7.4. Pipeline Processing

All HIFI data that arrives at the Herschel Science Centre will be processed through standard pipelines that are designed to take into account all known instrument properties in order to produce a fully calibrated set of data. The following two sections indicate the basic processing steps that are expected to be done for the Wide Band Spectrometer (WBS) and High Resolution Spectrometer (HRS).

7.4.1. WBS Pipeline Processing Steps

The following processing steps are performed, in modules, starting from raw HIFI data frames obtained from either polarization using a WBS.

  • 1. Bad pixels are corrected for.

  • 2. Dark levels are subtracted from the 4 CCDs used in taking a single data frame, This is based on the values of the first four un-illuminated channels of CCD.

  • 3. Non-linearity of intensity scaling in WBS is removed

  • 4. Zero frame subtraction. A zero second "bias" frame is removed.

  • 5. Fitting of internal frequency comb with polynomial

  • 6. Apply frequency fit to frames to provide channel frequency scaling

  • 7. Apply attenuator setting correction

  • 8. Apply hot/cold flux calibration

  • 9. Subtract reference spectra (as needed)

  • 10. Apply sideband gain correction (based on ground-based test information)

  • 11. Apply antenna temperature efficiency

The end product is a set of double sideband spectra for an observation with flux, measured as antenna temperature, versus frequency.

An additional possible component to the pipeline is a module to remove known ripple structures from the data.

7.4.2. HRS Pipeline Processing Steps

The following processing steps are performed, in modules, starting from raw HIFI data frames obtained from either polarization using a HRS. Note that the data frames from the HRS are correlation functions.

  • 1. Re-organize data into subbands, depending on the configuration. Output is the autocorrelation functions re-organized in subbands.

  • 2. Compute power and offset of the subband values

  • 3. Remove bad channels

  • 4. Normalize the raw correlation function. Normalizes by channel 0 so that the power is set to 1. Output is a normalized autocorrelation function.

  • 5. Apply a quantization distortion correction. Uses a calibration table to correct for the analogue to digital quantization effects. Output is a corrected autocorrelation function.

  • 6. Apply power correction. The signal power is corrected for non-linearity effects.

  • 7. Apply window (default = Hanning smooth used)

  • 8. Apply symmetrization to the autocorrelation functions

  • 9. Apply FFT to place spectrum in frequency domain

  • 10. Apply frequency scaling based on the HRS LO frequencies values to place appropriate frequency information in the frame (placement within 4GHz IF frequency range.

  • 11. Multiply the normalized spectrum (power=1) by the corrected power computed in 6.

  • 12. Apply hot/cold flux calibration

  • 13. Subtract reference spectra

  • 14. Apply sideband gain correction (based on ground-based test information)

  • 15. Apply antenna temperature efficiency