An efficient channelizer tree for portable software defined radios

Power consumption is one of the most critical issues in the portable software-defined radio devices. A software radio receiver has the need to downconvert, bandwidth limit, and downsample a single narrowband channel from a span of frequencies in the Nyquist zone collected by the input analog to digital converter. In this paper, we present two techniques that perform the receiver function more efficiently than the standard Gray chip architecture formed by its direct digital synthesizer (DDS), and two stages of downsampling with a cascade integrator comb (CIC) filter and a pair of half-band filters. We compare the workload of this conventional architecture to two new architectures by applying them to the task of extracting a single, 30 kHz wide, channel from a 30 MHz band sampled at 90 MHz. One proposed structure replaces the CIC filter with a 10-stage cascade of 2-to-1 downsampling half-band filters with successively narrower transition bandwidths. In the second proposed structure, the DDS is moved to the output of the filtering stages which perform a sequence of 2-to-1 downsampling operations in half-band filters that perform incidental spectral translation by aliasing. We enlarge the set of half-band filter that reside at 0 and fs/2, to also include the Hilbert transform half-band filters residing at ±fs/4. At every stage in the cascade, the selected band resides in one of the four half-band filters. The 2-to-1 downsampling with that filter reduces the bandwidth and aliases the desired center through a sequence of known center frequencies. The desired channel is recovered from the output of the final stage by a complex heterodyne applied (at the low output rate) to obtain the desired spectral shift to base band. The paper provides a detailed workload analysis of the proposed structure along with simulation results that prove its effectiveness.