You are on page 1of 3

How ever, exact simple expressions can be derived in the special case of interleaved sub carrier assignment schemes

, as can be seen in the next sub section. On the other hand, in order to facilitate the computation of Mk (z,k) in the case of block-sub carriers assignment schemes, we present accurate approximations and tight lower bounds which are based on the Jensen’s inequality.

A. Exact Expression of Mk(z,k) in case of Inter leaved sub carrier Assignment:
For the sake of performance analysis, we assume here that all sub carrier in the vicinity of sub carrier M are allocated to different users. This implies that at most one sub carrier from each set Ak may interface with sub carrier M. Let the index of this sub carrier be denoted by k. Then (14) reduces into a simpler expression shown in(15) which involves averages wrt two symbols only. Let Q(z)=[] , where he average is taken over the information symbols bk,k and bk ,k. Then(15) can be expressed as shown in (16). In the case of square M-QAM with Gray coding, Reb k n and Imb km are selected independently over the set {}

B. Conditional Gaussian Approximation:
Exact evaluation of Mk (z,k) in case of block sub carrier assignment schemes involves a large number of discrete Rvs {bnk } for which exact averaging would require a large number of operations partiwlanly in case of high order modulations. In order to reduce the computational complexity required to compute (14) in case of block-sub carrier assignment schemes, we propose a new improved Gaussian approximation method. Let V,c= Now, when k is given, then Vkc become a sum of independent RVS. However, the set of c RVS{}

Therefore, we approximate he set of RVS{Vc,c=0,1,----c-1} by joint complex Gaussian RVS(conditioned on k) having zero becomes and (conditional) correlation given by

That is, exponent of (14) is approximated by a sum of L, Gaussian quadratic form. Now, using known results for the moment generating function of complex Gaussian quadratic form(e.g [21]), we arrive at the following closed form approximate expression for (14)

is readily obtained by switching that the function exp(-z x) is convex ∀z. C.----Vk.δk) which is valid for arbitrary sub carrier assignment schemes. It is to be emphasized at this point that the proposed Gaussian approximation is distinctive from the classical Gaussian approximation methods which are commonly used in interface analysis in wireless systems.c-1} with elements defined in(19). . x>0.δk) A lower band on Mk (z. Upper bounds on the exact bit and symbol error probabilities are then obtained when (z1) Is used in (13) instead of (14). A Lower Bound On Mk(Z.Where Ic is the cxc identity matrix and Ak is the correlation matrix (conditioned on k) of the jointly complex Gaussian vector {Vk. and therefore Jensen’s inequality can be involved to show the result in (z1).

respectively. Both theoretical and simulation results are displayed in Figs .2 for interleaved sub carrier assignment scheme. fig 2 and 3 reveal that the proposed upper bound on BER is in fact a tight upper bound. for the exact. we consider a 50% loaded system with q=0.2 and 3. On the other hand. Tk=c/c-1 Tmax∀k where the maximum delay spread is set at Tmax =t/10 and the path-loss exponent β=4.when Tw=Tg=2Tmax. and therefore can be employed to provide an efficient and accurate method for performance evaluation of OFDMA with arbitrary sub carrier assignments. The theoretical results are evaluated by using (26). .1}. respectively. which show excellent match to the relevant simulation result.01.ρk) which are given in (16). and for several modulation orders. 2 and 3 for interleaved and blocked sub carrier assignment schemes. improved Gaussian approximation. where γc=γ0e---.2 and 3. The total number of sub carriers is fixed at N=1024. We consider an OFDMA system without extended CPs.and the number of sub carriers in a block in the case of block-sub carrier assignment scheme is 10. but with three different expressions of µk (Z. Firstly the accuracy of the new exact and approximate expressions of BER is investigated in Figs. and for three different values of the normalized maximum propagation delay ∆∈{0.01. that the difference between the approximate and exact (or simulation). Here the average BER of QPAK is plotted against the normalized range of the useful signal ρ0.NUMERICAL EXAMPLES In this section we present some numerical and simulation result for the sake of numerical examples. and plot the cutoff rate against the normalized maximum propagation delay (∆) for both interleaved and block sub carriers assignment schemes.5 in a multipath Rayleigh fading channel with exponential power-delay profiles. Result is diminished even in the case of the interleaved sub carrier assignments. and investigate the impact of he maximum propagation delay. we fix he CP timeinterval at Tg=Tmax. In fig 4. (20) and (21). On the cutoff rate. Here.and c=10. one can see from figs. The exact theoretical results are shown in fig. and the bound. Where | Ak| =1∀k. as for as the accuracy of the proposed approximations and bounds is concerned. More interestingly.