Estimating the noise in the spectrum T of S8, using repeated observations. Plus symbols show the observed standard deviation of ΔT₁₂ from channel by channel comparisons of two consecutive spectra, which ought to have very similar stray radiation and baselines. The measurements agree closely with the prediction (dotted line) from Eq. (6) for line noise alone. Open circles give the standard deviation of ΔT_pol; these lie slightly above the prediction because of baseline errors. The filled circles give the standard deviation of ΔT = T − ⟨T⟩, which reflects all errors. As a prediction of this, the crosses show the result of adding 7% of the rms spectrum of the stray radiation correction in quadrature to the standard deviation of ΔT_pol. The diamonds result from further addition of the effects of tiny 0.3% gain changes. The S8 data are consistent with errors in stray  ~7% and scaling (e.g., gain) errors of  <0.3%.