Analogue techniques for very large scale integrated circuits

The increased penetration of digital signal processing in integrated circuit (IC) technology has relegated analogue circuits to the role of interfacing with the analogue `outside world'. This situation has created new challenges since the analogue circuit frequently has to perform to higher standards with an IC process optimised for digital applications. This thesis addresses this dilemma in two ways. In Part A, an active-filter technique is presented which uses the resistors and capacitors commonly found in VLSI processes. Tuning of these highly linear passive components is achieved by digital means. In the technique, local measurement is made of an IC time-constantand a digital code is generated for programming binary weighted capacitor arrays to achieve self-tuning. After discussing practical design considerations, a filter is designed and integrated in a CMOS process. The measured results demonstrate a fully functioning technique with promise of extremely high linearity for filters requiring modest precision. In Part B, a new technique for sampled-data signal processing, called `switched-currents', is presented. It performs all the common functions of the switched-capacitor technique but without the need for linear floating capacitors and so it can be implemented in the most basic digital VLSI CMOS process. Basic modules (delay cells, integrators, differentiators) are developed and filter synthesis is illustrated. The effect on analogue performance of MOS transistor circuit imperfections (mismatch, output impedance, setting behaviour, charge injection and noise) is analysed and circuits are described which minimise their effect. An IC implementation demonstrates performance and confirms earlier analyses.