On the Uplink Channel Estimation in WCDMA

全文

(1)應用於寬頻分碼多工擷取上鏈路之信道估測 On the Uplink Channel Estimation in WCDMA 彭寶杞 Bao-Chi Peng 達宙通訊系統股份有限公司 Acer Mobile Networks Inc. 新竹市科學工業園區力行路 23 號 23 Li Hsin Rd., Science-Based Industrial Park, Hsinchu 300, Taiwan bcpeng@acermn.com.tw 摘要由於數位信號處理技術的進步以及有限的頻寬資源，第三代移動通訊採用同調偵測來提昇系統效能。信道估測的目的在於量測信道的相位和振幅響應。本文中我們比較了適用於寬頻分碼多工擷取的信道估測技術，並提出我們模擬的結果。關鍵字 : 寬頻分碼多工擷取，信道估測，同調偵測. Abstract Due to the improvements of DSP technology and restricted bandwidth resources, coherent detection is applied in 3G systems to improve system performance. The goal of channel estimation is to estimate the amplitude and phase of the channel response. In this paper we compare different approaches for WCDMA uplink channel estimation. We also show some simulation results of our design. Keywords: WCDMA, Coherent Detection. 1.. Channel. goal of the channel estimation is to estimate the amplitude and phase of the channel response. Fig. 1 shows the constellation diagram of a typical QPSK modulated signal. Fig. 1(a) shows the transmitted constellation. As the signal propagates through the channel, the amplitude and phase are changed which is shown in Fig. 1(b). We can compensate the phase error if we can accurately estimate the phase response. The resulting constellation is shown in Fig. 1(c). Note that the amplitude is not compensated directly. In a direct-sequence spread-spectrum system, the multipath is thought of as the transmission diversity. At the receiver end, the multipath signal is recombined before the decisions are made. We need the information to optimally combine the different paths. The maximum ratio combining (MRC) was proved to be the optimum multipath combining method in rake receiver. Therefore we have to estimate the signal gain in the channel that is used as the weighting of each path.. Estimation,. Introduction. The WCDMA has been selected as part of IMT-2000 radio technology. The demands on high data rates and vehicular velocities from 3km/h to 250km/h[1] require efficient and accurate channel estimation. In traditional CDMA systems, non-coherent detection is applied to demodulate the received signals. Due to the improvements of DSP technology and restricted bandwidth resources, coherent detection is applied in 3G systems to improve system performance. The phase information is the key to coherently demodulate the signal. For coherent CDMA systems with the rake receiver, the amplitude information is also required to combine different output of rake fingers. The. 1.1. Frame structure Fig. 2 shows the frame structure of the uplink dedicated physical channels [2]. Each radio frame of length 10 ms is split into 15 slots, each of length Tslot = 2560 chips, corresponding to one power-control period. The DPDCH and the DPCCH are I/Q multiplexed within each radio frame . The uplink DPDCH is used to carry the DCH transport channel. The uplink DPCCH is used to carry control information generated at Layer 1. The Layer 1 control information consists of known pilot bits to support channel estimation for coherent detection, transmit power-control (TPC) commands, feedback information (FBI), and an optional transport-format combination indicator (TFCI). The exact number of bits in the uplink DPDCH and the different uplink DPCCH fields (Npilot ,.

(2) NTFCI, NFBI, and NTPC) are configured and can also be reconfigured by higher layers. The spreading factor of DPCCH is always equal to 256.. where N is the noise including the signal coupled from DPDCH, and SF is the spreading factor of DPCCH.. 2.. Channel Estimation Techniques. 1.2. Context The dual channel QPSK modulation is applied in the uplink direction. In the transmitter part, the Q-channel and I-Channel convey the data bits of DPCCH and DPDCH respectively. After up-converted to radio-frequency band, the signal is transmitted through the multipath-fading channel. The receiver down converts the signal and passes it to the baseband. The rake architecture is commonly used in CDMA systems. After the DPCCH and DPDCH are separated using different spreading code, they are sent to the channel estimator which extracts the channel gain information. In general both of the physical channels are available to estimate the channel gain. However the pilot field exists only in DPCCH. Decision-feedback or some hybrid techniques are applied to make DPDCH useful. If DPDCH is not applied to simplify the design, the resulting rake finger structure is shown in Fig. 3. There is a ?2 phase difference between the I- and Q-channel. The transmitted symbols of dual channel QPSK are complex values,. D( n) = d d (n) + jd c (n) where d d (n) and d c(n) are the data bits in DPDCH and DPCCH respectively. The received symbol for one path can be exp ressed as, in the baseband equivalent point of view,. β (d d (n) s d ( n) + jd c (n) s c ( n)) + N ′ where N’’ is the noise and the complex channel gain is equal to Achejθch, in which Ach is the magnitude response and θch is the phase response. The spreading code sc(n) and sd(n) are the product of scrambling and channelization codes of each channel respectively. Assume the local scrambling and channelization code are ideally synchronized with the ones in transmitter. We also lump the front-end gain into the channel gain. Therefore the despreaded symbol of DPCCH feeding into the channel estimator can be approximated as  d (n )  RC ( n ) = β (n ) d + jd c ( n)  + N ′( n)  SF  = Ach (n )e j (θ ch ( n) +π / 2) d c ( n) + N ( n). There are many channel estimation techniques ranging from simple average to adaptive filtering. Different methods are suitable for different channel environment and require different computing power. We summarize the characteristics of some of the techniques in the following. Multipath diversity is maintained by combining different path components in the rake receiver. This requires one channel estimator for each path. Therefore the channel estimator must be simple enough and be able to operate in low-SIR environment. 2.1. Simple Average This is a simple but effective method to estimate the channel complex gain. Many other channel estimation techniques exploit it to reduce the sampling rate from symbol rate to slot rate before the algorithms are applied to reduce the complexity. For a simple average method, the pilot fields in DPCCH are correlated with the local pilot bits generated in the receiver. Refer to Eq.(1), the correlator output βa(n) is down-sampled to produce the estimation output.. RC ( n ) = =. Np. 1 Np. ∑R. 1 Np. ∑ (A. C. ( i) p (i) (2). i =1 Np. i =1. ch. ( i)e j (θ ( i ) +π / 2) + N (i ). ). where Np is the number of pilot bits in each pilot field and Rc(n) is the channel estimator input in Eq.(1). The averaging suppresses the noise term N(i). The channel gain can be figured out as. β a ( n) = RC ( n)e − jπ / 2 =. 1 Np. ∑ (A Np. ch. (i) e j θ ( i ) + N sa ( i). i =1. ) (3). The product of the received signal and the conjugate of βa(n) is the compensating factor feeding into MRC. 2.2. WMSA. (1). Weighted-multislot. averaging. method.

(3) (WMSA)[4] is a special case of FIR filters. It smoothes the outputs of the simple averaging method. The structure is shown in Fig. 4. The output values of the WMSA are equal to. β WMSA ( n) =. K. ∑α β i. a. (n + i ). i = − K +1. where K controls the bandwidth of the FIR filter. 2.3. Wiener Filter The Wiener filter approach[3][6] requires the knowledge of the Doppler spread and the signal-to-noise ratio (SIR). For low velocities, the Wiener filter approach is quite robust against model errors. However for high velocities, the performance of Wiener filter degrades. In the interpolation perspective, linear interpolation performs well compared with Wiener filter for high velocity environment. Power Control also worsens the model error. 2.4. Forward Linear Prediction Forward prediction[5][6] reduces the memory requirement to store outputs of rake fingers. However it degrades if non-AWGN exists which do exist in practical environment.. 3.. Simulation Results. In the simulations we employ slot format #0[2] which is suitable for 12.2Kb voice service. There are 6 pilot bits in each slot. The chip rate is 3.84MHz, and the DPCCH symbol rate is 15KHz. We follow the multipath fading propagation condition in [1] including the speeds and number of paths. Fig. 5, Fig. 6, and Fig. 7 show the block error rate of ideal channel estimation, simple average, and our system with ideal delay estimation and 1/3 Viterbi decoder which meet the 3GPP requirements[1].. 4.. Summary. Pilot field-aided coherent detection is adopted in 3GPP WCDMA uplink. To make the receiver cost effective, a simple and robust. channel estimation structure is required. The simple average approach reduces the data rate. Therefore many advanced techniques employ it as a front-end stage. After the estimation process is completed, the linear interpolation is usually adopted to match the data rate for DPCCH and DPDCH. Adaptive filtering is effective for low vehicular velocity environment without fast power control. As the environment changes, different channel estimation technique should be applied. Therefore velocity estimation is required. There are also many other algorithms rely on velocity estimation, eg. multipath tracking and SIR measurement for handover. Channel estimation affects the baseband performance severely. According to our design experience, there are 2dB or more Eb /N0 gap between simple and properly designed channel estimation mechanisms to achieve the same block error rate.. 5.. Reference. [1] 3GPP TS 25.104: “UTRA (BS) FDD; Radio transmission and Reception.” [2] 3GPP TS 25.211: “Physical channel and mapping of transport channels onto physical channels (FDD).” [3] Bengt Lindoff, Christer Ostberg, and Hakan Eriksson, “Channel Estimation For The W-CDMA System, Performance and Robustness Analysis From A Terminal Perspective”, IEEE 49th Vehicular Technology Conference, Vol. 2, 1999. [4] H. Andoh, M. Sawahashi, and F. Adachi, “Channel Estimation Using Time Multiplexed Pilot Symbols for Coherent Rake Combining for DS-CDMA Mobile Radio,” IEEE PIMRC’97, Vol. 3, 1997 [5] S. Cacopardi, F. Gatti, and G. Reali, “Channel estimation using linear prediction for wireless indoor communications,” IEEE ICC '95 Seattle, Vol. 2, 1995. [6] Simon Haykin, Adaptive Filter Theory, 3rd edition, Prentice Hall, 1996.

(4) Q. Q. Q. I. I. (a). I. (b). (c). Fig. 1, Constellation diagram for QPSK. a) transmitted, b) received, and c) compensated signal. Data Ndata bits. DPDCH. k. T slot = 2560 chips, Ndata = 10*2 bits (k=0..6) Pilot Npilot bits. DPCCH. TFCI NTFCI bits. FBI NFBI bits. TPC NTPC bits. Tslot = 2560 chips, 10 bits. Slot #0. Slot #1. Slot #i. Slot #14. 1 radio frame: Tf = 10 ms. Fig. 2, Frame structure for uplink DPDCH/DPCCH. sc. to DPCCH MRC Channel Estimator from down converter. DPCCH-toDPDCH Convertor to DPDCH MRC. sd. Fig. 3, Simplified rake finger with channel estimator. DPCCH symbol. Simple Averaging. .... .... Output. Fig. 4, Block diagram of WMSA.

(5) Fading_1 DCCH BLER. Fading_1 DTCH BLER. 1. 1 2. 4. 6. 8. 10. 0. 2. 4. 6. 8. 10. SA. SA 0.1. Ideal CE7. BLER. BLER. 0. 0.1. Ideal CE7. 0.01. 0.01. Eb/N0. Eb/N0. Fig. 5, Block error rate of FADING_1. Fading_21 DCCH BLER. Fading_2 DTCH BLER. 1. 1 0. 2. 4. 6. 8. 10. 0. 0.1. 6. 8. 10. SA. BLER. BLER. CE7. 4. 0.1. SA Ideal. 2. Ideal CE7. 0.01. 0.01. 0.001. 0.001. Eb/N0. Eb/N0. Fig. 6, Block error rate of FADING_2. Fading_3 DCCH BLER. Fading_3 DTCH BLER. 1 0. 2. 4. 6. 8. 1. 10. 0. 2. 4. 6. 8. 10. 0.1 0.1 Ideal CE7. SA BLER. BLER. SA. 0.01. Ideal CE7. 0.01. 0.001. 0.0001. 0.001 Eb/N0. Eb/N0. Fig. 7, Block error rate of FADING_3.

(6)