Selective Spectrum Analysis for Analog Measurements

Abstract

This paper proposes selective spectrum analysis (SSA) technique that uses two multiplier/accumulators to estimate the spectrum. It requires much less area overhead than Fast Fourier Transform (FFT) and offers better performance in terms of area overhead, dynamic range, and accuracy than analog spectrum estimation techniques. As a result, SSA is a good output response analyzer (ORA) implementation for mixed-signal Built-In Self-Test (BIST). The SSA technique downconverts the device under test output at the interested frequency to dc by multiplication and filters out the undesired ac components through accumulation. But the ac components cannot be completely removed and introduce calculation errors. While these errors can be minimized by controlling the accumulation time, the rate of convergence is low such that long test time is required to achieve reasonable accuracy. An alternate approach is to choose integer multiple periods (IMPs) of the frequency under analysis to stop the accumulation. Performance of the SSA-based ORA is analyzed in a systematic way and it is shown that the proposed IMP circuits can further improve the efficiency of the ORA in terms of test time, area overhead, and measurement accuracy. Experimental results are presented for simulation as well as implementations of the SSA technique in field-programmable gate arrays and standard cell application specific integrated circuits.