ACT (DR6.02) Passbands

Passbands for PA4, PA5, PA6, PA7 along with systematic and statistical uncertainties NOTE: These passbands follow the convention that they are the response to a Rayleigh-Jeans source (the FTS measurement has not been divided by nu^2).

AdvACT_passbands.h5 is an HDF5 file with the passbands and uncertainties. The first level of keys includes array and frequency information (“PA4_f150”, “PA6_f090", “PA7_f030”, etc). The second level of keys includes information about each array specifically, with theoretically enough information to be able to compute band attributes and propagate the various errors.

Specifically, ‘frequencies’ has the passband frequencies (low frequencies are cutoff from this data since near 0 it blows up due to a systematic correction made). When computing centers or other band data, we use integration bounds given as a tuple of frequencies as “integration_bounds”. They are generally defined near percent-level cutoffs around the band regions.

The mean passband measurement is at “mean_band” while systematic error bound passbands are given by ‘lower_systematic’ and ‘upper_systematic’. The systematic errors are taken to be uniformly distributed between the lower and upper bound bands.

In order to take into account statistical data errors which are correlated throughout the band, a statistical ensemble with 100 resampled averages of mean corrected bands from the measurements (at “statistical_sample”) is provided. The mean of these should match the “mean_band” well, and the spread of measurements should show statistical uncertainty when computing items such as band centers, etc.

Finally, there is a systematic frequency shift uncertainty from the FTS calibration which is given as a single 1-sigma value at “frequency_shift_uncertainty”. This is the horizontal frequency uncertainty in the band region, i.e. the amount the band could possibly be shifted to the left or right.

The systematic errors are uniformly distributed around the mean between the upper and lower bands. To propagate the errors one performs the following via Monte Carlo:

1. Sample a systematic error band.

2. Sample a statistical error. With the statistical sample you can simply take a random band from the distribution and subtract from the mean band to obtain the error.

3. Add the systematic and statistical errors to obtain another band.

4. Sample a frequency shift error from a Gaussian distribution and apply it to the band (shift the frequencies by this error), with standard deviation given in the file.

Perform enough Monte Carlo samples and the error distribution should converge.

The second file (AdvACT_color_corrections.h5) has a similar key format as the first file. Its values are a table of color correction values for beam-filling sources. Dataset fields include "beta_values", "color_correction_values", and "color_correction_uncertainties". The dataset in "beta_values" should remain the same between all bands but is repeated for ease of access.

The passband_dr6_{array}_{freq}.dat files correspond to the passbands that have been used in the DR6 power spectrum analysis and likelihood. These passbands correspond to those in the AdvACT_passbands.h5 file after applying the "integration_bounds". No uncertainty is provided for the proccessed passbands.

Back To ACT DR6.02 Passbands Products