Most magnetic recording and many other digital communication channels exhibit statistical dependencies among errors. The design of error-control schemes for such channels .requires proper models and tools that can be used to evaluate error performance after decoding. In this paper a simplified partitioned model of a channel with memory suggested by B. D. Fritchman is considered. This model is used to derive expressions for block and bit error probability bounds for major block burst-error-correcting codes: interleaved, single burst-correcting, and nonbiuary codes. The model-based bounds are compared to the experimental ones using the data obtained for helical scan magnetic tape recorders. The comparison showed an agreement between experimental and analytical data within one order of magnitude, with the average difference being as small as 16 percent in some cases. A simple and effective implementation of a multiple burst-error-correcting scheme based on the majority-logic decoding of interleaved binary codes is suggested. The scheme requires about ten off-the-shelf IC's for both encoder and decoder with the interleaving degree up to 512.

An improved method of acquisition of pseudonoise signals (RARASE) is described and analyzed. This method is an outgrowth of the RASE method, rapid acquisition by sequential estimation. In the new method, the known recursion relation of the PN signal is used to determine if a short estimate of the state of the received PN signal is probably correct and if, consequently, an attempt should be made to track with that estimate. A high proportion of the incorrect initial state estimates can be discarded with relatively simple logic. Analyses are provided for a range of implementations from the simplest, a single 3-input modulo-2 gate, to multiple logic gates and for both a simple in-lock detector and a sequential detector. Significant improvement in acquisition time is shown for long PN signals and for the signal-to-noise range of -13 dB to +3 dB. Acquisition time reduction by a factor of 7.5 is calculated for a PN signal of length2^{15}-1bits compared to the sequential estimation (RASE) method using a simple in-lock detector. Reduction by factors of 1.5 to 3.5 is calculated using the sequential detector.