At yesterday's low-emittance tuning meeting it was pointed out
that the "model-design" plots for the dispersion correction do not
show the effectiveness of the correction. So I was charged with
producing the "model-data" plots for the correction.

Indeed it turned out that the correction I was showing only marginally
improved the dispersion anomaly, reducing it from about 80 cm to about 60 cm. However, by adjusting the weights to emphasize the dispersion
over the phase and the k1 quad values, it is possible to correct most
of the anomaly, reducing the peak difference to about 30 cm.
In particular, the quads can be used to eliminate the
overall curve in the anomaly. They do not correct the short-distance
'alternating bpm' effect which is a general feature of the dispersion
function itself. The phase function is preserved at the level of 10 degrees. The biggest player is quad 22W.

Here are the plots and text files for the correction using all quads:
Before correction:
http://www.lepp.cornell.edu/~critten/cesrta/let/24apr08/mode00120184.gif
After correction:
http://www.lepp.cornell.edu/~critten/cesrta/let/24apr08/mode00120186.gif
Weights, optimizer result, and quad changes:
http://www.lepp.cornell.edu/~critten/cesrta/let/24apr08/allquads.txt

The correction using only quads 10w-30w works just as effectively.
The biggest player is again quad 22W.
After correction:
http://www.lepp.cornell.edu/~critten/cesrta/let/24apr08/mode00120200.gif
Weights, optimizer result, and quad changes:
http://www.lepp.cornell.edu/~critten/cesrta/let/24apr08/quads10w-34w.txt

For future reference, I also include the CESRV script itself:
http://www.lepp.cornell.edu/~critten/cesrta/let/24apr08/hqdisp.in

-- Jim