Skip to Main Content
A method for the optimization of the efficiency of alternate tests for adjustable RF mixers is presented in this paper. Alternate tests provide a cost- and time-effective substitute for their conventional specification-based counterparts by attempting to predict rather than directly measuring a circuit's performance from its response to suitable test stimuli. In order to provide post-manufacture yield recovery through calibration, integrated RF circuits-especially nanometric circuits-are often designed to present some form of adjustability. Such a property offers a set of discrete states of operation, from which a performance-compliant state is selected by calibration. In general, an alternate test can be conducted for each discrete state of operation, thus providing a large set of test observables from which regression models can be constructed to predict performance in all available states. However, test time and cost concerns impose that the derivation of the predictive models should be performed through an optimization procedure that aims to select a subset of the test observables that minimizes a certain cost criterion. In this paper, alternate tests for adjustable RF mixers are considered where the test response consists of dc voltage levels that appear at certain circuit nodes while the mixer operates in homodyne mode. Selection algorithms are applied to determine the optimum observables from the test response. Simulations on a typical RF mixer, designed in an 0.18- μm CMOS technology, have shown significant improvement in the corresponding alternate test efficiency.