Magnetoelastic NanoMagnet Logic Circuits
thesisposted on 2015-07-21, 00:00 authored by Davide Giri
Among the alternative technologies proposed for the post-CMOS scenario, Quantum-dot Cellular Automata (QCA) is one of most promising. The magnetic based implementation of QCA, named NanoMagnet Logic (NML), is the only one that can both work at room temperature and is feasible with current fabrication processes. The base cells of NML technology are nanomagnets, which can be arranged on a plane to create any logic circuit. There is no standby power consumption and the energy required for magnets switching is several orders of magnitude lower than latest CMOS transistors. However the network for controlling the cells' magnetization can nullify the advantages in terms of power losses. This is the case of the Magnetic Clock NML, which has been extensively and thoroughly studied in literature. A novel implementation of NML technology, the Magnetoelastic NML (ME-NML), drives the nanomagnets through an electric field instead of a magnetic field, highly reducing the power consumption. This solution has already been proved theoretically and experimentally, however up to now only elementary circuits have been studied. The Magnetoelastic NML is the subject of this work. To fully understand its potential it is mandatory to analyze complex architectures keeping into account all the physical constraints related to the fabrication process. Firstly, we developed a RTL model for handling ME-NML circuits. The model also embeds the capability of evaluating area occupation and power consumption. We were also able to define a Standard Cell library, which is a big step toward the creation of a design tool. Secondly, we developed an accurate comparison of ME-NML with the Magnetic Clock NML and the CMOS transistor. ME-NML performances were excellent, enough to largely overcome both the other technologies. This also provided general information on circuit design. Once the validity over other technologies was proven, it was mandatory to understand which kind of architectural organization maximizes the performance of the ME-NML. Therefore through a second case study we performed the first step of this investigation, comparing three different versions of a MAC unit: parallel, serial-parallel and serial. The parallel approach guarantees the best results, but it requires a certain level of interleaving.