Some Aspects on Adiabatic Switching

In adiabatic switching 2] up to now the major focus has been on proving the usefulness of this kind of circuit technology by constructing various kinds of digital systems in full custom design. This is, however, not a desirable way to treat more complex systems. Therefore, it is necessary to provide some systematic design procedure that can be automized and is capable to handle all kinds of logic functions without loosing the adiabatic properties of the circuitry. The main condition for eeective adiabatic operation will be stated in this article. Further on an algorithm is proposed that will realize an arbitrary boolean function as a fully adiabatic gate. First the function will be represented by a binary decision diagram (BDD), which is, by itself, directly feasible as a transmission gate structure. However, this structure will contain too many transistors. So in a second step symmetries in the root branches of the BDD will be used to replace parts of the BDD structure by elementary pass-transistor gates, thus reducing the number of transistors considerably. To support this reduction, the representation of the boolean function as a BDD, which is not unique, must be chosen advantageously. This will lead to a combina-torial optimization problem interwoven with the reduction of the BDD structure. 1 Energy consumption in adiabatic circuits For low-power applications, adiabatic switching 2] seems to have promising capabilities to lower the losses in digital circuitry. It is, however, a completely new circuit paradigm with speciic problems that diier a lot from the known problems in standard CMOS circuits. The main idea of adiabatic charging is based on a circuit like shown in Fig. 1. Whenever the capacitor is charged or discharged through a resistor abruptly, an unavoidable 1