From Pascal Willis 
 
 Dear Jean-Paul,
 
 following my latest message, I have extended the previous study to
 other DORIS satellites to verify if the problem that I saw yesterday
 was inherent to the DORIS data or only present in the DORIS/JASON
 data.
 
 Here are the new results (each satellite computed independently,
 using all available data at CDDIS in 2002) :
 
 satellite       R_OFF           X_OFF           Y_OFF
             (toward Earth)
 
 jason           -10.5 +- 2.2    -1.7 +- 4.0      0.5 +- 3.6
 topex            -2.4 +- 2.1     0.1 +- 4.8     -1.8 +- 3.2
 
 spot2             1.9 +- 1.7    -1.2 +- 2.1     sigma > 1400
 spot4             1.5 +- 1.6    -1.6 +- 2.0     sigma > 1400
 spot5             3.2 +- 1.2     0.7 +- 2.2     sigma > 700
 
 envisat           2.2 +- 2.0     1.6 +- 2.8     sigma > 1400
 
 To compute the mean, I have only used the days for which the
 parameters were correctly determined, estimating R_OFF, X_OFF and
 Y_OFF simultaneously in the same runs.
 
 As you can see, all estimations except the JASON R_OFF parameter are
 within the error bars of the estimation and compatible with a 0.0
 value. At a lower level, there may be a small value for all SPOT
 R_OFF parameters around + 2 cm (towards the Earth). As stated in the
 previous message, the JASON R_OFF data also present a clear 120-day
 that would need to be confirmed when additional data will be
 available.
 
 So something looks strange in the DORIS/JASON data (or in the way I
 do my current estimation). This is not a small problem as it can
 shift the orbit by almost a constant value of 3 cm in the radial
 component for JASON. We absolutely need to investigate this aspect
 for GPS and DORIS independently, otherwise we would all get
 compatible JASON orbits (within 1 to 2 cm) but the difference with
 the "true" orbit could be between 1 to 3 cm, using either GPS or
 DORIS systems.
 
 I would be very much interested if other groups would try to do the
 same type of estimation to confirm these puzzling results.
 
 In my opinion, a similar effect could be artificially introduced in
 the DORIS data if the satellite long-term frequency drift would not
 be properly corrected in the CDDIS data. It is probably not in the
 time-tagging as this is controlled with the laser data but it could
 be in the measurement itself. Could this be a possible lead to
 investigate?
 
 Best regards
 Pascal