Performance of Space-Combining Transmission Diversity with OFDM

Performance of Space-Combining Transmission Diversity with OFDM for Broadband Wireless Access Systems Nobuaki MOCHIZUKI, Yusuke ASAI, Osamu KAGAMI and Hiroshi HOJO NTT Access Network Service Systems Laboratories 1-1 Hikari-no-oka, Yokosuka-shi, Kanagawa, 239-0847 Japan ph: +81 468 59 3254, Fax: +81 468 55 1752 E-mail: Mochizuki.Nobuaki@ansl.ntt.co.jp ABSTRACT This paper proposes a new diversity technique for broadband wireless access systems that uses orthogonal frequency division multiplexing (OFDM) and describes its performance. This technique consists of two components. One is optimal antenna selection in the reception phase and the other is signal transmission over each OFDM subchannel. Simulations and experiments are used to verify that the proposed technique offers excellent performance. In addition, to offset the quantization noise caused by large differences in the received power levels on the different antennas, a new automatic gain controlling (AGC) method is proposed and its performance is described. 1. INTRODUCTION Broadband wireless access systems have been gaining attention since they support ubiquitous multimedia communication. The demand for high-speed multimedia communication, which supports both voice and motion picture, is increasing, and users want to utilize various communication services. Mobile Internet access services, such as i-mode in Japan, are already gaining wider acceptance, but the data speeds of current wireless systems do not always meet user demands. Standardization bodies such as IEEE, ETSI and MMAC have established broadband wireless access systems [1]-[3] that use the 5-GHz band to offer aggregate user service rates of more than 20 Mbps [4]. In general, high bit rate wireless transmission incurs multipath delays that may exceed ten times the clock period, even in in-door environments [5]. The key to successfully deploying a broadband wireless access system is to employ a transmission technique that achieves low packet error rate (PER) over such frequency-selective fading channels. Numerous papers have revealed that orthogonal frequency division multiplexing (OFDM) works well in severe fading channels while offering reasonable complexity [6]. As a result, IEEE, ETSI and MMAC apply OFDM as the transmission technique in their respective systems, i.e., IEEE802.11a, HiperLAN/2, and HiSWANa. Even with a moderate number of subcarriers, OFDM exhibits satisfactory PER performance; the rootmean square (rms) delay spread is held below 100 ns (approximately). If the rms delay spread exceeds several hundreds of nanoseconds, however, the severe inter-symbol interference (ISI) that results excessively degrades the PER performance of OFDM. In this case, diversity techniques are most effective in overcoming the detrimental fading effects, since each subchannel in OFDM can be seen to incur time-variant dispersive fading. The diversity techniques applied to OFDM transmission at an access point (AP) that have been investigated so far are frequency domain sub-band diversity [7] and clustered OFDM with transmission diversity and coding [8]. The former is a diversity technique that employs space diversity reception separately in each OFDM subchannel. This substantially improves PER performance at the cost of increasing the implementation complexity of the receiver. The latter is a diversity technique that clusters the OFDM subchannels into several small blocks to increase the effectiveness of coding across frequencies and reduce the peak-to-average power ratio. This, however, may require a rather large number of clustered blocks, namely, antennas, because the diversity gain of this technique increases with the number. To solve the above-mentioned problems, the authors proposed the OFDM-incorporated space-combining transmission diversity technique [9]. This is a diversity technique in which each OFDM subchannel is independently transmitted by the antenna that is determined to offer the best performance from among all antennas in the reception phase. This paper verifies the PER performance of the proposed diversity technique by computer simulations and experiments. The following section describes the operating principle of the proposed technique. Sections 3 and 4 assess its performance through computer simulations and experiments, respectively. The last section confirms the effectiveness of adding an individual automatic gain controlling (AGC) method to the proposed technique. 2. OPERATING PRINCIPLE OF TRANSMISSION DIVERSITY The configuration of the proposed space-combining transmission diversity technique is depicted in Fig. 1. Note that, guard interval (GI) insertion and removal circuits have been omitted from this figure for simplicity; two antennas are used and the number of subcarriers is set to N. In the reception frame timing at the AP, the downconverter (D/C) changes the OFDM signals received at each antenna into zero-IF (zero-intermediate frequency) signals. These zero-IF signals are individually fed into the N-point DFT (discrete Fourier transform) circuits through serial-to-parallel (S/P) converters; they are then altered into N-subcarrier signals. The antenna selection (AS) circuit determines which of the two antennas currently offers the best performance (i.e. the antenna associated with the best condition radio channel) in a subcarrier-by-subcarrier manner. Here the power level measured at the last symbol in the received packet can be used, although additional measurements on some form of eye-opening characteristic are recommended if co-channel interference is a problem. The AS information indicating which antenna was selected at each subcarrier is transmitted from the receiver side to the transmitter side. The N-subcarrier signals are combined at each subcarrier in the reception diversity manner. In the transmission frame timing at the AP, the encoded and modulated signals are fed into the modulated signal mapping (MM) circuit. The MM circuit distributes the current signal to the appropriate inverse DFT (IDFT) circuit according to the AS information. Consequently, the output signal of the IDFT circuit selected contains the subcarriers that should experience the best radio channel and is transmitted through the antenna selected (#1 or #2). The frame format is illustrated in Fig. 2. This format is the same as that of HiSWANa and HiperLAN/2. The frame consists of a broadcast control channel (BCCH), a frame control channel (FCCH), a down-link phase (DL), an up-link phase (UL) and a random access channel (RACH). The BCCH is transmitted every 2 ms. The interval between DL and UL is defined as the transmission interval, Ttr. Here DL (UL) represents the transmission path from (to) the AP to (from) the mobile terminal (MT). 3. PERFORMANCE OF PROPOSED TECHNIQUE 3.1 Simulation Parameters Performance of the proposed transmission diversity technique in the presence of frequency-selective fading channels was evaluated by computer simulations. An exponentially-decaying multipath fading channel model was used and the number of delay waves was set to 18. The rms delay spread was set to 150 ns, a value seen in various environments such as moderately large indoor applications [5]. Other assumptions were that the symbol timing detection was perfect and that the frequency offset error was 0 Hz. The other major simulation parameters are listed in Table 1. Most simulation parameters are based on ongoing discussions in several standardization bodies such as ETSI-BRAN and MMAC-HSWA. Finally, communication channel changes which occurs during Ttr were considered using the normalized parameter fDTtr (fD: the maximum Doppler frequency). For the case of fDTtr = 0.1, Ttr becomes 2 ms, if fD = 50 Hz (10 km/h in the 5-GHz band). 3.2 Performance Evaluation PERs for the averaged carrier-to-noise power ratio (CNR) are shown in Fig. 3 as a function of fDTtr =0.1. Two combinations of modulation scheme and FEC (forward error correction) coding rate (R) were considered: QPSK with R=1/2 and 16QAM with R=3/4. In addition, PERs obtained without diversity and with reception diversity only are also plotted for reference. From this figure it is clear that the diversity gains at PER=10-2 are, respectively, 3.5 dB for QPSK with R=1/2 and 3.2 dB for 16QAM with R=3/4. Table 2 shows the diversity gains obtained with other combinations. These show that transmission diversity offers gains of more than 2.0 dB while reception diversity achieves gains of more than 3.5 dB. Consequently, it is obvious that the proposed transmission diversity technique significantly improves PER performance compared to the scheme without diversity. PER values versus normalized parameter fDTtr are plotted in Fig.4. CNR was either 18 dB or 19 dB. The modulation scheme - coding rate combination was 16QAM with R=3/4. The PERs obtained without diversity and with reception diversity only are also plotted for reference. Reception diversity is equivalent to the proposed transmission diversity with fDTtr =0. In this figure, when fDTtr becomes less than 0.3, the proposed technique improves the PER performance compared to the case without diversity. For the case of fDTtr =0.3, Ttr becomes 15 ms, if fD corresponds to walking speed, i.e., 20 Hz (4 km/h in the 5-GHz band). In addition, the PER performance of the proposed technique approaches that of reception diversity, as fDTtr becomes small. 4. EXPERIMENT 4.1 Model The performance of the above-mentioned transmission diversity technique was verified by experiments. The circuit tested, depicted in Fig. 5, basically mirrors the advanced wireless access (AWA) system [10] developed by NTT. This system is based on the HiSWANa and HiperLAN/2 specifications. The delay waves were independently added to the signals transmitted from the two antennas of the AP in the fading simulators. Each fading simulator used the same delay profile. The fading channel model used in the experiment was a 6-wave tapped-delay-line channel model as shown in Fig. 6. In addition, the path loss between the AP and MT was changed by the variable attenuator. The other parameters are listed in Table 3. 4.2 Experiment results The experiment results are shown in Fig. 7. Three combinations of modulation scheme and coding rate were examined: QPSK with R=1/2, 16QAM with R=9/16 and 16QAM with R=3/4. In this figure, the transmission diversity gain at PER=10-2 is 3.5 dB for QPSK with R=1/2. This result equals that found in the simulation. With either 16QAM with R=9/16 or 16QAM with R=3/4, the proposed transmission diversity technique halves the PER error floor compared to the case of no diversity. This error floor is caused by waveform distortion in the analog parts, i.e., HPA (High Power Amplifier), LNA (Low Noise Amplifier) and so on. The waveform distortion, however, has no effect on the gain achieved by transmission diversity. These measurements confirm that the proposed diversity technique offers improved PER performance compared to no diversity. 5. IMPACT OF AGC 5.1 Problem of quantization noise In the basic implementation of the proposed reception diversity technique, AGC is performed using an AGC control value estimated from the largest power signal. Using just one AGC value, however, is counterproductive if the different antennas receive widely different signal powers. When this difference becomes too large, the lowest power signal can experience excessive quantization noise which can degrade PER performance. To suppress this quantization noise, it is preferable to use different AGC values at each antenna. Unfortunately, the noise power may increase at reception side in the signal with small received-power and degrade the PER performance after combining. In addition, the difference in the received power level between the antennas, needed by the proposed transmission diversity technique, is no longer available. 5.2 Proposed: individual AGC with power revision Given the above-described problems, we propose a transmission/reception diversity technique that offers individual AGC with power revision. The power revision is performed using the different AGC values at each antenna before combining. This technique is expected to suppress both the quantization noise (due to the individual AGC circuits) and the CNR degradation (due to power revision). The configuration of the enhanced-AGC diversity technique is shown in Fig. 8. Its performance was evaluated by computer simulations. Most simulation parameters were the same as those in Table 1. The combination of modulation scheme and coding rate was 16QAM with R=3/4. The number of quantizing bits was 12 bit and the power difference between the two antennas was fixed to 10 dB. PERs versus enhanced-AGC CNR are shown in Fig. 9. PERs obtained by common AGC and by individual AGC without power revision are also plotted in this figure for reference. The horizontal axis plots the averaged CNR in the signal that offers the maximum received power. With reception diversity (common AGC), a packet error floor appears at PER=2x10-3. PER performance with individual AGC without power revision is about 3 dB lower than that of the common AGC method. The enhanced-AGC shows no error floor and no CNR degradation compared to the common AGC method. In addition, with transmission diversity, the proposed technique offers the same performance as common AGC, since the difference in received power levels between the two antennas can be obviously obtained on the reception side. 6. CONCLUSION This paper proposed a space-combining transmission diversity technique in which each OFDM subchannel is independently transmitted through the antenna that is found to offer the best condition radio channel. Simulation and experiment results verified the excellent performance of the proposed technique. In addition, this paper indicated the problem that the quantization noise becomes excessive if the difference between the received powers on different antennas becomes too large. To solve this problem, we proposed the technique of individual AGC with power revision; its excellent performance was verified. REFERENCE: [1] “Draft Supplement, High Speed Physical Layer (PHY) in the 5 GHz band,” IEEE 802.11a, 1999. [2] “Broadband Radio Access Networks (BRAN): HIPERLAN type 2 Functional Specification Part 1; Physical Layer (PHY),” ETSI-BRAN, 2001. [3] “Low Power Data Communication Systems: Broadband Mobile Access Communication System (HiSWANa); ARIB STD T-70,” ARIB, 2000. [4] Amendment of the Communication’s Rules to Provide for Operation of Unlicensed NII devices in the 5 GHz Frequency Range, FCC 97-5, January 1997. [5] H. Hasemi, “The Indoor Radio Propagation Channel,” Proceeding of the IEEE, vol. 8, No. 7, pp. 943-968, July 1993. [6] J.T.E. McDonnell, and T. A. Wilkinson, “Comparison of Computational Complexity of Adaptive Equalization and OFDM for Indoor Wireless Networks,” PIMRC’96, pp. 1088-1091, 1996. [7] H. Hamazumi, Y. Ito and H. Miyazawa, “Performance of Frequency Domain Sub-band Diversity Combination Technique for Wide-band Mobile Radio Reception –An Application to OFDM Reception,” IEICE Trans. On Comm. B-II Vol. J80B-II, No. 6, pp. 466-474, 1997. [8] L. J. Cimini, B. Daneshrad, and N. R. Sollender, “Clustered OFDM with Transmitter Diversity and Coding,” IEEE GCOM’97, pp. 703-707, 1997. [9] Y. Matsumoto, N. Mochizuki and M. Umehira, “OFDM Subchannel Space-combining Transmission Diversity for Broadband Wireless Communication Systems,” IEEE ICUPC’98, pp. 137-141, 1998. [10] A. Ohta, F. Nuno, N. Mochizuki, O. Kagami and M. Umehira, “Design of Advanced Wireless Access System,” APCC2000, pp. 157-162, 2000. Access point (AP) S/P #1 S/P #2 IDFT #1 IDFT #2 P/S #1 P/S #2 ANT #1 U/C #1 #2 U/C #2 U/C Mobile terminal (MT) Input Enc.& Mod. Mod. signal Mapping (MM) P/S IDFT S/P Enc.& Mod Input Antenna selection Info. Antenna Selection (AS) Output Dec.& Demod. Combiner P/S #1 P/S #2 DFT #1 DFT #2 S/P #1 S/P #2 D/C #1 D/C #2 D/C S/P DFT P/S Dec.& output Demod. : parallel signal : serial signal U/C : Up conversion D/C : Down conversion Fig. 1 Configuration of proposed transmission diversity BCCH : Broadcast Control Channel FCCH : Frame Control Channel RACH : Random Access Channel DL 2ms Up-link data from #n Down-link data to #n DL : Down-link UL : Up-link UL RACH Transmission interval : T tr BCCH FCCH DL 2ms UL RACH BCCH FCCH Fig. 2 Frame Format at AP Table 1 Simulation Parameters Parameter Modulation/detection Value BPSK, QPSK, 16QAM and 64QAM/ Coherent detection Convolutional coding/Viterbi decoding (R=1/2, 9/16 and 3/4, K=7) 20 MHz 5 GHz 52 (including 4 pilots) 800 ns 4 µs 54 byte 150 ns 0.05 ~0.6 (Doppler frequency 25 Hz to 300 Hz at Ttr=2 ms) Table 2 Diversity gains Modulation and Coding rate BPSK R=1/2 Gain of transmission diversity [dB] 2.0 3.5 2.8 3.2 3.0 Gain of reception diversity [dB] 3.5 4.0 5.0 5.9 5.9 FEC Clock frequency Radio frequency Number of subcarriers Guard interval OFDM symbol length Transmitted packet length RMS delay spread Normalized transmission interval fDTtr QPSK R=1/2 16QAM R=9/16 16QAM R=3/4 64QAM R=3/4 100 Transmission Div. Reception Div. Without Div. 100 CNR=18dB CNR=19dB fDTtr=0.1 Without diversity 16QAM R=3/4 10-1 PER 16QAM R=3/4 QPSK R=1/2 10-2 PER 10-1 10-2 Reception diversity 10-3 4 8 12 16 20 24 CNR [dB] 10-3 0 0.1 0.2 0.3 0.4 fDTtr 0.5 0.6 0.7 Fig. 3 PER vs. averaged CNR Fig. 4 PER vs. normalized fDTtr OFDM Modulation part D/A U/C HPA TDD SW Fading simulator HYB TDD SW Fading simulator OFDM modulation part Input Input Encoder OFDM modulation part D/A U/C HPA Antenna Info. Output OFDM demodulation part Decoder A/D D/C LNA variable attenuator TDD SW HPA U/C D/A Encoder OFDM demodulation part A/D D/C LNA LNA D/C A/D OFDM demodulation part Decoder Output Access Point (AP) Mobile Terminal (MT) Fig. 5 Circuit tested Relative Power (dB) 0 100 Without Div. Transmission Div. 10 20 0 30 110 230 10 -1 PER 490 Relative Delay (ns) 1050 16QAM R=3/4 Fig. 6 Fading model Table 3 Experiment Parameters Parameter Fading model RMS delay spread Doppler frequency Packet length Value 6-wave/Exponentially decaded Rayleigh fading 150 ns 20 Hz 3 PDU/frame 3.5dB 10 -2 16QAM R=9/16 QPSK R=1/2 10 -3 -85 -80 -75 -70 -65 -60 -55 -50 -45 Reception power [dBm] Fig. 7 Experiment results IDFT GI insertion Interpolation D/A U/C HPA Slave Input Encoder Interleaver Modulator (BPSK,QPSK ,16QAM) Antenna Info. Data Mapping IDFT GI insertion Interpolation D/A U/C HPA Master AGC value of master side Receiving data Antenna Selector Power revision Coherent detection DFT GI removal A/D AGC D/C LNA Master Power revision Coherent detection DFT GI removal A/D AGC D/C LNA Slave AGC value of slave side Fig. 8 Configuration of proposed enhanced AGC diversity circuit 1 Common AGC Individual AGC Proposed AGC 0.1 PER Transmission Div. 5dB 0.01 Reception Div. 3dB 0.001 18 20 22 24 26 CNR [dB] 28 30 Fig. 9 PER vs. averaged CNR