Effect of Noise on Analog Circuit Testing

The effect of test environment noise (tester noise) on test waveforms is considered. We show that tests generated ignoring the tester noise characteristics are prone to failure when actually applied to the circuit-under-test (CUT). The failure may result in the good circuit being declared faulty or the faulty circuit being declared good. This failure is independent of the fault model and nature of the test, i.e., AC or DC, time domain or frequency domain. We characterize the total noise at the primary outputs (PO's) of the circuit using second order statistics. We use the noise power spectrum and root mean square (RMS) values to make decisions about the test waveforms and recommend more noise-robust tests. For non-linear circuits we use the Central Limit Theorem of statistics to approximate narrow band noise at a primary input (PI) by a sum of sinusoidal distributions, and we use Monte-Carlo simulations to determine the noise at the PO's in the time domain. Results of experiments on an instrumentation amplifier, a biquadratic filter, and a Gilbert multiplier are presented, which prove that valid tests in a noise-free environment are invalid when tester noise is considered.