 Research
 Open Access
Scale effect analysis of earlytermination fixedcomplexity sphere detector
 Rongrong Qian^{1},
 Yuan Qi^{2}Email author,
 Tao Peng^{3},
 Jun Yang^{1} and
 Wenbo Wang^{3}
https://doi.org/10.1186/168761802013125
© Qian et al.; licensee Springer. 2013
 Received: 8 January 2013
 Accepted: 12 June 2013
 Published: 3 July 2013
Abstract
In this paper, we show that the earlytermination fixedcomplexity sphere detector (ETFSD) has a scale effect when being employed in multiantenna systems that have to detect signals of multiple users under the constraint of sum complexity. The socalled scale effect stands for a kind of effect in which the complexity per user can be reduced as the number of users increases without loss of diversity order. The main contributions of this study are as follows: (1) we establish a mathematical model based on the large deviation principle to apply the concept of scale effect to ETFSD; (2) we prove the existence of the scale effect of ETFSD in a multiuser scenario; and (3) we demonstrate that within multiuser systems, the scale effect can be exploited to realize ETFSD that has the optimal performance in the point view of diversity order under polynomial complexity constraint per user.
Keywords
 Error Probability
 Scale Effect
 Detection Outage
 Diversity Order
 Channel Matrix
1 Introduction
1.1 Research on MIMO detection
In the context of high signal/noise ratio (SNR), the optimality of the ML detector can be interpreted as the full diversity which means that the detector is capable of achieving the diversity order n _{R}[1]. Despite the optimal performance, however, the ML detector is not a promising scheme of MIMO detection in practice because of its extremely high complexity. It is evident from (1) that the ML detector has to perform an exhaustive search of ${m}^{{n}_{\mathrm{T}}}$ possible candidates and, thus, has the exponential complexity^{a} ${C}_{\text{ML}}=O\left({m}^{{n}_{\mathrm{T}}}\right)$.
Ever since MIMO techniques became widely accepted as among the few emerging key technologies for wireless communications, realizing the performance of the optimal or nearoptimal MIMO detection without exhaustive search has been an interesting topic for decades [2–9]. It is known from [2–7] that the problem of Equation 1 can be formulated as a lattice search problem [2–4], an integerleastsquares problem [5], a semidefinite relaxation problem [6], and a shortest path problem [7] which are shown to have their own efficient solution methods (algorithms). Many lowcomplexity strategies can also be used to restrict computational complexity or make the search more efficient, including restricting the search or invoking early termination. Even in recent years, MIMO detection has remained an attractive topic [10–15].
A lot of prior works have been done. However, to the authors’ knowledge, it is still hard to realize the MIMO detector with optimal performance (regarding both error probability and diversity order) with the polynomial complexity constraint^{b}. It is shown in [5] that, for integerleastsquares problem with small problem size, the expected complexity of the sphere decoder (SD) at high SNR can be approximated by a polynomial function. The closetlattice search and shortest path algorithm can efficiently cut down the complexity of solving the integerleastsquares problem [4, 7]. However, there does not exist a polynomial upper bound on SD complexity [16], closetlattice search, or shortest path algorithm for any fixed SNR. With some fixed performance degradation (14) [16], the SD has an exponential complexity (in both the worst and the average cases) of ${C}_{\text{SD}}=O\left({m}^{\gamma {n}_{\mathrm{T}}}\right)$ with γ∈(0,1] [16]. The fixedcomplexity sphere decoder (FSD) is proved to maintain full diversity with a complexity of ${C}_{\text{FSD}}=O\left({m}^{\sqrt{{n}_{\mathrm{T}}}}\right)$ that is ‘subexponential’ [17]. On the other hand, the linear receivers, e.g., zeroforcing (ZF) and minimum mean square error schemes, have low complexity but quite poor diversity order n _{R}−n _{T}+1 [1]. The situation is discouraging as several profound studies show that many MIMO detectors have their own inherent limitations [16, 18, 19]. Under this situation, the authors believe that exploiting more potential positive factors in MIMO detectors might be one of the appropriate ideas to make the next breakthrough. In this paper, we focus on exploiting the potential positive factor of MIMO detection in a multiuser scenario.
1.2 MIMO detection in a multiuser scenario
It can be found that most studies on the complexity of MIMO detection are under a singleuser assumption [2–13], while this study focuses on the complexity of MIMO detection in a multiuser scenario. To understand the motivation behind, one can think of the factor that MIMO receivers in base stations would have to detect signals from multipleaccess users. In previous studies [20, 21], the scale effectoriented MIMO detector (SEOD) was developed for multiuser system, which was also shown to have a kind of effect called the scale effect in which the complexity per user can be reduced as the number of users increases without loss of diversity order. Moreover, from [21], we know that scale effect would play an important role in a cloud base station because the cloud base station is a largescale multiuser system where the baseband processes of a great number of base stations (each could have a lot of accessed users) are performed together within a virtual base station pool under the limit of overall runtime.
However, a big question left to us is whether the scale effect only exists in SEOD. If the answer is ‘yes’, it will mean that the scale effect has limited practical meanings since the SEOD is a very special MIMO detection scheme which has been carefully designed. Whereas if other schemes of MIMO detection also have the scale effect, we would expect that the scale effect could become an important feature of MIMO detection in multiuser scenarios.
1.3 Contributions
In this study, we focus on investigating the earlytermination fixedcomplexity sphere detector (ETFSD) that is an extended version of FSD by incorporating earlytermination mechanism. The existence of scale effect in ETFSD is proven (Theorems 2 and 3). Due to the scale effect, the complexity per user of ETFSD can be kept under a polynomial constraint while maintaining fulldiversity performance (Theorem 3). We note that the original FSD has no scale effect; thus, in a multiuser case, it still achieves full diversity with subexponential complexity. This result implies that the scale effect not only exists in SEOD but also in ETFSD and, thus, encourages us to extend and deepen the studies on scale effect in the future. Moreover, an analytical framework based on the large deviation principle is established to investigatethe scale effect in this paper, while [20] does not provide such a framework.
1.4 Organization
The rest of this paper is organized as follows: In Section 2, the ETFSD scheme is specified, and its underlying mathematical model is presented. In Section 3, we prove that the ETFSD in a singleuser scenario is able to achieve full diversity; meanwhile, the complexity behavior of ETFSD is studied considering both singleuser and multiuser scenarios. With respect to the multiuser scenario, the scale effect of ETFSD is rigorously defined, which in Section 4 is proven to exist. Consequently, the numerical results are provided in Section 5. We conclude our paper in Section VI.
2 Earlytermination fixedcomplexity sphere decoder
2.1 ETFSD scheme
Among the existing MIMO detectors, the FSD [9] is an attractive scheme; it is highly suitable for hardware implementation due to its fixed complexity and parallelcomputingfriendly features. By simply incorporating an earlytermination mechanism into the FSD, we can establish the ETFSD scheme. The main process of the ETFSD is a constrained tree search through a tree with n _{T} levels where m branches originate from each node. The whole tree search can be realized in two stages: full expansion (FE) and single expansion (SE) search stages, which can be summarized as follows:

The first stage is the FE search in which a full search is performed in p levels, expanding all m branches per node. In this stage, the transmit symbol with larger noise amplification appears at an earlier level of the tree. Note that a larger noise amplification can also be understood as a smaller postprocessing SNR.

The second stage is the SE search in which a single search needs to be performed in the remaining n _{T}−p levels. Analogous in principle to previous studies (such as in [17, 22]), the SE search expands one branch per node followed by the ZF estimate.
Besides the constrained tree search, the complete realization of the ETFSD also includes the preprocessing of the channel matrix (as FSD [9, 17, 22] does), which determines the detection order of the transmit symbols and has two stages as well: the preprocessing for the FE search and that for the SE search. To be specific, the preprocessing of the channel matrix can be expressed by Algorithm 1 using the following notations. (·)^{ † } denotes the MoorePenrose pseudoinverse. FlagET is a flag that determines whether the early termination is enabled in the algorithm or not, i.e., FlagET = True means enabled, otherwise disabled. The function denoted by find(b≠x) finds the indices of the elements of the vector b that do not equal x. The preprocessing is realized iteratively, and within the j th iteration, the effective channel matrix H ^{ j } is reconstructed by removing some column from H ^{ j−1}, where the initialization condition is H ^{0}=H. Accordingly, P ^{ j } denotes the MoorePenrose pseudoinverse of H ^{ j }. Algorithm 1 returns a as the ordered indices of the transmit symbols. The elements of a with indices 1 to p are the indices of the transmit symbols to be detected in the FE search. Then the remaining n _{T}−p elements of a are the indices of the transmit symbols to be detected in the SE search. Furthermore, $\mathbf{b}=\phantom{\rule{0.3em}{0ex}}[\phantom{\rule{0.3em}{0ex}}{b}_{1},{b}_{2},\cdots \phantom{\rule{0.3em}{0ex}},{b}_{({n}_{\mathrm{T}}j)}]$ is just an intermediate vector with its elements satisfying $\{{b}_{1},{b}_{2},\cdots \phantom{\rule{0.3em}{0ex}},{b}_{({n}_{\mathrm{T}}j)}\}=\{1,2,\cdots \phantom{\rule{0.3em}{0ex}},{n}_{\mathrm{T}}\}\setminus \{{a}_{1},{a}_{2},\cdots \phantom{\rule{0.3em}{0ex}},{a}_{j}\}$. Also, $\rho \triangleq {n}_{\mathrm{T}}{E}_{s}/{\sigma}^{2}=1/{\sigma}^{2}$ is the average SNR at the receiver (a.k.a. the average receive SNR), and γ∈(0,1) is a real value.
Algorithm 1 (a,p) = CHANNELPREPROCESSING(H,p,n _{ T }, FlagET,γ)
Compared with the channel preprocessing algorithm of the FSD [9, 17, 22], Algorithm 1 of the ETFSD has a newly introduced functionality, as seen in lines 9 to 12, which can terminate early the loop of the preprocessing for the FE search depending on the channel matrix and update p before breaking the loop when FlagET := True is set. Moreover, we have $\overline{p}$ as one input of Algorithm 1 to guarantee full diversity. Early termination implies that the number of levels of the FE search stage could be less than $\overline{p}$. If we set FlagET := False, Algorithm 1 would execute identically as [17, 22]. Thus, FSD could be considered as a special case of ETFSD. In line 7, the index of the largest diagonal entry of P ^{ j }(P ^{ j })^{ H }, i _{max}, is found. In fact, the diagonal entries of P ^{ j }(P ^{ j })^{ H } reflect the postprocessing noise amplification (see (5) as well). In line 8, the original index of the transmit symbol, a _{(j+1)}, is obtained based on i _{max}. It is clear that in the j th iteration of channel preprocessing, the a _{(j+1)}th transmit symbol has the largest noise amplification. In this work, for both the convenience of mathematical analysis and the reduction of computational complexity, ZF is employed in the SE stage. Thus, the preprocessing for the SE stage is performed as in line 17.
Algorithm 2 (${\stackrel{~}{s}}_{\text{FSD}}$,d _{ max }) = TREESEARCHFSD(y ^{ j },$\stackrel{\u0304}{p}$,j,d _{ max },$\stackrel{~}{s}$)
2.2 Mathematical model of the ETFSD
Applying the ZF estimate to (2),${\stackrel{~}{s}}_{\text{SE}}$would be obtained as${\stackrel{~}{s}}_{\text{SE}}={\left({\mathbf{H}}_{\text{SE}}\right)}^{\u2020}{\mathbf{y}}^{p}$.
Under such signal model, it should be clear that after the ZF estimate in the SE stage,${\stackrel{~}{s}}_{\text{SE}}={\mathbf{s}}_{\text{SE}}+{\left({\mathbf{H}}_{\text{SE}}\right)}^{\u2020}\mathbf{v}$, where${\mathbf{s}}_{\text{SE}}=\phantom{\rule{0.3em}{0ex}}{[\phantom{\rule{0.3em}{0ex}}{s}_{{a}_{(p+1)}},{s}_{{a}_{(p+2)}},\cdots \phantom{\rule{0.3em}{0ex}},{s}_{{a}_{{n}_{\mathrm{T}}}}]}^{\mathrm{T}}$.
Meanwhile, lines 8 to 13 of Algorithm 2 perform the SE search in which m ^{ p } candidate paths would be equalized by the ZF estimate, while one path with minimum Euclidean distance from y that corresponds to${\stackrel{~}{s}}_{\text{FSD}}$ is returned as the detected vector. In line 11, a specifies the indices of the elements of${\stackrel{~}{s}}_{\text{FSD}}$.
2.3 A first look at the complexity of ETFSD
because it is proven in [17, 22] that setting the number of the FE stage to$\overline{p}$ is a sufficient condition for the FSD to achieve full diversity.
3 ETFSD in a singleuser scenario
3.1 Error probability analysis
where p _{ e ET−SE} is the probability of the error event${\stackrel{~}{s}}_{\text{SE}}\ne {\mathbf{s}}_{\text{SE}}$ with FlagET = True.
where λ _{ k }(·) is the k th smallest eigenvalue of the matrix.
where the second and third inequalities are derived by following Equations 17 and 14 of [22], respectively, and β is some real constant according to [26].
Thus, the complementary event${\overline{A}}_{j}$ is equivalent to η _{(j+1)}≥ρ ^{ γ }.
Lemma 1
Proof
Taking both the relation of (9) and the definition of A _{ j } as (10) into consideration, as γ−1<0, (11) can be obtained directly. □
In addition, let E _{ SE } and E _{ ET−SE } denote the event that${\stackrel{~}{s}}_{\text{SE}}\ne {\mathbf{s}}_{\text{SE}}$ where s _{ SE } consists of the symbols of the transmit vector, which are detected in the SE search stage with the early termination disabled and enabled, respectively. Hence, p _{ e SE}=P(E _{ SE }) and p _{ e ET−SE}=P(E _{ ET−SE }).
Lemma 2
In considering the error probability of the ETFSD, we can first get${p}_{e\mathrm{ET}\mathit{\text{SE}}}\le p({n}_{\mathrm{T}}p){\overline{N}}_{e}\phantom{\rule{1pt}{0ex}}exp\phantom{\rule{1pt}{0ex}}\left(\frac{{\rho}^{\gamma}{d}_{\mathrm{min}}^{2}}{4}\right)+{p}_{e\text{SE}}$. Thus, recalling from (4), we are further able to obtain${p}_{e\mathrm{ET}\mathit{\text{FSD}}}\le p({n}_{\mathrm{T}}p){\overline{N}}_{e}\phantom{\rule{1pt}{0ex}}exp\phantom{\rule{1pt}{0ex}}\left(\frac{{\rho}^{\gamma}{d}_{\mathrm{min}}^{2}}{4}\right)+{p}_{e\text{SE}}+{p}_{e\text{ML}}$ , where ${\overline{N}}_{e}$and d _{ min } are constants that can be found in the proof of this lemma.
Proof
See Appendix 1. □
3.2 Diversity order analysis
To simplify the expression, evaluation of the performance of MIMO detectors makes use of diversity order, which is a metric constructed from the error probability and average received SNR [27] (an introduction to the diversity order metric and its associated properties can be found in Appendix 2).
Based on the above discussion, for the diversity order of the ETFSD in a singleuser scenario, we can obtain the following theorem.
Theorem 1
As the condition of triggering the early termination is appropriately constructed, i.e., η _{(j+1)} ≥ ρ ^{ γ } with 0<γ<1, the ETFSD specified by Algorithms 1 and 2 retains full diversity, i.e., d _{ ET−FSD }=n _{ R }.
Proof
See Appendix 3. □
3.3 Complexity behavior
Recalling from Algorithm 1, one should be aware that the number of FE search of the ETFSD, p, varies with the channel matrix H. Thus, in the reminder of the paper, we will use p(H) to denote p so as to emphasize that p(H) is a function of H. Due to the early termination, the complexity of the ETFSD, ζ _{ p(H)}, is a random variable when the channel matrix is random. For any p(H), it fulfills that${\zeta}_{p\left(\mathbf{H}\right)}\in \{{\zeta}_{0},{\zeta}_{1},\cdots \phantom{\rule{0.3em}{0ex}},{\zeta}_{\overline{p}}\}$, where ζ _{ i }=O(m ^{ i }),$0\le i\le \overline{p}$. Thus, ζ _{ 0 }=O(1). Note that m ^{ i } implies the number of all possible candidates of s. This complexity measure is the same as in a previous study [17]. We now investigate the properties of ζ _{ p(H)} in terms of the probability and expectation of ζ _{ p(H)}.
Lemma 3
where$\stackrel{\u0307}{\le}$ is defined similarly as$\doteq $ in Appendix 2.
Proof
See Appendix 4. □
It is apparent that, given Lemma 3, in the highSNR regime, the early termination employed by the ETFSD can efficiently reduce the expected complexity (compared with the FSD) but seems to have no improvement on the worstcase complexity, which is more significant for practical hardware implementation.
4 Scale effect of ETFSD
As a long run concept in microeconomics, ‘economies of scale’ refers to reductions in unit cost as the size of a facility and the usage levels of other inputs increase [28]. In previous studies [20, 21], we proposed extracting the scale effect so as to realize the reductions in complexity per user of SEOD as the number of users increases. In this section, we investigate the scale effect of ETFSD with a large deviation principlebased analytical framework. For notational brevity, we use SCCETFSD to represent the sum complexity constrained ETFSD in a multiuser scenario.
4.1 Complexity behavior in a multiuser scenario
In the point of view of worstcase complexity, this study will show that in a multiuser scenario^{ c }>, the scale effect is one positive factor that should be beneficial to the realization of the ETFSD (the corresponding analysis can be found in the next subsection). Since the scale effect exists in a multiuser scenario, we now focus on discussing the complexity behavior of the ETFSD in a multiuser scenario where the MIMO receiver has to detect the signals of n users with the ETFSD performed for each user.
To simplify the analysis, we assume that different users in a multiuser scenario suffer i.i.d. channel fading and have the same average received SNR. Therefore, if we let the random variable, ζ _{ p(H),j }, denote the complexity of the ETFSD of the j th user, we will naturally get that ζ _{ p(H),j } and ζ _{ p(H),k } are i.i.d. random variables as long as j≠k. This assumption should be reasonable in practice. For instance, different users within a base station might stay in different locations, and the wireless channels they encounter could have no spatial correlation. Moreover, current base station systems always employ the openloop power control; hence, the average received SNR of each user at the base station side can be adjusted to an identical value.
Definition 1
In a multiuser scenario, we define${S}_{n}\triangleq \sum _{j=1}^{n}{\zeta}_{p\left(\mathbf{H}\right),j}$ as the sum complexity of the ETFSD of n users, while$\frac{1}{n}{S}_{n}$ is just the complexity per user.
Obviously, S _{ n } and $\frac{1}{n}{S}_{n}$ are not constants but random variables.
Assumption 1
where C _{ n } is called the sum complexity constraint and its subscript n denotes the number of users. Besides, we assume that$n{\zeta}_{0}<{C}_{n}<n{\zeta}_{\overline{p}}$, which means that the sum complexity constraint is less than the worstcase sum complexity$n{\zeta}_{\overline{p}}$ while larger than n ζ _{ 0 } that is the minimum possible value of the sum complexity of ETFSD with n users.
Because the runtime limit is always demanding, we are interested in how the ETFSD behaves in a multiuser scenario with the sum complexity constraint C _{ n } that satisfies$n{\zeta}_{0}<{C}_{n}<n{\zeta}_{\overline{p}}$. Since$n{\zeta}_{0}<{C}_{n}<n{\zeta}_{\overline{p}}$, S _{ n } shall exceed C _{ n } with certain probability. Once S _{ n }≥C _{ n } occurs, to meet the runtime limit, the ETFSD of some users will have to unconditionally quit before completely executing the tree search of Algorithm 2, which can be seen as a detection outage that would cause performance degradation to the corresponding users. Such detection outage is undesirable; however, it is inevitable. The performance impact brought by the outage event due to the occurrence of S _{ n }≥C _{ n } to the ETFSD of each user will be quantified by means of diversity order loss. The probability P(S _{ n }≥C _{ n }) attracts us because it can reveal the complexity behavior of the ETFSD in a multiuser scenario. To make a fair comparison between the results derived in a multiuser scenario and those traditionally in a singleuser scenario, it is better for us to focus the analysis on the behavior of complexity per user. Therefore, we will restrict our attention to studying$P(\frac{1}{n}{S}_{n}\ge \frac{1}{n}{C}_{n})$.
In the following, we will firstly investigate$P(\frac{1}{n}{\overline{S}}_{n}\ge \frac{1}{n}{C}_{n})$ and then extend to derive an upper bound of $P(\frac{1}{n}{S}_{n}\ge \frac{1}{n}{C}_{n})$ by using (19). Several theoretical results could be derived as follows.
Lemma 4 (Cramer’s theorem for empirical average [29])
Lemma 5
where${\u25b5}_{\zeta}\triangleq {\zeta}_{\overline{p}}{\zeta}_{0}$, and P(A _{ 1 })=1−P(X _{ p(H),j }=ζ _{ 0 })=1−P(ζ _{ p(H),j }=ζ _{ 0 }).
Proof
See Appendix 5. □
To make the result of Lemma 5 more comprehensible, we investigate$I\left(\mathcal{E}\left({X}_{p\left(\mathbf{H}\right),j}\right)+\delta \right)$ in both highSNR regime. When SNR ρ grows to +∞, P(A _{ 1 })→0 and 1−P(A _{ 1 })→1,$I\left(\mathcal{E}\left({X}_{p\left(\mathbf{H}\right),j}\right)+\delta \right)$can thus be approximated by$I\left(\mathcal{E}\left({X}_{p\left(\mathbf{H}\right),j}\right)+\delta \right)\approx \frac{\delta}{{\u25b5}_{\zeta}}log\frac{\delta}{{\u25b5}_{\zeta}\delta}\frac{\delta}{{\u25b5}_{\zeta}}logP\left({A}_{1}\right)log\left(\frac{{\u25b5}_{\zeta}}{{\u25b5}_{\zeta}\delta}\right)$ . Hence, it is reasonable to conclude that at high SNR, $I\left(\mathcal{E}\left({X}_{p\left(\mathbf{H}\right),j}\right)+\delta \right)$ is a positive number dominated by $\delta logP\left({A}_{1}\right)/{\u25b5}_{\zeta}$.
Furthermore, by utilizing Lemma 3, it can be obtained that${\mathrm{lim}}_{\rho \to +\infty}\frac{I\phantom{\rule{1pt}{0ex}}\left(\mathcal{E}\left({X}_{p\left(\mathbf{H}\right),j}\right)+\delta \right)}{log\rho}\ge \frac{\delta (1\gamma )({n}_{\mathrm{R}}{n}_{\mathrm{T}}+1)}{{\u25b5}_{\zeta}}$. Assuming that g(n) is a function of n, we use the expression g(n)≍e ^{−n γ } with the specific symbol ≍ to represent the mathematic meaning that, for any ε>0, there exists a number N with which e ^{−n(γ+ε)}<g(n)<e ^{−n(γ−ε)},∀n≥N.
Lemma 6
From (19) and (21), it follows that$P\left(\frac{1}{n}{\overline{S}}_{n}\ge \mathcal{E}\left({X}_{p\left(\mathbf{H}\right),j}\right)+\delta \right)\asymp {e}^{\mathit{\text{nI}}\left(\mathcal{E}\left({X}_{p\left(\mathbf{H}\right),j}\right)+\delta \right)}$ , where the exact expression of $I\left(\mathcal{E}\left({X}_{p\left(\mathbf{H}\right),j}\right)+\delta \right)$ is given in Lemma 5.
4.2 The existence of the scale effect
Assume that the sum complexity of the SCCETFSD is restricted by the constraint condition in (17). Since the event$\frac{1}{n}{S}_{n}<\frac{1}{n}{C}_{n}$ is identical to S _{ n }<C _{ n },$\frac{1}{n}{C}_{n}$ can be regarded as the complexity constraint imposed on every user.
Definition 2
If there exits C _{ n } that satisfies $\frac{1}{n}{C}_{n}<{\zeta}_{\overline{p}}$ , the scale effect is said to exist if the SCCETFSD for each user can maintain full diversity.
Then let us continue to explore the diversity order of the SCCETFSD in the presence of the constraint as in (25).
The meaning behind (25) is that the complexity per user,$\frac{1}{n}{S}_{n}$, would have the worstcase value$\frac{1}{n}{C}_{n}^{\ast}=\mathcal{E}\left({X}_{p\left(\mathbf{H}\right),j}\right)+\delta $ that tends to ζ _{ 0 }+δ at an adequately high SNR.
which could lead to the detection outage. Moreover, let${\overline{\mathcal{A}}}_{n}$ be the complement of${\mathcal{A}}_{n}$such that${\overline{\mathcal{A}}}_{n}=\left\{{S}_{n}\right\}\setminus {\mathcal{A}}_{n}$.
where the probability$P\left({E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\ast}{S}_{n}\in {\mathcal{A}}_{n}\right)$ is directly affected by the detection outage.
Lemma 7
Let${d}_{{E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\ast}}$ be the diversity order corresponding to$P\left({E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\ast}\right)$such that$P\left({E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\ast}\right)\doteq {\rho}^{{d}_{{E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\ast}}}$.
Proof
See Appendix 6. □
Moreover, according to the proof procedure of Lemma 7, we find that, if$\frac{\mathrm{n\delta}(1\gamma )({n}_{\mathrm{R}}{n}_{\mathrm{T}}+1)}{{\u25b5}_{\zeta}}\le {n}_{\mathrm{R}}$, the detection outage due to the occurrence of event S _{ n }≥C _{ n } might cause diversity order loss to the SCCETFSD for every user; otherwise no loss of diversity order would happen. Therefore, one of the most important results of this study can be obtained as follows.
Theorem 2
Proof
Theorem 2 stems from Lemma 7 in a straightforward manner. □
where$\tau \in ({\zeta}_{0},\phantom{\rule{0.3em}{0ex}}{\zeta}_{\overline{p}})$ is fixed, regardless of ρ or other system parameters.
Then let${\mathcal{\mathcal{B}}}_{n}$ denote ${\mathcal{\mathcal{B}}}_{n}=\left\{{S}_{n}\left\right.\frac{1}{n}{S}_{n}\ge \frac{1}{n}{C}_{n}^{\u266f}\right\}=\left\{{S}_{n}\left\right.\frac{1}{n}{S}_{n}\ge \tau \right\}$ , and correspondingly, let ${\overline{\mathcal{\mathcal{B}}}}_{n}$ be the complement of ${\mathcal{\mathcal{B}}}_{n}$ , i.e., ${\overline{\mathcal{\mathcal{B}}}}_{n}=\left\{{S}_{n}\right\}\setminus {\mathcal{\mathcal{B}}}_{n}$.
where the detection outage impacts$P\left({E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\u266f}\right)$ by affecting the probability$P\left({E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\u266f}{S}_{n}\in {\mathcal{\mathcal{B}}}_{n}\right)$.
where generally g _{ k } ρ)≥1 since the ML detector has the optimal performance. Furthermore, g _{ k }(ρ) has the relation as (34).
As long as the number of users, n, satisfies (31) in which case$P\left({S}_{n}\in {\mathcal{A}}_{n}^{\tau}\right)\stackrel{\u0307}{<}{\rho}^{{n}_{\mathrm{R}}}$, the above will lead to${\mathrm{lim}}_{\rho \to +\infty}\frac{P\left({S}_{n}\in {\mathcal{\mathcal{B}}}_{n}\right)}{{p}_{e\text{ML}}}=0$ since$P\left({S}_{n}\in {\mathcal{\mathcal{B}}}_{n}\right)$ tends to zero at a faster rate than p _{ e ML}.
This means that by following the analytical framework of [13],$P\left({E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\u266f}\right)$has a vanishing gap to p _{ e ML}. To be more specific, the performance degradation caused by the detection outage due to the occurrence of the event${S}_{n}\in {\mathcal{\mathcal{B}}}_{n}$ is insignificant in comparison with the error probability of the ML decoder.
Theorem 3 (Extension of (35))
Given the sum complexity constraint in the form${C}_{n}^{\u266f}=\mathrm{n\tau}$ (see (32) as well), let${d}_{{E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\u266f}}$ denote the diversity order related to$P\left({E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\u266f}\right)$, i.e.,$P\left({E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\u266f}\right)\doteq {\rho}^{{d}_{{E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\u266f}}}$. It would hold that${d}_{{E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\u266f}}={n}_{\mathrm{R}}$ if n is sufficiently large to meet (31).
Both Theorems 2 and 3 show the existence of the scale effect. Moreover, from Theorem 3, one shall find that if the complexity per user of the SCCETFSD is limited by some fixed$\tau \in ({\zeta}_{0},\phantom{\rule{0.3em}{0ex}}{\zeta}_{\overline{p}})$, since ζ _{ 0 }=O (1) is polynomial complexity, τ can be chosen as close as possible to ζ _{ 0 } such that the complexity constraint per user of the SCCETFSD could have a polynomial ^{ d } value while each user of the SCCETFSD still maintains full diversity.
4.3 The benefit of scale effect
It is not easy to (theoretically) ensure fulldiversity MIMO detection with polynomial complexity, while in a multiuser scenario, we propose to take advantage of the scale effect as a positive factor to improve MIMO detection. How can the scale effect improve MIMO detection? Theorems 2 and 3 give us the answer. Especially, in view of Theorem 2, the diversity order of the SCCETFSD is affected by n as (30). As long as n is large enough to meet (31), the SCCETFSD can maintain full diversity under the constraint of (25),${C}_{n}^{\ast}$, that tends to n ζ _{ 0 }+n δ when the average received SNR approaches +∞. Furthermore, in view of Theorem 3, when the sum complexity constraint of the SCCETFSD satisfies${C}_{n}^{\u266f}=\mathrm{n\tau}$ as (32), the detection outage incurs performance degradation that vanishes considerably fast with increasing ρ, in which case the error probabilities of the SCCETFSD for each user and the ML detector have a vanishing gap as (35). Herein, τ is fixed and can arbitrarily approach ζ _{ 0 } where ζ _{ 0 }=O(1) is polynomial. It is also obvious that, under the constraints of both (25) and (32), the scale effect can always be exploited.
5 Numerical results
This section is devoted to presenting numerical results that help make the analytic outcomes achieved above more intuitive. Throughout the rest of this section, without loss of generality, we assume that the complexity of a single ETFSD is ζ _{ p } if the initial number of levels of the FE stage is p. Meanwhile, the specification of MIMO system can be referred to Section 1.1 such that the path between each transmitter and receiver antenna suffers Rayleigh fading. Within the Monte Carlo simulation, 10^{ 6 } realizations of channel are used.
5.1 Diversity order of ETFSD
5.2 Asymptotic behavior of P(A _{ 1 } )
5.3 Asymptotic behavior of$\mathcal{E}\left({X}_{p\left(\mathbf{H}\right)}\right)$
5.4 Statistics of$P\left(\frac{1}{n}{\overline{S}}_{n}\ge \mathcal{E}\left({X}_{p\left(\mathbf{H}\right),j}\right)+\delta \right)$
6 Conclusion
This paper provides the insights into the scale effect of ETFSD. Under the scale effect, the worstcase complexity per user of ETFSD could decrease as the number of users increases without loss of the diversity order. Within the study, the large deviation principle is used as an appropriate and powerful tool to prove the existence of the scale effect.
Endnotes
^{ a } Within this study, unless otherwise stated, the complexity of MIMO detection is measured by the number of all possible candidates of s.
^{ b } The polynomial complexity constraint stands for the complexity constraint of MIMO detection being no more than a polynomial function in either n _{ T }or m.
^{ c } For the multiuser scenario of MIMO system, we assume that all users have the same number of antennas and different users operate on orthogonal resources via time division multiple access or frequency division multiple sccess technologies.
^{ d } Even if we take the length of s candidate or the complexity of ZF estimate into consideration, the complexity related to ζ _{ 0 } is still polynomial.
Appendices
Appendix 1
Proof of Lemma 2
Proof
where P(·,·) denotes the joint probability of two joint events.
In the above, with p+1≤i≤n _{ T } , ${\stackrel{~}{s}}_{{a}_{i}}$is the element of${\stackrel{~}{s}}_{\text{SE}}$, while the corresponding${s}_{{a}_{i}}$ is the element of s _{ SE }.
By combining (9) and (44), we can conclude that setting the judging criterion for early termination as (8) is reasonable since ρ ^{ γ } with 0<γ<1 can make the relations in (9) and (44) able to simultaneously establish tight upper bounds (which decay rapidly as ρ increases) for the associated probabilities, P(η _{(j+1)}<ρ ^{ γ }) and p _{ e ET−SE}.
Therefore, Lemma 2 is verified. □
Appendix 2
Background knowledge of diversity order
Definition 3
Suppose there is a function f(ρ). The symbol$\doteq $ is always employed to simplify the expressions of exponential equality as$f\left(\rho \right)\doteq {\rho}^{d}\iff {\mathrm{lim}}_{\rho \to +\infty}\frac{logf\left(\rho \right)}{log\rho}=d$. If f(ρ) is the error probability of some MIMO detector and ρ denotes average received SNR, d will be the diversity order [27].
Meanwhile, the notation$\stackrel{\u0307}{\le}$ and$\stackrel{\u0307}{<}$ follow after replacing = with ≤ and < in above the definition, respectively.
Lemma 8
where min(·,·) returns the minimum of two inputs.
Proof
□
Lemma 9
Proof
We can prove (46) by${d}_{1}={\mathrm{lim}}_{\rho \to +\infty}\frac{log{f}_{1}\left(\rho \right)}{log\rho}\ge {\mathrm{lim}}_{\rho \to +\infty}\frac{log{f}_{2}\left(\rho \right)}{log\rho}={d}_{2}$. □
Appendix 3
Proof of Theorem 1
Proof
where d _{ ET−FSD }≤d _{ ML } is true because the ML detector has the optimal performance. Thus, the diversity order of the ETFSD will not go beyond that of the ML detector.
Since p _{ e SE} corresponds to the probability of the error event${\stackrel{~}{s}}_{\text{SE}}\ne {\mathbf{s}}_{\text{SE}}$in the FSD (or equivalently, the ETFSD without early termination), by referring to [9], we get${\mathrm{lim}}_{\rho \to +\infty}\frac{log{p}_{e\text{SE}}}{log\rho}=({n}_{\mathrm{R}}{n}_{\mathrm{T}})(\overline{p}+1)+{(\overline{p}+1)}^{2}$. With the special setting of$\overline{p}$ as (3), it is true that$({n}_{\mathrm{R}}{n}_{\mathrm{T}})(\overline{p}+1)+{(\overline{p}+1)}^{2}\ge {n}_{\mathrm{R}}$.
We can conclude that the ETFSD is able to achieve full diversity. □
Appendix 4
Proof of Lemma 3
Proof
For (14), as the event ζ _{ p(H)}=ζ _{ 0 } is identical to another event η _{ 1 }≥ρ ^{ γ } according to (10), we can get$P({\zeta}_{p\left(\mathbf{H}\right)}={\zeta}_{0})=P\left({\overline{A}}_{1}\right)=1P\left({A}_{1}\right)$. Due to Lemma 1, P(A _{ 1 })→0 or$P\left({\overline{A}}_{1}\right)\to 1$as ρ→+∞.
Then taking the logarithm of (9) and then dividing by logρ, we could derive (15).
Then by applying Lemma 1, (16) can be verified.
This completes the proof of Lemma 3. □
Appendix 5
Proof of Lemma 5
Proof
Incorporating φ(t) of (51) into (53) and organizing the items, we have $\delta [\phantom{\rule{0.3em}{0ex}}1P({A}_{1}\left)\right]+\mathrm{\delta P}\left({A}_{1}\right){e}^{t({\zeta}_{\overline{p}}\phantom{\rule{0.3em}{0ex}}\phantom{\rule{0.3em}{0ex}}{\zeta}_{0})}=P\left({A}_{1}\right)[\phantom{\rule{0.3em}{0ex}}1P({A}_{1}\left)\right]({\zeta}_{\overline{p}}{\zeta}_{0})(1{e}^{t({\zeta}_{\overline{p}}\phantom{\rule{0.3em}{0ex}}\phantom{\rule{0.3em}{0ex}}{\zeta}_{0})})$.
It should be clear from a look at the items on the righthand side of the second equality in (55) that $[\phantom{\rule{0.3em}{0ex}}1P({A}_{1}\left)\right]{\zeta}_{0}t+P\left({A}_{1}\right){\zeta}_{\overline{p}}\phantom{\rule{0.3em}{0ex}}t+\mathrm{\delta t}=\phantom{\rule{0.3em}{0ex}}P\left({A}_{1}\right)({\zeta}_{\overline{p}}{\zeta}_{0})t+{\zeta}_{0}t+\mathrm{\delta t}$.
It can also be observed that $log\phantom{\rule{0.3em}{0ex}}\left([\phantom{\rule{0.3em}{0ex}}1P({A}_{1}\left)\right]\right.+P\left({A}_{1}\right)\left.{e}^{t{\u25b5}_{\zeta}}\right)=log\left(1+P\left({A}_{1}\right)({e}^{t{\u25b5}_{\zeta}}1)\right)$.
By replacing t with the critical point of (54), we get $log\left(\right[\phantom{\rule{0.3em}{0ex}}1P\left({A}_{1}\right)]+P({A}_{1}\left){e}^{{t}_{\mathit{\text{cp}}}{\u25b5}_{\zeta}}\right)=log\left(\frac{{\u25b5}_{\zeta}[\phantom{\rule{0.3em}{0ex}}1P({A}_{1}\left)\right]}{{\u25b5}_{\zeta}[\phantom{\rule{0.3em}{0ex}}1P({A}_{1}\left)\right]\delta}\right)$.
Therefore, the exact expression of the maximum value of f(t), i.e., f _{max}, is eventually obtained by (55). By recalling (22), (52), and (55), we obtain $I\left(\mathcal{E}\left({X}_{p\left(\mathbf{H}\right),j}\right)+\delta \right)={f}_{\mathrm{max}}$.
Thus, the validation of Lemma 5 is established. □
Appendix 6
Proof of Lemma 7
Proof
Based on the conclusion of (49), it follows that $P\left({E}_{\mathrm{ET}\mathit{\text{FSD}},k}\right)\doteq {\rho}^{{n}_{\mathrm{R}}}$.
Then in view of (28), (45), and (46), it suffices to show that $P\left({E}_{\mathrm{ET}\mathit{\text{FSD}},k}{S}_{n}\in {\overline{\mathcal{A}}}_{n}\right)P\left({S}_{n}\in {\overline{\mathcal{A}}}_{n}\right)\stackrel{\u0307}{\le}\phantom{\rule{0.3em}{0ex}}\phantom{\rule{0.3em}{0ex}}{\rho}^{{n}_{\mathrm{R}}}$.
Provided that, when ${S}_{n}\in {\overline{\mathcal{A}}}_{n}$, no detection outage occurs, which means that the sum complexity constraint has no effect on the detected result, it is true that $P\left({E}_{\mathrm{ET}\mathit{\text{FSD}},k}{S}_{n}\in {\overline{\mathcal{A}}}_{n}\right)P\left({S}_{n}\in {\overline{\mathcal{A}}}_{n}\right)=P\left({E}_{\mathrm{SCC}\mathit{\text{ET}}\mathit{\text{FSD}},k}^{\ast}{S}_{n}\in {\overline{\mathcal{A}}}_{n}\right)P\left({S}_{n}\in {\overline{\mathcal{A}}}_{n}\right)$.
Therefore, by combining (56) with (58) while having (45), the proof of Lemma 7 is complete. □
Declarations
Acknowledgements
The authors wish to thank the anonymous reviewers for their helpful comments and their thoughtful reviews. They greatly improved the quality of the paper. A short version of this paper was presented at the IEEE PIMRC 2012. This work is sponsored by the BUPT Innovative Funds for Young Scholars (2012RC0507 and 2013RC0210), National Science and Technology Major Projects under grants 2012ZX03003011, 2012ZX03003007 and 2013ZX03003012, National Natural Science Foundation of China under grant nos. 60572120, 61001117, and 60602058, the National Key Basic Research Program of China (973 Program) under grant no. 2009CB320400, and the Joint Funds of NSFCGuangdong under grant U1035001.
Authors’ Affiliations
References
 Paulraj A, Nabar R, Gore D: Introduction to SpaceTime Wireless Communications. Cambridge: Cambridge University Press; 2003.Google Scholar
 Fincke U, Pohst M: Improved methods for calculating vectors of short length in a lattice, including a complexity analysis. Math. Comp 1985, 44: 463471. 10.1090/S00255718198507772788MathSciNetView ArticleGoogle Scholar
 Viterbo E, Boutros J: A universal lattice code decoder for fading channels. IEEE Trans. Inform. Theory 1999, 45(5):16391642. 10.1109/18.771234MathSciNetView ArticleGoogle Scholar
 Agrell E, Eriksson T, Vardy A, Zeger K: Closest point search in lattice. IEEE Trans. Inform. Theory 2002, 48(8):22012214. 10.1109/TIT.2002.800499MathSciNetView ArticleGoogle Scholar
 Hassibi B, Vikalo H: On the spheredecoding algorithm I. Expected complexity. IEEE Trans. Signal Process 2005, 53: 28062818.MathSciNetView ArticleGoogle Scholar
 Wiesel A, Eldar YC, Shamai S: Semidefinite relaxation for detection of 16QAM signaling in MIMO channels. IEEE Signal Process. Lett 2005, 12(9):653656.View ArticleGoogle Scholar
 Lee K, Chun J: ML symbol detection based on the shortest path algorithm for MIMO systems. IEEE Trans. Signal Process 2007, 55(11):54775484.MathSciNetView ArticleGoogle Scholar
 Guo Z, Nilsson P: Algorithm and implementation of the Kbest sphere decoding for MIMO detection. IEEE J. Select. Areas Commun 2006, 24(3):491503.View ArticleGoogle Scholar
 Barbero LG, Thompson JS: Performance analysis of a fixedcomplexity sphere decoder in highdimensional MIMO systems. In Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing 2006. New York: IEEE; 2006:1419.Google Scholar
 Gestner B, Ma X, Anderson DV: Incremental lattice reduction: motivation, theory, and practical implementation. IEEE Trans. Wireless. Communi 2012, 11(1):188198.View ArticleGoogle Scholar
 Wubben D, Seethaler D, Jalden J, Matz G: Lattice reduction. IEEE Trans. Wireless. Communi 2011, 28(3):7091.Google Scholar
 Choi JW, Shim B, Singer AC, Cho NI: Lowcomplexity decoding via reduced dimension maximumlikelihood search. IEEE Trans. Signal Process 2010, 58(3):17801793.MathSciNetView ArticleGoogle Scholar
 Jalden J, Elia P: Sphere decoding complexity exponent for decoding full rate codes over the quasistatic MIMO channel. IEEE Trans. Inform. Theory 2012, 58: 57855803.MathSciNetView ArticleGoogle Scholar
 Abediseid W, Damen M: Lattice sequential decoder for coded MIMO channel: performance and complexity analysis. IEEE Trans. Inform. Theory 2010. http://dx.doi.org/10.1109/ISIT.2010.5513497 Google Scholar
 Goldberger J, Leshem A: MIMO detection for highorder QAM based on a Gaussian tree approximation. IEEE Trans. Inform. Theory 2011, 57(8):49734982.MathSciNetView ArticleGoogle Scholar
 Jalden J, Ottersten B: On the complexity of sphere decoding in digital communications. IEEE Trans. Signal Process 2005, 53(4):14741484.MathSciNetView ArticleGoogle Scholar
 Jalden J, Barbero LG, Ottersten B, Thompson JS: The error probability of the fixedcomplexity sphere decoder. IEEE Trans. Signal Process 2009, 57(7):27112720.MathSciNetView ArticleGoogle Scholar
 Jalden J, Ottersten B: On the limits of sphere decoding. In Proceedings of International Symposium on Information Theory 2005. New York: IEEE; 2005:16911695.View ArticleGoogle Scholar
 Taherzadeh M, Khandani AK: On the limitations of the naive lattice decoding. IEEE Trans. Inform. Theory 2010, 56(10):48204826.MathSciNetView ArticleGoogle Scholar
 Qian R, Qi Y, Peng T, Wang W, Yang J: On the scale effects oriented MIMO detector: diversity order, worstcase unit complexity and scale effects. Elsevier Signal Process 2013, 93(1):277287. 10.1016/j.sigpro.2012.08.006View ArticleGoogle Scholar
 Qian R, Qi Y, Peng T, Wang W: Exploiting scalar effects of MIMO detection in cloud basestation: feasible scheme and universal significance. In Proceedings of the IEEE Globecom 2012 Workshop on Cloud BaseStation and LargeScale Cooperative Communications. New York: IEEE; 2012:216221.Google Scholar
 Jalden J, Barbero L G, Ottersten B, Thompson JS: Full diversity detection in MIMO systems with a fixedcomplexity sphere decoder. In Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processin 2007. New York: IEEE; 2007:III49.Google Scholar
 Gore D, Heath Jr. RW, Paulraj A: On performance of the zero forcing receiver in presence of transmit correlation. In Proceedings of the IEEE International Symposium on Information Theory 2002. New York: IEEE; 2002:159159.Google Scholar
 Narasimhan R: Spatial multiplexing with transmit antenna and constellation selection for correlated MIMO fading channels. IEEE Trans. Signal Process 2003, 51(11):28292838. 10.1109/TSP.2003.818205View ArticleGoogle Scholar
 Lutkepohl H: Handbook of Matrices. Chichester: Wiley; 1996.Google Scholar
 Khoshnevis A, Sabharwal A: On diversity and multiplexing gain of multiple antenna systems with transmitter channel information. In Proceedings of the Allerton Conference on Communication, Control and Computing. Monticello: University of Illinois; 2004.Google Scholar
 Zheng L, Tse D: Diversity and multiplexing: a fundamental tradeoff in multipleantenna channels. IEEE Trans. Inform. Theory 2003, 49(5):10731096. 10.1109/TIT.2003.810646View ArticleGoogle Scholar
 Economies of scale Wikipedia; 2013.http://en.wikipedia.org/wiki/Economies_of_scale . Accessed 1 Feb 2013
 Hollander F: Large Deviations. Providence: American Mathematical Society; 2008.Google Scholar
 Proakis J G: Digital Communications. New York: McGrawHill; 1995.Google Scholar
 Ross SM: A First Course in Probability. Upper Saddle River: Pearson Education; 2009.Google Scholar
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.