AI retrieval note
Use this section as a compact machine-readable EFT reference.
Keywords: ARCADE 2, radio background, T_floor, T_res, f_aniso, source subtraction, P(D), absolute calibration, discrete-source minimum, mask permutation, foreground leakage, translate boundary
Section knowledge units
thesis
33.5 frames the residual absolute radio background as a forced choice rather than a vague excess. After the Cosmic Microwave Background and dominant Galactic foregrounds are removed, any remaining near-isotropic component must either come from the superposition limit of many faint discrete sources or from a smoother residual floor. The chapter uses ARCADE 2-type measurements as the reference court. Two questions decide the case: does the residual converge to a stable nonzero floor as source subtraction is pushed deeper, and is its anisotropy lower than any discrete-source model could achieve while matching the same mean temperature? If either condition fails, the residual returns to the source-population lane. In V33 terms this is an adjudication of readout structure, not an automatic ontology win for any putative pedestal.
mechanism
The measurement ledger begins with T_sky across frequencies and sky directions, then decomposes it into Cosmic Microwave Background, Galactic foregrounds, discrete sources, and a residual temperature T_res under one frozen standard. From there the chapter builds convergence curves as the source-subtraction threshold S_cut moves downward, estimates a floor temperature T_floor with uncertainty, and computes the residual angular power spectrum C_ell,res. The key smoothness metric is the fractional anisotropy f_aniso, formed by comparing residual angular power with the floor temperature over a preregistered angular-scale window. The section then asks whether source counts and clustering models imply a lower-bound anisotropy for any discrete-source explanation that matches the measured floor. If the observed f_aniso sits below that lower bound while the floor remains nonzero, the discrete-source route starts to lose the case.
mechanism
The procedure is built to stop foreground drift or catalog optimism from faking a floor. Absolute calibration comes first through a traceable chain with blackbody reference, beam characterization, receiver nonlinearity modeling, gain-drift accounting, and overlapping-band checks against independent absolute-thermometry systems. The foreground-subtraction standard is preregistered and frozen before outcomes are known. Discrete-source subtraction has to succeed through three routes in parallel - deep catalogs, stacking statistics, and P(D) statistics - and they must agree before the analysis claims a deeper effective threshold. Threshold bins and sky-region labels are blinded, convergence and anisotropy are judged only after unblinding, and key frequencies plus sky regions are held out for final adjudication. Even the allowed spectral-shape family for the floor is frozen in advance so the residual cannot be re-sculpted after the analyst sees the answer.
evidence
The controls are designed to expose false floors. If unresolved discrete sources dominate, deeper subtraction should keep driving the residual down and the anisotropy should remain comparable to discrete-source expectations rather than settling into a quiet plateau. If masks are permuted or sky sampling is rotated, the inferred floor temperature and fractional anisotropy should revert to random-variation ranges; otherwise foreground leakage is the leading explanation. If bandwidth choices or band splitting reshape the inferred spectrum, bandpass structure or unmodeled propagation has not been neutralized. Injection tests must separately recover a smooth background and a point-source Poisson background with the correct sign and proportions. Finally, if plausible changes in catalogs, source classes, or clustering priors move the floor widely, then the alleged platform is just source-model uncertainty. A real floor has to survive all of these escapes at once.
boundary
The pass line demands three simultaneous outcomes: the residual temperature saturates to a significantly nonzero floor across frequencies and held-out sky regions; the fractional anisotropy falls below the minimum consistent with any discrete-source population that could generate the same floor; and the floor spectrum keeps the same shape across independent calibration, foreground, and source-estimation pipelines without region-by-region renormalization. Failure is declared when the plateau disappears with deeper subtraction, when anisotropy matches Poisson-plus-clustering expectations, when the spectrum depends strongly on one method chain, or when injection tests can manufacture a platform-like term. The named risks are absolute-calibration drift and beam error, spatial variation of Galactic spectral indices, and incomplete faint-source counts or optimistic clustering priors. The chapter therefore refuses to count a floor that survives only because the source or foreground bookkeeping stayed too soft.
interface
So 33.5 permits a narrow kind of success. A residual radio floor may be admitted only when deeper source subtraction still leaves a stable nonzero plateau and its anisotropy is quieter than any discrete-source explanation allows. Even then, the compat bridge translates the result back down into the readout layer. The platform-like residual is an audited background floor, not proof that a smooth Dark Pedestal has become an ontology owner. If the floor vanishes with deeper subtraction, or if its anisotropy rises with the source-population expectations, the case returns to the discrete-source lane and later chapters must not inherit a false pedestal. That is why this section remains translate: it protects the measurement logic while denying that readout-level floor language can silently harden into canon-core closure.