In order to cover the 3 600 square degrees area contained between 
, four different data sets were
used. The strip delimited by 
and 
was sampled by
the surveys of Weaver & Williams (1973) from 
to
, and Strong et al. (1982) from 
to
. The region outside this band was surveyed by Heiles & Habing \
(1974) and Colomb et al. (1977, 1980). In
Fig. 1 (click here), the observed points in each survey are plotted. Regions of
overlap among the different surveys were used to control the quality of the
match. Since several regions appear quite under-sampled, unpredictable
errors can be produced after the processing and combination of the data.
Therefore, we avoid the analysis of small scale structure. In what follows,
a brief description of the different surveys is given:
Figure 1: Distribution of the observed points in
each survey. Open triangles correspond to the survey of Heiles &
Habing (1974); open circles, to Colomb et al. (1977, 1980); filled circles, to
Strong et al. (1982); and filled triangles, to Weaver & Williams (1973)
a) Weaver & Williams (1973): this atlas
covers a regular grid in 
, with an interval of 
in
longitude and 
in latitude. Observations were performed over
an area of 
, 
, using the 26 m telescope of Hat Creek, Berkeley, with an
angular resolution of 
. The noise level was 0.38 K in
antenna temperature. The relation between antenna and brightness
temperature is 
The
profiles consist of 200 velocity channels spaced in 
\
(except for the extreme channels, not used in the present study, for which
is different), and the
reference velocity, which in general varies for each profile, is referred to
channel 122.
b) Heiles & Habing (1974): observations were also made with the
26 m telescope of Hat Creek, Berkeley, in the region observable from the
local latitude and for 
. In this case, the sampling
was irregular in 
, with 
and
. The southern limit of the sample was
. The noise level was in general less than 0.4 K. The
data are given in antenna temperature, with 
. The profiles, with 100 velocity channels, have a velocity
interval of 
, except for the first eight and last
five channels, for which 
. The reference velocity, 0
in almost all the profiles, is refered to channel 52.
c) Colomb et al. (1977, 1980): observations were taken
with the 30 m radiotelescope of the IAR, with a resolution of 
at 1420 MHz. This survey has been performed with the purpose of being
combined with the survey of Heiles and Habing, giving a complete coverage of the sky at
. The data were obtained by scanning in Right Ascention for
fixed declination, with 
between scans. The velocity
resolution is 2 . The velocity range is about 80 , and the first
channel velocity varies from one profile to the other. The data
are given in antenna temperature, with 
.
d) Strong et al. (1982): observations were taken
with the 64 m radiotelescope at Parkes, with an angular resolution of
at 1420 MHz. The spectral resolution was 0.824 , and the
number of velocity channels, 512. This survey covers the area 
, 
, with a regular
sampling in galactic coordinates with 
and 
. The reference velocity, at the central channel, is 0 ,
except for the strip delimited by 
,
, where this velocity is +50 . The
noise level is 1 K. Due to the rather low signal-to-noise ratio, a
Hanning smoothing was applied, which lowered the velocity resolution of the
profiles to 1.648 and the rms noise level to 0.6 K. The data were
corrected for a systematic shift of 2.6 in velocity (Dubner et al. \
1992). No beam smoothing was applied.
All data were converted into brightness temperature and fitted into one
single field, within the velocity range 
, in 2
velocity intervals, with a uniform grid in galactic coordinates.
Each grid point profile was calculated as the weighted mean among the four
nearest profiles (one per quadrant). In the zone of overlap between the
Strong et al. and the Weaver & Williams surveys, only profiles from the former were used.
The final product was a set of spectral data placed each 
in l
and 
in b. The images were processed using the AIPS software. The
rms noise level of each channel is better than 1 K.