Strip defect recognition in electrical tests of silicon microstrip sensors

This contribution describes the measurement procedure and data analysis of AC-coupled double-sided silicon microstrip sensors with polysilicon resistor biasing. The most thorough test of a strip sensor is an electrical measurement of all strips of the sensor; the measured observables include e.g. the strip's current and the coupling capacitance. These measurements are performed to find defective strips, e.g. broken capacitors (pinholes) or implant shorts between two adjacent strips. When a strip has a defect, its observables will show a deviation from the "typical value". To recognize and quantify certain defects, it is necessary to determine these typical values, i.e. the values the observables would have without the defect. As a novel approach, local least-median-of-squares linear fits are applied to determine these "would-be" values of the observables. A least-median-of-squares fit is robust against outliers, i.e. it ignores the observable values of defective strips. Knowing the typical values allows to recognize, distinguish and quantify a whole range of strip defects. This contribution explains how the various defects appear in the data and in which order the defects can be recognized. The method has been used to find strip defects on 30 double-sided trapezoidal microstrip sensors for the Belle II Silicon Vertex Detector, which have been measured at the Institute of High Energy Physics, Vienna (Austria).