.

A short dipole is being decollared and re-assembled with different thicknesses of the midplane insulation and of pole shims to measure the influence on field quality. The experiment started in October 2002 under the supervision of D. Tommasini, and is being carried out by N. Dalexandro, H. Kummer, G. Molinari, G. Trachez. Strategy of the test has been defined by L. Bottura, H. Kummer, W. Scandale, E. Todesco, D. Tommasini. Simulations to evaluate the influence of these changes on field quality have been carried out by P. Ferracin and G. Gubello through a coupled magneto-mechanical model. An Excel file (you need a nice password) is constantly updated with the results of the experiment.

A summary of the results obtained so far:

Reproducibility of the magnetic field harmonics after a 'soft' decollaring (extraction of the locking rods without coil disassembling). Two soft decollarings have been performed and then magnetic measurements have been taken (test II and III), plus the initial state (test I). Variability of multipoles due to soft decollaring: c1 less that 2 units, b3 less than 0.3 units, b5 less than 0.04 units, b7 less then 0.02 units. These variabilities are small enough to ensure a precise measure of the effect we are looking for (3 units of b3, 1 units of b5, 0.3 units of b7).

Reproducibility of the magnetic field harmonics after a 'hard' decollaring (extraction of the locking rods and coil disassembling, test IV). In one aperture the variations are rather small (0.1 units of b3, 0.1 units of b5, 0.01 units of b7). In the other one, pretty large variations are observed (0.5 units of b3, 0.3 units of b5, 0.05 units of b7). The collared coil is measured ten days later (test V), and is capsized (test VI), giving the same results. A second hard decollaring (test VII) is done. In this case the variability of multipoles is small in both apertures (0.2 units of b3, 0.05 units of b5, 0.01 units of b7). These variabilities are small enough to ensure a precise measure of the effect we are looking for (3 units of b3, 1 units of b5, 0.3 units of b7).

The first two tests (0.1 mm or 0.2 mm more on the midplane inner layer) show a good agreement with the models (see file). In this case we can use either a rigid model (assuming that the coil is absorbing all the additional insulation, keeping the same collar shape), or a realistic model that takes into account of the additional deformation of the coil and of the collars (ANSYS). This second model agrees measured sensitivities within 7% for b3 b7 and within 20% for b5. Some variability is observed in the results of the rigid model, depending on the hypothesis used for compressing the coil; they all anyway feature a similar agreement with experimental data. The linearity of the effect on the multipoles is very well verified.

The third tests (0.1 mm more on the midplane inner layer) shows only a fair agreement with the models (see file): one finds 0.95 units of b3 instead of 0.65; we remind that the measured maximal variation of b3 after a decollaring is of 0.2 units. For the b5, the situation is slightly worse, since we measure -0.02 units against an expected value of -0.10 units, and the reproducibility is of 0.05 units. Summarizing, in must be noted that the expected low influence of the outer layer midplane insulation thickness on multipoles is confirmed by measurements. With this low effect, the influence of multipole variations due to recollaring can strongly affect the measure. Indeed, the fair agreement between model and measurements is not justified by the reproducibility due to one recollaring. We therefore plan to decollar the magnet and go back to the nominal lay-out to measure our starting point again.

The collared coil has been reassembled with nominal insulation on the midplane (test XI). With respect to the data of the nominal setting measured in test VII we find a difference of 0.5 units of b3, 0.1 units of b5 and 0.04 units of b7. This change of multipoles in the baseline is enough to explain the bad agreement between model and experiment quoted in the previous paragraph. Indeed, if one considers the difference between the new baseline and the case with 0.05 mm more on the midplane, we find sensitivities that are in good agreement with the model (25% for b3, and less than 5% for b5 and b7).

The collared coil has been reassembled with 0.05 mm more on the midplane inner layer and with 0.05 mm less on the pole, inner layer (same prestress). No change has been performed on the outer layer. Results are in agreement with model within 20% for b3 and b7 and within 10% for b5 (see file).

The collared coil has been reassembled with 0.05 mm more on the midplane and with 0.05 mm less on the pole, on both layers (same prestress). Results are in agreement with model within 20% for b3 b5 and b7 (see file). We observe a systematic overestimate of 20% of the model with respect to the experiment.

The collared coil has been reassembled with the previous lay-out plus 0.125 mm more in the midplane, both on inner and outer layer, glued on one side only of the ground insulation. This is the solution that will be implemented in long magnets. Results show that there is a much higher effect on main field (7 units insted of 3 units) and, with respect to the rigid model, the effect on b5 is 50% less. If the model with deformations is used, b3 b5 and b7 are shifted of 15% to 25% less with respect to the model. Data can be found here.

(finished, total of 14 collarings, thanks to Hans, Gilles, Gianni, Noel, Davide ... and Luca)

Nine collared coils of series dipoles are being assembled with additional midplane insulation of 0.125 mm. Updated results can be found here.