Development and Implementation of an Advanced Airport Data Link Based on Multi-Carrier Communications

Transcription

1 IEEE 7 th Int. Symp. on Spread-Spectrum Tech. & Appl., Prague, Czech Republic, Sept. 2-5, 2002 Development and Implementation of an Advanced Airport Data Link Based on Communications E. Haas, H. Lang, and M. Schnell German Aerospace Center (DLR), Institute of Communications and Navigation DLR Oberpfaffenhofen, D Wessling {Erik.Haas, Helmuth.Lang, Michael.Schnell}@dlr.de Abstract In this paper, both the concept and the demonstrator implementation of an advanced airport data link is described which is used for data exchange between pilots and controllers in future surface movement guidance and control systems on airports. To fulfil future airport capacity requirements, the proposed data link has to be a high capacity, high rate data link. Therefore, the data link design is based on multi-carrier code-division multiple-access which is a flexible and highly efficient multiple-access transmission technique suitable for mobile communications in fading conditions. Both simulation results with typical aeronautical channel scenarios and bit error rate measurements using the demonstrator hardware show the suitability of multicarrier code-division multiple-access for aeronautical communications within the airport and its vicinity. Moreover, implementation effects which degrade the performance are identified and countermeasures are proposed and investigated. The results show that a reduction of the implementation loss of up to 2 db is achievable by applying appropriate countermeasures. I. INTRODUCTION The air traffic will continuously increase over the next years. Compared to today s air traffic load, it is expected, that air traffic will approximately be doubled in This tremendous increase will lead to bottlenecks in air traffic handling. Especially, the airports are identified to be one of the most capacity restricting factors in the future if no countermeasures are taken. Thus, improving the overall throughput of airports is an important goal for efficient air traffic handling in the future. Currently, there are several research activities which aim at improving aircraft traffic handling, especially on and around airports. One example is the project TARMAC ( Taxi and Ramp Management and Control ) [1] which is the German research contribution to the Advanced Surface Movement Guidance and Control System (A-SMGCS) development. For efficient operation, A-SMGCS requires a communication link between tower and aircraft, which can guarantee the necessary data transfer between controllers and pilots. Existing communication standards for air traffic control like the VHF ( Very High Frequency ) Digital Link Standards are not capable of handling the additional communication load for efficient A-SMGCS operation. Therefore, development and implementation of a new communication link for operation within the airport environment is necessary. Within this paper, both the concept and the demonstrator implementation of an Advanced Airport Data Link (ADL) is described, which is designed to fulfil the requirements of A-SMGCS with respect to communications. The paper is organized as follows. In Section II, the main requirements of A-SMGCS for an ADL are summarized. The ADL concept as well as simulation results are presented in Section III. Implementation effects and appropriate compensation methods are investigated in Section IV and finally conclusions are given in Section V. II. ADVANCED AIRPORT DATA LINK The ADL has to provide the information exchange necessary to establish A-SMGCS on airports. Therefore, a transmission bit rate of at least 128 kbit/s per user should be made available to handle the expected data traffic between controllers, airport authorities, and airline companies on the one side and the pilots on the other side. Moreover, at least 100 simultaneously active users (aircraft and ground vehicles) should be served by the ADL in order to fulfil future airport capacity requirements. Since such a high capacity, high rate data link requires relatively large bandwidth, it can not be realized within the VHF band which is already extensively used. Thus, alternative frequency bands have to be identified. One possibility is the frequency band MHz which currently is allocated to aeronautical radio navigation, especially to the Microwave Landing System (MLS). Since MLS does not seem to be adopted by a considerable number of airport authorities, this frequency band might be used by other

2 aeronautical applications in the future, such as wireless airport data links. The main ADL requirements are: High transmission bit rate of at least 128 kbit/s per user to provide the necessary transmission capability for surveillance, taxi management and additional services. High capacity of at least 100 simultaneously active mobile users to fulfil future airport capacity requirements. Large coverage area which extends km around the airport in order to establish communication also during landing and take-off. Connection to the airport intranet in order to allow for an expansion towards additional services, such as catering orders, airline instructions, or aircraft attendance information exchange. Different priorities for different kinds of services. Flexibility of the physical layer in order to provide simple exchange between user capacity and transmission bit rate. Moreover, the ADL has to be thoroughly adapted to the characteristics of the mobile transmission environment on and around the airport. Since the ADL has to deal with mobile users travelling at relatively high speeds up to about km/h, the wireless transmission channel is time-variant and introduces a significant Doppler frequency shift. Moreover, severe signal fading is very likely to occur since reflections from the ground, from parking or moving aircraft, and from buildings on the airport are expected. Taking into account the channel characteristics as well as the requirements listed above, a broadband transmission scheme for the ADL which makes use of spreadspectrum techniques is very promising. Therefore, for the Ground/Air (G/A) link we propose MC-CDMA [2] which is a combination of Orthogonal Frequency Division Multiplexing (OFDM) [3] with the CDMA technique [4]. Whereas OFDM enables a high rate data link within a mobile environment and is standardized for DAB ( Digital Audio Broadcasting ), DVB-T ( Digital Video Broadcasting - Terrestrial ), and HIPERLAN ( High Performance Local Area Network ), the CDMA technique allows for multipleaccess with high capacity. As a consequence, MC- CDMA provides very high spectral efficiency and is a flexible broadband transmission scheme. Our research and development for the ADL currently focuses on the G/A link, where MC-CDMA is implemented in digital signal processing technology. For the Air/Ground (A/G) link of the operational ADL no final decision has been made for a proposal. Currently, we investigate the suitability of MC-CDMA, differential coherent OFDM and other broadband transmission schemes, such as DS-CDMA ( Direct-Sequence CDMA ) [4], OFDMA ( Orthogonal FDMA ) [5], and IFDMA ( Interleaved FDMA )[6]. III. ADL CONCEPT BASED ON MC-CDMA In the following, the parameter choice for the MC- CDMA G/A link is motivated. For a detailed description of MC-CDMA refer to [7]. To keep the resulting IntersubCarrier Interference (ICI) low, the subcarrier spacing F s should be less than 10 of the maximum Doppler spread. Based on the time-variant aeronautical channel models described in [8] the relevant Doppler spread at 5 GHz is about 250 Hz. Therefore, the subcarrier spacing should be larger than 2.5 khz. In order to have very low ICI the subcarrier spacing is set to F s = 4 khz. For the operational ADL the number of subcarriers N c is chosen to be 2048, resulting in a total bandwidth of B = 2048*4 khz = MHz. The OFDM symbol duration is 1/F s = 250 µs with an additional 10 µs guard interval to avoid intersymbol interference of consecutive OFDM symbols. The modulation type is chosen to be Quaternary Phase Shift Keying (QPSK) [4]. The spreading length is L = 8. The number of maximum users K max and the raw bit rate per user R b depends on the chosen constellation of symbols per user M and user groups Q. Due to hardware restrictions, the bandwidth of the demonstrator system and thus the number of subcarriers is lower. The summarized system parameters can be found in Table 1. TABLE 1 MAIN SYSTEM PARAMETERS AND CHARACTERISTICS Main Parameters: Operational ADL Demonstrator ADL Carrier frequency GHz GHz Bandwidth MHz MHz Number of subcarriers Subcarrier spacing 4 khz 3.85 khz OFDM symbol 250 µs 260 µs duration Guard interval 10 µs 10 µs duration Modulation type QPSK QPSK Spreading length 8 8 Main Characteristics: Number of users Raw bit rate per user kbit/s kbit/s The time-variant aeronautical fading channel necessitates channel estimation and equalization on the receiver side for coherent detection of QPSK. Pilot

3 & & 1 /,.- symbols have to be multiplexed into the data stream for this purpose. For the current channel conditions, pilot symbols with distance N f = 7 in frequency direction and N t = 3 in time direction are sufficient [7]. Additional synchronization symbols facilitate OFDM frame/symbol timing, carrier frequency offset and sampling frequency offset synchronization. The resulting frame structure for the demonstrator system, consisting of N s = 28 OFDM symbols with N c = 128 subcarriers each, is shown in Figure 1. 0 *,+* '() left empty. Additionally, for filtering purposes, some subcarriers on both sides of the spectrum have to be left empty as well. Using L = 8, a total of Q = 11 user groups is possible as shown in the bottom part of Figure 1. The spreading distance of the chips from one group is 14. The maximum number of bits that can be transmitted in one OFDM frame of the demonstrator is therefore limited to 25*11*8*2 = 4400 bits. To verify the suitability of OFDM and MC-CDMA for aeronautical communications, simulations have been performed using the following three channel scenarios proposed in [8] the arrival, the taxi, and the parking scenario. Whereas the parking scenario shows no Line Of Sight (LOS) path, the taxi and the arrival scenario are characterized by a medium and very strong LOS component, respectively. According to the different velocities associated to the different scenarios, the parking scenario shows the smallest, the taxi scenario a medium, and the arrival scenario the largest Doppler frequencies. The results are presented in Figures 2-5. Results marked with SIM are simulations performed on a 64- bit floating point system, whereas results marked with DSP are obtained from simulations with the 32-bit floating point Digital Signal Processing (DSP) system. Timing synchronization, carrier frequency and phase synchronization on the LOS path and sampling rate synchronization are assumed to be ideal. Furthermore, the SIM results with ideal channel equalization assume that the Doppler shifts, carrier phases and excess delays of the LOS path as well as the echo paths are known in the receiver. The curves show the average Bit Error Rate (BER) P b compared to the used Signal to Noise Ratio (SNR) per bit γ b. The results for the arrival scenario in Figure 2 indicate that under ideal conditions, the performance is close to the optimum case with only AWGN due to the strong LOS path. Since the large Doppler frequencies generate ICI, real channel estimation introduces a performance degradation. Fig. 1. OFDM frame structure for MC-CDMA ground-air link demonstrator. The first OFDM symbol is left empty followed by two synchronization symbols [9]. The remaining 25 OFDM symbols consist of data symbols multiplexed with pilots. The subcarrier with 0 Hz offset compared to the carrier frequency is in the center of each OFDM symbol at l = 64. For simplicity, the pilot symbol positions in OFDM symbols without pilot symbols are Fig. 2. Performance results for OFDM transmission using the arrival scenario.

4 For the taxi scenario of Figure 3, a high error floor remains for the case that no equalization is used. With ideal equalization, the error floor is significantly lower. In the real equalization simulation, ICI has been neglected, resulting in a continuously decreasing BER. Considering ICI, it can be expected that the remaining error floor is slightly higher than in the ideal case. A loss of ~2 db between the ideal and the real case remains since additional noise effects are introduced by the channel estimation process. subcarriers. For the not fully loaded system with K Q = 1 or K Q = 2 active users per group, the system using Maximum Ratio Combining (MRC) equalization and single user detection performs significantly better than the system without spreading which is OFDM QPSK. For a fully loaded system with K Q = 8 using Zero Forcing (ZF) equalization and single user detection, the performance is comparable to OFDM QPSK, but can be further improved by applying minimum mean square error equalization or joint detection as described in [7]. Fig. 3. Performance results for OFDM transmission using the taxi scenario. The parking scenario in Figure 4, even though it has the lowest Doppler frequencies, delivers the worst results, since a direct LOS path is missing and the signal has to be demodulated with the echoes alone. Again ICI in the real channel equalization process is neglected. For comparison, the theoretical flat Rayleigh fading performance is shown as well [4]. Fig. 4. Performance results for OFDM transmission using the parking scenario. The performance of MC-CDMA is now considered for the parking scenario to see whether performance improvements are achievable by implementing this technique. Figure 5 shows the corresponding results. Note that in this case the power for one QPSK symbol and thus the power for one bit is distributed on 8 Fig. 5. Performance results for MC-CDMA transmission using the parking scenario. IV. IMPLEMENTATION EFFECTS In the practical implementation of an MC-CDMA system, additional effects due to the hardware have to be considered. First, there may be Direct Current (DC) offsets in or a DC gain mismatch between the I and Q channels of the low pass area. Additionally, there can be a filter bandwidth mismatch due to different tuning between the I and Q channels. The sampling theorem considers ideal Dirac-pulses, whereas a D/A-converter produces rectangular pulses of finite length T sample. Thus, the edges of the OFDM spectrum are attenuated by ~3.9 db on both sides. On top of that, the used I/Q-modulator/demodulator and other Intermediate Frequency (IF) and Radio Frequency (RF) components cause additional distortions in the OFDM subcarrier spectrum. The received subcarrier squared amplitudes and phases in the low pass area without compensation and with I/Q DC-gain mismatch for elliptical 5th order Low Pass Filters (s) on the transmitter side are shown in Figure 6(a) where QPSK modulation has been applied. A frequency-independent variation in the received subcarrier amplitudes is observed depending on the phase of the information-carrying data symbol. For an mismatch, a frequency-dependent variation would be observed. The filter type is now slightly adapted, the DC-gain mismatch is removed and s are applied on the transmitter and receiver side. The D/A-

5 converter and transmitter s are pre-compensated in the transmitter by multiplying each subcarrier with an individually estimated, real-valued compensation factor. The receiver s are post-compensated in the receiver. The result is shown in Figure 6(b) where an equal distribution of the subcarrier amplitudes and thus an equal protection against errors on the channel is observed. A phase bending due to filter phase rotations can be seen as well. If phase bending is not tolerable, complex-valued compensation factors can be used instead of the real-valued ones. It has to be mentioned though that the computational complexity is considerably increased. pass filter with 2.5 MHz bandwidth for the 70 MHz IF band and a mixer that converts the signal up to the GHz area with a local oscillator followed by another 2.5 MHz band pass filter. A subsequent Power Amplifier (PA) with 15 W output power at 1 db compression point is connected to a further band pass filter to suppress possible spectrum images due to non-linear effects. The signal is finally sent to the antenna, or to a connecting cable as in this laboratory set-up. s ϕ (a) Fig. 7. Performance improvement achieved through compensation of implementation effects. s Transmitter Data Signal Source Generation Modulator I Digital/Analog Conversion Q 70 MHz I/Q Modulator IF RF Part with PA RF ϕ Data Sink Receiver Signal Extraction Demodulator Synchronization Analog/Digital Conversion I Q 70 MHz IF I/Q Demodulator Noise Generator RF Part with AGC Fig. 8. OFDM demonstrator set-up for radio frequency area test. (b) Fig. 6. Received low pass subcarriers (a) without and (b) with compensation. The benefits of accurate compensation are shown in the performance graph of Figure 7 for the IF area. For this measurement, DQPSK is used instead of QPSK or MC-CDMA, to circumvent the effects of channel estimation and equalization on the performance. By applying compensation for DC-offset/gain, D/Aconverter, filter amplitude, and I/Qmodulator/demodulator, the performance is improved by about 2 db at a BER of The remaining implementation loss is reduced to less than 1 db. The complete OFDM demonstrator system laboratory set-up, including the RF part and a wired connection without antennas, is shown in Figure 8. The transmitter RF part basically consists of a band On the receiver side the signal from the antenna or from the connecting cable, respectively, is filtered with a band pass filter and then amplified with a low noise amplifier followed by the mixer for downconversion to the 70 MHz IF band. After image rejection with a band pass filter an Automatic Gain Control (AGC) amplifier with a slope time of 100 ms ensures a constant signal level at the input of the 70 MHz I/Qdemodulator. For BER performance analysis, an IF band noise generator is inserted between the RF part output and the I/Q-demodulator input. Thus, an AWGN channel is investigated. In the transmitter as well as in the receiver, additional isolators for protecting the circuit from possible signal reflections, attenuators for signal conditioning, and 10 db power splitters for signal monitoring are integrated. The OFDM symbols of the received MC-CDMA frame from Figure 1 without zero and synchronization symbols are shown in Figure 9.

6 s ϕ balance. Furthermore, it is derived that the synchronization algorithm is stable and has only a minor influence on the performance. More important is the fact that with the practical channel equalization, additional noise is introduced in the system and the performance is degraded by 2-3 db compared to the theoretical AWGN performance. The remaining hardware implementation loss is ~1.7 db at a BER of Fig. 9. MC-CDMA (K Q=1, full power) demonstrator received RF subcarrier squared amplitudes and phases with estimated subcarrier amplitude RF compensation, f g=248 khz elliptic 5 th order s and linear interpolated ZF equalization. There is only one active user per user group, i.e. K Q = 1, transmitting MC-CDMA-spread QPSK symbols with the same power as the pilot symbols. The estimated RF distortions on the subcarriers are pre- and post-compensated in the transmitter and receiver, respectively. The estimated channel at the pilot symbol positions is linearly interpolated at the data symbol positions applying two-times onedimensional filtering. The signal is finally equalized using ZF equalization. The four possible QPSK phase states are clearly visible in the positions of the valid data symbols. The transmitter data symbol power is reduced now, so that the average power with eight active users per user group, i.e. K Q = 8, is equal to the pilot symbol power. The equalized I/Q-constellation diagrams for K Q = 1, K Q = 2, K Q = 5, and K Q = 8 active users per user group are displayed in Figure 10. All subcarriers of several OFDM symbols are multiplexed on the oscilloscope screen, where the horizontal direction is the I-axis and the vertical direction is the Q-axis. The equalized pilot symbols are visible as well on the positive I-axis. In a final step, the IF band noise generator is connected to the system and the BER performance analyzed. The corresponding results are depicted in Figure 11. Whereas the symbol/frame timing synchronization is running free, the carrier frequency synchronization for f c and the sampling rate synchronization for F sample can be coupled with the transmitter to investigate the sensitivity of the synchronization algorithm. From Figure 11 it can be seen that the BER performance for one active user is worse than for eight active users. This is explained by the fact that the synchronization and pilot symbol power stays constant and thus for the case of one user, this user has to deliver the full synchronization and pilot symbol energy, resulting in a degradation of its energy (a) (b) (c) (d) Fig. 10. Received RF MC-CDMA subcarrier constellation diagrams for spreading code length L=8 and K Q=1 (a), K Q=2 (b), K Q=5 (c), and K Q=8 (d) active users per user group.

7 The demonstrator ADL is the world s first implementation of MC-CDMA for mobile applications and has been successfully demonstrated during the 3rd International Workshop on Spread Spectrum & Related Topics (MC-SS 2001) which was held in November 2001 at Oberpfaffenhofen, Germany. Fig. 11. Average BER for OFDM QPSK MC-CDMA with wired transmission in RF band, linear ZF channel equalization and different synchronization conditions; symbol/frame timing synchronization freerunning. V. CONCLUSIONS The development and implementation of an ADL based on MC-CDMA techniques is presented in this paper. The ADL is designed to provide the information exchange necessary to establish A- SMGCS on airports and, therefore, has to fulfil certain requirements. Both the concept and the demonstrator implementation of the ADL are carried out with respect to these requirements. Especially, the transmission environment is taken into account for the MC-CDMA design. Moreover, implementation effects and appropriate countermeasures are investigated. It turns out, that the loss in performance due to implementation effects can be reduced from 3 db to less than 1-2 db at a BER of 10-3 if appropriate countermeasures are applied. REFERENCES [1] DLR project TARMAC, [2] K. Fazel and L. Papke: On the Performance of Convolutionally-Coded CDMA/OFDM for Mobile Communication System. In IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC 93), pages , September [3] S.B. Weinstein and P.M. Ebert: Data Transmission by Frequency-Division Multiplexing Using the Discrete Fourier Transform. IEEE Transactions on Communications, Vol. COM-19, No. 10, pages , October [4] J.G. Proakis: Digital Communications. McGraw-Hill, New York, [5] H. Sari and G. Karam: Orthogonal Frequency-Division Multiple Access and its Application to CATV Networks. European Transactions on Telecommunications (ETT), Vol. 9, No. 6, pages , July/August [6] M. Schnell, I. De Broeck, U. Sorger: A Promising New Wideband Multiple-Access Scheme for Future Mobile Communications Systems. European Transactions on Telecommunications (ETT), Vol. 10, No. 4, pages , November/December [7] S. Kaiser: CDMA Mobile Radio Systems Analysis and Optimization of Detection, Decoding, and Channel Estimation. VDI Fortschrittberichte, Reihe 10, Nr. 531, VDI-Verlag GmbH, Düsseldorf, [8] E. Haas: Aeronautical Channel Modeling. IEEE Transactions on Vehicular Technology, Vol. 51, No. 2, pages , March [9] T.M. Schmidl and D.C. Cox: Robust Frequency and Timing Synchronization for OFDM. IEEE Transactions on Communications, Vol. 45, No. 12, pages , December 1995.