Developments in the high precision control of magnet currents for LHC

Abstract

The LHC will require over 1700 magnet power converters, some of which will need an unprecedented precision of about 1 ppm (of 13 kA). This paper presents the approach taken, prototype methods, initial results and charts future design directions. These results confirm that such performance can be obtained reliably and at a reduced cost compared to conventional methods. Developments of a real-time controls infrastructure needed to support on-line beam feedback are outlined. BACKGROUNDFor the LHC machine to achieve its full potential, the power converter system needs to attain a peak performance of about 1 part per million (ppm) in terms of resolution, stability and reproducibility. This represents an improvement over current practice of approximately a factor of ten. In addition, the very large electrical time constants presented by super-conducting magnets, coupled with the need to remove dynamic errors required a new approach. In order to meet this challenge a number of studies and practical tests have taken place over the last few years aimed at proving that such increased performance can be obtained reliably. A strategy for obtaining such improvement was presented in an earlier paper [1]. In brief this strategy was :• Employ digital regulation methods rather than analogue methods.• Apply digital corrections of known errors.• Employ real-time feedback mechanisms (both magnetic and beam related).• Incorporate in-situ calibration techniques (this subject is not covered further in this paper).This approach required an extensive revision of present practice for the regulation and remote control. This paper presents the design of the prototype system, the results obtained and outlines future design directions. PROTOTYPE IMPLEMENTATIONThe digital regulation hardware, shown in figure 1, has been built using a commercial digital signal processor (DSP) card as the heart of the system. This processor, a Texas TMS320C32, 32 bit floating-point device computes the current reference value every milli-second, compares this to the actual measured current and then computes, using a regulation algorithm, the required signal to drive the power converter to the correct value. This signal is converted into analogue form via a digital to analogue converter (DAC). The output current is measured by a purpose designed, Sigma-Delta, analogue to digital converter (ADC), which has better than ±1 ppm performance and is directly interfaced to the DSP. The interface between the DSP, the remote control system and the power converter hardware is provided by a second commercial micro-controller card. This card employs a Motorola MC68HC16Z1 which acts as the 'master' for the DSP.The software for both processors is written in 'C', using commercial software development tools. Only minor use of assembly language has proven necessary. Both systems employ 'direct to processor' connections to the PC (BDI and JTAG), allowing code down-load and extensive debug capabilities. The reference and regulation algorithms operate as a single repetitive t...

The design of the superconducting wiggler power supply system for HEPS-TF

“…For any accelerator magnet power supply, its load characteristics are the basis of power closed-loop control. For the normal conducting magnet, we generally adopt proportional integral differential (PID) control as the closed-loop control algorithm, and the core design of the current loop integral time constant mainly depends on the load time constant [6,7]. For the superconducting magnet, its load time constant is much larger the normal one.…”

LHC inner triplet powering strategy

“…For LHC power converters, the high-precision current loop is a digital loop implemented in a dedicated digital controller [6]. The K and D decoupling matrices could be implemented either in digital or in analogue hardware.…”

Optimisation of the current ramp for the LHC

“…In order to address these issues, a digital approach has been adopted [4]. These methods can generate smooth ramping functions as opposed to the traditional straight line segment method, and can ensure that resolution, overshoot and following errors will be less than one part per million of the maximum current.…”

“…This work resulted in the definition of a current ramp function composed of mathematically defined, smoothly joining segments. A prototype digital controller based on a DSP has been developed and built [4]. In this equipment the current ramp is computed from the segment equations in real time.…”