To: Distribution 17 September 97

From: Martin Nordby

Subject: IR Engineering and Physics Meeting Minutes: 12 September 97

Andy Ringwall presented the Q1C rotating magnet design and analysis. The design was suitably advanced in very good shape, that we agreed this could be considered its Final Design Review. Q1C is made from two slices of Q1B-type rings of SmCo, potted in brass collars which rotate. By counter-rotating the two Q1C slices, the total integrated gradient of Q1 can be changed +4% to –12% of its nominal. Rotation is done by hand-driven worm gears, which allow full 360 degree range. The worm has a 96:1 gear ratio, which prevents back-driving, and allows for good precision of rotational motion. Magnetic thrust loads are taken by a carbon-graphite thrust bearings which are sandwiched between the rotating collar and the outer housing. Rotational torque is taken by the worm and pillow block. The two Q1C rings are mounted off the back end of the Q1 magnet after Q1 is assembled and the threaded rods installed and torqued.

The worm drive is turned manually through slots in the Support Tube, using a custom dowel tool. This engages holes in the worm shaft, and allows 0.75" of rotation for each re-set of the tool.

3-D Amperes analysis by David Humphries shows that the maximum magnetic thrust load between magnet slices is +/- 1400 lbs. Maximum magnetic torque is 186 ft-lb. Added to this is the torque due to rotational friction, given the expected friction coefficient of 0.2. This produces a max torque on the worm gear of 280 ft-lb. Tangential force on the worm due to this torque is 560 lb, producing a shaft torque of 330 in-lb (40 lbs on an 8" long wrench).

The out-of-plane magnetic forces on a slice can bow it, which deflects its outer collar. This unevenly loads the thrust washer, with the estimated peak bearing stress of 2000 psi. This is well below the material limit of 20,000 psi. These forces are worsened by the interaction with the 15 kG solenoid field, which produces a 470 in-lb torque (peak) on each block. Given a conservative mechanical model of slice bending due to this torque, the peak bending stress in the potting epoxy is 3000 psi around the outside collar. This is less than half the yield stress for epoxy of 8000 psi.

Action items arising from this review:

--Determine resolution needed for rotational alignment of the rings (Mike Sullivan, Andy Ringwall).

--Look into adding a clamp to prevent back-driving of worm drive (Andy Ringwall).

--Develop method to read rotation angle while tuning (Andy Ringwall).

James Osborn reported on conclusions of action items stemming from the Final Design Review.

First, James went to single-turn cooling to avoid uncomfortably high exit temperatures. He also is using Syn-Flex hose everywhere, with no ceramic manifolds needed. Bend radii of the hoses are larger than the manufacturer’s recommended minimum. Manifolds are made from stainless steel bar, and the busses are water-cooled 0.25" copper conductor.

The in- and outlets on the underside of the magnet will connect to hard-plumbed stubs, so the hose can be replaced without removing the magnet.

Finally, James plans to mount klixons at both ends of each coil circuit, to protect against possible reverse flow, as well as normal low-flow problems. Klixon tests show that the small klixons trip quickly enough to respond to any fault scenario, as long as there is water in the magnet.

James is also building the flower-petal shaped mirror plate proposed by Stefan Mikhailov to passively compensate for the induced octupole from the BaBar solenoid field. Since this produces other harmonics, the magnet must be tuned with this mirror plate in place (it can not be added later without re-tuning the magnet). We must decide which mirror plate to use before tuning begins.

James Osborn and Jack Tanabe presented results of analysis of expected harmonics, given the as-built tolerances of laminations and coils. This responds to concerns about the effect of out-of-tolerance lamination profiles, and provides a first look at tuning methods for the final magnet.

Q2 laminations will be laser-cut, but the tolerances of prototype lams is twice what is specified. James’ analysis showed that +/- 0.0005" tolerance on pole tip profiles is needed, but the laser cutting was producing +/- 0.001"-0.0015" tolerances. James and Jack have been investigating the effect of this tolerance on magnet quality, and how to deal with it. They used Halbach perturbation theory to both predict expected harmonics produced by the errors, and to investigate solutions. This method has been used to tune the Q4 magnets, LER quads, and the ALS quads. The Halbach coefficients are used to correlate X, Y, and rotational perturbations of the pole tips to changes of any multi-pole.

To understand the effect of as-built tolerances on magnet quality, CMM measurement data of the lams was fit to the hyperbolic pole tip shape. This was done after the data was conformally-mapped into dipole space, so the fitting was to a straight line. The fit was very good, with data fitting a line to better than 0.001". All prototype lams show systematic errors in the shape and orientation, showing that the poles flare towards each other. The fit curve is then used to produce the perturbation coefficients, which indicate that the as-fabricated lams would produce 2-3 x 10^-4 harmonics for the lower-ordered multipoles. The n = 6 term dominates, showing that the poles are effectively at too small a radius.

This analysis method was reversed to look at the effect of trimming the field quality by translating or rotating the core half. Rotating a core half produces all odd normal harmonics, while sliding in X produces even skew, and moving up/down in Y produces even normal harmonics. This trimming method can reduce almost all harmonics to less than 10^-4. Coupling this with chamfering the ends will provide enough knobs to tune the magnet.

The effect of mis-alignment of the coils was also investigated. Since the magnet steel is nowhere near saturation, the mis-alignment effect can be studied by looking only at the difference-field produced by moving the coil, without modelling the coil itself. This assumes the magnet steel behaves linearly (which it should).

Seven coil motion scenarios were investigated. Below is a summary of the cases, and the effect on magnet harmonics:

--Case #1: Moving all 8 coil packages radially inward. 0.005" motion produces 1x10^-4 n=6 harmonic, with other even harmonics less than 5x10^-5.

--Case #2: Moving all 4 coils transversely towards the vertical plane. 0.005" motion produces 1.5x10^-4 n=6 term, and other even harmonics less than 1x10^-4.

--Case #3: One coil each on top and bottom moved transversely together. 0.005" motion produced 2x10^-4 b5 and -1x10^-4 b3. Both of these cannot be trimmed out simultaneously by displacing the core, so this particular placement error scenario must be avoided or corrected.

--Case #4: Moving one out of four coils transversely. 0.005" motion produces 1x10^-4 b1 and b3 of opposite sign. This produces essentially half the effect as Case #3.

--Case #5: One out of eight coils windings moved. 0.005" motion produces 1x10^-5 level harmonics. Not a problem.

--Case #6: Single conductor on mid-plane moved inside coil pack. 0.005" motion produces 1x10^-5 b6 and other terms. This accounts for most of the effect of Case #5, showing that the first mid-plane winding in each coil accounts for essentially all harmonics produced by displacement of the package.

--Case #7: Single conductor next to mid-plane conductor. 0.005" motion produces 1x10^-6 harmonics. Not a problem.

The first production coil has been made, with placement accuracies estimated to be less than 0.005". This, and the perturbation analysis show that we are within reach of producing the 10^-4 accuracy needed, and that continued attention to fabrication accuracy and coil placement will yield magnets which can be tuned to the quality needed.

These minutes, and agenda for future meetings, are available on the Web at: