This paper compares the performance of Uniform Linear Arrays (ULA), Minimum Redundancy Arrays (MRA) and Co-prime Sensor Arrays (CSA) in terms of the Peak Sidelobe Level (PSL) of their beampatterns. A ULA distributes its sensor elements equidistantly on a line, achieving a PSL of -13.5dB [Van Trees, 2002]. Sparse arrays span the equivalent aperture as a fully populated ULA with fewer sensors providing cost and computational advantages but with higher PSLs. To span a given aperture, MRAs [Moffet, 1968] require the fewest sensors to include all the spatial correlation lags in its co-array [Johnson and Dudgeon, 1993]. A CSA interleaves a pair of ULAs undersampled by co-prime factors [Vaidyanathan and Pal, 2011]. A CSA can be conventionally processed as a single non-uniform array or by product processing of its subarrays. This paper shows only the product processed CSA sharply decreases its PSL with increasing aperture, eventually matching ULA PSLs [Adhikari et al, 2014]. The MRA and linearly processed CSA PSLs remain unaffected by aperture extension, nearly equal to each other and much higher than the ULA PSLs. Thus, the product processed CSA has the best PSL performance among the considered extended sparse arrays. [Work supported by ONR grant N00014-13-1-0230.]