A critical evaluation is made of the spectral bias which occurs in the use of a laser doppler velocimeter(LDV). Two processing algorithms are considered for spectral estimates: the sample and hold interpolation method(SH) and the nonuniform Shannon reconstruction technique(SR). Assessment is made of these for varying data densities $(0.05{le}d.d.{le}5)$ and turbulence levels(t.i.=30%, 100%). As an improved version of the spectral estimator, the utility of POCS (the projection onto convex sets) has been tested in the present study. This algorithm is found useful to be in the region when $d.d.{gep}3.$