MOAR LECTURES power spec density Sy = mag^2(P)/Tb * sum(Rncos(n2fpi*Tb)) Rns are discrete autocorrelations of the line symbols ak mmmm...bianary, southern for binary did on-off signaling, eg. unipolar RZ pulse or no pulse averages to 1/2, average product of 2 random pulses is 1/4 (autocorrelation term) massaging gives us Sy = mag^2(P)/Tb * (1/4 + 1/4 * sum(e^(-jn2fpi*Tb)) Sy = mag^2(P)/4Tb * (1+1/Tb * sum(delta(f-n/Tb))) thus, p=rect((t-Tb/4)/(Tb/2)) -> P=Tb/2 * sinc(fpi*Tb/2) * e^(-jfpi*Tb/2) considering AMI: 0 -> no pulse 1 -> +/- pulse, alternating signs R0 = 1/N * (0N/2 + (+/-1)^2 * N/2) = 1/2 Rn = 1/N * sum(akav(k+n)) n=1: akav(k+1) = 00 01 10 11 with vals 0,0,0,-1, yields 1/4 on avg n=2: akav(k+2) = 00 01 10 11 100 101 110 111 with vals 0,0,0,0,0,-1,0,1 yields 0 on avg . . . Rn = 0 for n>=2 Sy = mag^2(P)/Tb * (1/2 + 2(-1/4)cos(2fpi*Tb)) Sy = mag^2(P)/2Tb * (1 - cos(2fpi*Tb)) Sy = mag^2(P)/2Tb * sin^2(fpi*Tb) p is rect -> Sy = Tb/4 * sinc^2(fpi*Tb/2) * sin^2(fpi*Tb) bipolar signaling has sevarl advantages, spectrum has a DC null, BW is not excessive, single error detection capability two disadvantages: need to detect 3 levels instead of 2, other problem is locking of sync for large sets of 0s european uses HDB3, replaces 4 0s with special seq of 000V or B00V, B is a 1, correct polarity, V is a bipolar violation which one is used depends on the number of 1s following the last special sequence ex: 1100000000110000010 AMI - ________- _____-_ - - HDB3 - ___ -__- - __ _-_ - - - - - B8ZS - ___ -_- - _____-_ - - - - B3ZS - -_ -_ __- -_ __-_ - - - - - BNZS, binary N zero sub B8ZS is common, replaces 8 0s with 000VB0VB B3ZS used in DS3 lines, 000 replaced with B0V