In this paper, a scan-based programmable, configurable, and scalable architecture is proposed. This architecture is suitable for a wide range of applications in signal processing requiring programmability and presenting high bandwidth and real-time requirements beyond the capacity of off-the-shelf DSPs or FGPAs. The architecture is specifically targeting a very common type of signal processing operation: sliding window operations (SWOs). Through various examples, the "programmability, configurability, and scalability" of the proposed architecture are illustrated. Our approach is then compared to traditional programmable architectures with coefficient registers in terms of gate count, speed (delay), and other implementation-related issues. This comparison reveals that our architecture leads to less complex solutions with comparable performance. In general, this approach can be seen as an alternative offering reduced recurrent costs at the expense of potentially higher nonrecurrent costs, which makes it very attractive for high volume production.

The decoding of long memory high-rate punctured convolutional codes by sequential decoding algorithms is investigated. Both the stack and the Fano algorithms have been thoroughly tested through computer simulation with coding rates ranging from R=2/spl sol/3 to R=/spl frac78/. Error and overflow probabilities and variability of decoding effort are similar for both algorithms. With hard quantization, plateaus appear in the cumulatives of decoding effort for both algorithms. Comparing the punctured approach of decoding to the more traditional technique for high-rate codes, it is found that punctured decoders perform a larger number of simpler computations, so that the overall decoding effort is on the average more important for the usual decoder than it is for its punctured counterpart. Finally, computational variability, error and overflow probabilities are no worse for punctured decoders than they are for normal decoders.>

The weight spectra of high-rate lpunctured convolutional codes are evaluated under the hypothesis of a low-rate structure. This interpretation yields results slightly different from those obtained when weight spectra are evaluated assuming a true high-rate structure for punctured codes. The search for long memory punctured codes is extended by providing new punctured codes of rates 4/5, 5/6, 6/7, and 7/8 with memories ranging from 9 to 19.>

The authors present some properties of punctured convolutional codes, providing a construction method and a list of new, good, high-rate, long-memory punctured codes. The structure of punctured codes is examined and an upper bound on the free distance of punctured codes is derived, indicating that punctured codes are good codes. A construction method that generates the low-rate original codes which duplicate given known high-rate codes through perforation is proposed. Tables of punctured codes that duplicate the best known nonsystematic codes of rates 2/3 and 3/4 with memory lengths ranging from 3 to 23 and from 3 to 9, respectively, are given, together with the best known systematic codes for rates ranging from 2/3 to 7/8 with very long memory, M=44 through 48.>

An investigation is conducted of the high-rate punctured convolutional codes suitable for Viterbi and sequential decoding. Results on known short-memory codes M>