Figure 1: Configuration of the square Virgo Fields A through D into which each
plate was subdivided. Each field was the subject of a separate scan yielding a
separate 4096 by 4096 pixel grid of transmission readings, though plates were
not readjusted between scans of their different fields. Note that there was
a 192-pixel-wide overlap between Fields A and B, and between C and D. Angular
dimensions of the corresponding fields of view are indicated. Note that the
positions of the peripheral areas containing the sensitometer spot sequences
varied from plate to plate and have therefore not been shown, even though they
were scanned
The plates were digitised by the ROE's Image and Data Processing Unit using its
COSMOS measuring machine in mapping mode with a spot/pixel size of 32 microns.
The central -square field of each plate was subdivided into four
separate scanning fields to ensure that the resulting data files would be of a
manageable size. The configurations of the scanning fields, which were labelled
"A'' through "D'', are depicted in Fig. 1 (click here). Each field yielded a data
array of dimensions 4096 by 4096 2.1-arcsec-square pixels. As the position
of each plate relative to the measuring-machine carriage remained unchanged
throughout the scans of its four different fields, the (x,y) coordinate
systems of the different fields were not rotated with respect to one another;
only the origins were shifted. Transformations between the coordinate systems
of the same field on different plates, did however always require a degree of
both shift and rotation as it was not possible to scan identical regions of
different plates.
In addition to the main scans, the peripheral areas of the plates containing the sensitometer calibration spots were scanned too. As there were two sets of calibration spots per plate, two such scans were needed per plate.
The original scans of plates J4882 and R2936, that had been made in 1984, were found to be unsatisfactory (see Sect. 3.3) and re-scans were made in 1990 with a pixel size of 1.05 (though the resulting higher resolution data arrays were later binned up to the 2.1 pixel size of the earlier scans). Several changes had been made to COSMOS during the intervening period such as increasing the cathode-ray beam strength and upgrading the density-data format from 8 bit to 12 bit. These improvements also increased the machine's dynamic range.
The initial relative-intensity calibrations were based on the sensitometer-spot scans. The UKST is presently fitted with two different step-wedge sensitometers; one sixteen-step device of Kitt Peak National Observatory (KPNO) design and one seven-step device. Originally the UKST carried two identical seven-step devices, but one of them was replaced by a KPNO sensitometer at the end of 1979. The step-wedge combinations and configurations of the plates relevant to this study are listed in Table 2 (click here). The seven-step wedges were illuminated by tungsten light bulbs whilst the KPNO projector uses a quartz-halogen lamp in conjunction with a BG34 colour-correction filter.
plate | scan used | step wedges projected![]() |
U9362 | 1984; 8-bit | N edge:7-step, SE corner:16-step |
J4882 | 1990; 12-bit | N edge:7-step, E edge:7-step |
J9229 | 1984; 8-bit | N edge:7-step, SE corner:16-step |
R2936 | 1990; 12-bit | N edge:7-step, E edge:7-step |
|
Figure 2: The characteristic curve generated for plate R2936 by combining both
of the plate's seven-spot sequences
A median emulsion-density value was obtained for each calibration spot from the digitised scan data. A characteristic curve was then generated for each spot sequence by fitting a polynomial function to plots of log spot density against log relative light intensity; values of the latter having been obtained directly from knowledge of the relative transmittances of the individual wedge components. The two characteristic curves for an individual plate could then be combined by sliding one in log relative-intensity space until the curves' residuals were minimised [in log relative-intensity space]. 3rd-degree or 5th-degree polynomial functions were then fitted to yield the final characteristic curves that were adopted. An example of one of the adopted curves is shown in Fig. 2 (click here).
In order to generate a two-dimensional polynomial surface function to model the
mode sky density levels over
each of the main scans, the central 4080-pixel square area of each 4096
pixel square grid was divided into an array of 48 by 48 square bins each
containing 7225 pixels. A separate histogram of pixel-density values was
constructed for each of the 2304 bins per scanning field. An optimum sky value
was then obtained for each bin by kappa-sigma
clipping (as first described by Godwin 1976 and Newell & O'Neil 1977) of
the data within each bin.
After the initial rejection of pixel values exceeding 3
the typical sky density values (estimated roughly in the first instance by
inspection of the digitised plate-scan data)
the data for each bin were "clipped'' four times, on
each occasion discarding pixels deviating from the sample mean by more
than a specified number of standard deviations: 1
on the first
iteration, 2
on the second and third iterations and 2.5
on
the fourth. The increasing severity of the rejection criteria was necessary
because the sample standard deviations were inevitably reduced by each
iteration. After the fourth iteration the new mean pixel value was adopted as
the best estimate of the sky; on the grounds that it would correspond closely
to the mode value provided that only a minority of the pixels within the bin
were
influenced by stellar or galaxy images. Obviously the latter assumption did not
hold for those bins containing extended galaxy images and so further clipping
of the resulting 48 by 48 array of sky estimates was necessary.
The same kappa-sigma clipping algorithm with the same rejection criteria was
subsequently applied to the 2304 sky estimates in order to preclude those bins
containing the images of extended objects from influencing the final function
used to model the sky. For each of the four scanning fields; a 4th degree
[28-term] two-dimensional orthogonal polynomial function was fitted to the remaining
estimated mode-density values (which corresponded typically to 93% of the
bins).
Individual pixel-density values were converted to relative intensities by means
of the relevant characteristic curve and each was then divided by the local
sky-value estimate (also in the same relative-intensity units) after this local
sky-value estimate had first been subtracted. The subtraction was necessary in
order to remove the night-sky contribution whilst the division served to
counter "field effects'' and "vignetting''; effects which are discussed in
Sect. 7 (click here). The results were 4096-pixel-square arrays of intensity values in
sky units (i.e. normalised to the sky brightness). It therefore remained to
estimate the true sky brightness function for each plate in absolute units in
order to provide a zero point for the relative-intensity scale.
Photoelectric aperture measurements, including simulated aperture CCD photometry of galaxies within the fields covered by the scan data, were used in order to determine the zero points of the relative-intensity scales. Ideally, all of the usable photoelectric measurements would have been compared directly with their photographic counterparts; the latter being obtained by the summation of pixel-intensity values within circles of appropriate radius centred on the image centres. However, owing to saturation of the emulsion density within the cores of the majority of the galaxy images for which photoelectric photometry was available, and to limitations in the dynamic range of COSMOS, such a direct comparison was only possible in a handful of cases. In the vast majority of cases it was therefore necessary to compare intensities within annuli, and so provided the inner boundary of each annulus lay exterior to the saturated core region of the relevant image, the problem of core saturation could be overcome.
A total of 387 U-, 658 B-, 658 V- and 36 -band published
photoelectric-photometry measurements were collated for 65 galaxies within the
field of interest from Longo & de Vaucouleurs (1983)
Longo & de Vaucouleurs
(1985) de Vaucouleurs & Longo (1985) Burstein et al. (1987) and
Poulain
(1988). The B and V measurements were used to derive
values via
the transformation:
as used by e.g. Buta & Corwin (1986).
A radial integrated-light profile was generated for each galaxy image for which
there existed usable aperture photometry in the literature. As the scan data
had already been reduced to a relative-intensity scale in sky units, these
profiles could be computed in mag brighter than the sky. Each sum
of an aperture measurement (in mag) and the modulus of the profile value
(in mag above the sky) at the same radial distance, , from the image
centre therefore provided an estimate of the sky brightness at the position of
the particular galaxy. The radial extent of each image's saturated region (if
present) could therefore be estimated from plots of estimated sky brightness
against
.
Unreliable published aperture-photometry measurements could often be
spotted at this stage: as data-points deviating significantly from the curve
defined by the majority of the others. Outliers were only excluded however, if
the other points had been obtained from several independent sources. Once the
radius of the saturated core,
, had been established for a
plate image, all measurements made with apertures of less than 2
were
culled from the lists of photometry allocated to that image.
Figure 3: Sky-brightness estimates in (rounded to the
nearest one decimal place) for the positions of galaxy images
on Plate U9362. The circle sizes are an indication of the number of
measurements used per galaxy (see key)
Cored sky-surface-brightness estimates were then generated for each
usable pair of published aperture measurements provided that the members of
each pair were separated from one another in by at least 5 arcsec.
The inner and outer radii of the annuli varied according to the availability
of the published measurements. A mean sky
brightness (in
) and a standard deviation on
the mean were then computed in intensity space for each image separately along
with totals of the number of measurements used for that image. A separate plot
was then generated for each plate (see Fig. 3 (click here) for an example) so that
the degree of flatness of the sky could be assessed. Once it had been
established that the skies were flat (i.e. field effects and
vignetting had been adequately compensated for) a grand weighted mean (again in
intensity space) was computed for each plate for all of the annuli considered
for that plate. These means are listed in Table 3 (click here). The annuli were
weighted according to the number of intensity units they spanned. Note that the
actual standard deviation on the mean for Plate R2936 was 0.020
, but
it was felt that a value of
would be more appropriate
because this sky estimate was dominated by a
single observer's photometry (that of Poulain 1988) which could be expected to
be more self-consistent than the same number of measurements made by as many
observers.
In the case of Plate R2936, additional zero-point estimates were derived from
independently calibrated CCD observations of NGC 4551 and NGC 4478 (see
Sect. 3.4). The final zero point adopted for this plate was the weighted
mean of the values given in Tables 3 (click here) and 4 (click here); the value of which turned out
to be .
plate | estimated sky intensity![]() | number of |
(![]() | measurements used![]() | |
U9362 | 21.207 ![]() | ![]() |
J4882 | 22.209 ![]() | ![]() ![]() |
J9229 | 21.875 ![]() | ![]() ![]() |
R2936 | 21.526 ![]() | ![]() |
|
Whilst it was possible to calibrate the U and plates solely with galaxy
aperture photometry measurements obtained from the literature, a different
approach was necessary for Plate R2936, owing to the scarcity of published
R-band, and in particular
(Cousins R) photometry. In this subsection we
describe the reduction of
-band CCD frames of NGC 4478 and NGC 4551
to radial integrated-light profiles that could be used to improve the absolute
calibration. The sky-brightness estimates
that were obtained from these profiles were used to
supplement those estimates obtained using the published
aperture photometry described in Sect. 3.3.
Our CCD frames originated from two different sources. We obtained one usable Jacobus Kapteyn Telescope (JKT) CCD frame of NGC 4551 exposed for 300 s together with frames of flat fields and Landolt (1983) equatorial standard stars from the data archives of the Royal Greenwich Observatory. These observations were made by Peletier and Huizinga over the period 1987 May 22-26. Three additional KPNO 2.1-m telescope CCD frames were kindly provided to us by Reynier Peletier. These latter frames had already been independently calibrated with aperture-photometry measurements corresponding to saturated regions of the galaxy images on the UKST plates (i.e. measurements that could not be used in the calibrations described in Sect. 3.3). Two of the KPNO frames were 300-s exposures of NGC 4551, whilst the third was a 600-s exposure of NGC 4478.
We used essentially the same reduction procedure as described in Young & Currie (1991) and the same software with only minor modifications. In the case of the JKT frames, the illumination of the flat fields was probably not very uniform (Peletier, private communication) and this may be the origin of the small residual non-linear gradient across the sky that had to be subtracted with a second-degree two-dimensional polynomial function even after flat-fielding process. Despite the difficulties encountered in estimating the sky level beneath the JKT image of NGC 4551 precisely, the probable error on the derived sky surface-brightness value was small enough for this estimate still to be extremely useful. The colour equation we derived from the standard-star frames was:
in which represents the total number of counts due to the star,
t is the exposure time in s and z is the zenith angle.
This enabled us to generate a radial integrated-magnitude profile adopting
(B-V)=0.96 from the Third Reference Catalog of Bright Galaxies
of de Vaucouleurs et al. (1991) (RC3).
In the cases of the KPNO 2.1-m telescope frames, after cleaning them, we subtracted the backgrounds (adopting the same background count and sky surface-brightness values as quoted in Peletier et al. 1990) in order to generate radial integrated-magnitude profiles.
Figure 4: Diagrammatic representation of the "coring'' procedure adopted. An
idealised graph of computed sky-brightness versus innermost coring-radius is
shown. See Fig. 5 (click here) for an example of one of the curves derived
experimentally
Figure 5: Computed sky brightness versus innermost coring radius for
the R2936 image of NGC 4551 combined with the JKT profile for that
galaxy. The minor departure from the ideal curve (which should level
off at large radial distances) is presumably due to slight inadequacies
in the adopted background functions
telescope | galaxy [frame] | estimated sky brightness |
(![]() | ||
JKT | NGC 4551 | 21.56 ![]() |
KPNO 2.1 m | NGC 4551 [1] | 21.55 ![]() |
KPNO 2.1 m | NGC 4551 [2] | 21.62 ![]() |
KPNO 2.1 m | NGC 4478 | 21.65 ![]() |
|
A coring procedure as described in Sect. 3.3 was necessary in order to
estimate sky-brightness values for the positions of NGC 4478 and NGC 4551
on Plate R2936. This time however, CCD radial integrated-magnitude data were
used in place of aperture-photometry measurements. Also, in order to estimate
the radial extent of core saturation, , within the plate images, the
radius of the central area excluded from the coring was varied. An ideal
graph of the variation of estimated sky brightness against innermost
coring radius is depicted in Fig. 4 (click here). Four such
curves were
generated (one for NGC 4478 and three for NGC 4551). All of these
curves were reasonably well-behaved, exhibiting only slight negative gradients
beyond
; see Fig. 5 (click here) for an example.
The sky-brightness estimates for the plate could therefore be read
off directly from these plots, and the adopted values are listed in
Table 4 (click here).
Errors in zero points derived from the KPNO frames are likely to be of the order of 0.05 mag but definitely less than 0.10 mag (Peletier, private communication). Considerably smaller errors would normally be expected for zero points derived from the JKT frame as this frame was calibrated with standard-star observations rather than with aperture photometry. However, the flat fielding problems encountered during the JKT reductions reduced the accuracy somewhat. The turnovers in the graphs such as Fig. 5 (click here) yielded handles on the likely sizes of those errors due to inadequacies in the background subtraction. These errors were then propagated together with those originating from earlier stages of the reductions; which were computed according to the same procedure as used in Young & Currie (1991) for the JKT frame, and using a value of 0.07 mag in the case of the KPNO observations. The resulting error estimates are tabulated in Table 4 (click here) together with the corresponding sky-brightness estimates.
Prior to sorting galaxy images from stellar ones, the plate-scan data were
first smoothed with a centre-weighted 3-square smoothing algorithm. This
algorithm applied a Hanning filter using a 3-by-3 kernel. The resulting
smoothing function had a full-width half maximum (FWHM) of 2 pixels.
Initial sets of image parameters were then generated by means of Jon Godwin's
image-recognition and segmentation programme LLLION
(see Godwin 1976 and Dixon 1979 for details). The parameters
generated for each image at this stage included: isophotal magnitude,
reduced radius
of the limiting isophote, mean surface brightness within that limiting
isophote, central surface brightness, surface brightness of the brightest
overlapping isophote (in cases where nearby images overlapped with the one in
question) ellipticity, position angle and the isophote used for the
determination of the position angle.
The surface brightness at which saturation effects (due to the limitations in the dynamic ranges of both the emulsion and of the microdensitometer) became significant was estimated for each plate. This was achieved by the assembly of several composite stellar profiles per plate (by the same method as those described in Sect. 3.6 were). By varying the upper surface-brightness limits on the portions of the stellar profiles to be included in the composite profiles, the onset of saturation could be observed. Composite stellar profiles based only on the unsaturated portions of images yielded plots similar to Fig. 6 (click here), but at a certain point, the inclusion of brighter isophotes caused some data points [from previously excluded saturated images] to appear below the narrow stellar locus. Once the surface-brightness level at which the onset of saturation occurred had been established for each plate, galaxy images with saturated cores could then be identified and flagged.
Figure 6: The composite stellar profile from plate J4882 on which the
star-galaxy sorting was based. The best-fitting eighth-degree polynomial
function (solid line on graph) was adopted as the template for stellar images
Metcalfe et al. (1995) have demonstrated that in previous survey work based on both the COSMOS and APM measuring machines, there is increased scatter in [both isophotal and total] galaxy magnitudes for objects with higher central surface brightnesses. They showed that this effect (which should not be confused with true saturation effects) is almost certainly due to limitations in the machines. They conclude that several precautions ought to be taken to minimise this problem when working with COSMOS or APM galaxy data. All of their concerns have been met during the compilation of the VPC.
Except in the band, between forty and fifty calibrating galaxies
of widely varying central surface brightnesses have been used per plate, enabling us to fix
our U and
zero points with confidence. All of the plates were taken under
conditions of poor to very poor seeing, which should reduce the possibility of scale
errors in our magnitude scales. Also, the brightest galaxy images with unsaturated
VPC photometry tend to correspond to objects that are of relatively low surface
brightness (which abound in the direction of the Virgo Cluster). These images tend
not to have profiles that fall off very steeply near their centroids, and
are thus very much less susceptible to the effects discussed by
Metcalfe et al.
It should also be mentioned that even if some minor [true] saturation effects were
present in the VPC's [integrated] isophotal magnitude measurements,
they would be unlikely to have any significant effect on the VPC's t-system total
magnitudes (see Sect. 4 (click here)). This is because in the t system the central surface brightness
is in fact an extrapolated quantity.
plate | faint limit | bright limit | seeing FWHM |
(![]() | (![]() | (arcsec) | |
U9362 | 25.0 | 20.9 | 2.7 |
J4882 | 25.0 | 21.9 | 3.0 |
J9229 | 25.0 | 21.9 | 1.7 |
R2936 | 24.0 | 20.9 | ![]() |
|
The faintest usable limiting isophotes did of course vary from plate to plate,
even between plates exposed in the same band (on account of differences in the
levels of plate noise and quantization of the intensity levels) though for
consistency's sake the same limiting isophote of 25.0 mag arcsec -2 was
adopted for both plates.
Although it had been decided to obtain as complete a
galaxy sample as possible down to the integrated isophotal magnitude of
18.5 all images down to
20.0,
U25.0
20.5 and R24.0
20.0 were segmented in
anticipation of differences (both systematic and random) between parameters
derived from the different plates. Details of the isophote limits adopted
are listed for all four plates in Table 5 (click here).
Note that although Plate J9229 was made under better seeing conditions than Plate J4882
was (which would normally yield plate scans of smaller dynamic range),
Plate J4882 was quite a dark plate and was not nitrogen flushed. As a result,
the dynamic ranges of the plate-scan data from these two plates were actually
found to be very similar.
Although the tasks of star-galaxy sorting and image selection are often still performed wholly by eye (the entries in the VCC, for example, were selected by visual inspection alone) many different automated techniques have been developed for discriminating between images of stars and galaxies ever since those of Oemler (1974), Godwin (1976) and MacGillivray et al. (1976). These methods all involve isolating images that approximate well to the point spread function and are therefore probably stellar, from those with shallower profiles that are probably not stellar; though it should be noted that none of these automated procedures can be relied upon to yield a pure galaxy sample.
Godwin's method which was followed here assumes that the shape of a stellar
image profile is essentially independent of apparent magnitude except of course
when saturation effects are at play. A separate set of composite stellar-image
profiles was assembled for each plate by sliding over a hundred stellar
profiles (from stars of widely different apparent magnitude) onto one another
in log surface-brightness space, omitting only those isophotes brighter than
the saturation limit. Whilst a composite profile was used to quantify the
point-spread function for each smoothed plate-scan, a single composite [and
unsaturated] J4882 profile was selected
for use as a template in the sorting procedure; see Fig. 6 (click here).
As the original intention was to generate a -limited dataset, a
plate had to be used for the sorting. It was decided to use the J4882
plate because upon initial inspection of galaxies in Field A it appeared that
this plate was slightly deeper than Plate J9229.
This turned out to be a rather unfortunate choice because J4882 was later
discovered to be the inferior plate of the two by quite a considerable margin.
Figure 7: The slope parameter, , plotted as function of
magnitude for all images within Field B (one quarter of the total area scanned)
for which it could be evaluated. All images of
lying
below the dashed line (representing
) were inspected visually
as part of the star-galaxy separation procedure
Carter & Godwin (1979) introduced the image-profile parameter, . This
was defined as the slope of the linear regression of the logarithmic
surface-brightness residuals against log radius. It was evaluated after first
minimising the sum of the residuals, by sliding the image profile onto the
adopted template in negative log surface-brightness space. Stellar images tend
to have
in the vicinity of zero. As galaxies tend to have shallower
profiles, and the residuals were evaluated in the sense: image minus template,
their
values tend to be negative. At very faint magnitudes however,
the separation between stars and galaxies becomes less pronounced. This is
because fainter galaxy images approximate more closely to the stellar template
and the profiles of both stars and galaxies get noisier as the images get
fainter. This effect can be seen in Fig. 7 (click here).
It was decided to inspect visually all images with both
and
on plate J4882 bearing in mind that galaxies with
steeper profiles almost invariably had very starlike appearances [on these
Schmidt plates at least] and would probably escape an independent visual
selection process anyway. A glass copy of plate J9229 was used in conjunction
with a light table and binocular microscope (of maximum magnification 25 times)
for this purpose. The total number of images that were visually inspected was
almost 4000; and of these roughly one third were confirmed to be galaxies.
Stellar contamination of the galaxy sample is almost certainly less than
0.25% (i.e. less than three objects) as only very occasionally was there any doubt
as to whether an object that satisfied the automatic galaxy-candidate selection
process was a star or a galaxy. The number of galaxies with
s in excess
of -0.065 (and therefore omitted from the final catalogue) is not certain.
Nevertheless this number is expected to be very small for galaxies brighter than
the isophotal-magnitude completeness limit (see Sect. 6.2 (click here)).
Equatorial coordinates were generated from the plate (x,y) system by means of the Starlink package ASTROM. 25 reference stars of mB > 10.0 were selected from Heckmann et al.'s (1975) AGK3 astrometric catalogue for this coordinate conversion.
Those galaxy images that were unsaturated on Plate J4882 and which were successfully
resolved from all adjacent images during image segmentation, of which there were over
a thousand, were matched with
their counterparts on the other plates by means of -matrix transformations in
their plate (x,y) coordinates (i.e. allowing for both shift and rotation). The
colour-generating programme KOLORE (see Godwin 1976 and
Dixon 1979) was then
run to compute (
) and (
) equal-area colours together with
corresponding reduced radii, position angles, ellipticities and isophote
information. Means were then taken of parameters derived independently from
each of the two
plates, assigning equal weights to the plates. At a
later stage however, it was discovered that those parameters generated from
Fields C and D of plate J4882 were unreliable due to inadequate
hypersensitisation of those regions. It was therefore decided only to use
those parameters derived from plate J9229 for those two fields.
Also, when an image was affected by a plate flaw, only the scan data from the
unaffected plates were used, resulting in the loss of some information. Luckily
though, only a handful of images was affected. Note that
total colours
were also generated at a later stage (as described in Sect. 4.2 (click here)).
It was not possible to derive reliable image parameters for saturated objects owing
to the loss of information about the image cores. However, as many of the
saturated objects are amongst the most prominent galaxies in the Virgo field,
the majority were already catalogued in the VCC and/or the RC3, whence certain
parameters could be extracted. Some of the objects were not in either catalogue
however, and eye estimates of their apparent magnitudes were
necessary. In such cases the values arising from the segmented plate scans
could only be used as a faint limit, as saturation effects would inevitably
cause the image brightnesses to be underestimated.
Certain galaxy images were contaminated by adjacent stellar images and/or by
other galaxies. The image-recognition and segmentation programme LLLION was
able to separate most overlapping images from one another. However, in 51
cases LLLION could not resolve them cleanly, because the degree of overlap was
so large and/or the disparity in brightness between the objects was so great.
Under such circumstances, apparent magnitude
estimates were based on the values derived from the image segmentation
for the unresolved objects combined, minus an eye-estimated value for
the contaminating object. When the contaminating object was a star, the
magnitude of the star was estimated by finding by eye an isolated stellar image
of similar size and looking up its magnitude in the original unsorted list of
segmented objects. Subtracting contaminating-galaxy images was a less precise
affair, and no attempt was made to separate the closest pairs of double
galaxies, unless the segmentation algorithm had already succeeded to a certain
extent.
Those galaxy images that were saturated on the plate scan of J4882 were
treated separately. In many cases, right ascensions, declinations, NGC or IC
designations, VCC numbers, ellipticities, position angles as well as
colours could be extracted from the RC3.
colours
were derived from the
values by means of Fig. 6b on page 35 of
de Vaucouleurs et al. (1976); the required
apparent radii values having been obtained from the RC3.
values were then estimated from those
values quoted in
the VCC, by subtracting a typical offset of 0.1 mag:
, from
and
. These equations are only intended as provisional
approximations, and are not applicable to dwarfs (which tend not to be
saturated in the VPC).
Radial velocities were extracted from the RC3 whenever possible, but original
sources had to be consulted for most objects known to be receding at velocities
in excess of 15 000 km s-1, as that was the velocity limit of the RC3
galaxy sample. Morphological types were extracted from the VCC whenever listed
therein, but visual inspection of the glass copy plate was necessary for those
objects with not catalogued in the VCC. It was decided
not to attempt the morphological typing of galaxies fainter than
, as beyond this limit the galaxy images were generally too small to be
typed reliably.