Date: Mon, 03 Jun 2002 11:43:21 +0200
From: Jean-Paul Berthias <jean-paul.berthias@cnes.fr>
To: Gilles Tavernier <Gilles.Tavernier@cnes.fr>,
    Jean-Jacques Valette <valette@cls.fr>, John Ries <ries@csr.utexas.edu>,
    Pascal Willis <willis@ensg.ign.fr>,
    Werner Gurtner <gurtner@ubeclu.unibe.ch>
CC: Laurent Soudarin <lsoudarin@cls.fr>,
    Jean-Michel Lemoine <Jean-Michel.Lemoine@cnes.fr>,
    Alfred Piuzzi <alfred.piuzzi@cnes.fr>
Subject: Concern with new format

Dear colleagues,

There is one issue with the new DORIS format that I want you to be aware of
before our meeting next week in Biarritz. It underlines a difference 
between the old and the new format that I had underestimated until now.

The new export Doris data is generated by

V = c/fbea [(fbea-fsat) - D/dt] = [c/fbea] [-D/dt] + [c/fbea] [fbea-fsat]

where D is th cycle count, fbea is the nominal beacon frequency and fsat is a
polynomial  approximation of the on-board frequency. The only difference in
format is that the beacon frequency in the new format is the nominal value
instead of the pass by pass estimate computed in our orbit determination 
process (used in the old format).

The corresponding processing equation is

V = c (Dfbea/fbea - Dfast/fsat)  + c (1+Dfbea/fbea) * D/dt

where Dfbea (resp. Dfsat) is the difference between the actual beacon 
(on-board) frequency and the value used to generate the data. With the
old format, the beacon frequency offset was always small, and the 
last term could be neglected leading to

V = bias + c D/dt

For the new data type, I thought that this was still true overall until last
week ! Then, coming from a completely different context, J-M Lemoine told me
about station altitude errors induced by neglecting the last term for beacons
with large frequency offsets. Laurent confirmed that they have observed this
effect. So even if neglecting the last factor probably does not impact orbit
determination, it can lead to wrong results for station positioning.

Thus old and new DORIS formats cannot be processed in exactly the same
way. More precisely, the last term should not be forgotten in the new
data processing, while it is of no importance in the old one. For
software which always contained this correction, there is no problem, 
but for others there is a risk which would now translate into wrong
station altitudes. We will have to keep that in mind when dicussing
format issues next week. After all, the only advantage of using
the nominal frequency instead of the pass estimate is to be able 
to recover frequency offsets for beacons ...

Best regards,
Jean-Paul