarisation (Figure 3). A gateway will receive the down-link replicas (Ka- or Ku-band) of its own up-link S-band beacon and its DSP subsystem will compare such replicas in terms of both amplitude and phase. Assuming that amplitude and phase of the up-link beacon signals, as measured at the on-board S-band feed 6

7 outputs, are all identical, the DSP subsystem then has sufficient information to correct the precalculated DBF coefficients, such as to take account of the transmission paths behaviour. Figure 3 Reference Signal Allocation The second approach (on-board reference) envisages the on-board generation of a single unmodulated carrier placed in the middle of each 10-MHz segment (and not a comb of carriers as shown in Figure 3), and the injection of such carrier just after each on-board S-band receive feed. In practice there will be a single (redundant) carrier generator followed by a power splitter. The splitter output signals will be fed, by coaxial cables, to directional couplers placed directly in between the S-band feeds and the LNAs. In this way all the mutual amplitude and phase relationships among the reference carriers can easily be controlled (in practice, though not strictly mandatory, efforts will be paid to have the same amplitude and the same phase for all reference signals, as measured after the coupler; conversely it is very important that all the mutual amplitude- and phase-relationships remain constant with time). Figure 4 On-board Reference Carrier Injection For the wanted signal the coupler causes a loss which is to be minimised not to unacceptably impact the receive system noise figure. Such loss is mainly due to two contributions: the ohmic loss and the coupling loss which is directly dependent on the coupling factor. Depending on the LNAs stability, another scheme could foreseen the injection of the reference signal after the LNAs themselves thus avoiding noise figure deterioration. Whichever is the selected solution, injected reference carriers for beamforming calibration can be on-ground extracted and used to compensate the phase rotation of the signal. Carrier extraction can be done by a DPLL (Digital PLL) or a feedforward carrier estimator. 7

8 CONCLUSIONS Growth of mobile communications is expected to demand increased system capacity for mobile communications satellites. DBF is considered as a valid mean to comply with this requirement. Nevertheless, the exploitation of the state-of-the-art adaptive techniques do not seem viable onboard due to the increase of payload complexity, especially in broadband multiple-access communication missions. The alternative to entrust to the gateway the exploitation of the more sophisticated adaptive techniques was proposed. The study performed demonstrated both the feasibility of such concept and its attractiveness in allowing on-board unreachable processing loads. The traditional separation of satellite communications payload between antennas and repeater was just disappearing with the upcoming of active antennas where performances can not be simply separated into the contributions of its constituent components. This separation is even more ephemeral in a system whose performance are determined by the chain: satellite-antenna, repeater propagation characteristics and on ground digital beamforming. A system engineering approach is hence deemed essential for the successful design and development of a similar system. REFERENCES [1] J. Litva, T. K.J. Lo, Digital Beamforming in Wireless Communications, Artech House, 1996 [] J. H. Winters, "Smart Antennas for Wireless Systems", IEEE Personal Communications, Vol. 1, pp. 3-7, Feb 1998 [3] G.Björnström, "Digital Payloads: Enhanced Performance Through Signal Processing", ESA Journal, Vol.17, pp. 1-9, 1993 [4] B. D. Van Veen, K. M. Buckley, "Beamforming: A Versatile Approach to Spatial Filtering", IEEE ASSP Magazine, pp. 4-4, Apr 1988 [5] D.L. Brandel, W.A. Watson, A. Weinberg, NASA's advanced tracking and data relay satellite system for the years 000 and beyond, Proceedings of the IEEE, Vol. 78, No. 7, pp , Jul 1990 [6] P. Angeletti, M. Lisi, "An Integrated Link Budget Method for the Capacity Assessment of Third Generation Mobile Satellite Systems", Proceedings of the EMPS 000 conference. 8