Built-in performance characterization of embedded mixed-signal circuits
Recent advances in semiconductor technologies enable the integration of previously disparate designs into a single system-on-chip (SOC). While the SOC offers significant benefits in reducing overall design cost, it poses signifi- cant challenges for testing. A traditional specification-based test method is no longer attractive in testing deeply embedded analog and mixed-signal circuits due to limited I/O accessibility, limited tester resources and signal integrity is- sues. Built-in Self-Test (BIST) has been considered as a promising solution to overcome such difficulties; however its widespread adoption has been hindered for several reasons. The aim of this thesis is to develop efficient self-test techniques for em- bedded analog and mixed-signal circuits, which provide test accuracy equiv- alent to a traditional specification-based test, but with minimal overhead in terms of performance, area and additional test cost associated with DFT cir- cuits. The outcome of the self-test is a set of performance parameters, allowing us to evaluate DUTs with respect to its specification, and to guide a self-repair mechanism efficiently. A compact representation of analog signals and its sim- ple recovery algorithm are developed to replace a traditional analog signal test. With this alternative representation, the hardware overhead associated with analog signal generation and measurement is significantly reduced, thereby the requirements for BIST implementation are substantially relaxed. To overcome accuracy and precision limitations posed by on-chip designs for test or exist- ing hardware used for test, a spectral prediction technique and a statistical digital equalization technique are studied. These techniques are incorporated into a loopback test scheme where analog cores can be tested with pure digi- tal methodologies, but where test accuracy is yet considerably limited due to fault masking problem and precision limitation. An efficient fault diagnosis technique based on the BIST techniques and circuits for self-repair, is also in- vestigated. This study constitutes the first attempt at the diagnosis of analog faults which accounts for existing BIST and self-repair circuits.