The stabilization of discrete system zeros is commonly achieved by allowing the input pattern to deviate from zero-order hold (ZOH). In this paper, we show that this design technique may not be practical at small sampling times when the continuous system has unstable zero(s). Deviation of the input pattern from ZOH may also cause undesirable fluctuation in the inter-sampling-time output. In order to reduce this effect, we present a method which uses linear quadratic optimization for the determination of suitable zero locations.