Coefficient accuracy for CDD packet splitting techniques

The accuracies of various charge packet splitting techniques used in charge domain devices is investigated. Experimental circuits are designed and fabricated to investigate the effects of split barrier placement (dynamic vs. static), electric field patterns in the vicinity of the split barrier (blunt vs. sharp splitters), a completely symmetrical electric field pattern (dummy splitters), channel width and split ratio on coefficient accuracy. It is found that the coefficient accuracy is essentially independent of frequency up to the highest value tested, 10 MHz. The limiting factor is the high frequency characteristic of the clock drivers. Dynamic splitting results in higher coefficient accuracy than the static splitting. The blunt barrier shape strongly increases coefficient error in the dynamic splitting but has a negligible effect in the static case. Dummy splitters substantially increase the accuracy of split ratios. Optimum packet splitting combines dynamic splitting with dummy splitters, leading to coefficient error as low as 0.07%.