The optical survey of BTS was based on a set of long (typically 2 hour) exposure, fine-grain IIIaJ emulsion Palomar Schmidt plates which were systematically inspected for all dwarf galaxy-like images. A total of 32 "dwarfs'' and dwarf candidates were found in the region of our blind HI-survey strip, 23 of which were judged to be "members'' and possible members of the CVn cloud. The morphological types of these objects are almost equally distributed between very late-type spirals (Sd-Sm), magellanic irregulars (Im), early-type dwarfs (dE or dS0), and ambiguous types (dE or Im; for dwarf morphology see Sandage & Binggeli 1984). Not included here are bright, normal galaxies like Sc's, which are certainly bound to be rediscovered by the HI-survey.
The "depth'' of the BTS survey is given through the requirement that
the angular
diameter of a galaxy image at a surface brightness level of 25.5
B mag arcsec-2 had to be larger than . This completeness
limit turned out to be close to,
but not identical with, a total apparent blue
magnitude
(cf., Fig. 1 of BTS). For galaxies closer than about
Mpc
(or
kms-1), i.e., the distance range of the CVn cloud,
in principle all galaxies brighter than
should be included in the catalogue; out to
Mpc (or
2300 kms-1), the volume limit of our survey, this value is close
to
. It should be
mentioned that these limits are valid only for "normal'' dwarf galaxies which
obey the average surface brightness-luminosity relation (the "main sequence''
of dwarfs, cf., Ferguson & Binggeli 1994). Compact dwarfs, like BCDs and
M 32-type systems, can go undetected at brighter magnitudes, though they seem
to be rare in the field at any rate.
What HI detection rate could be expected for such an optically selected
dwarf sample of the mentioned optical depth? Our blind HI survey was
designed to be maximally sensitive for nearby (V < 1000 kms-1), low HI mass
dwarf galaxies with HI masses of a few times 107 . Assuming
the complete sample of
dwarf irregular galaxies in the Virgo cluster outskirts as representative,
an HI mass of 107
corresponds - on average - to a total
blue luminosity,
, of
(Hoffman et al. 1988), which
in turn roughly corresponds to
.
Based on the optically defined completeness limit, the depth of our
blind HI survey is comparable to the optical depth of the BTS survey
for "normal'' gas-rich dwarfs.
Thus our blind survey would uncover only new galaxies if the faint end
of the HI mass function is rising steeply and/or if there exists
a population of LSB dwarf galaxies with high
/LB ratios which
fall below the optical completeness limit of the CVn survey.
As it turned out (see below), and in contradiction to what was found in a similar survey in the CenA group region (Banks et al. 1998), no extra population of extremely low-surface brightness (quasi invisible) but HI-rich dwarf galaxies was found in the volume surveyed here.
During each observation, the flat telescope mirror was first tilted towards the target declination and the focal carriage was then blocked in place on its rail, near the middle of the track where the illumination of the mirror is optimal; tests in windy conditions showed that its position remained stable to within a millimeter or so, quite satisfactory for our purpose.
The procedure required fixing the telescope pointing coordinates each day to point at a chosen starting coordinate in the sky as tabulated in Table 1. These coordinates for the start of the scan are given for the 1950 equinox. But since the observations were made over the period of January 1996 to Novembre 1996, the observed survey strips lie along lines of constant declination of epoch 1996.5. In order to accumulate integration time and increase the sensitivity along the survey strips, each strip was scanned 5 times on different days.
We adopted a search strategy with a full 4' beamwidth sampling rate in R.A. (integrating 16 s per spectrum, followed by an 0.3 s second read-out period). The 1024 channel autocorrelation spectrometer was divided to cover two slightly overlapping 6.25 MHz wide bands centered on V=324 kms-1 and V=1576 kms-1, respectively, in two polarizations. Since each declination strip was traced 5 times, and there were 17 declination strips observed (spaced by the full 22' beamwidth in declination), a total of more than 250 000 individual spectra were recorded in the course of the survey phase of this project. A complete strip obtained on one day consisted of 40 "cycles'' with each cycle comprising 20 integrations of 16 s. In many instances, the telescope scheduling prevented us from obtaining the full number of cycles, as is summarized in the Table 1, where the survey's sky coverage is listed with "Dec'' indicating the centre declination of each strip and the number of cycles obtained on each day is listed.
(1) The raw spectra were converted from the telescope format to binary compatible files for processing in the ANALYZ package (written and used at the Arecibo Observatory).
(2) Using ANALYZ and IRAF, the data from
each day's driftscan were formatted into an IRAF image format, one
image per day, with successive spectra filling successive
lines in the image. In this way, each of the 85
driftscans in Table 1 could be viewed and manipulated as a single
data block. The technique had been used earlier in the Arecibo survey
reported by Sorar (1994) (see also
Briggs et al. 1997;
Zwaan et al. 1997).
(3) The first important processing step is calibration of the spectral passband. This was accomplished by averaging all the spectra from a single day's driftscan (after editing spectra corrupted by radio interference, strong HI emission from discrete galaxies, or instrumental problems) into one high signal-to-noise ratio spectrum. Each line of the data image was then divided by this "passband''.
(4) The 5 passband-calibrated data images from each declination strip were blinked against each other in order to verify system stability. Since continuum sources appear as lines in these images, the timing and relative strength of these lines confirm the timing of the integrations throughout the driftscan and the gain stability of the receivers. Examples of these images are shown by Briggs et al. (1997, Fig. 1).
(5) Continuum substraction was performed by fitting a quadratic baseline separately to each line of the data image, after masking out the portion of the spectrum containing the local Galactic emission. Further editing was performed as necessary after inspection of the continuum-subtracted images.
(6) The data images for each declination were averaged, the spectral overlap removed, the two polarizations averaged, and the data Hanning smoothed in the spectral dimension to remove spectral ringing. This resulted in a velocity resolution of 10 kms-1 and an rms of typically 10 mJy.
(7) At this stage, the calibrated data could be inspected and plotted in a variety of ways: Simple image display and optical identification of galaxy candidates (accompanied by full use of the image display tools in IRAF and GIPSY for smoothing and noise analysis), plots of individual spectra with flux density as a function of frequency, or display as a three dimensional cube of R.A., Dec. and velocity.
Visual image inspection is highly effective at finding galaxy candidates,
particularly in uncovering extended (in postion as in velocity spread)
low-signal features. But
vigorous application of a statistical significance criterion is necessary
to avoid selection of tempting but not statistically significant
candidates. In this survey, the threshold was set at , which
led to a large number of non-confirmed features once the follow-up
observations were made (cf., Sect. 3.2.2).
Copyright The European Southern Observatory (ESO)