access to data in
https://suratlant.locean-ipsl.upmc.fr/AX1-xbt20002015.tar.gz
The validated and 'cleaned' XBT profiles are in section files,
such as xbt122010.asc,which sequentially list profiles
with a header 
 2010 12 24  5  6  59.567 -42.867  487
including date (year, month, day, hour, minute),
position (latitude and longitude) and the number of lines (n)
in the profile.
then n records of estimated z and T
     6.7    3.99
...

In 2015, FastDeep have also been dropped with a different equation
(as recommanded by Sippican)
in xbt012015.asc  with first level at 6.2 m
in xbt042015f.asc and xbt062015f.asc with first level at 9.5 m

All other profiles are from  Deep Blue/T7 probes
As far as we know, all the Deep Blue/T7 profiles 
are reported using Hanawa et al (1995) fall rate equation


No bias-corrections have been applied on fall rate or temperature.

Profiles in files with names ending as f.asc presented excessive noise
and profiles have been filtered: 
- over 11 successive points, the average of the three closest
values to the median is retained and attributed to the mid-point depth.
- data provided as 
z,t,err
where err is the largest differences between the
three points averaged, thus providing an estimate of residual noise 
in the filtered profile

Data have been collected using a
Protecno EBST-CLS-ARGOS unit until 06/2007
and a Devil unit since 11/2007