ECE 461 Fall 2006 Probability of Bit Error for M -ary Modulation • Assuming that M = 2ν for some positive integer ν, we can map the symbols of any M -ary signaling scheme to ν-bit vectors. To compare modulation schemes with different constellation sizes, it is useful to plot the average bit error probability for M -ary modulation versus the bit Eb Es SNR γb = N = νN . 0 0 • For M -ary orthogonal signaling, exploiting the perfect symmetry in the constellation, it is easy to show that irrespective of how bits are assigned to symbols, we have Pb = 2ν−1 2(log2 M −1) Pe = Pe M −1 M −1 We can hence get Pb as a function of γb , based on expressions for Pe in terms of γs . • For linear modulation, finding an exact expression for P b as a function of γb is difficult except in special cases such as BPSK and QPSK. Furthermore, P b depends on how bits are assigned to symbols (except of course in the special case of BPSK). • Gray Coding. Assign bits to symbols so that nearest neighbors differ by exactly 1 bit. Gray coding is possible for many linear modulation schemes (e.g., PSK, PAM), but not all (see HW 7). • Nearest Neighbor Approximation (NNA) for P b . Let the symbol m be represented by the bit vector bm = [b1,m · · · bν,m ]> , and define: Ndmin (bm , i) = # NN’s of bm that differ from bm in the i-th bit position . Then and s P ({i-th bit position in error} | {m sent}) ≈ N dmin (bm , i) Q Pb,i d2min (m) 2N0 s M 2 (m) d 1 X min . = P {i-th bit position in error} ≈ Ndmin (bm , i) Q M 2N0 m=1 Finally Pb = ν 1X Pb,i . ν i=1 For Gray coded constellations, the NNA approximation for P b is at most equal to Pe /ν (why?). c V.V. Veeravalli, 2006 1