Erratum for: A&A, 695, A260 (2025), https://doi.org/10.1051/0004-6361/202453056

This is an erratum for Livingston et al. (2025) that corrects and clarifies text based on comparisons between the rPicard/CASA v7.8.1¹ VLBI calibration pipeline and Astronomical Image Processing System AIPS² after consultation with the author of rPicard (Janssen et al. 2019). We erroneously stated that we had discovered an issue with the rPicard/CASA polarization calibration, which produced a 45° electric vector position angle (EVPA) offset in our observation due to flipping Stokes Q and U (Paragraph 4 of Sect. 2.1). After consultation with the author of rPicard, we understand that Stokes Q and U were not flipped and instead an observed EVPA shift of 45° resulted from how cross-term delays, i.e., Right-Left polarization feed delays or R-L delays, were solved for in this particular experiment. There is no bug in the rPicard/CASA processing pipeline. When solving for R-L delays using rPicard, we selected the option for a single solution across all spectral windows (spws) using the rldelay_axis_combine parameter, which led to a multiband R-L delay solution. This is justified by the data as there are no delay differences between the spws. Taking out this delay amounted to a 45° EVPA shift. The same shift would also be observed if a multiband R-L delay were solved for using AIPS. Without solving for a multiband R-L delay, the cross-term phase (R-L phase) solutions used to set the absolute EVPA must be applied per spw to avoid coherence losses over the full band in the cross-hand data. rPicard applies such R-L phase solutions per spw if no multiband R-L delay has been solved.

Based on discussions with the author of rPicard, we believe that the 9% difference in amplitudes between rPicard/CASA and AIPS reported in Paragraph 2, Sect. 2.1 of Livingston et al. (2025) has two potential drivers: (1) a difference in bandpass normalization: AIPS amplitudes are calibrated with a complex bandpass normalized over the inner 75% of channels when following the MOJAVE calibration steps. MOJAVE regularly checks its amplitude calibration against single-dish observations from the University of Michigan Radio Observatory and the Owens Valley Radio Observatory (OVRO) 40 m Telescope and achieves an absolute flux density accuracy to about 5%, which accounts for kiloparsec emission (Lister et al. 2018).

rPicard amplitudes are instead calibrated with a scalar bandpass normalized over all channels. The next major rPicard release (v8.0.0) will allow the user to specify the fraction of inner channels to be used for the bandpass normalization via the solnorm_cmplx_bandpass parameter.

(2) Technical bugs in the python astropy package: instead of using the raw FITS Interferometry Data Interchange (or FITS-IDI) file from the National Radio Astronomy Observatory (NRAO) archive, astropy was used to attach telescope gain and system temperature information from the Effelsberg dish to the FITS-IDI before loading the input data. astropy changes the header data unit (HDU) list structure, converting the custom FITS-IDI HDU into a Group HDU; this changes the value of NAXIS and adds a NAXIS 1 keyword. This occurs when writing the FITS-IDI out, and the changes are evident when reading the FITS file back. CASA does not recognize such files as raw FITS-IDI files and so digital correlator effects are not corrected. AIPS applies these corrections regardless, as long as DIGICOR is set in the FITLD step. The latest CASA version prevents the user from loading broken FITS-IDI files by raising an error when these file corruptions are detected. FITS-IDI files can be edited with astropy and used by CASA as long as the user creates a custom HDU class:

import astropy.io.fits as pyfits
class IdiHDU (pyfits. PrimaryHDU):
    @classmethod
    def match_header (cls, header):
        try:
                keyword = header. cards [0]. keyword
        except:
                keyword = header. ascard [0]. key
                pass
        return (keyword == ‘SIMPLE’
                and ‘GROUPS’ in header
                and header [‘GROUPS’] == True
                and ‘NAXIS’ in header and
                header [‘NAXIS’] == 8)
pyfits.register_hdu (IdiHDU)

References

© The Authors 2025