Design tools for synthesis and optimization of circuit-electromagnetic systems

With ever shrinking technologies and increased integration of complex systems on a chip, mature computer-aided design (CAD) techniques for design automation of radio frequency (RF) components remains incipient. In the multi-GigaHertz (multi-GHz) range, on-chip passives must be modeled electromagnetically for accurate characterization of their high-frequency behavior. Most RF circuits consist of active and passive components, and it is necessary to perform multi-stage parasitic extraction for accurate circuit simulation. The parasitics associated with the on-chip passives are captured using a full-wave electromagnetic (EM) simulator. On the other hand, parasitics attributed to the active components are extracted using a circuit parasitic extractor. Thus, an RF circuit is essentially a circuit-electromagnetic system. This dissertation proposes an active-passive co-synthesis methodology and framework for rapid sizing (synthesis) of RF circuits. Synthesis of multi-GHz RF circuits brings together difficult challenges related to simulation, extraction and multidimensional space search. A stochastic combinatorial optimization technique is employed for design space exploration. The benchmark RF circuits are low-noise amplifiers (LNAs). Two parametric macromodeling approaches are proposed in order to develop a synthesis strategy without the EM solver inside the optimization loop. EM simulations can be prohibitively expensive during design space exploration and the active-passive co-synthesis methodology successfully overcomes this computational bottleneck through macromodeling. The first approach constructs parametric macro-models of on-chip passives on a frequency-by-frequency basis and assumes a priori knowhow of the center frequency of the LNA under consideration. The second approach develops broadband parametric macromodels of on-chip passives and lends frequency scalability to the methodology. This facilitates LNA design for a wide range of center frequencies. A framework for statistical analysis of RF circuits has also been developed for yield estimation of RF circuits. A decoupling strategy leverages separate characterization of active and passive components in terms of their admittance parameters. Finally, the active and passive components are re-combined using their admittance parameters in order to perform yield estimation of the complete circuit-electromagnetic system. Another class of circuit-electromagnetic systems that are rapidly gaining prominence in our daily lives are radio frequency identification (RFID) tags. An RFID tag consists of an antenna and a tag integrated circuit (IC) with the antenna as an EM component. This dissertation addresses range maximization of RFID tags in order to alleviate the asymmetry of uplink and downlink ranges. Design solutions for the tag IC are proposed to maximize range of passive RFID tags. Recently, attempts have been made to include multiple RFID tags on the same object in order to improve range. However, this induces crosstalk between tags. Almost no prior research has been undertaken in this regard for RFID system deployment, and this dissertation studies the feasibility of inclusion of multiple RFID tags to boost range.