A Low-Overhead Interference Canceller for
High-Mobility STBC-OFDM Systems
Hsiao-Yun Chen, Wei-Kai Chang, and Shyh-Jye Jou
Abstract—This paper proposes a low-overhead space-time block code (STBC) interference canceller for high-mobility STBC-or-thogonal frequency division multiplexing (STBC-OFDM) systems. The proposed STBC interference canceller combined with the two-stage channel estimator can be applied to wireless metropolitan area network (WMAN), like IEEE 802.16e system. At the vehicle speeds of 240 km/hr for 16 quadrature amplitude modulation (16 QAM), the bit error rate (BER)can be improved about 10 times of that just using the two-stage channel estimator. The proposed design is implemented in 90 nm CMOS technology. The gate count is 109.3 K, and the power dissipation is 1.45 mW at 83.3 MHz operation frequency with 1 V power supply. However, up to 61% hardware can be reused from the existed two-stage channel estimator design. After reusing, the proposed STBC interference canceller requires only 42.2 K gates, which is 4.9% overhead of the two-stage channel estimator.
Index Terms—Data detection, interference canceller, STBC-OFDM, WMAN.
I. INTRODUCTION
W
IRELESS metropolitan area network (WMAN), al-lowing end-users to travel throughout a hot zone cell without losing connectivity, has been a very important tech-nique in wireless communication. The high-quality services provide portability and mobility to make users more convenient to access information. The orthogonal division multiple access (OFDMA) technique is adopted in WMAN standards to support multiple-input multiple-output (MIMO) systems and multiple access scheme over multipath fading channels. However, the mobile channel often varies rapidly, which is caused by a large Doppler spread, particularly when the mobile station (MS) moves at the vehicular speed. A fundamental phenomenon that makes credible wireless transmission expensive and difficult is time-varying multipath channels. In order to improve the trans-mission quality in fast and selective fading channels, transmit diversity is an effective technology for reducing fading effect in mobile wireless communication, especially when receive diversity is expensive or impractical to acquire. In recent years,Manuscript received August 24, 2012; revised December 27, 2012; accepted January 17, 2013. Date of publication March 18, 2013; date of current version September 25, 2013. This work was supported by the UMC, CIC, and the Na-tional Science Council of Taiwan, R.O.C. This paper was recommended by As-sociate Editor M. Sanduleanu.
The authors are with the Department of Electronics Engineering, Na-tional Chiao Tung University, Hsinchu 30010, Taiwan, R.O.C. (e-mail: jerryjou@mail.nctu.edu.tw).
Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TCSI.2013.2248831
space-time block code (STBC) has been shown to give high code rate and good performance. It can be applied in orthog-onal frequency division multiplexing (OFDM) systems with multiple antennas to provide better performance by exploiting transmit diversity, and it was also supported by WMAN stan-dards. Nevertheless, STBC-OFDM systems are sensitive to the temporal channel variation within a code word since the sym-bols within one code word interfere with each other. Besides, time-varying multipath channels introduce carrier inter-ference (ICI) among the OFDM subcarriers. These interinter-ference noises degrade the STBC-OFDM system performance. Hence, an STBC interference cancellation scheme is required for better performance when the detailed channel state information (CSI) variation is unavailable.
In this paper, the design and implementation of an STBC interference canceller for high-mobility WMAN are proposed. The implementation uses IEEE 802.16e standard as a test plat-form. IEEE 802.16e is an extension of IEEE 802.16-2004 for providing high data rate transmission and mobility of WMAN [1], [2]. It enables mobile speed up to 120 km/hr but is also backward compatible to support the fixed mode. This design has lower computational complexity and better performance as compared with other previous researches [3]–[6]. By using 90 nm CMOS technology, this design has about 109.3 K gates and dissipates 1.45 mW at 83.3 MHz operating frequency. How-ever, most of the hardware in the proposed design can be reused from the existed two-stage channel estimator design [8]. After reusing, the area can be reduced to 42.2 K gates. It is only 4.9% overhead of the two-stage channel estimator. The proposed data detection design includes the following achievements:
• development of an low-complexity STBC interference cancellation algorithm for high-mobility STBC-OFDM systems;
• implementation of an efficient STBC interference can-celler;
• integration of the proposed STBC interference canceller andthe existed two-stage channel estimator with only 4.9% overhead.
This paper is organized as follows. Section II describes the system architecture. Section III briefly reviews the two-stage channel estimator and introduces the proposed interference can-celling data detection method. Section IV presents the proposed STBC interference canceller. Then, the simulations and the de-sign results are provided in Section V. Finally, Section VI is the conclusions.
Notation: The superscript stands for complex conjugate; denotes transpose; denotes Hermintian transpose. The notation denotes the absolute value. The notation denotes
Fig. 1. Proposed STBC-OFDM system with two transmit antennas and one receive antenna. TABLE I
MAJORPARAMETERS OF THEPROPOSEDSTBC-OFDM SYSTEM
the floor function. The notation performs modulo- op-eration. In general, an italic letter stands for a number; a bold letter stands for a matrix or a set; a bold and italic letter stands for a vector; an upper case letter denotes a frequency-domain signal; a lower case letter denotes a time-domain signal. stands for the -th column of matrix while stands for the b-th row. represents the pilot sub-carrier index set, and denotes the number of elements inside the pilot index set . Based on the same principle, represents the data subcarrier index set.
II. SYSTEMARCHITECTURE
The OFDMA specification of IEEE 802.16e that supports the multi-antenna technology is adopted in this paper. The subcar-rier allocation of partial usage of subchannels (PUSC) in down-link (DL) transmission is supported in this proposed system. The major parameters of the proposed STBC-OFDM system are summarized in Table I. The quadrature phase shift keying (QPSK) and 16 quadrature amplitude modulation (16 QAM) are supported for data subcarriers, while binary phase shift keying (BPSK) is adopted for pilot subcarriers and preamble symbols. Each frame is composed of one preamble symbol and 40 OFDM data symbols. The cyclic prefix (CP) length is 128 sampling pe-riods, i.e., 1/8 of the useful symbol time.
The proposed STBC-OFDM system with two transmit an-tennas and one receive antenna is shown in Fig. 1. In the trans-mitter, Alamouti’s STBC encoding method is used to encode two transmitted symbols, and , within a time slot which is the duration of two OFDM symbols. Note that the guard in-terval insertion, guard inin-terval removal, digital-to-analog con-verter, analog-to-digital concon-verter, carrier recovery, timing
re-covery, and front-end components are omitted in Fig. 1 for sim-plicity. The receiver architecture consists mainly of a two-stage channel estimator and an STBC interference canceller along with other blocks. Without loss of generality, the signal pro-cessing of the received data is focused on each time slot, and the time slot index is omitted hereafter except otherwise stated.
Let be the -th subcarrier of an OFDM symbol, which can carrier the pilot or data symbol. After -point inverse fast Fourier transform (IFFT) and appending cyclic prefix with length , the transmitted time-domain data symbol at the -th sample is denoted as
(1) Assuming the multipath fading channel between the a-th transmit antenna, for , and the receive antenna is made of discrete paths. The received signal can be expressed as [9]
(2)
where represents the complex path gain for the -th path at the -th sample within the -th symbol interval of a time slot, for , and denotes the delay of the -th path. denotes a circularly symmetric zero-mean white Gaussian random noise. By removing the cyclic prefix and taking the fast Fourier transform (FFT), the received signal in frequency do-main is given by
(3)
where represents the frequency-domain channel matrix from the -th transmitted subcarrier to the -th
re-ceived subcarrier for Note that
stands for the ICI caused by Doppler spread, and the diagonal entries , denotes the channel frequency response (CFR).
If the system operates under stationary environment, the temporal variation of wireless multi-path channel is not sig-nificant within any two successive OFDM symbol durations. This quasi-static condition is the basic assumption of Alam-outi’s STBC scheme. However, if the system operates under highly-mobile environment, the quasi-static condition no longer holds because of rapid channel variation, and severe in-terference is introduced in Alamouti’s STBC decoding method. More details of Alamouti’s STBC scheme under these two environments are described as follows.
A. Quasi-Static Channel
If the channel is quasi-static within a time slot, the complex path gain is almost fixed within two OFDM symbol time, and the ICI terms are negligible.
(5)
(6)
where denotes the -th approximated path gain of a time slot. stands for the channel frequency response over one time slot duration.
Then, the received signal with two received OFDM symbols and in a time slot can be arranged as
(7) where (8) (9) (10) (11) According to STBC decoding method, the decoded symbols
and can be expressed as
(12) where is the decision process and
(13)
As a result, the two encoded symbols can be recovered without any coupling interference if the quasi-static condition holds during a time slot.
B. Highly-Mobile Channel
If the system operates under highly-mobile environment, the quasi-static condition in (5) and (6) fails. The received signals in (7) should be re-expressed as (14) where (15) (16) (17) (18) (19) (20) (21) (22) As compared with (7), the time-variant channel introduces three interference noises on the right-hand side of (14). acts as the co-carrier interference (CCI) because it couples be-tween two encoded symbols of the same subcarrier. and are both ICI, but their difference is that couples between and symbol vectors while does not. The presence of these three interference components in time-varying channels results in severe performance degradation.
It has been studied that CCI problem is more important than ICI problems, and simulation results provided in [4] shows that the CCI power is larger than the ICI power by 7 8 dB regard-less of the channel variation rate.
III. INTERFERENCECANCELLINGDATADETECTIONMETHOD
In order to mitigate the impact of CCI and ICI, various approaches have been studied in [3]–[6]. In [4], a successive interference cancellation (SIC) and least squares (LS) method were proposed. In [6] the maximum likelihood (ML) method was used to deal with CCI problem. In [3], a modified ML method was proposed to reduce the complexity of traditional ML method. Most of the previous works do not use STBC decoding technique; instead, they use other sophisticated data
Fig. 2. Proposed data detection flow.
detection methods. However, these data detection methods result in a significant hardware overhead.
One design target of this paper is to integrate the interfer-ence cancellation algorithm with an existed two-stage channel estimator [8]. Under limited clock cycles and hardware budget, we use the Alamouti’s STBC decoding technique and focus on modeling interference noise components. Based on the method in [5], the data detection flow is described in Fig. 2, where stands for the iteration count of the interference cancellation and
stands for the iteration count of the data decision. The param-eters of , , , and represent four control values in the proposed data detection flow.
A. Channel Estimation
Various DFT-based channel estimation method has been studied using either minimum mean square error (MMSE) cri-terion or ML cricri-terion for OFDM system with preambles [10], [11]. Since CSI and signal to noise ratio (SNR) are unavailable at receiver in real implementation, the ML scheme is easier to implement than MMSE scheme. Moreover, the decision-feed-back (DF) scheme can be adopted in DFT-based channel estimation to use decision data as pilot to track channel vari-ations for providing sufficient tracking information. Recently, Ku and Huang [12] presented a two-stage channel estimation method for STBC-OFDM systems under fast time-varying multipath channels. They concluded that a refined two-stage channel estimator is more robust than classical DFT-based method. An initialization stage uses a multipath interference cancellation (MPIC)-based decorrelation method to identify the significant paths of channel impulse response (CIR) in the beginning of each frame. However, the CIR estimated by the preamble cannot be directly applied in the following data bursts since the channel is time-variant. Thus, a tracking stage is then used to track the path gains with known CIR positions. B. STBC Interference Canceller
1) Co-Carrier Interference Modeling: is resulted from channel variation between two symbol intervals within one time
Fig. 3. Relationship between the estimated CFRs.
slot. Various studies had demonstrated that CCI is the major impact on STBC-OFDM systems [3]–[6], so CCI cancellation shall be done before ICI cancellation in our proposed algorithm. To calculate CCI noise, we need to know the CFRs of different transmission antenna pairs, i.e.
. Nevertheless, most channel estimators designed for STBC-OFDM systems are aimed to esti-mate the averaged CFR over two symbol intervals such as , in (22). The goal of this paper is to provide an efficient interference cancellation solution for STBC-OFDM systems. Therefore, we propose an algorithm using only the time-slot based CFR estimation to approximate the CFRs of two symbol intervals within a time slot.
The study in [7] had proved that the averaged CFR over a duration approaches to the CFR taken in the middle of the dura-tion. Based on [7], Fig. 3 demonstrates the relationship between CFR tracking estimations in time domain, where denotes the estimated CFR of -th transmit antenna during the -th time slot, and denotes the tracking-update value calculated by the two-stage channel estimator [8]. represents our approximated CFR of the -th transmit antenna during the -th symbol interval. We notice that no matter which method is used to obtain , but the following constraint must be held
(23) A low cost solution for getting is to interpolate the existed information, and . The estimation of , is non-causal for real implementation, so several predictive approaches have been studied [13].
In the study of [13], three predictive models using only previous estimated information are proposed for time-varying channel. They concluded that the 1st-order predictive model and linear extrapolation have better performance under time-varying channel. Moreover, by applying the 1st-order predictive model on , we can interpolate simultane-ously without violating the constraint in (23). As a result, our proposed approximation is given by
(24) (25) 2) Inter-Carrier Interference Modeling: In time-varying channel, CCI component is more significant than ICI compo-nent in STBC-OFDM systems. As a result, other studies for STBC interference cancellation target only on CCI and ignore
Fig. 4. Diagram of one observation block.
Fig. 5. BER performances versus the decoding iteration number for 16 QAM at the vehicle speed 240 km/hr and 360 km/hr.
ICI problem [3]–[6]. In this section, we propose a low-com-plexity algorithm to model ICI by exploiting the information used for CCI modeling. In this way, CCI and ICI noise can be cancelled simultaneously with little overhead. The ICI com-ponent is much smaller than and is very difficult to be estimated accurately from the available time-slot based estimated CFR ; hence, we ignore and focus on modeling only.
The estimation of defined in (19) requires the temporal path gain variation in each symbol interval, but the channel estimator provides only the temporal averaged CSI of a time slot. We adopt a LS fitted polynomial method [14]–[16] to model the detail of the -th complex path gain as follows.
(26) where represents the modeled complex path gain at time sample index , is the -th path and the -th order LS fitting polynomial coefficient in time domain, , stands for modeling error, and denotes the order of the LS fitting polynomial. It requires observations to solve this -th order LS fitting problem, so we consider the estimated CFRs in time slots and group them as an observation block for each ICI modeling computation. As illustrated in Fig. 4, an observation block is composed of time slots where the latest estimated CFR is in the -th time slot. Then, by combining(26) with the frequency-domain channel matrix (4), we have
(27)
Fig. 6. BER performances versus the ICI subcarrier number for 16 QAM at the vehicle speed 240 km/hr and 360 km/hr.
where and are the estimated channel matrix of and , respectively. Two parts decomposed from are defined as
(28)
(29) where is a constant matrix that can be pre-computed, and . denotes the LS fitting polynomial coeffi-cient matrix in frequency domain and is given by
(30) where
(31) (32)
and . is the observation block as
illustrated in Fig. 4.
Fig. 5 shows the BER performance of the proposed algorithm with the different iteration number of the interference cancella-tion , as defined in Fig. 2, for 16 QAM modulations at the ve-hicle speed of 240 km/hr and 360 km/hr. In the case of , the interference cancellation is not applied. Fig. 6 shows the BER performance of the proposed CCI and ICI modeling with the different number of the neighboring ICI subcarriers . For example, represents ten neighboring subcarriers are considered for ICI modeling. In the case of , ICI noises are ignored, and only CCI cancellation is applied. In order to focus on the effect of CCI and ICI noises, the ratio of received bit energy to the noise power spectral density is set to be 30 dB in these simulations. Fig. 5 and Fig. 6 show that the curves get into saturation at and . Moreover, the
Fig. 7. Architecture of the proposed STBC interference canceller.
2nd-order ICI modeling doesn’t outperform the 1st-order mod-eling. By considering the design cost and the performance, we adopt the 1st-order ICI modeling with four neighboring subcar-riers ( ; ) and execute two interference cancella-tion iteracancella-tion in the proposed design.
IV. PROPOSEDSTBC INTERFERENCECANCELLER
Fig. 7 shows the architecture of the proposed STBC inter-ference canceller with the two-stage channel estimator. Based on the two-stage channel estimation method studied in [12], a low complexity and robust implementation was proposed by [8]. The initialization stage is decomposed to a preamble match, an IFFT, a straight MPIC (SMPIC)-based decorrelator, and an FFT. The tracking stage is decomposed to a STBC decoder, a demapper, an LS estimator, a FFT, a path decorrelator, a Hes-sian matrix calculator, and an IFFT. Moreover, the IFFT and FFT are shared between the initialization stage and the tracking stage.
In the proposed STBC interference canceller, the STBC re-encoder exploits the estimated CFR variation and the decision symbols to generate the variation of noise-free received signal within a time slot. Based on the received signal variation of a subcarrier, the interference shaping filter approximates the CCI noise introduced into this subcarrier and the ICI noise spreading to other subcarriers. Then, the inter-ference noises of this subcarrier are accumulated. Finally, the estimated interference is cancelled from the actual received signal , and this refined received signal is fed into the STBC decoder to get better decision symbols.
The signal processing flow is presented in Fig. 8. The white blocks in Fig. 8 (e.g.the memory used to store the latest es-timated CFRs, the memory holding the received signals, the pilot read-only memory (ROM), the STBC decoder, and the demapper)can be reused from the channel estimator hardware [8]. Only the gray blocks are required for implementation. This process only needs the neighboring subcarriers that induce the significant ICI noise into the central subcarrier. Since the constellation values are known, we store the demapped binary bit values of the decision symbols instead of their complex
Fig. 8. Operation flow of the proposed STBC interference canceller.
values to save the memory size. The details of key blocks will be described in the following sections.
A. Simplification of the Proposed Algorithms
1) CCI Noise: According to the approximation models de-rived in (15) and (24), (25), the CCI noise estimation can be formulated as
(33)
(34) (35) where denotes the modeled CCI noise and .
2) ICI From Data Subcarriers: According to (27), the estimated CFR matrix is composed of constant matrix which models the shape of Doppler spread and the variable part which is transformed from the observed CFRs. We further combine the linear transformation with to obtain a new matrix which has the constant entries and can be precomputed. The final expression of averaged CFR matrix is given by
(36) where
(37) (38) for . As a result, the ICI from data subcarriers can be estimated as
(39) 3) ICI From Pilot Subcarriers: Both the transmitted data subcarriers and the pilot subcarriers contribute ICI noise to the received data subcarriers. As a result, our proposed design takes pilot ICI noise into account. The detail of pilot insertion in our system was described in [8]. The pilots do not pass through the STBC encoding before transmission, so the pilot ICI estimation is given by
(40) where represents the estimated pilot ICI noise, de-notes the pilot symbol at the -th pilot subcarrier, .
4) Joined Interference Modeling: It can be observed that depends on the value of . If we define the inter-ference shaping coefficients as
(41) (42) then (33) and (39) can be combined together as
(43) Based on [9], ICI noise only needs to consider the neighboring subcarriers when the normalized maximum Doppler
Fig. 9. STBC re-encoder circuit design.
frequency is small (less than 0.1). Therefore, we modify (43) to
(44) Similarly, can be reformulated as follows
(45) where the index is in the range ,
and . In (45), the first and second summation parts are the pilot ICI noise from the subcarriers belong to the even and odd pilot index set, respectively.
B. STBC Re-Encoder
Complex multiplications are involved in (44) and (45). The implementation of a complex multiplier can be reduced from four multipliers and two adders to three multipliers and five adders. Since the data and pilot values have a constant value set, the complex multiplications in (44) and (45) can be im-plemented as constant multiplications. Hence, the STBC re-en-coder contains 12 constant multipliers and 20 adders as shown in Fig. 9. Each constant multiplier can applythe canonic sign digit (CSD)expression and can be implemented by several adders. As shown in Fig. 9, the input and are the decision data or pilot symbols, and the outputs are selected according to the subcarrier status of the input signals.
Fig. 10. Shaping filter circuit design for the first transmit antenna branch.
C. Interference Shaping Filter
We apply the CSD code on and to minimize the nonzero digits in the representation. The Doppler spread and the CCI noise can be approximated by multiplying the pre-computed weighting coefficients and , as described in (41), (42), with the variation of noise-free received signal reconstructed by STBC re-encoder. Therefore, the complex multiplications in the interference shaping filter can be imple-mented as the constant CSD multiplications using only several adders. The shift registers are used to accumulate the ICI noises from the neighboring subcarriers and the CCI noise from the same transmitted subcarrier. Fig. 10 illustrates the proposed circuit design.
D. Other STBC Signal Detection Methods
In this section, we briefly describe some other frequently referenced STBC signal detection methods including the ML method [3], the simplified maximum likelihood (SML) method [3], a LS method [4], a SIC method [4], and a diagonalized maximum likelihood decoder (DMLD) method [17].
1) ML Method and SML Method: From (14), the received signal can be expressed as
(46)
where denotes the sum of , , and .
The ML method can be described as follows:
(47) where denotes the constellation points. The ML method computes every possible combination of decision symbols and to select the most probable one.
The SML method stands for the signal detection method pro-posed by [3], and it does not estimate every combination of to solve (47) directly. Instead, it takes all constellation points as and subtract from , and then it
(48) It takes the CFRs of two symbol intervals into account, so it has better performance than Alamouti’s STBC method.
The SIC method introduces the ordered successive interfer-ence cancellation (OSIC) concept to improve the LS method. The complexity of the SIC is low, but it suffers error propaga-tion if the first symbol decision is not correct.
3) DMLD Method: The DMLD method decouples two trans-mitted symbols by multiplying the received signal with a daigonalizing matrix as
(49) It detects symbols as
(50) where denotes the negative determinant of the daigonal-izing matrix.
The daigonalizing matrix reconstructs the orthogonality of the STBC code word, so the two symbols can be decoded separately. The DMLD method applies ML criterion to do the decoding. Therefore, it has good performance but high complexity.
V. SIMULATIONS ANDDESIGNRESULTS
The simulation results demonstrate the performance of the proposed STBC interference canceller based on the system described in Fig. 1. Since this paper is focused on the data detection as shown in Fig. 2, we assume that both timing and carrier frequency synchronization are perfect. The multipath channel adopts the International Telecommunication Union (ITU) Veh-A channel model with relative path power profiles of 0, 1, 9, 10, 15, and 20 (dB), and the path excess delays are uniformly distributed from 0 to 50 sampling periods which are smaller than the length of the CP. Jakes model is also used to generate Raleigh fading environment.
Several data detection schemes are defined in Table II. Both the software scheme (S2) and hardware scheme (H2)are based on the proposed CCI and ICI interference cancellation methods with different control parameters as de-scribed in Fig. 2. The performance of the two-stage channel estimation (T0) is also provided for comparisons. Moreover, the performances of the perfect conditions P0, P1 and P2 are in-cluded for benchmarking. All schemes are simulated in floating point expect H2-fix. H2-fix is H2 scheme simulated in fixed point. Note that the output SNR at the STBC decoder defined in [8] is used as a gauge of the system performance to optimize the fixed-point word lengths in H2-fix. The word lengths of several
TABLE II
SIMULATEDDECODINGSCHEMES
TABLE III
WORDLENGTHS OFSEVERALKEYSIGNALS IN THEPROPOSEDSTBC INTERFERENCECANCELLER
key signals in the proposed STBC interference canceller are summarized in Table III.
Fig. 11 shows the performance of 16 QAM modulation at the vehicle speed of 240 km/hr. The two-stage channel estimator provides very accurate CSI which performs almost equal to the perfect channel estimation, but the bounding floor at high SNR caused by interference noises is very obvious. With the aid of the proposed CCI and ICI interference cancellation methods in S2 and H2, the error floor phenomenon in T0 is significantly im-proved by about ten times. The performance of S2 shows that BER of can be achieved in of 30 dB. The H2 scheme has less data decision iterations than the S2 scheme. Al-though a high iteration operation in the proposed data decision scheme can have better performance, the timing requirement of a less iteration operation can be released to achieve low-cost design. The very small performance gap between H2 and S2 scheme indicates that the H2 scheme is an economic choice for hardware implementation. Moreover, the performance gap be-tween H2 and H2-fix is less than 0.5 dB. We can conclude that the implementation method proposed in Section IV does not de-grade the performance of the proposed algorithm. As compared with P1 and P2, S2 has only 0.4 dB and 3.5 dB performance gap in BER of . Fig. 12 shows the performance of QPSK modulation at the vehicle speed of 240 km/hr. Our proposed S2 reduces the error floor of the two-stage channel estimator by about two times. BER of can be achieved in of 30 dB. Because QPSK modulation is very robust in noise in-terference, the differences among the performances of S2, H2, and H2_fix are very small. As compared with P1 and P2, S2 has only 2.0 dB and 2.8 dB performance gap in BER of .
Several STBC data detection methods [3]–[6]have been briefly described in Section IV. Fig. 13 shows the BER perfor-mance of our proposed method and other methods at of
Fig. 11. BER performance for 16 QAM at the vehicle speed 240 km/hr.
Fig. 12. BER performance for QPSK at the vehicle speed 240 km/hr.
Fig. 13. BER performances versus the vehicle speed for 16 QAM.
16 dB. Note that the symbol-interval based CFRs is obtained by (24), (25) instead of the ideal channel estimation assumed in [3]. As shown in Fig. 13, our proposed scheme outperforms the other methods except the ML method at the vehicle speed of 240 km/hr and has similar performance at higher speed.
Table IV compares the computational complexities of one subcarrier data detection by using the proposed method and
other methods [3]–[6], and [17]under our proposed system, where denotes the number of constellation points. The complexity does not include that of the demapper and the channel estimation. A constant complex multiplier can be implemented as shift operations and additions, so it has con-siderably lower complexity than a variable complex multiplier. Although the ML data detection method has high performance, the complexity of the ML method is also significantly higher than other methods.
The detection methods [3]–[6], and [17]require the CFRs information of two symbol intervals of each time slot, e.g., . In contrast, the proposed method requires only the average CFRs of each time slot. Hence, the channel estimator’s computational complexity can be reduced by two times.
Moreover, other works are designed to reduce CCI, and they require extra decision feed-back technique, i.e., the sequential decision feedback sequence estimation (SDFSE) proposed in [6], to calculate ICI noise. However, our proposed method solves CCI and ICI problem jointly based on the same CFRs information. Therefore, our proposed method can has less computational complexity. As shown in Fig. 13 and Table IV, our proposed method requires low computation complexity and can provide high performance.
The CCI modeling in the proposed method is approximated from the averaged CFRs estimation in (24), (25). The extrapola-tion or interpolaextrapola-tion results may be unreliable under fast fading channel since the temporal variation from symbol to symbol may be violent. Therefore, the performance of the CCI cancel-lation will be limited if the vehicle speed is very high.
The synthesis results are listed in Table V. The pro-posed STBC interference canceller is implemented in 90 nm 1P9MCMOS technology. The overall area of the proposed STBC interference canceller is 0.44 and is equivalent to 109,299 gates. The proposed STBC-OFDM system samples the received signal at 11.2 MHz, and the STBC interference canceller operates at 83.3 MHz. The power is equivalent to 1.45 mW from 1 V supply voltage.
The ratio of combinational and non-combinational area is about 1:4; in other words, the usage of memory in this design occupies about 80% of the design. As shown in Fig. 8, the STBC decoder, the demapper, the received signal memories, and the memories storing the latest CFR estimations can be reused from the two-stage channel estimator. The overhead of our proposed interference canceller are the STBC re-encoder,
the shaping filter, the dual-port memories for decision symbols, and the single-port memories used to store the CFR estima-tions of previous time slot. The STBC re-encoder and shaping filter take 11% of the STBC interference canceller. The over-head memories take 28% of the STBC interference canceller. Up to 61% of the hardware can be reused from the existed two-stage channel estimator design. After reusing with the two-two-stage channel estimator, the area requirement can be reduced to only 42,227 gates. As implemented in [8], the total gate count of the two-stage channel estimator is 859,604 gates. Our proposed STBC interference canceller takes only 4.9% overhead of the two-stage channel estimator, but the BER performance of this data detection scheme can behighly improved.
VI. CONCLUSION
In this paper, an STBC interference cancellation algorithm for ICI and CCI is proposed. We implement an efficient STBC in-terference canceller for an STBC-OFDM system in outdoor mo-bile channels. The proposed joint ICI and CCI canceller is suc-cessfully applied in an STBC-OFDM system with two transmit antennas and one receive antennas.
As compared with the performance of the two-stage channel estimator, the BER can be improved about 10 times by applying the proposed STBC interference cancellation for 16 QAM at the vehicle speed of 240 km/hr with beyond 30 dB. The BER can achieve without using channel coding.
The design of the proposed STBC interference canceller has an equivalent gate count of 109,299 gates, and 61% of which can be reused from the two-stage channel estimator. The design dissipates 1.45 mW at 83.3 MHz operating frequency using 90 nm CMOS technology with 1 V supply voltage. With verifica-tions through design and simulation results, the proposed STBC interference canceller can provide a performance-improving so-lution for the STBC-OFDM systems in WMAN mobile wireless communication.
REFERENCES
[1] Local and Metropolitan Area Networks Part 16: Air Interface for
Fixed Broadband Wireless Access Systems, IEEE Std 802.16-2004,
Oct. 2004.
[2] Local and Metropolitan Area Networks Part 16: Air Interface for Fixed
and Mobile Broadband Wireless Access Systems, IEEE Std
[3] K. I. Lee, J. Kim, and Y. S. Cho, “Computationally efficient signal detection for STBC-OFDM systems in fast-fading channels,” IEICE
Trans. Commun., vol. 90, no. 10, pp. 2964–2968, Oct. 2007.
[4] J. W. Wee, J. W. Seo, K. T. Lee, Y. S. Lee, and W. G. Jeon, “Succes-sive interference cancellation for STBC-OFDM systems in a fast fading channel,” in Proc. IEEE Vehicular Technology Conf., Jun. 2005, vol. 2, pp. 841–844.
[5] J. Kim, B. Jang, R. W. Heath, Jr., and E. J. Powers, “A decision directed receiver for Alamouti coded OFDM systems,” in Proc. IEEE Vehicular
Technology Conf., Oct. 2003, vol. 1, pp. 662–665.
[6] J. Kim, R. W. Heath, Jr., and E. J. Powers, “Receiver designs for Alamouti coded OFDM systems in fast fading channels,” IEEE Trans.
Commun., vol. 4, no. 2, pp. 550–559, Mar. 2005.
[7] H. Hijazi and L. Ros, “Rayleigh time-varying channel complex gains estimation and ICI cancellation in OFDM systems,” Eur. Trans.
Telecommun., vol. 20, pp. 782–796, 2009.
[8] H. Y. Chen, M. L. Ku, S. J. Jou, and C. C. Huang, “A robust channel estimator for high-mobility STBC-OFDM systems,” IEEE Trans.
Cir-cuits Syst. I, vol. 57, no. 4, pp. 925–936, Apr. 2010.
[9] W. G. Jeon, K. H. Chang, and Y. S. Cho, “An equalization technique for orthogonal frequency-division multiplexing systems in time-variant multipath channels,” IEEE Trans. Commun., vol. 47, no. 1, pp. 27–32, Jan. 1999.
[10] L. Deneire, P. Vandenameele, P. van der Perre, B. Gyselinckx, and M. Engles, “A low-complexity ML channel estimator for OFDM,” IEEE
Trans. Commun., vol. 51, no. 2, pp. 135–140, Feb. 2003.
[11] J. H. Park, M. K. Oh, and D. J. Park, “New channel estimation ex-ploiting reliable decision-feedback symbols for OFDM systems,” in
Proc. IEEE Int. Conf. Commun., Jun. 2006, vol. 7, pp. 3046–3051.
[12] M. L. Ku and C. C. Huang, “A refined channel estimation method for STBC/OFDM systems in high-mobility wireless channels,” IEEE
Trans. Wireless Commun., vol. 7, no. 11, pp. 4312–4320, Nov. 2008.
[13] T. A. Lin and C. Y. Lee, “Predictive equalizer design for DVB-T system,” in Proc. IEEE Int. Symp. Circuits Syst., May 2005, vol. 2, pp. 940–943.
[14] H. Hijazi and L. Ros, “Polynomial estimation of time-varying multi-path gains with ICI mitigation in OFDM systems,” in Proc. IEEE Int.
Symp. Commun., Control and Signal Process., Mar. 2008, pp. 905–910.
[15] H. Hijazi and L. Ros, “Polynomial estimation of time-varying multi-path gains with intercarrier interference mitigation in OFDM systems,”
IEEE Trans. Vehicular Technol., vol. 58, no. 1, pp. 140–151, Jan. 2009.
[16] M. L. Ku, W. C. Chen, and C. C. Huang, “EM-based iterative receivers for OFDM and BICM/OFDM systems in doubly selective channels,”
IEEE Trans. Wireless Commun., vol. 10, no. 5, pp. 1405–1415, May
2011.
[17] H. Kanemaru and T. Ohtsuki, “Interference cancellation with diago-nalized maximum likelihood decoder for space-time/space-frequency block coded OFDM,” in Proc. IEEE Vehicular Technology Conf., May 2004, vol. 1, pp. 525–529.
Hsiao-Yun Chen received the B.S. degree in elec-tronics engineering from Feng Chia University, Taichung, Taiwan, in 2002, the M.S. degree in elec-trical engineering from National Central University, Chung-Li, Taiwan, in 2004, and the Ph.D. degree in electronics engineering from National Chiao Tung University, Hsinchu, Taiwan, in 2009, respectively.
Her research interests include baseband signal processing, integrated circuit and system designs for wireless and mobile communications.
Wei-Kai Chang was born in Taipei, Taiwan, R.O.C. He received the B.S. and M.S. degrees in electronics engineering from National Chiao Tung University, Hsinchu, Taiwan, in 2008 and 2010, respectively.
His research interests include algorithm designs for wireless and mobile communications, integrated circuit design, and physical layer baseband signal processing.
Shyh-Jye Jou received the B. S. degree in electrical engineering from National Chen Kung University, Tainan, Taiwan, in 1982, and the M.S. and Ph.D. degrees in electronics from National Chiao Tung University, Hsinchu, Taiwan, in 1984 and 1988, respectively.
He joined Electrical Engineering Department of National Central University, Chung-Li, Taiwan, from 1990 to 2004 and became a Professor in 1997. Since 2004, he has been Professor of Electronics Engineering Department of National Chiao Tung University and became the Chairman from 2006 to 2009. Since August 2011 he has been the Vice President of Office of International Affairs, National Chiao Tung University. He was a visiting research Professor in the Coordi-nated Science Laboratory at University of Illinois, Urbana-Champaign during 1993–1994 and 2010 academic years. In the summer of 2001, he was a visiting research consultant in the Communication Circuits and Systems Research Lab-oratory of Agere Systems, USA. He has published more than 100 IEEE journal and conference papers. His research interests include design and analysis of high speed, low power mixed-signal integrated circuits, communication and bio-electronics integrated circuits and systems.
Prof. Jou was the Guest Editor of the IEEE JOURNAL OF SOLID-STATE
CIRCUITS in November 2008. He received the Outstanding Engineering Professor Award from the Chinese Institute of Engineers in 2011. He served as the Conference Chair of IEEE International Symposium on VLSI Design, Automation and Test (VLSI-DAT) and International Workshop on Memory Technology, Design, and Testing. He also served as Technical Program Chair or Co-Chair in IEEE VLSI-DAT, International IEEE Asian Solid-State Circuit Conference, IEEE BIOMEDICALCIRCUITS ANDSYSTEMS, and other interna-tional conferences.
