Once the errors have been assigned a first turn steering algorithm (``threader'') is used to find the initial closed orbit. After setting the sextupoles to 50% of their strength a closed orbit correction is performed which is based on the information of beta functions and phases for the ideal optics. This is followed by another correction loop at full sextupole strength until the monitors have zero readings.
For the orbit correction two schemes are considered. One is based on the Singular Value Decomposition (SVD) algorithm. The other involves interleaved three corrector bumps ``sliding'' around the machine.
The global SVD scheme has the advantage of being able to handle an unequal number of monitors and correctors in the case of faulty monitors and/or saturated correctors and is therefore very flexible. On the other hand the SVD scheme requires a good knowledge of the linear machine optics in order to determine the inverse of the corrector/monitor correlation matrix A-1.
For the envisaged monitor and corrector layout (72 monitors and 72 correctors at the same locations in both planes) (see Figure 1) the SVD and the sliding bump orbit correction scheme converge to the same correction state. This can be explained by the fact that A-1 is a sparse tridiagonal matrix containing the kick ratios of interleaved three corrector bumps. It should be noted that the properties of A-1 have implications on the implementation of the fast global orbit feedback [4].
m (zero monitor readings) are observed in both planes. As an example Figure 2 visualizes the mean, rms and maximum orbit in the vertical plane. The maximum corrector kicks needed are 50% below the design maximum of 
1 mrad. In the vertical direction about 20% more corrector strength is needed than in the horizontal plane although the rms horizontal kick is about 30% larger. This can be explained by a 50% less efficient correction in the vertical plane.
