Substrate noise in mixed-signal integrated circuits

The continued scaling of CMOS technology, together with the progress in the design of high frequency analog and mixed-signal CMOS circuits, has enabled the integration of many of the functions needed to implement a broadband communications transceiver. One of the significant challenges in such system-on-a-chip design is to implement broadband analog circuits on the same die with the large complex digital signal processing functions that are required for many modern communications applications. Owing to various parasitic coupling mechanisms, there is a distinct possibility that the substrate noise caused by the transients in the digital circuitry will corrupt the low-level analog signals and seriously compromise the achievable performance. This may ultimately preclude the practical integration of a complete broadband communication system on a single CMOS chip. As a result, there is a need for a deep understanding of how the substrate noise is related to specific characteristics of the digital circuitry and how that noise might affect the behavior of analog circuits.; In this research, a frequency domain approach to analyzing and modeling the substrate noise is developed. It is shown that substrate noise can affect the analog circuits not only by direct coupling into the signal band but also by intermodulation with analog inputs. The spectral distribution of substrate noise as a function of digital circuit characteristics is also studied, and a filter bank model for predicting the substrate noise spectra as a function of the digital circuit characteristics is proposed. This model provides insight into substrate noise mitigation while saving the need for time consuming transient analyses of the entire digital circuit. Finally, the effects of substrate noise on the performance of the low noise amplifier (LNA) for a GPS receiver have been measured, and the experimental results agree well with theoretical predictions. It is shown that although the substrate noise has a power several orders of magnitude higher than the received GPS signal, only those noise components located in specific frequency ranges actually degrade the performance of the LNA.