This paper discusses a failure analysis case study in a highly integrated mixed signal device caused by inductive coupling of on-chip signals. The techniques utilized and the approach to root cause analysis are discussed in depth. The interactions between the device design and failure mechanism are identified in detail. Focus is placed on drawing conclusions from the sum of individual data points, and the discussion provides an analytical path by which similar failures can be isolated and specific device sensitivities can be identified.

We analyzed the gain error issue and non-linearity issue of a precise Analog-to-Digital Converter (ADC) and found the root cause to be Random Telegraph Signal (RTS) noise in bipolar devices. The RTS noise produced nearly 1mV abrupt changes in a band-gap reference voltage, and affected ADC and other circuits where the reference voltage was used. We developed a new measurement method enabling us to detect RTS noise in bipolar with higher signal-to-noise ratio than traditional methods. We also developed an algorithm to extract RTS occurrence frequency and average magnitude. The RTS characterization capability has helped us invent a new bipolar structure and develop new processes to minimize RTS noise.

Accurately measuring parameter mismatch for analog MOSFETs, such as the threshold voltage (Vt) or W/L ratio, is often required in analog circuit failure analysis. The challenge in probing analog MOSFETs using atomic force probing (AFP) is contact resistance. Contact resistance between AFP tips and tungsten contacts can cause large error at high current. This paper discusses measurement error caused by contact resistance and the techniques to identify and reduce the contact resistance effect.