How to Control a 27-km-long Machine

Nipun Mathur
National Instruments Corp.
Austin, Tex.

The European Organization for Nuclear Research, more commonly known as CERN, hopes the LHC will soon reveal the secrets of the universe. The machine will recreate the conditions right after the Big Bang on a miniature scale by accelerating and colliding two beams of microscopic particles traveling at 99.9999991% of light speed.

Particles traveling at close to the speed of light can have high levels of energy. The LHC creates beams with trillions of particles with energy equivalent to that of a 400-ton train traveling at 150 km/hr. Any stray particles deviating from the beam core can severely damage the machine. So collimators eliminate these particles to keep that from happening.

There will eventually be over 150 collimators in the LHC. But controlling them presents a challenge. The problem isn’t how fast the collimators move or react — collimator parts only move a few millimeters at about 1 kHz at most. The real difficulty is synchronizing collimators that can be 13 km away from each other. In all, there are 500 axes of motion involved over this large distance. The axes are made up of motors that position the collimators’ jaws. These jaws are synchronized to within 100 μsec and move with accuracies of ±10 μm.

The LHC collimators mainly comprise several layers of jaws that absorb stray particles. Each jaw consists of a pair of blocks 1-m long, 80-mm wide, and 25-mm deep. The blocks can be CFC, graphite, copper, or tungsten, depending on their proximity to the beam halo. Primary and secondary blocks see high thermal loads and stresses and are made of lowdensity material. Tertiary blocks are metallic and stop low-energy particles. But regardless of block material, jaw positioning must be accurate to one-tenth the beam core diameter, so the required accuracy is 20 μm.

Also, the jaws must be positioned at a precise angle with respect to the particle beam trajectory. Two stepper motors, one at either end of each jaw face, handle the positioning. Any vibration during motion can damage the blocks, some of which are graphite brazed on copper and quite fragile. So both motors on each block are synchronized to within 1 msec of each other. And jaws must be almost perfectly parallel to each other; the tolerable orientation angle is below 2 milliradians.

CERN engineers designed the control system using National Instruments Corp. LabView software. It runs on a PXI programmable automation controller platform, also from NI. In all, there are 120 PXIs. Each controls between one and three collimators. PXI platforms can be synchronized using external or internal clocks. But for the LHC, a central clock coordinates motion control. Specifically, a central computer sends synchronization signals over dedicated optical fibers to a motion-control supervisor. This signal then goes to one of the PXI platforms which houses a high-precision timing source. Signals from this source, in turn, go to external timing inputs on other PXI platforms.

There is a high level of radiation near the collimators. This means electronics can’t be placed anywhere near them. So the PXI controllers sit a few hundred meters motionfrom the collimators, protected in an underground area.

For reliability, CERN divided the collimator controls between two sets of PXI platforms. The first set controls the stepper motors and reads resolver feedback from motor shafts at a 400-Hz rate. In the standard configuration, one PXI platform controls up to 15 stepper motors mounted on three different collimators.

The second set of PXI platforms reads LVDT sensors, backups for the resolvers. These get read at a rate of 1 kHz to check real-time positioning of the collimator jaws. If the PXI platform controlling the motors fails, CERN can still determine where the collimators’ jaws are using LVDTs. And if the PXI platform reading the LVDTs fails, CERN can bring the collimators to a known safe position.

Both kinds of PXI chassis run LabView Real-Time on the controller. And CERN used the NI SoftMotion development module and reconfigurable modules to create a custom motion controller.

CERN computers interface with collimator actuators and sensors through reconfigurable I/O modules. LabView FPGA runs on these devices. All inputs/outputs are completely reprogrammable with LabView. The I/O card controlling each collimator carries a 3M-gate FPGA, providing eight analog inputs and outputs, and 96 digital input/outputs. The digital outputs are through NI C Series modules which provide isolation — necessary because the control electronics are some distance from the motors. The I/O cards send step-anddirection pulses to the stepmotors and read limit switches, interlocks, and triggers. Resolver signals are read through cards made in-house by CERN engineers.

Collimator Control
In normal operation, collimator jaws go through a 20-min motion profile. The LHC collimator motion- control process starts when the central controller issues a command. The motion-control system checks for these commands every millisecond, acknowledges their receipt, and signals that actions have taken place.

Collimator jaw settings are defined in normalized terms around the beam center. So the motion-control computers must convert these into real jaw positions for each location. The computers also interpolate movement profiles from the central control so they are smooth and synchronize all axes of the same collimator. The computer generates trajectory setpoints that take into consideration the speed, acceleration, and jerk specified for each motion profile. These setpoints then go to the FPGA controller. In addition, the motion controller generates interlocks if it detects problems.

A control loop runs on the FPGA at 1 MHz. It generates step and direction signals controlling the stepmotors once it gets data from the trajectory generator. The I/O task for both digital and analog signals runs on the FPGA every 5 msec. The resolver signals are decoded in the FPGA and read at 400 Hz. The real-time processor also reads the status of the FPGA every millisecond to check for data failures.

Meanwhile, as the motion controller executes commands, the collimator-position sensors double- check that collimator jaws are positioned properly. These sensors use the LVDTs to gauge absolute position of each jaw. Two LVDTs read each jaw position, one for each stepmotor. Two additional LVDTs measure the distance between jaws of the same collimator. PXI platforms compare LVDT readings with error thresholds from LHC central control. These thresholds may be different at different points in the motion profile.

The position reading takes place at 100 Hz. Readings from the seven LVDTs on each collimator go to two data-acquisition cards with eight 16-bit, 250 ksamples/sec simultaneous differential analog inputs. The acquired data then transfer via DMA to the controller. A reconfigurable I/O card with eight analog outputs generates the stimulus for the LVDTs.

As with the motion-control system, the PXI position computers receive commands from the LHC central controller every millisecond. They translate these commands into error thresholds for each of the LVDTs. The resulting positional accuracies are measured in micrometers. The LHC is in its final shakedown stage and should be central controller every millisecond. They translate these commands into error thresholds for each of the LVDTs. The resulting positional accuracies are measured in micrometers. The LHC is in its final shakedown stage and should be operating this summer.

Make Contact
National Instruments Corp.
NI.com
Video of CERN LHC collimators
tinyurl.com/4nwfrb