We published compilations of HI-data in 1982 and 1990 (Bottinelli et al.
1982; Bottinelli et al. 1990) but the data are regularly updated from
literature. The reduction of raw measurements is the same. The 21-cm line
widths are reduced to two standard levels (20% and 50% of the peak)
and to zero-velocity resolution using the following formula:

where *ws*(*l*,*r*=0) is the standard 21-cm line width at level *l*=20 or *l*=50,
while *w*(*l*',*r*) is a raw measurement at a level *l*' made with a velocity
resolution of *r* . The constants *a*, *b* and *c* are (Bottinelli
et al. 1990): *a*=0.014, *b*= -0.83, *c*= -0.56.

The resulting standard 21-cm line widths *ws*(*l*=20,*r*=0) and *ws*(*l*=50,*r*=0)
are corrected for systematic errors by intercomparison
reference by reference (program INTERCOMP, Bottinelli et al. 1982) leading to
standard widths *w*20 and *w*50 and their
actual uncertainties *sw*20 and *sw*50 respectively.

*w*20 and *w*50 are used to calculate the log of the maximum velocity
rotation following the expression.

where incl is the inclination (in degrees)
between the polar axis and the line of sight
calculated from the classical formula (Hubble 1926):

where log*ro* = 0.43+0.053.*t*, if (or log*ro*=0.38 if *t*>7),
has been obtained from the most flattened galaxies.

is the weighted mean of the logarithm of
the line widths *w*20 and *w*50 corrected for internal velocity dispersion.
The adopted weight of level 20% is twice the
weight of level 50% because it is less sensitive to the definition
of the maximum and also because it corresponds to larger fraction of
the disk.
The correction for internal velocity dispersion is taken
according to Tully & Fouqué (1985).

where *w* is either *w*20 or *w*50 and
, assuming an isotropic distribution of the
non-circular motions and a nearly
Gaussian velocity
distribution (i.e. *k*(20)=1.96 and *k*(50)=1.13).

Mean maximum velocity rotation log*vm* is available for 6415 galaxies,
from 34 436 individual measurements *w*20 or *w*50.

The actual uncertainty on log*vm* can be approximated by
(For the detailed calculation see Bottinelli et al. 1983):

where *sw* and *w* are used for (*sw*20 or *sw*50) and (*w*20 or *w*50),
respectively.
The histogram of *s*log*vm* is presented in Fig. 9 (click here).

**Figure 9:** Histogram of the actual uncertainty on maximum velocity rotation
log*vm*

A preliminary compilation of central velocity dispersions log*s* was
published in 1985 (Davoust et al. 1985) and included in our database.
This compilation has been regularly updated from literature (including
compilations made by Whitmore et al. 1985; McElroy 1995;
Prugniel & Simien 1995).
Measurements from various references have been homogeneized using the
INTERCOMP program (Bottinelli et al. 1982).
The mean central velocity dispersion log*s* is available for 1816 galaxies
resulting from 3402 individual measurements.
The actual uncertainty slog*s* in log scale is shown in Fig. 10 (click here).

**Figure 10:** Histogram of the actual uncertainty on central velocity dispersion
log*s*

In Fig. 11 (click here) we present the completeness of kinematical parameters
log*vm* or log*s* in comparison with the total completeness curve.
The completeness is fulfilled up to about *m*=12.0 mag.

**Figure 11:** Completeness curve for *m*. The completeness is satisfied up
to the limit (solid line).
This limit drop to if we impose that the
radial velocity is known (dashed line), and to *m*=12.0 if we impose that
either the maximum velocity rotation or the central velocity dispersion
is known (dotted line)

web@ed-phys.fr