New Spill Structure Analysis Tools for the VME Based Data Acquisition System ABLASS at GSI T. Hoffmann, P. Forck, D. A. Liakin * Gesellschaft f. Schwerionenforschung, Planckstr. 1, D-64291 Darmstadt * ITEP Moscow, Russia Abstract. During the last years, a comprehensive VME-based data acquisition system for counter applications was developed. This package called ABLASS (A Beam Loss measurement And Scaling System) is used at the GSI heavy ion synchrotron (SIS18), at the high energy beam transfer lines (HEBT) and at the connected experiments. To achieve a maximum of experimental rate capability and to protect sensitive targets from significant intensity peaks, the particle distribution within the slowly extracted bunched beam has to be qualified. To analyze this spillmicrostructure by means of scintillator pulses, new sophisticated tools were invented, such as Q- Analysis, which generates evaluated data in the µs-region. To measure the time distribution of the particles relative to the bunching RF-phase and the probability curve of consecutive particle hits in the ns-regime, a VME Multihit TDC with 25ps time resolution was implemented into ABLASS. The principle and outcome of these new methods substantiated by actual ion beam data will be presented. Keywords: Data acquisition, beam instrumentation, microstructure, bunched beam, heavy ions, slow extraction, TDC, VME, ABLASS PACS: 29.85.+c, 29.40.Mc, 29.27.Fh, 29.27.Eg, 29.27.Ac, 29.20.Lq INTRODUCTION The data acquisition system ABLASS (A Beam Loss measurement And Scaling System), which was assembled to count particles and pulse modulated signals all over the SIS18 and the HEBT, is well established in the control system of GSI and is permanently used [1]. The VME based system uses a CES-RIO3 CPU, 4x 32-Bit- 200MHz SIS Multiscaler and some types of NIM modules [2]. All the time, analog signals produced by common detectors are digitized and counted in one central electronics room. Users all over the GSI campus have the possibility to launch the ABLASS applications simultaneously with their own set of signals selected out of 128 available channels. It allows determining beam losses, correlations between detector signals and transformer- and/or RF-signals (e.g. cavity-rf), gives long time trending information and supports beam optimization. The main features and principles of ABLASS are presented in [3,4].
Within the last year the system was completely duplicated to have a detached expert system for highly resolved spill structure analysis and to create the possibility to install new sophisticated analysis tools, which do not overload ABLASS and which are not required for daily operating. The new expert system is called ABLAX (ABLAss-eXpert). FIGURE 1. ABLAX main GUI and the main feature applications: spill structure (middle), crossbar (lower left) and trending (lower right). Due to the sensitive resonance driven slow extraction at the SIS18, the extracted particle flux exhibits significant fluctuations. These fluctuations can be flattened by extracting a bunched beam or by the proper choice of several SIS-parameters, e.g. the chromaticity [5]. For several experiments, a smooth beam delivery is essential for target protection and maximization of experimental rate capability. Through ABLAX and new installed hardware it was possible to develop the tools for observing spill structure in the time region of µs and ns.
Q-ANALYSIS Online optimization of the spill structure is achieved by using the new feature Q- Analysis (quality-analysis). It allows a fast readout of the scalers down to the µs sampling region and a variable parameter adoption for the data analysis. The count rate of a detector, such as a scintillator, is used for this measurement. The related functionality is described for the exemplary measurement shown in Fig. 2: The rate from a particle counter is recorded with 10 khz (max. 1 MHz) sampling frequency. Within a 10 ms time span the minimal, maximal and average count rate is displayed and the ratio maximum-to-average is used for an expressive characterization of the spill quality. Due to the beam bunching during the extraction, this ratio is decreased by a factor of ~3 compared to the un-bunched case. As this tool is a real online measurement and is updated spill by spill, the distance between the flattops of the three graphs can be watched closely to optimize the beam. Diminishment of these distances by changing synchrotron parameters leads to a better micro-structure. The gradient of the graphs should be almost flat and stable, like parallel horizontal lines. In addition data may be averaged over a defined amount of cycles. For a selected region of interest in the Max-Avg-Min graph histogram data can be obtained. This is shown in the lower right side of Fig 2. The distribution of events, i.e. micro-structure, affects directly the histogram shape. For the processes without fine structure this distribution should follow the Poisson statistical distribution. FIGURE 2. Online evaluation of a single spill of 1750 MeV/u Ar 18+ delivered to the GSI HADES experiment. The particle counter is read with 100 µs time increment for the 10 s spill. Upper left: Maximum-to-average ratio during the spill; Lower left: Minimum (bottom), average (middle) and maximum (top) within a 10 ms time slice; lower right: histogram of the count rate. For the adoption to the time requirements of the experiment, the max-to-average and min-to-average rate is calculated for different readout times, realized as a function
of averaging time (i.e. bin size), as displayed in Fig. 3. The factor F is calculated from the counts per unit of time by the formula: 2 x F = (1) 2 x +σ 2 x This formula gives a helpful duty factor, where a convergence to unity corresponds with a smooth beam delivery. For the given example, beam current breaks during beam delivery are present for readout times below ~500 µs as depicted by the ratio min/avg 0. For an un-bunched beam this would be typically several ms. FIGURE 3. Minimum-to-average ratio (triangle), average-to-maximum ratio (cube) and duty factor (cross) as a function of time binning (from 10 µs to 1 s) for the spill of Fig. 2. A further characterization is yielded for the probability distribution of the count rate. In the displayed case for a bunched beam extraction (Fig. 2, right), the maximum of this distribution histogram is slightly lower than the average rate. For an unbunched beam, the maximum of this distribution would be about one order of magnitude lower, reflecting the occurrence of breaks during beam delivery. TDC ANALYSIS To reach high resolution in the ns time scale for spill structure measurements and optimization with bunched beam extraction a CAEN V1290N multi-hit TDC with
21Bit and 25ps (LSB) resolution was incorporated into ABLAX [6]. Performance and jitter measurements were undertaken to confirm the manufacturer s specifications. This TDC mainly measures the arrival time of the ions relative to the bunching RFphase. A schematic of the acquisition principle is shown in Fig. 4. FIGURE 4. Schematic of the TDC measurement. To produce meaningful information histograms with time information are generated. To investigate the fine data structure, a reference clock synchronized with the RF field generator (SIS-RF-Master) and one dedicated detector (e.g. beam loss monitor, halo counter) signal are fed into the TDC. The obtained information of these two TDC channels is stored in the internal FIFO. This data already includes the complete required time information. To build meaningful time histograms for phase delay and pulse interval ordinary arithmetic evaluation is used. The FIFO data is achievable via VME bus, so the controller (RIO3) has an access to the accumulated data. The output of the FIFO is divided by software into two streams reference timing (Channel A) and measured particle time structure (Channel B) filling two histograms. The first one contains the time interval from pulse to pulse (Data histogram I) and shows the process intensity. The other histogram (Data histogram II) contains the time difference from RF to the particle pulse and presents the fine structure of the process. All data as well as the histogram parameters, like bin size and number of bins, are accessible by network. A dedicated terminal program was implemented into ABLAX. It provides a suitable interface to control this TDC application and online data visualization as it is shown in Fig. 5. The high performance RIO3 processor collects the histogram data online into its own memory. The input-throughput rate is limited only by the VME bus cycle and the TDC's VME access delay time. On TDC's FIFO overflow the RIO3 generates a software 'gate' by clearing the TDC FIFO to allow the continuation of the data acquisition process.
As displayed in Fig. 5, this particular distribution is strongly correlated with the RF. The standard deviation of the time difference is only 6 ns in this case (Fig. 5, bottom left), which is about one order of magnitude shorter than the width of the circulating bunches in the synchrotron. The TDC data also delivers the time spacing of consecutive particle hits (Fig. 5, top right) and its display should be used to match the delivered particle current to the maximal count rate capability of the experiment. These TDC based measurements in connection with the scaler readout cover the time range from ns to s and are online available for beam optimization. FIGURE 5. TDC-time-spectra recorded with same beam parameters as in Fig. 2. The bunch repetition frequency is 5.191 MHz, corresponding to 192 ns. Upper left: Time distribution of the particles relative to the bunching RF-phase, about half of a RF-period is shown. Lower left: The standard deviation σ of this distribution during the SIS cycle. Upper right: The probability distribution of consecutive particle hits composed of ~6 ns broad peaks and 192 ns distances. Lower right: Intensity distribution during the spill. Multipurpose VME module A multipurpose VME module was designed in the scope of a performance enhancement of the ABLASS system. The initial idea to provide just a flexible timing module controlled via VME was transformed to a comprehensive electronic board design with user defined firmware which allows using this board in different applications without redesigning the board layout.
FIGURE 6. Block-scheme of the multipurpose VME module The block-scheme of the module is shown in Fig. 6. The powerful combination of the two-core 600 MHz BlackFin DSP from Analog Devices and the Xilinx Spartan FPGA makes this board suitable for many VME-based solutions which are not limited to beam diagnostics applications. The VME interface is particularly implemented within the FPGA. The current firmware builds a slave VME module with A16..A32 address modes and 32-bit data access. The VME FPGA component also includes some data registers directly accessible from VME, DSP or application specific FPGA modules. The VME access to these registers is processed by normal VME cycles and does not produce any additional wait or delay states on the VME bus. The VME servicing code for DSP also allows the VME bus to get an access to the unlimited amount of data collected and evaluated by any DSP core. The two-core DSP may be used in different ways. It can be programmed as a true real-time high performance filter or as data format converter to release more process time and resources on the central VME processor (RIO3) running the 'real-time' multitasking LynxOS. Many other applications may be developed using this multipurpose VME module. For example, a feedback system in the way that measured data is used to change settings of devices to optimize the beam quality automatically. ABLASS Timing The ABLASS/ABLAX timing is a good example for using this multipurpose VME module, shown in Fig. 7. The timing information for the GSI accelerators is distributed by a special type of network, i.e. MIL1553 derivate. The timing signal consists of information about the current virtual accelerator and the correspondent machine events encoded in Manchester code. The VME module provides full decoding of the timing information and generation of synchronized pulses or bursts of
pulses on the dedicated outputs. Each of the eight outputs may be set by the VME controller and assigned independently to the specific start and stop or intermediate event. The output waveform can be changed from gate pulse to short synchronization pulse or gated frequency output with programmed pulse period. FIGURE 7. The multipurpose VME board. It provides 8 in/outputs, a Xilinx Spartan FPGA and an Analog Devices BlackFin DSP. For ABLASS it is used as a timing module. In this application the FPGA provides initial filtration and decoding of the Manchester signal. It also produces the defined output signal shape according to the timing signal and controls all onboard in/outputs. Core A of the DSP (Fig. 6) manages the list of the deterministic events with correspondent timing and machine information which is stored in the DSP memory and is offered to the RIO3 processor via VME bus. Core B (Fig. 6) of the DSP is responsible for the VME bus interfacing (high priority interrupts) and provides the signals for the front panel LED display. As a demand for the FAIR project to effectively fill the new ring SIS 100, a SIS18 cycle repetition rate of 4 Hz is planned [7]. This fast mode reduces spill pauses to the few ms region, the reliability of the DAQ timing had to be optimized with this timing module. REFERENCES 1. T. Hoffmann, D. A. Liakin, P. Forck, Proc. 10th Beam Instrumentation Workshop, BIW, Upton, New York, USA, p. 329 (2002) 2. http://www.struck.de, www.ces.ch 3. D. A. Liakin, T. Hoffmann, P. Forck, Proc. 6th European Workshop on Beam Diagnostics and Instrumentation for Particle Accelerators, DIPAC, Mainz, Germany, p. 164 (2003) 4. T. Hoffmann, D. A. Liakin, P. Forck, P. Moritz, Proc. 11th Beam Instrumentation Workshop, BIW, Knoxville, Tennessee, USA, p. 294 (2004) 5. P. Forck et al., Proc. 7th European Particle Accelerator Conference, EPAC, Vienna, Austria, p. 2237 (2000) 6. http://www.caen.it 7. P. Spiller, Proc. of the IEEE 2005 Particle Accelerator Conference, PAC, Knoxville, Tennessee, USA, p. 294 (2005)