for POWER LINE COMMUNICATIONS

We discuss the application of coding and modulation for power-line communications. The results are based on models derived from realistic channel measurements in existing networks and are in agreement with existing standards. Introduction This papers discusses several aspects of communications over the low voltage distribution lines. We first discuss a modulation scheme know as M-tone signaling. After a general formulation, we review Spread-FSK (Frequency Shift Keying) [I] and present a M-tone signaling scheme with error control coding for packet transmission. In an earlier paper [2], we discussed the issue of channel capacity. Tlie topic of data security is also of great interest, since our PLC channel is a "busy'-type of channel, i.e. all network participants can listen, enter or disturb the communications. Capacity and Modulation One of the main obstacles in the development of PLC systems is the maxitnum allowed transmitter , output voltage. According to the CENELEC norms, EN 50065.1, part 6.3.2, the maximum allowed peak voltage for narrow band signals (i.e. a 20-dB bandwidth of less than 5 khz in width) at 9 lcHz equals 5 V, exponentially decreasing to 1 V at 95 W, and for broad-band transmitters (i.e. a 20-dB bandwidth of more than 5 khz in widtli) equals 5 V = 134 dB ( pV). As a consequence, the transmitters are output voltage limited and bandwidth limited. Therefore, from a cornrnunication point of view Mary modulation is only interesting for ratlier small values of M. Furthermore, M-ary modulation for M > 2, ahnost always ends up in adding different frequencies in such a way that the maximum amplitude is not under control (see OFDM signals). Tlie candidate modulation schemes tlmt remain interesting are simple FSK and M-tone signaling combined with error correction andor interleaving. FSK and M-tone signaling Iuve the advantage of leading to a constant envelope signal modulation and a detnodulation in a coherent as well as a noncoherent way. In general, FSK is not an optimal way of co~lununication, but it gives rise to low complexity transceivers. In an M-tone modulation system, symbols are modulated as one of the sinusoidal waves described by si(t> = E c o s ( 2 d i t); 0 S t r is (1) where i = 0,1, ..., M-1 and Es is the received signal energy per symbol and The signals are orthogonal and for non-coherent reception the frequencies are spaced by 11% Hz, being the transmission rate. The demodulation may be accomplished using 2M correlators, 2 per signal waveform. In a coherent system, the frequency spacing can be reduced to 1/2Ts. The bandwidth efficieny in this case is For larger M, M-ary FSK is spectrally inefficient. Coherent Binary-FSK is characterized by having a two-dimensional signal space, see Fig.la. The distance between two message points is &. The symbol error probability for transmission over an Additive White Gaussian noise .channel with single sided noise power spectral density No, at lug11 values of EJNo, for Binary Phase Shifi keying, coherent M-FSK and noncolierent M-FSK can be approximated as In spread-FSK, gie receiver is assumd to liaveehe possibility of detecting whether one of tlie two frequencies is disturbed. It then ignores tlus particular frequency in the detection. The decision rules are to be modified, see Fig. lb, in order not to degradate tlie system too much. Tlle distance between two signal points in the modified detection is reduced to 6, which gives rise to a loss of 3dB in an AWGN cllannel. -One of the problems in using a FSK based systemis the sensitivity to so called narrow band noise. Often, this type of noise is caused by television sets or computer terminals. S-FSK is part of a more general solution to this problem. Consider an M-ary FSK signal set that consists of an orthogonal set of M frequency-shifted signals. We may assume noncoherent detection and thus the bandwidth required to transmit M-ary FSK signals is I ,decision boundary Fig. la. Signal space for coherent S-FSK . Fig. lb. Modified decision boundary for coherent ' S-FSK. In Fig. 2. we give a modified decision boundary for S-FSK when the detector is not informed about the disturbance. Note tlmt the asymptotic performance is the same as for the informed detector. ------I * decision boundary 1 0