Signal source generation for mixed-signal embedded test structures

This dissertation discusses research on signal source generation schemes for embedded test circuits. Conventional signal generation methods are briefly reviewed. A system application example using the embedded signal sources with a signal detection circuit, an on-chip substrate characterization method, is proposed. Without the need of final layout information or an assumed substrate model for electrical behavior of substrate, this method can extract the signal propagation behavior of substrate itself. This enables circuit designers to consider substrate noise effects in the early design stage for reducing the design turn-around time. A memory-based periodic bitstream approach is presented. Also, a comparison of this proposed approach with existing techniques is done in the frequency domain. The memory-based periodic bitstream approach uses a minimum set of frequency bins and requires a tunable noise-shaping filter. However, the post-processing analog filter design is not straightforward and becomes an on-chip test design bottleneck. Therefore, a linear feedback shift register (LFSR) approach using finite impulse response (FIR) filtering is presented to extract better quality signals from square-wave pulses generated by simple digital hardware. Since the FIR filtering reduces the unwanted harmonic components, a noise-shaping filter can be implemented using passive elements without the necessity of being frequency-tunable. The fabricated sample measurement shows up to 45dB and 40dB spurious-free dynamic range (SFDR) for single-tone and dual-tone signal source generation, respectively.; The proposed memory-based approach generates a higher frequency signal source than the original periodic bitstream method by minimizing the number of frequency bins, and it enables a simple test setup and quick system launching by eliminating the external initialization requiring additional automatic test equipments (ATEs). Using the LFSR approach with FIR filtering, the signal source generation system is simpler than the memory-based approach by simplifying the digital signal processing circuits and relaxing the analog filtering specifications.