On the Swept-threshold Sampling in UWB Medical Radar

The non-invasive techniques to measure vital signs have received much attention lately. They exhibit several advantages compared to the traditional invasive techniques. One such technique can be based on radar principles. In this paper we scrutinize the statistical properties of a medical radar developed on a single CMOS chip, which operates in the ultra wideband from 3.1 GHz to 10.6 GHz. A key part of the chip is based on a new technique for sampling at very high frequencies called swept- threshold sampling. This is based on multiple pulse emissions, accumulation and range-gating. We derive expressions for the bias and variance of swept-threshold sampling and show that the bias is a strong function of noise power and input value, but independent of the swept-threshold parameters for a fixed input range. The variance is shown to be proportional to the quantization step size and standard deviation of the noise process. Finally, simulation results are provided as proof of concept, and they show that the derived theoretical equations for bias and mean square error are valid.