2 Observations and data reduction

The 21 cm observations were carried out in 1992-1996 with the Nançay 300-m and Green Bank 43-m radio telescopes.

The Nançay radio telescope has a half-power beam width (HPBW) of 3.7' (EW) $\times 22'$ (NS) at the declination $\delta$ = 0$^\circ$.We used a dual-polarization receiver, with a system temperature of $\approx$ 40 K in the horizontal linear polarization and $\approx$ 40 K and 60 K in the vertical linear polarization in the periods 1992-1994 and 1995 respectively. As all the candidate BCGs had known optical redshifts from the 6m spectroscopic observations, we split the 1024-channel autocorrelator into two, covering a bandwidth of 6.4 MHz each, and centered at the frequency corresponding to the optical redshift. In this configuration, each segment covered a velocity range of 1350 km s^-1. The channel spacing was 2.6 km s^-1 before smoothing and the effective resolution after averaging pairs of adjacent channels and Hanning smoothing was $\approx$ 21 km s^-1. The gain of the telescope was 1.1 K/Jy at the declination $\delta$ = 0$^\circ$. The observations were made in the standard total power (position switching) mode with 2-minute on-source and 2 minute off source integrations. Typically, we aim to achieve a rms noise of 2.5 mJy per channel, after smoothing the channel spacing to 21 km s^-1. This leads to a typical integration time of 1 hour on the galaxy and 1 hour on the comparizon field. For BCGs with the faintest HI flux densities, the ON integration time may go up to 3-4 hours. Since the calibration noise diode power has changed significantly over the years, and since its power is frequency-dependent, we have exercised great care in improving the standard Nançay radio telescope calibration method to correct the data for the time and frequency dependencies. This procedure is detailed in Appendix A.

The NRAO 43 m radio telescope has a HPBW of 22'. We use a two channel dual-polarization prime-focus receiver with a system temperature of $\approx$ 20 K in both polarizations. A bandwidth of 20 MHz covering 4225 km s^-1 was used with the 1024-channel autocorrelator split in two. The channel spacing was $\approx$ 8 km s^-1 and the effective resolution after Hanning smoothing was $\approx$ 16 km s^-1. The observations were also made in the position switching mode, with 6 minutes off-source and 6 minutes on-source integrations. The calibration signal levels and telescope pointing errors were checked by observing standard calibrators listed in the 43 m Telescope Observer's Manual. The calibration noise diode temperature was measured in the laboratory before the observing run. The gain of the telescope was 0.29 K/Jy.

The Green Bank data was reduced using NRAO's standard LINE program, while the Nançay data was reduced using the software developed by the telescope's staff. For each integration, the two polarizations which were detected independently were averaged to improve sensitivity. The baselines were generally well fitted by a polynomial of third order or lower and subtracted out.

To check that the data taken from the two telescopes are on the same flux scale, we have observed several galaxies with both telescopes. To within the errors, the Nançay and Green Bank fluxes agree with each other.