Pseudorandom Phase Precoded Spatial Modulation
Abstract
Spatial modulation (SM) is a transmission scheme that uses multiple transmit antennas but only one transmit RF chain. At each time instant, only one among the transmit antennas will be active and the others remain silent. The index of the active transmit antenna will also convey information bits in addition to the information bits conveyed through modulation symbols (e.g., QAM). Pseudorandom phase precoding (PRPP) is a technique that can achieve high diversity orders even in single antenna systems without the need for channel state information at the transmitter (CSIT) and transmit power control (TPC). In this paper, we exploit the advantages of both SM and PRPP simultaneously. We propose a pseudorandom phase precoded SM (PRPPSM) scheme, where both the modulation bits and the antenna index bits are precoded by pseudorandom phases. The proposed PRPPSM system gives significant performance gains over SM system without PRPP and PRPP system without SM. Since maximum likelihood (ML) detection becomes exponentially complex in large dimensions, we propose low complexity local search based detection (LSD) algorithm suited for PRPPSM systems with large precoder sizes. Our simulation results show that with 4 transmit antennas, 1 receive antenna, pseudorandom phase precoder matrix and BPSK modulation, the performance of PRPPSM using ML detection is better than SM without PRPP with ML detection by about 9 dB at BER. This performance advantage gets even better for large precoding sizes.
Keywords – Multiantenna systems, spatial modulation, pseudorandom phase precoding, local search based detection.
I Introduction
The link reliability in singleinputsingleoutput (SISO) fading channels is poor due to lack of diversity. One way to improve the link reliability is to get diversity gains through the use of multiple antennas. Diversity gains can be achieved even in singleantenna systems using rotation coding [1] or transmit power control [2]. Transmit power control requires channel state information at the transmitter (CSIT). Whereas, rotation coding does not require CSIT. The idea in rotation coding is to use multiple channel uses and precode the transmit symbol vector using a phase precoder matrix without requiring more slots than the number of symbols precoded. A phase precoder matrix with optimized phases is shown to achieve a diversity gain of two in SISO fading channels [1]. In [3], the rotation coding idea has been exploited for large precoder sizes. Instead of using optimized phases in the precoder matrix (solving for optimum phases for large precoder sizes is difficult), pseudorandom phases are used. Also, the issue of detection complexity at the receiver for large precoder sizes has been addressed by using the low complexity likelihood ascent search (LAS) algorithm in [4]. It has been shown that with pseudorandom phase precoding (PRPP) and LAS detection, nearexponential diversity is achieved in a SISO fading channel for large precoder sizes (e.g., precoder matrix).
Recently, spatial modulation (SM) is getting increasingly popular for multiantenna communications [5],[6]. SM is a transmission scheme that uses multiple transmit antennas but only one transmit RF chain (thus requiring less RF hardware size, cost and complexity). At each time instant, only one among all the transmit antennas will be active and the others remain silent. The index of the active transmit antenna will also convey information bits in addition to the information bits conveyed through modulation symbols (e.g., QAM). An advantage of SM over conventional modulation is that, for a given spectral efficiency, conventional modulation requires a larger modulation alphabet size than SM. For example, conventional modulation (with 1 transmit antenna and 1 transmit RF chain) requires 8QAM or 8PSK to achieve 3 bpcu spectral efficiency. Whereas, in SM (with 4 transmit antennas and 1 transmit RF chain) the same spectral efficiency can be achieved using BPSK. This is because while the BPSK symbol can convey 1 bit, 2 additional bits can be conveyed through the index of the chosen transmit antenna. This possibility of using a smaller modulation alphabet size in SM, in turn, results in SNR gains (for a given probability of error performance) in favor of SM over conventional modulation [7],[8].
In this paper, we exploit the advantages of both SM and PRPP simultaneously. Our contributions in this paper are as follows.

We propose a method to precode both the modulation bits and the antenna index bits in SM systems using pseudorandom phases. We refer to this system as PRPPSM system. The novelty here is that while conventional PRPP system uses a square precoding matrix of size (where is the number of channel uses), in PRPPSM system, in order to precode the antenna index bits in addition to the modulation bits, we use a rectangular precoding matrix of size (where is the number of transmit antennas).

For small precoder sizes (e.g., ), we demonstrate using ML detection that the PRPPSM system significantly outperforms PRPP system (without SM) and SM system (without PRPP), for the same spectral efficiency. This performance advantage is because of the SNR gain due to the use of smaller alphabet size and diversity gain due to phase precoding.

For large precoding sizes, we propose a low complexity detection algorithm based on local search. The novelty here is a suitable neighborhood definition that takes into account the antenna index bits in the PRPPSM signal set.
Ii PRPP and SM systems
In this section, we briefly introduce PRPP and SM systems.
Iia PRPP system
Figure 1 shows the PRPP transmitter. It takes modulated symbols and forms the symbol vector , where is the modulation alphabet. The symbol vector is then precoded using a precoding matrix to get the transmit vector . The th entry of the precoder matrix is , where the phases are generated using a pseudorandom sequence generator. The seed of this random number generator is preshared among the transmitter and receiver. The precoded sequence is transmitted through the channel, which is assumed to be frequencyflat fading. The channel fade coefficients are assumed to be i.i.d from one channel use to the other. At the receiver, after channel uses, the received symbols are accumulated to form the received vector , given by
(1)  
where , , s are i.i.d. complex Gaussian fade coefficients with zero mean and unit variance, and is the noise vector whose entries are distributed as . The entries of the matrix are uncorrelated and . This creates a virtual MIMO system. It has been shown in [3] that as the precoder size becomes large (e.g., ) the performance of PRPP in SISO fading, using the likelihood ascent search (LAS) detection algorithm in [4] with MMSE initial solution, approaches exponential diversity performance (i.e., close to SISO AWGN performance). This point is illustrated in Fig. 2 which shows the performance of PRPP with BPSK modulation for and 400 in SISO fading channels.
IiB SM system
The SM system uses transmit antennas but only one transmit RF chain as shown in Fig. 3. The number of transmit RF chains, . In a given channel use, the transmitter selects one of its transmit antennas, and transmits a modulation symbol from the alphabet on the selected antenna. The number of bits transmitted per channel use through the modulation symbol is , and the number of bits conveyed per channel use through the index of the transmitting antenna is . Therefore, a total of bits per channel use (bpcu) is conveyed. For example, in a system with , 8QAM, the system throughput is 4 bpcu.
The SM alphabet set for a fixed and is given by
(2) 
For example, for and 4QAM, is given by
(3)  
Let denote the transmit vector. Let denote the channel gain matrix, where denotes the channel gain from the th transmit antenna to the th receive antenna, assumed to be i.i.d complex Gaussian with zero mean and unit variance. The received signal at the th receive antenna is
(4) 
where is the number of receive antennas, is the th symbol in transmitted by the th antenna, and is the noise component. The signals received at all the receive antennas can be written in vector form as
(5) 
For this system model, the ML detection rule is given by
(6) 
Iii Proposed PRPPSM system
The proposed PRPPSM transmitter consists of transmit antennas and transmit RF chains as shown in Fig. 4. It takes modulated symbols and forms the symbol vector , where is the modulation alphabet. Let the matrix of size denote the transmit antenna activation pattern, such that , where is the SM signal set given by (2). The matrix consists of submatrices , , each of size , such that . The submatrix is constructed as
(7) 
where is a vector of zeroes, and is a vector constructed as
(8) 
where is the index of the active antenna during the th channel use. Note that, with the above definitions, . For example, in a system with and , to activate antennas 1, 2 and 1 in three consecutive channel uses, respectively, the matrix is given by
(9) 
Note that the indices of the nonzero rows in matrix gives the support of the spatially modulated vector . For example, in (9), the support given by is .
The vector is then precoded as , using a rectangular precoder matrix of size . The th entry of the matrix is , where the phases are generated using a pseudorandom sequence generator, whose seed is preshared among the transmitter and receiver. The output of the precoder is transmitted on the selected antenna in each channel use^{1}^{1}1Remark: In the proposed PRPPSM system, precoding is applied to both the modulation bits as well as the antenna index bits, i.e., the transmit vector is . Instead, if only the modulation bits are precoded, the transmitted vector will be . When the antenna index bits are not precoded, the system fails to provide the diversity gain advantage of PRPP to the antenna index bits, and hence has a poor BER performance. The proposed PRPPSM system, on the other hand, achieves very good performance because of the precoding of the antenna index bits as well..
Let denote the number of receive antennas. The received signal vector at the receiver is given by
(10) 
where , is the channel matrix of the th channel use, the elements of are i.i.d. complex Gaussian with zero mean and unit variance, is the noise vector where the entries of are distributed as . Note that . This creates a virtual MIMO system. For this system model, the ML detection rule is given by
(11) 
The indices of the nonzero rows in and the entries of are demapped to obtain the information bits.
Note that the ML solution in (11) can be computed only for small precoder sizes because of its exponential complexity in , i.e., ). In Section IV, we will establish the superiority of the PRPPSM over conventional PRPP and SM systems using ML detection. For large precoder sizes, we propose a low complexity detection algorithm in the following subsection.
Iiia Proposed PRPPSM detector
In this subsection, we propose a local search based detection (LSD) algorithm that achieves nearML performance in PRPPSM systems with large at a low computational complexity. The local search detector obtains a local minima in terms of the ML cost in a local neighborhood. In the proposed PRPPSM system, a key requirement for local search is a suitable neighborhood definition that takes into account the antenna index bits also. We propose the following neighborhood definition for the local search. The set of neighbors of a given pair of , denoted by , is defined as the set of all pairs that satisfies one of the following three conditions:

and for exactly a single index .

and differs from in exactly one entry.

for exactly a single index , and for that index , .
For a PRPPSM system with , , and , an example of a neighborhood is
The proposed LSD algorithm starts with an initial solution , which is also the current solution. Using the defined neighborhood, the algorithm considers all the neighbors of and searches for the neighbor with the least ML cost which also has a lower ML cost than the current solution. If such a neighbor is found, then this neighbor is designated as the current solution. This marks the completion of one iteration of the algorithms. The iterations are repeated until a local minima is reached (i.e., there is no neighbor better than the current solution). The solution corresponding to the local minima is declared as the final output . This algorithm is listed in Algorithm 1 below.
Computing the initial support and initial solution : The algorithm needs the initial support matrix and the initial solution vector . The initial support matrix is obtained as follows. Obtain a vector through an MMSE estimator as
The vector consists of subvectors , , each of size , such that . The indices of the elements with the largest magnitude in each are taken as the indices of the nonzero rows of . The initial solution vector is obtained through an MMSE estimator as
where and denotes the Euclidean distance quantizer such that .
Iv Simulation results
In this section, we present the simulation results on the BER performance of the proposed PRPPSM system with ML detection (for small ) and LSD detection (for large ).
Figure 5 compares the BER performance of PRPPSM against the performance of SM without PRPP at a spectral efficiency of 3 bpcu using ML detection. BER plots for , and BPSK modulation for different precoder sizes are shown. It is observed that that the performance of PRPPSM is better than SM without PRPP by about 9 dB at and BER. This performance advantage in favor of the PRPPSM system is due to the diversity gain offered by the pseudorandom phase precoding.
Figure 6 compares the performance of PRPPSM against the performance of PRPP without SM (i.e., , ) at a spectral efficiency of 3 bpcu using ML detection. Here, the PRPPSM system has , BPSK modulation, and the PRPP system without SM has , 8QAM with varying precoder sizes . It is observed that the performance of PRPPSM is better than the PRPP system without SM by about 4 dB at and BER. This performance advantage in favor of PRPPSM system is mainly due to the SNR gain in using BPSK in PRPPSM against using 8QAM in PRPP with out SM, for the same spectral efficiency of 3 bpcu.
Figure 7 compares the performance of PRPPSM using the proposed LSD algorithm against the performance of PRPP without SM using LAS detection from [4]. MMSE initial solution is used in both LSD and LAS algorithms. For PRPPSM, we have used , BPSK modulation. For PRPP without SM, we have used 8QAM. Large precoder sizes are used here; the precoder sizes used in both the systems are . The spectral efficiency in both the systems is 3 bpcu. It is observed that the performance of PRPPSM system is better than the PRPP system without SM by about 10 dB at =70 and BER. This indicates the potential of PRPPSM system to perform very well when large precoder sizes are employed.
V Conclusions
We proposed a novel pseudorandom phase precoder based spatial modulation (PRPPSM) scheme for uncoded transmissions over fading channels. In the proposed scheme, we precoded both the modulation bits as well as the antenna index bits using a rectangular precoder matrix. Our simulation results showed that the proposed PRPPSM system achieved diversity gains and SNR gains compared to conventional PRPP and SM systems. To facilitate the detection when large precoder sizes are used in the proposed PRPPSM systems, we proposed a low complexity local search algorithm which used a neighborhood definition suited for taking into account the antenna index bits as well. We note that the spatial dimension in large antenna arrays can be exploited by transmitting the length precoded vector through antenna elements. This would lead to a tradeoff between the number of transmit RF chains and number of channel uses. Such a generalized PRPPSM scheme is an interesting topic for further study.
References
 [1] D. Tse and P. Viswanath, Fundamentals of Wireless Communication, Cambridge University Press, 2005.
 [2] V. Sharma, K. Premkumar and R. N. Swamy, “Exponential diversity achieving spatiotemporal power allocation scheme for fading channels,” IEEE Trans. Info. Theory, vol. 54, no. 1, pp. 188208, Jan. 2008.
 [3] R. Annavajjala and P. V. Orlik, “Achieving near exponential diversity on uncoded lowdimensional MIMO, multiuser and multicarrier systems without transmitter CSI,” Proc. ITA 2011, Jan. 2011.
 [4] K. V. Vardhan, S. K. Mohammed, A. Chockalingam, and B. S. Rajan, “A lowcomplexity detector for large MIMO systems and multicarrier CDMA systems,” IEEE J. Sel. Areas Commun., vol. 26, no. 3, pp. 473485, Apr. 2008.
 [5] R. Mesleh, H. Hass, S. Sinaovic, C. W. Ahn, “Spatial modulation,” IEEE Trans. Veh. Tech., vol. 57, no. 4, pp. 22282241, Jul. 2008.
 [6] M. Di Renzo, H. Haas, A. Ghrayeb, S. Sugiura, and L. Hanzo, “Spatial modulation for generalized MIMO: challenges, opportunities and implementation,” Proceedings of the IEEE, vol. 102, no. 1, pp. 5355, Jan. 2014.
 [7] N. Serafimovski1, S. Sinanovic, M. Di Renzo, and H. Haas, “Multiple access spatial modulation,” EURASIP J. Wireless Commun. and Networking 2012, 2012:299.
 [8] P. Raviteja, T. Lakshmi Narasimhan, and A. Chockalingam, “Multiuser SMMIMO versus massive MIMO: uplink performance comparison,” available online: arXiv:1311.1291 [cs.IT] 6 Nov 2013.