ATTENDEES: S. Benson, G. Biallas, C. Bohn, D. Douglas, J. Karn,
B. Legg, W. Oren
Items of Discussion
The twisted quad problem was restated for those not at the commissioning meeting. A summary of the commissioning meeting presentation is attached to these minutes.
Dave Douglas reported that his analysis of the effect of pitch and yaw errors due to twists on the order of 10 mrad showed that, for thick quads, there was a problem due to the fact that the apparent center derived from a quad centering algorithm and the actual center were displaced by greater than 100 µm (this effect basically arises from the thick lens nature of these quads).
S. Benson reported that, by choosing only quads with less than 1 mrad of twist, the twist problem was avoided for the quads mounted on the wiggler bench. This same technique might be used for the other two double quad girders as well. The one rejected quadrupole from the wiggler bench should be used as the third magnet after the wiggler. This is mounted alone on a girder and the electron beam does not have to be accurately centered in it.
Al Guerra has provided a technique for untwisting the quads which reportedly worked quite well with several quads in the main machine. This could be used for quads with very large twists.
Dave Douglas commented that twists much less than 10 mrad but larger than the pitch and yaw spec. of 1 mrad. (say 2 or 3 mrad) are probably OK but that one should measure the quads in the telescopes around the optical chicanes to make sure that none of them are severely twisted. He also remarked that the back leg quads were much weaker than the telescope quads and could be considered as thin lenses. Even a large twist in them would not cause problems.
The problem of aligning the magnetic centers of the wiggler and the quads on the wiggler girder was brought up. W. Oren has noted that, due to transfer errors from the alignment fiducials to the magnet center, one could easily have systematic errors in the quad positions of greater than 200 µm. S. Benson pointed out that one can remove the transfer errors by bore sighting the quads and wiggler (the alignment targets for this are just about ready for signoff). W. Oren conceded that one could probably align the mechanical centers of the quads and the wiggler to ±100 µm but that there was no guarantee that the magnetic centers were aligned to this precision. S. Benson admited that, at this point, the assumption of the mechanical and magnetic centers was just a leap of faith, though there is no good evidence to refute the assumption.
The proposal was made to measured the relative magnetic center
of the quadrupoles and the wiggler using a pulsed wire setup from
Northrop Grumman. At this time is it not known whether this method
is sufficiently accurate to determine the relative positions to
±100 µm but we will explore it. We have to do pulsed
wire measurements on the wiggler anyway to record its pedigree
before it is exposed to radiation. This will allow us to determine
whether it is damaged or not. If the wire is lengthened to measure
the entire girder, it may allow all five magnets to be aligned.
Measure the twist of the remaining 7 quads in the telescopes around the optical chicanes. If they are very large (greater than a couple of mrad) try to straighten using A. Guerra's algorithm. G. Biallas
Arrange to measure the wiggler girder using a pulsed wire. S. Benson
Arrange for bore sighting the wiggler girder. W. Oren.
Summary of commissioning meeting from June 24
In trying to align the double quadrupole girders, a potential problem has
come to my attention. First some background.
The QB and QG quadrupoles are made up of stacks of precision stamped
laminations which define the pole surface and the reference surfaces.
Four grooves exist in the middle of the four laminations.
Each quadrupole has an "aisle side" and a "wall
side" which are determined by the location of the plug strip
to which the electrical connections are made. The plug strip is
on the wall side and should therefore always be on the outside
of the accelerator ring. Note that the CEBAF linac quads were
installed backwards and do not follow this convention but that
all the quadrupoles in the FEL will follow this convention.
The quads are measured upside-down while clamped in a fixture
which contacts the middle of the reference groove on the plug
strip side and two points near the ends of the reference groove
opposite the plug strip.
The alignment fixture mounts on the same three points as the magnet
measurement fixture. The position and angle of the QB or QG is
adjusted and measured by measuring the position and angle of the
alignment fixture. There may be some errors in transferring the
fixture position to the magnet position.
Though the laminations are precisely made they are not well constrained
in all directions. They can flex and slew with respect to each
other. They are welded to bars at eight points which constrains
the motion somewhat but also can create a twist in each quadrant
of the magnet.
Due to the magnet clamps and the good lamination tolerances each
two reference grooves should occupy locations in parallel planes.
There can be a rotation in each plane however. Given these constraints
it is easy to show that the parallelepiped which is formed by
the four reference grooves will be distorted so that it is twisted.
Each "face" defined by four adjacent grooves will be
twisted with respect to the opposite face.
If the support for the quadrupole mirrors the alignment and magnet
measurement fixture, the roll and pitch of the quad should be
set correctly by setting the roll and pitch of the fixture. The
vertical axis of the quadrupole will be an approximately linear
function of the distance along the quadrupole, i.e. there will
be a screw dislocation in the quadrupole.
The quads are quite elastic and so the twist cannot be easily
taken out by trying to straighten them out. According to W. Oren,
straightening out laminated magnets can be done by twisting them
past their elastic deformation limit, but they return to their
original twist after the field is cycled.
The screw error will not affect the beam as long as the quadrupole can be considered as a thin lens. The skew quad moment at the entrance of the quad will cancel the skew quad at the exit. It is clear that for a thick lens the entrance and exit do not cancel. Is this a problem?
(Note: Dave Douglas answers that it is only a problem for a twist
of greater than 10 mrad.)
The yaw angle of the quadrupole is set using the yaw of the alignment
fixture. This means that the top of the quad will be properly
aligned but that half of the twist will still be present in the
quadrupole axis. The amount of this twist is not known at this
time. The pitch angle is set with respect to the aisle side of
the magnet. This means that there will be a pitch error as well,
approximately equal to the yaw error. Could this be a problem?
(Note that this error is present in all the main machine quads
and that the measured field is about an axis which is also off
in yaw and pitch).
It would be nice to align double quad girders by making the two
quads parallel using a mandrill. This will "mis-align"
the quads as far as yaw and pitch goes. Is this a problem? One
could also align the bottoms of the quads and reverse the yaw
error. What might be the result of this?
Some of the comments from the conversation follow(apologies for
errors - stuff was flying pretty fast):
D. Douglas - What is the tolerance for the magnetic center?
S. Benson - +/- 100 µm for the quads around the wiggler.
D. Douglas - What drives that spec.?
S. Benson - By constraining the quad alignment to set a fixed,
known axis through the wiggler, Steve hopes to cut the number
of search variables for the FEL from 12 to 4. This should have
a major impact on "Turn-on" times and will be worth
the effort in the long run.
Q. If the field measured corresponds to the axis aligned to, and it's adequate, why change it? By using different alignment system, may be able to improve the field integral without changing the magnet. But is this really a problem for the bulk of the FEL quads?
A. We don't really know.
Actions- D. Douglas to try simulating the alignment errors Steve is describing. Once we have some analytic answers, we will schedule a meeting for Dave, Steve, J. Karn, C.L. Bohn, G. Neil, W. Oren and
Leigh to decide how alignment should be done.