Construction of an accelerator or transport system requires specification of error tolerances for system components and beamline elements. Ideally, this process is an exploratory and iterative one in which both analytic and numerical tools, as well as previous machine experience, are utilized in concert with engineering design to drive the evolution of a component error budget. For the IR FEL Demo, the short project time scale forced use of a different methodology. In order to meet an aggressive construction schedule, we have made detailed exploration of error effects in this machine using analytic tools only. Rather than performing extensive (and time consuming) numerical studies to establish error sensitivities, analytic scaling relations have been established and an error budget developed. Numerical simulation tools have then been applied to certify that this error budget is sufficient (though perhaps not necessary) to meet machine performance targets.

This process is more time-efficient than the more traditional simulation-based tolerance specification process. It does, however, entail greater technical risk, inasmuch as the analytic treatment on which it is based is necessarily less detailed and precise than studies based on simulation. Analytic methods provide only estimates of error effects and, unlike simulation, do not easily treat the interaction of multiple errors. To compensate for these deficiencies, we have made considerable use of prior experience and have been quite conservative in the specification of error tolerances. Information developed from CEBAF (at Jefferson Lab) has been employed, and valuable input has been provided by Dr. Jay Flanz, a designer of the MIT-Bates recirculator (on which much of this design is based).

This philosophy leads to high probability of success and provides an extremely fast and cost effective construction project, even though individual error specifications may be tighter than those developed using more traditional methods. This is much like the old computer hardware vs. software conundrum - one must optimally allocate hardware and software resources to assure success. It would be foolish to spend huge amounts on a computer platform to run poorly written software; similarly, it's dumb to expend tremendous resources to finely hone software to run on a lousy platform! Here, because of the short time scale, we would be unable to develop complete specifications for the system using traditional methods, while insuring that the machine have accepatble performance. Using numerically-certified analytic methods and exercising reasonable technical prudence, we can save significant optics design effort, time, and expense while maintaining machine performance, all for modest increases in engineering and component cost.

Using analytic methods confirmed by numerical studies, we find that the system response to errors is generally similar to, or weaker than, that in the CEBAF linac. This confirms the suitability of using standard Jefferson Lab components whenever possible. A notable exception to this rule are the end-loop magnets; in this region the bend angles are large, the dispersions are large (~ 2 m), and the beam size is extremely large (> ~ 10 cm) so that good field quality and control is needed in all components. Considerable effort has been expended to assure that dipole fields are uniform (to ~ 10^-4) thoughout the magnet working aperture, end fields are well controlled, characterized and modeled, and power supply regulation is adequate to avoid ripple-driven beam quality degradation. A program of magnet prototyping and measurement has led to a design of magnet coils, iron mechanical design, and end field clamp configurations that provide stray field control and well defined end field profiles. Information from this prototype effort has been incorporated into the transport system optical design and beamline layout.

Last modified: 10 March 1997
http://www.jlab.org/~douglas/ is maintained by: douglas@jlab.org