CMB Broad-Band Power Spectrum Estimation

Abstract

The natural outcome of theoretical calculations of microwave background anisotropy is the angular power spectrum ${\cal C}_\ell$ as a function of multipole number $\ell$. Experimental ${\cal C}_\ell$'s are needed for direct comparison. Estimation procedures using statistics linear in the pixel amplitudes as well as the conventional but less useful quadratic combinations are described. For most current experiments, a single broad-band power amplitude is all that one can get with accuracy. Results are given for the Capri-meeting detections. Mapping experiments, sensitive to many base-lines, can also give spectral ``colour'' information, either with a series of contiguous narrow-band powers or as parameterized by a local ``colour'' index $n_{\Delta T}$ (scale invariant is -2, white noise is 0). Bayesian analyses of the full first year DMR and FIRS maps give very similar band-powers (\eg $Q_{rms,PS}=17.9 \pm 2.9 \mu K$ {\it c.f.} $18.6 \pm 4.7 \mu K$ for $n_{\Delta T}=-2$) and colour indices (with 1 and 2 sigma error bars) $n_{\Delta T}+3=2.0^{+0.4;+0.7}_{-0.4;-1.0}$ and $ 1.8^{+0.6;+0.9}_{-0.8;-1.3}$ ({\it c.f.} the value 1.15 for a ``standard'' scale invariant CDM model). The 53 and 90 GHz DMR maps, as well as the FIRS map, have residual short-distance noise which steepens $n_{\Delta T}$. Allowing the pixel error bars to increase absorbs much of the residual, but further exploration is needed to see if a second residual evident in the data which is responsible for the high $n_{\Delta T}$ is from systematic errors or is physical.