Modular coil design (Williamson)
	Vacuum vessel design (Brown)
	Separatrix study (Reiersen)

 

	Current Sept., 2001	PVR March, 2001	Comment on current design
Number of modular coils / coil shapes	18 / 3	21 / 4	Lower cost, better access, more consistent with required physics properties and flexibility.
Symmetry planes containing coils	none	v = 0	Allows tangential NBI access without extended coil.
Number / location of TF coils	18 / centered on modulars	21 / between modulars.	Better access.
Number of PF coil pairs	5	4	Provides required physics flexibility, Ohmic volt-seconds.

Everyone was on board with these changes.  The issue that immediately surfaced was defining the geometry of the reference coil designs.  There are three coil sets in the reference design: a modular coil set, a TF coil set, and a PF coil set.  At present, we do not have a buildable modular coil design.

Strickler reviewed progress in the modular coil design.  He has been using 18 background TF coils and the 3 ring coils (no solenoid coils) and has achieved about the same error levels as with a 1/R background TF and 5 PF coils (ring coils plus solenoid coils).  Strickler incorporated a new algorithm into CoilOpt.  The algorithm calculates the vector from a point on a coil to the nearest point on the plasma.  It calculates another vector from the same point on a coil to the nearest point on the neighboring coil.  The angle between these two vectors is calculated.  A penalty function is calculated that depends on the sine of the angle.  If the angle is small, it means the coils are being jockeyed radially to increase their separation - something we want to avoid.  If the angle is close to 90º, then the coils lie on a winding surface that conforms to the plasma - something we desire. Initial tests showed that the new algorithm improved the minimum angle a bit.  Strickler is continuing to test this new algorithm.

Other algorithms that have been discussed in the past include a direct calculation of current density.  Nelson has provided an algorithm to calculate this metric.  Strickler reported that this has not been incorporated yet for two reasons.  Firstly, the priority has been to establish a reference coil design.  Secondly, the existing algorithms for coil-to-coil and coil-to-plasma spacing should be able to achieve the same effect.  There was a strong consensus that establishing a reference coil design should be the first priority.

Strickler is continuing to investigate coil designs that are basically a continuation of the 0831 family of coil designs.  These are 18-modular coil designs with no coils on the symmetry planes.  Background TF coils carry current in the same direction as the modular coils.  The background TF coils were allowed to carry different currents.  Strickler should determine if this flexibility (unequal currents in the TF coils) is important because it carries a substantial cost penalty - independent power supplies and separate busing for each coil type.

Williamson discussed progress and issues in modeling the 18-modular coil options.  Williamson presented the 0917b1 modular coil design.  The minimum radius of curvature was just under 9cm (without manual correction).  The minimum coil-to-coil spacing was 10.7cm (although the spacing from coils to their mirror images across symmetry planes was not shown).  Williamson showed the Pro/E models of the windings.  His pictures clearly show the persistent problem that we have in the trough region, that is that the windings overlap.

The winding surface is typically quit lumpy in the inboard region.  Williamson discussed his efforts to manually smooth the winding surface by making radial cuts and reducing the extreme indentation.  He was pessimistic that the existing coils could simply be projected to this surface.  However, Strickler indicated that the revised winding surface could be approximated in Fourier harmonics and put in as a fixed surface in CoilOpt.  CoilOpt could then re-optimize the modular coil trajectories and currents.  This appeared to be a reasonable path forward to alleviate LOCAL concerns, especially coil-to-vessel interferences.

Williamson also discussed manual changes to the modular coil geometry on the outboard side.  He moved the M3 coil in the 0831 coil design to provide more space for tangential NBI.  Strickler indicated that this should not have a significant effect at the plasma.  This appears to be a reasonable path forward to alleviate LOCAL concerns with NB interferences.

Tom Brown then discussed progress he and Mike Cole had made in modeling the vacuum vessel.  He showed an expanded that that featured much larger radii of curvature.  To the eye, it looked like it would be easier to form the vessel.  Also, it looked like it could accommodate a larger plasma volume.  Additional modifications are planned to alleviate NB interferences and to assure adequate space for inboard RF launchers.  Brown pointed out another persistent problem - coil-to-vessel interferences.  These are clearly visible in the radial cuts shown below.  Coil-to-coil interferences were also apparent in Brown's model of the 0831 modular coil set.

Reiersen presented a study of the separatrix topology in the vacuum configuration.  This was work first done by Georgiyevsky and Rudakov and later confirmed by Brooks.  The methodology used was the same employed for divertor studies on Uragan-3M.  Field line tracing codes were used to trace field line trajectories.  The field lines were followed up to 300 field periods (~1km) or until the field lines were lost, i.e. stopped progressing toroidally.  The results were qualitatively similar to Uragan-3M:

 

	A well-defined separatrix (the boundary between field lines circulating toroidally and those that do not) appears to exist and is defined by the iota=0.6 (n/m=3/5) resonant surface
	Field lines outside the last closed (smooth) magnetic surface are long (>1km) even though they appear stochastic until the iota=0.6 surface is approached
	Smooth islands exist beyond that point on the iota=0.6 surface.  In fact two (!) out-of-phase 3/5 island chains are seen with smooth islands
	Lengths are long if the field lines close to form smooth islands
	Lengths are shorter (but still substantial) if the field lines lie in the stochastic region between islands
	Outside the iota=0.6 surface, field lines are quickly lost, i.e. stop circulating toroidally.  Instead, these field lines start wrapping around the coils.

It was suggested that the same methodology (calculate connection lengths and generate Poincare plots one surface at a time) be applied to analyzing the S3 state; and  that the stay-out zone be defined relative to the last surface with long field line lengths.