MOS Implementations of TTL architectures: A case study

This paper describes the architectural transformation undergone by a digital signal processor developed at MIT as it was adapted from a TTL MSI implementation to a monolithic MOS implementation. Though the choice of architectures in each technology was guided by the functional requirements of the Klatt vocal tract model (16-pole, 2-zero formant synthesis), the resultant architectures were dramatically different. The example demonstrates that good TTL architectures may not translate neatly into good MOS architectures and that, if a system in the first technology is to be reproduced in the second, a moderate investment in architectural redesign may yield great returns during the steps of circuit design, layout and testing. In fact, this case shows that as an architecture grows progressively simpler, it can also grow progressively more general-capable of serving a broader class of applications than first intended.