The energy-efficient capacity of the noiseless optical PPM channel is examined. It is shown that, although the capacity per photon can be made to increase without bound, the capacity per channel use (for best energy efficiency) is always less than 2 nats and approaches 2 nats/symbol as the bandwidth expansion factor goes to infinity.

Even in noiseless optical channels one must take into account the fact that the time resolution available is finite. An optimization scheme under the constraint of a given information rate (in nats/second) and minimum time-slot resolution is presented. It is shown that system efficiencies in excess of tens of nats/phonon will be extremely difficult to achieve due to fundamental time resolution limitations.

If power is abundant and components are ideal, communication using all linear components may be achievable. However, in most practical communication systems, power is severely limited and the inefficiencies associated with actual linear power amplifier stages \nare intolerable. For these reasons, most communication systems employ highly nonlinear power amplifiers. Whether the resulting channel is linear or non-linear depends on ones definition of the channel. If the channel is defined as being from the input of the transmitter to the output of the demodulator at the receiver, the channel is clearly nonlinear, even through the electromagnetic channel may be completely linear.

The combination of a bandpass limiter and an nthpower tracking loop is considered. It is shown that the performance of the combination does not depend on the form of the nonlinear function used to create the nthharmonic signal as long as some energy in that zone is produced. Furthermore, the order in which the limiter andnpower nonlinearity occur is unimportant so that the results apply equally well to then-phase Costas type of implementations of the nthpower loops. Closed form expressions for the signal suppression factors are obtained, and it is shown that the coherent and noncoherent SNR's out of the nthpower-limiter combination can be expressed in terms of two of these suppression factors. This latter point enables one to estimate the loop SNR of an nthpower-limited phase locked loop by simply calculating the suppression factors for nthand ((2n)^{th}) power loops. Finally,nthpower phase locked loops without limiters are compared with similar loops with limiters where it is shown that the presence of the limiter can actually enhance the output SNR at moderate to large values of input SNR.

A new model for predicting the computational performance of a sequential decoder operating in a noisy carrier reference environment is described. The major difference between this model and previous models is that the new model characterizes the number of computations per frame as the sum of the computations resulting from a number of independent searches. This number of independent searches can then be considered as an effective frame length. When this computational model is averaged over noisy reference phase errors using a medium-rate interpolation scheme, the results are found to agree quite favorably with experimental measurements.

Expressions for the output signal-to-noise power ratio of a bandpass soft limiter followed by a coherent detection device are presented and discussed. It is found that a significant improvement in the output signal-to-noise ratio (SNR) at low input SNR's can be achieved by such soft limiters as compared to hard limiters. This indicates that the soft limiter may be of some use in the area of threshold extension. Approximation methods for determining output signal-to-noise spectral densities are also presented.