A Flexible Labview”-based Data Acquisition And Analysis

Transcript

r # A FLEXIBLE LABVIEW”-BASED DATA ACQUISITION AND ANALYSIS SYSTEM FOR SCANNING MICROSCOPY Daniel H. Morse and Arlyn J. AntoMc Sandia National Laboratories, Livermore, CA 94551 USA Graham S. Bench and Mark L. Roberts Lawrence J&ermore National Laborato~, Livermore CA 94551 USA. Abstract A new data analysis system has been developed with computer-controlled beam and sample positioning, video sample imaging, multiple large solid angle detectors for x-rays and gamma-rays, and surface barrier detectors for charged particles. The system uses the LabVIEWM programming language allowing it to be easily ported between different computer operating systems. In the present cordlguration, digital signal processors are directly interfaced to a SCSI CAMAC controller. However, the modular software design permits the substitution of other hardware with LabVIEW-supported drivers. On-line displays of histogram and two-dimensional elemental map images provide a user-ftiendly data acquisition interface. Subregions of the two-dimensional maps may be selected interactively for detailed analysis or for subsequent scanning. Off-line data processing of archived data currently yields elemental maps, analyzed spectra and reconstructions of tomographic data. 1 .. . .... .... . ....... .,, . --.w-.. -.,,. e..-,, , . . m- ,..*=.-.-..-.., ------ , ,~-.-—--’r-~,- ~- - .- DISCLAIMER This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, make any warranty, express or implied, “or assumes any legal liability or responsibility for ‘the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or or service by trade name, trademark, manufacturer, otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof. DISCLAIMER Portions of this document may be illegible in electronic image products. Images are produced from the best available original document. . .. , , This work performed under the auspices of the U.S. Department of Energy under Sandia National Laboratories contract DE-AC04-94AL85000 and Lawrence Livermore National Laboratory contract W-7405-ENG-48. 1. Introduction A new nuclear microprobe, designed principally for rapid particle-induced x-ray emission (PIXE) analysis of individual particles on environmental samples, has been jointly constructed by Sandia National Laboratories and Lawrence Livermore National Laboratory [1,2]. A unique feature of this system is the very large solid angle (>1 sr), provided by four 200 mmz IGLEV detectors -- two located in front of the sample and two behind. Environmental falter samples are typically scanned with a 10 pm X 10 pm, 50 @, 3 MeV proton beam. Rapid scan rates (>1000 pixels/s) and large scan areas (>10 mmz ) reduce damage to the sample. The combination of large solid angle and relatively high beam current permits rapid location and classification of particles based on the presence of counts in speci.tied X-ray peaks. Heavy elements whose characteristic x-ray energies lie well above the bremsstrahlung peak can be ident.iiled with only a few counts per pixel. Pixels containing particles-of-interest can subsequently be rescanned at a lower current to provide better statistics and more quantitative results. The d~ta acquisition system also supports Rutherford backscattering spectroscopy (RBS), scanning transmission ion microscopy (STIM) and ion microtomography (JMT) analyses [3]. Although other software systems were taken into consideration [4,5,6], we developed our First, many of our experiments are very own software for the following reasons. specialized in nature and have specitic performance requirements. We also wanted to have flexibility in choosing the hardware. Finally, our previous software [7], written in C on a 2 t , Sun computer, would have required significant moddicat.ion. We looked for a solution that was less tedious than programming in C, relatively platform independent, and adaptable to changes in experimental demands or hardware. With these factors in mind, we based the data acquisition system on LabVIEW, a graphically oriented programming language. LabWEW is well-established language that is supported on UNIX, Windows and Macintosh operating systems and is readily ported between platforms. As a high level graphic-oriented language, LabVIEW has built-in graphics displays and other graphical objects, such as switches, numerical displays and text panels. Additionally, LabVIEW includes a library of mathematical subroutines and other utilities, including a growing library of drivers for many kinds of hardware. Other advantages are its inherent modularity, in that each subroutine is written and tested as a stand-alone program, and that it supports multiple tasking. 2. Hardware AIthough we origin~y intended to use analog ADC’S, we eventually chose to purchase a unit manufactured by X-ray Instruments, Inc. (XIA) which contains four sets of amplMers and digital ADC’S packaged in a single-width CAMAC module. According to the manufacturer, the digital ADC’s have nearly half the”dead time of an equivalent analog system. The cost per ADC was also lower compared to the analog system we considered (e $3000/channel in U.S. dollars for ADC and amplifier). A LabVIEW driver was supplied to provide software control of all hardware parameters as well as list mode data acquisition. The module also contained a built-in counter that could be used to output a pixel number for each pulse height measurement. We later discovered several limitations of the XIA module. First, there are no hardware gate inputs; gating must be done through software. Second, data acquisition must be stopped periodically to transfer the 16k ADC data buffer to the computer. This is not a serious problem except at extremely high count rates, since it only increases the dead time by a few percent. 3 Third, optimizing the ftiteen or so ● parameters for each ADC can be very tedious, since each parameter must be adjusted by editing its value in a text fle and then collecting a spectrum. Once the parameters are set, however, they seem to be stable. Finally, there are no “busy” outputs for the ADC’S. With our previous system, we used the ADC “busy?’ output to gate the output pulses from our current integrator before sending them to the dwell time counter in the scanning system. The result was a direct measurement of live-time corrected charge, so that, allowing for some statistical variation, each pixel received the same charge. Although the module returns an average live time for each data buffer, pixels with higher count rates will have more dead time and, therefore, fewer counts than expected. As long as the dead time is low or relatively constant the error can be tolerated. Figure 1 shows a schematic representation of the data acquisition hardware. Pulse height and pixel location data are read from the XIA module over a high-speed Jorway model 73A SCSII crate controller, capable of transferring data at the speed of the CAMAC bus. The sample stage, manufactured by Newport Corporation, uses vacuum compatible dc motors and a linear optical encoder with closed loop feedback to provide 1 ~m positioning accuracy. A multi-function PCI bus I/O card from National Instruments supplies all other the counters and DAC outputs needed for the scanning system. A PCI video board was included to store images of the sample through any of the four video microscopes mounted in the chamber. All of the hardware interfaces sdected were supplied with LabVIEW drivers. 3. Scanning One of six scanning modes can be selected. (1) Single-spot analysis is the simplest case with no scanning at all. (2) Rectangular two-dimensional scanning allows the selection of scan size, pixel size, scan center position, number of scans, and dwell time from a pop-up menu. In addition, a scan interlace ratio may be specified, so that adjacent lines are not 4 # scanned sequentially, thereby reducing beam damage to the sample. (3) Mechanical rastering (not yet implemented) maybe selected to raster the sample instead of the beam for scanning large areas at the expense of scan speed. (4) Tomographic scanning is the same as rectangular scanning except that the sample is rotated after each pass. (5) Mask scanning moves the beam sequentially to pixels located in a region-of-interest as specitied in a mask array. The mask is normally generated by taking a quick scan of a rectangular area, then automatically selecting only those pixels whose counts in a specified energy range exceed a specified threshold. The mask region may be enlarged slightly by passing it through an algorithm that includes all pixels bordeting the original mask region. (6) Finally, as an aid in focusing, a single pair of horizontal and vertical lines may be scanned across a grid or similar high-contrast target. A plot of signal intensity versus position in each axis is updated continuously to provide rapid feedback to the operator. Beam positioning is accomplished simply by generating a sequential list of X and Y coordinates and then passing the array to a LabVTEW driver supplied with the 170 board. In our hardware configuration, the outputs of the ADC’S are coupled to high speed, high voltage amplifiers with a fixed gain of 1000 (manufactured by Trek, Inc.) which rue connected to electrostatic deflection plates. The dwell time at each pixel is determined by a programmable divide-by-n counter that can be driven by current integrator pulses, clock pulses or AK events. For thin targets, charge is collected in a 37 mm diameter by 150 mm deep stainless steel Faraday cup, lined with graphite on the bottom to the reduce X-ray background in the chamber. entrance. The cup is fitted with an electron’ suppression ring at its However, we’find no sigtilcant change in beam current with the suppression voltage on or off, indicating that very few secondary electrons are entering or leaving the cup even without suppression. 4. Data displays 5 . LabVIEW offers a wide selection of graphical objects (controls and indicators) that can be dragged onto control panels as shown in figure 2. Each object placed on the front panel automatically appears as a symbol on a second “wiring diagram” window where its inputs and outputs me connected to other program elements. For example, an X-Y plot of a histogram can be implemented simply by dragging an icon onto the front panel and connecting the histogram array to the corresponding symbol on the wiring diagram. In a similar manner, PIXE map data can be displayed by connecting a 2D array to an intensity plot A plot representing the ratio of two elements maps. cm be displayed by selecting any two Other XY plots can be displayed to show the sums of rows and columns of a selected map. All of the plots except the histogram are displayed on separate pop-up windows to avoid clutter on the main screen. The function and appearance of any LabVIEW display can be set by the program or moditled interactively through pop-up menus. For example, it is possible to interactively change the maximum or minimum value on any scale, to select a logarithmic or linear display, to pan or zoom the display window, or to read the value of any pixel in a map by positioning a cursor. Windows may be plotted at any time either interactively or by program control. 5. Data processing and storage After initializing the hardware, we initialize the LabVIEW global variables by reading a parameter fde from a specified previous data set (by default the last data set). Next, we wait for the user to modify any parameters and press the start button. The scan is defined by four arrays: two to indicate the sequence of X and Y scanning voltages, and two to indicate corresponding X and Y coordinates in the map display. The scanning DAC’S and dwell time counter are loaded and enabled to begin scanning with no further software control. Each time the dwell-time counter counts down to zero it triggers the next DAC output and also triggers a pixel counter in the XIA module. At this point two separate parallel processing loops are initiated using the multi-tasking capability of LabVIEW -- one fi.mction reads data from the AK while the other processes data. The XIA module is programmed to set a CAMAC LAM signal when any of the four 16k buffers (one for each ADC) is half full. The module is polled every 0.1 second until the LAMis set or until 3 seconds has elapsed. Because the XL4 module uses the same microprocessor for data processing and transfer, data acquisition is interrupted and the dwell time counter is gated off during the actual data mmsfer. Data is stored temporarily in a LabVIEW array set up as a circular buffer until it can be processed. If the buffer becomes full, data acquisition will be interrupted until space is available. Fimilly, after the scan is completed or manually terminated, a detailed parameter file is written. The collected data are stored in three formats in every experiment. First, an image of the list data in each buffer is appended to a disk fde each time data are read. Second, a histogram is accumulated in memory for each ADC to be displayed ‘imdstored at the end of the acquisition period. Finally, map arrays are updated according to a list of pre- determined energy regions. A cube fde (a file containing a histogram and location coordinates for each pixel) is not saved, because cube arrays can be excessively large. Cube fdes or additional map files can be generated and displayed, if needed, by postprocessing the image fde with LabVIEW routines that are similar to the data acquisition program. More quantitative off-line analysis is performed using the Ion Micro-Analysis Package (lMAP) which has recently been ported from UNIX to the Macintosh [8,9]. Figure 3 shows an example of elemental maps generated by IMAP from PIXE data collected by scanning the cross-section of a stranded superconducting wire. 6. Performance The detector resolution of the IGLET-X detector measured with the XIA module set for a Gaussian time constant of 8 us was 155 eV at 5.9 keV and 1000 counts/s. The resolution measured under similar conditions with an analog system was nearly identical. We tested 7 the throughput of the data acquisition system at high count rates suitable for semiquantitative measurements needed for locating particles on a large area falter. The maximum count rate per detector with a 2 WSsecond Gaussian faltering time constant was about 20,000 count.ds. Even at the highest rates, the resolution was better than 210 eV and the pile-up for a spectrum with one dominant peak (aluminum) was less than 1.5 per cent. Our 266 Mhz PowerPC data acquisition computer was able to acquire, display and store data even at this rate (>60 kHz from three channels) without over-running its data buffer. We also tested the scanning system at rates as high as 5000 pixels/second, the maximum clocking frequency of our current integrator. SignKlcantly faster scanning frequencies are certainly possible, limited mainly by the settling time of the DAC and scanning amplifiers. 7. Conclusion The data acquisition system we have developed supports a wide selection of scanning, on- line data display and processing options at data rates in excess of 60,000 randomlyoccurring pulses per second and can run under Windows, Macintosh or UNIX platforms. Because of the growing library of LabVIEW drivers, the modularity of program, and the ability to test each module independently, the data acquisition system could easily be adapted to other hardware or other data processing requirements. (Modification of the software requires the purchase of a LabVIEW development license for about $2000 U. S.) Future plans are to incorporate additional post-processing software, complete the mechanical scanning capability, and possibly to implement analog ADC’s for other systems. References 1. A.J. Antolak, D.H. Morse, G.S. Bench, D.W. Heikkinen, M.L. Roberts, E. SiderasHaddad, Nucl. Instr. Meth. B 130(1997) 211-218. 2. M.L. Roberts, P.G. Grant, G.S. Bench, T.A. Brown, B.R. Frantz, D.H. Morse, and A.J. Antolak, (“The Stand-alone microprobe at Livermore”, these proceedings) 8 c 3. D.H. Morse, .A.J. Antolak, G.S. Bench, D.W. Heikkinen, M.L. Roberts, E. SiderasHaddad Nucl. Instr. Meth. B 130(1997) 740-745. 4. G.W. Grime, M Dawson, Nucl. Instr. Meth. B 104(1995), 107-113. 5. G.R. Moloney, P.M. Obrien, A. Saint, L.Witham, A. Sakalleriou, A. Bettiol, GJF.Legge, Nucl. Instr. Meth. B 104 (1995) 114-118. 6. M. Elfman, P. Kristiansson, K. Malmqvist, J. Pallon, A. Sjoland, R. Utui, C. Yang, Nucl. Instr. Meth. B 130(1997) 123-126. 7. D.H. Morse, Nuc1. Instr. Meth. B 85(1994) 693-698. 8. A.J. Antokdc and G.S. Bench, Nucl. Instr. Meth. B 90 (1994) 596-601. 9. A.J. Antolak, G.S. Bench, M. L. Hlldner, and D. H. Morse, Nucl. Instr. Meth. B85 (1994) 597. Figure 1. A schematic representation of the principal hardware and signal connections of our data acquisition and scanning system. Figure 2. A screen-captured image of the data acquisition system front panel partially obscured by two pop-up windows. One window is a map of copper X-rays and the other represents the sum of the map values in Y for each pixel in X. The sample is a copper grid with bars spaced at 12.5 pm used for tests of spatial resolution. Figure 3. Cross-section of a super-conducting wire made for the proposed ITER reactor. The stranded Nb-Sn core is sheathed in a thin layer of tanta.Ium and a thicker layer of copper. These scans were collected in LabVIEW and analyzed off-line with the Ion MicroAnalysis Package (IMAP). The scan area is 0.55 mm X 0.55 mm, and the pixel size is 3.5 Lm X 3.5 Lm. c -L g’ z 3 u a n ,.. ,,:.: .,.:. :,... .... w ,.., ;. .-: . + . t,