- Email: [email protected]
A high speed PC-based data acquisition and control system for positron imaging
ARTICLE IN PRESS Nuclear Instruments and Methods in Physics Research A 604 (2009) 355–358
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
A high speed PC-based data acquisition and control system for positron imaging T.W. Leadbeater , D.J. Parker School of Physics and Astronomy, The University of Birmingham, Edgbaston, Birmingham B15 2TT, UK
a r t i c l e in fo
abstract
Available online 10 February 2009
A modular positron camera with a flexible geometry suitable for performing Positron Emission Particle Tracking (PEPT) studies on a wide range of applications has been constructed. The demand for high speed list mode data storage required for these experiments has motivated the development of an improved data acquisition system to support the existing detectors. A high speed PC-based data acquisition system is presented. This device replaces the old dedicated hardware with a compact, flexible device with the same functionality and superior performance. Data acquisition rates of up to 80 MBytes per second allow coincidence data to be saved to disk for real-time analysis or post processing. The system supports the storage of time information with resolution of a half millisecond and remote trigger data support. Control of the detector system is provided by high-level software running on the same computer. & 2009 Elsevier B.V. All rights reserved.
Keywords: Positron imaging PET PEPT Data acquisition
1. Introduction Modern positron imaging devices utilise a large number of small gamma ray detectors arranged around the object under study. Position dependent event data are then derived from the detector response to collinear gamma ray emissions arising from positron annihilation within the field of view. The detection of both gamma rays in coincidence defines a line-of-response (LOR) joining the two detector elements along which the annihilation is said to have occurred. These LOR data can then be used to reconstruct a quantitative map of tracer concentration over a volume (Positron Emission Tomography, PET), or used to rapidly locate a single tracer particle present in the field of view by triangulation (Positron Emission Particle Tracking, PEPT) [1]. Fig. 1 illustrates these concepts; a number of LORs from a point source are shown interacting within a cylindrical detector array. LORs are then described by a signal comprising the transaxial crystal number (0–511) and the axial crystal number (0–15) for both end points of the LOR. At the University of Birmingham Positron Imaging Centre variants of the ECAT series of scanner produced by Siemens/CTI are used for the study of industrial and engineering apparatus. These scanners consist of 128 individual detector blocks, each containing an array of 8 8 Bismuth Germanate (BGO) scintillator crystals viewed by four photomultiplier tubes (Fig. 2, part A). Due to their modular construction; where the detectors are arranged together in a number of identical small sub-systems
(called buckets, Fig. 2, part B) it has proved possible to remove the detectors from the gantry frame and arrange them in different geometries, thus creating a modular positron camera suitable for PEPT studies [2]. Custom support gantries have been constructed to hold the modular detector elements around the apparatus under study. An example geometry is given in Fig. 3 showing 16 separate modules each containing four detector blocks. Some of the allowed lines of response are shown to illustrate the field of view. The sensitivity varies as a function of position, this is shown as the density of the lines. A small ring camera was developed using components from a Siemens ECAT-951 full body PET camera. Studies were performed on this system in order to ascertain the operating characteristics before it was used for a PEPT study of annular shear flow. Parker et al. report experience using the modular camera and give a full account of the performance of the initial system [3]. In summary of these results a maximum event rate of around 700 k s1 (prompt þ delayed) was seen for activities exceeding 80 MBq. Many of these events were random coincidence noise. The maximum true coincidence rate occurred with a tracer activity of 43 MBq; the recorded data rate at this point, 305 k s1 , was limited primarily by the data transfer speed using the original data acquisition system. Optimum tracer activity lies between 20 and 50 MBq for this initial trial system.
2. Positron emission particle tracking Corresponding author.
E-mail address: [email protected] (T.W. Leadbeater). 0168-9002/$ - see front matter & 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2009.01.184
The PEPT method uses an iterative approach to locate a single tracer particle within the field of view by triangulation over a
ARTICLE IN PRESS 356
T.W. Leadbeater, D.J. Parker / Nuclear Instruments and Methods in Physics Research A 604 (2009) 355–358
small number of LORs. Disregarding events corrupted due to scattering, a sequential set of LORs should cross each other at the location of the tracer particle to within the spatial resolution of the camera. It has been shown [1] that for a set of N detected events, with a fraction f used to locate the tracer, the precision D of locating a stationary particle is given approximately by w
D pffiffiffiffiffi
(1)
fN
Fig. 1. LORs from a point source within a positron camera.
where w is the spatial resolution of the camera. It can be seen that the value of D can be set to the desired precision by the appropriate choice of N. The value of N is proportional to the activity of the source, thus over a long enough time scale (increasing N) a stationary particle can be located arbitrarily well. Using a high activity particle with the correspondingly high data rate allows accurate and frequent location. For example; acquiring event data at a rate of 100 k s1 , with f equal to 10%, a stationary tracer can be located to within about 0.5 mm every 10 ms (N of 1000) using the modular camera (w of around 5 mm). During this time, however, a tracer particle moving at 2 m s1 will move 20 mm with the LORs spread along its trajectory. To deal with fast moving tracers it is necessary to locate more frequently with the corresponding loss of precision (reduced N). Typical values for N lie between 50 and 300 events. In the trial study described above only 16 out of a maximum 32 modules were used. Higher data rates are expected with larger
Fig. 3. Example modular camera geometry.
Fig. 2. Detector block (A) and detector bucket (B).
ARTICLE IN PRESS T.W. Leadbeater, D.J. Parker / Nuclear Instruments and Methods in Physics Research A 604 (2009) 355–358
systems using more detector modules. Recent trials with the initial system [4] have reported location rates of approximately 1.5 kHz with a measured standard deviation in three dimensions for a stationary particle of 2 mm. A higher location rate of 7 kHz resulted in a standard deviation of 10 mm for a stationary source.
3. Event data In a complete system (of 32 detector buckets) there are 8192 discrete detection elements allowing over 1:5 107 possible coincident combinations (LORs). These data are presented at the output of the coincidence processor in the form of a 32 bit parallel data word. There are 6 bits describing the pair of detector modules containing the event. Then, for both detectors, 2 bits describe the detector block in the bucket, 3 bits describe the row and 3 bits describe the column for the excited crystal segment in each block. Two bits identify the coincidence processor board which processed the event, and 2 event related bits flag multiple and delayed events. The remaining 6 bits in the word are not used by the coincidence processor. In the new data acquisition system these bits can be used to give additional functionality, for example defining an event time and flagging external event triggers. A relationship between crystal segment number and crystal location relative to a fixed origin is derived after measurement of the detector module position. When running the PEPT algorithm LOR data are converted into cartesian (x1 , y1 , z1 ; x2 , y2 , z2 ; t) coordinates giving the spatial position of the two interactions and the time the event occurred. This conversion is customised to the camera geometry used for the specific application.
4. Data acquisition electronics Following the initiating gamma-ray event and electronic handshake the maximum theoretical data rate from each coincidence processor (of four) is 4 MHz, resulting in a maximum of 16 MBytes per second requiring processing. The original ECAT system was custom designed using off-the-shelf components in a compromise between speed and cost. The design [5] allowed a maximum data rate of 1 MHz transmission to the storage and processing system. The primary limitations were due to memory access times, however, storage to disk in list-mode was significantly slower due to large disk access times. Data storage rates could be improved by using two acquisition systems operated in parallel but with increased costs and added complexity. The demand for high speed list mode data storage required for PEPT experiments has therefore motivated the development of an improved data acquisition system. This system is based around the PCI-7300A Digital I/O Card manufactured by Adlink technologies inc. It features 80 MBytes per second, 32 bit data transfer via direct memory access (DMA) along the PCI bus, with a further 8 bits auxiliary I/O and simple handshaking/trigger process capability. In order to interface with the existing coincidence processor system a small number of dedicated support circuits have been built. These use TTL technology and deal with the handshaking requirements, timing tag word insertion and remote trigger handling. Conventional PET imaging does not require high resolution temporal information to be stored, however, this is essential for PEPT experiments as locations are made on a frequent basis. Time information for each event is determined by the time-of-arrival of the event in the processing unit. A dedicated board adds a timestamp to individual events with a temporal resolution of 0.5 ms. This is sufficiently frequent for PEPT as the location time is
357
calculated from the average time of the trajectories used in the final solution. This can cover a number of milliseconds. The frequency of time events can easily be increased if the data rate from the new system is high enough that multiple locations occur between time events (within 0.5 ms). This would improve the temporal resolution of the PEPT study allowing for the accurate tracking of high speed particles. Other instruments with a wide range of possible applications (e.g. tachometer (industrial), cardiac gating (medical)) can be used to provide additional data for any study. With this in mind a remote trigger system has been included in the hardware to allow up to two extra devices to insert data into the list mode data stream. A bit is set when the remote input is triggered, and the data word inserted into the list-mode stream with a time stamp. Fig. 4 shows a block diagram for the acquisition system electronics. The coincidence processor is driven by a handshake driver and event data words are placed on the output FIFO (First In First Out memory stack) following a successful handshake and trigger. Time information and remote trigger information is inserted into the FIFO by triggering the data capture as events occur. The new data acquisition hardware has been designed to interface with the ECAT-931 coincidence processor made by Siemens/CTI. Recently it has been used to capture and process data from ECAT-951 series detectors and is able to process data from similar systems with little or no modification. It is also possible to connect the new data acquisition system to multiple coincidence processor units allowing a further 32 detectors to be used, however, there has not been a need to experiment with a system this large. Control of the coincidence processor and individual detectors is made possible by high level software running on the same computer. In this a serial RS-232 link is used to interface with the systems, the software converts user level commands into machine code used to communicate with the camera. In this manner detector calibration is performed and system checks and event data can be accessed. The entire system is illustrated in the block diagram given in Fig. 5. Here the detectors and coincidence processor are controlled by serial commands issued from the computers USB port. Data from the individual detector modules undergo coincidence processing and the resultant coincidence data word is captured following successful handshaking. Data is transferred as a block from the local FIFO to a circular buffer sitting in system memory via DMA along the PCI and system busses. The PC resources can be
Fig. 4. Acquisition system electronics.
ARTICLE IN PRESS 358
T.W. Leadbeater, D.J. Parker / Nuclear Instruments and Methods in Physics Research A 604 (2009) 355–358
bursts of above 1.7 MHz have been seen during periods where the tracer particle was located in areas of high sensitivity. This corresponds to writing between 1 and 15 MBytes per second to disk. The increased data rate has allowed more trajectories (increased value of N) to be used for each PEPT location without the corresponding loss of accuracy due to tracer motion. In these experiments accurate locations were made every 11.3 ms on average using 2000 events per location (N). For this application a laser tachometer was used to correlate tracer motion with that of the extrusion shaft by inserting data using the remote trigger. While the new system has been seen to operate with performance exceeding the maximum performance of the initial system, it has so far not been tested to determine its full limits; these are expected to be much higher than reported here. Fig. 5. System block diagram.
used for processing the data, performing data analysis, and displaying real-time LOR data. The circular buffer consists of eight buffers of 32 kbits (10k array of 32 bits) wide. Data are transferred asynchronously from system memory to disk as each buffer is filled. This is a potential bottleneck as disk write speed is again the slowest transfer rate in the system. Buffers which are dropped due to slow disk access have been flagged, in tests this has not occurred. If this is deemed to be a problem in the future data can be transferred from the circular memory buffer into conventional system memory for temporary storage. It can then be transferred to disk post acquisition or during periods where data acquisition rates are low. Raw event data can be saved to disk or removable storage media (CD/DVD) in binary format. These raw data files can then be processed using the PEPT algorithm to determine the particle locations. This allows different processing techniques to be used to optimise the location data. The output from the PEPT algorithm is an ASCII file containing the average time of the location, three Cartesian coordinates describing the particle position, and a value which determines the quality of the location by calculating the standard deviations of the trajectories used to locate each event. The new data acquisition system has been used with a small modular camera setup consisting of eight detector buckets. These have all been configured to operate in coincidence and the data rate is consequentially high. In trial PEPT runs on a plastic extrusion process using this system sustained data acquisition rates of approximately 120 kHz have been demonstrated. Short
5. Summary The use of positron imaging techniques applied to industrial subjects has been introduced. Recent developments with the detector systems have resulted in a modular camera with flexible geometry allowing an increase in potential applications. Additionally the modular camera is transportable and has allowed PEPT studies to be performed at a number of sites across the University and off campus. Improvements in computer power and reductions in cost have allowed the development of a new data acquisition system constructed to interface with the modular camera. It can be used for data acquisition on a wide range of positron imaging devices as it offers high speed data transfer with handshaking capability. While its initial performance on small scale systems has been reported to be in excess of 1.7 MHz data transfer rate, the ultimate limit is expected to be much higher. The system has proved be reliable and operates simply and effectively. It will be used for many future PEPT studies. References [1] D.J. Parker, C.J. Broadbent, P. Fowles, M.R. Hawkesworth, P. McNeil, Nucl. Instr. and Meth. A 326 (1993) 592. [2] A. Sadrmomtaz, D.J. Parker, L.G. Byars, Nucl. Instr. and Meth. A 573 (2007) 91. [3] D.J. Parker, T.W Leadbeater, X. Fan, A. Ingram, Z. Yang, this issue. [4] D.J. Parker, T.W. Leadbeater, X. Fan, M.N. Hausard, A. Ingram, Z. Yang, Meas. Sci. Technol. A 326 (2008) 592. [5] W.F Jones, M.E. Casey, L.G. Byars, S.G. Burgiss, in: IEEE Nuclear Science Symposium, 1985.
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
A high speed PC-based data acquisition and control system for positron imaging T.W. Leadbeater , D.J. Parker School of Physics and Astronomy, The University of Birmingham, Edgbaston, Birmingham B15 2TT, UK
a r t i c l e in fo
abstract
Available online 10 February 2009
A modular positron camera with a flexible geometry suitable for performing Positron Emission Particle Tracking (PEPT) studies on a wide range of applications has been constructed. The demand for high speed list mode data storage required for these experiments has motivated the development of an improved data acquisition system to support the existing detectors. A high speed PC-based data acquisition system is presented. This device replaces the old dedicated hardware with a compact, flexible device with the same functionality and superior performance. Data acquisition rates of up to 80 MBytes per second allow coincidence data to be saved to disk for real-time analysis or post processing. The system supports the storage of time information with resolution of a half millisecond and remote trigger data support. Control of the detector system is provided by high-level software running on the same computer. & 2009 Elsevier B.V. All rights reserved.
Keywords: Positron imaging PET PEPT Data acquisition
1. Introduction Modern positron imaging devices utilise a large number of small gamma ray detectors arranged around the object under study. Position dependent event data are then derived from the detector response to collinear gamma ray emissions arising from positron annihilation within the field of view. The detection of both gamma rays in coincidence defines a line-of-response (LOR) joining the two detector elements along which the annihilation is said to have occurred. These LOR data can then be used to reconstruct a quantitative map of tracer concentration over a volume (Positron Emission Tomography, PET), or used to rapidly locate a single tracer particle present in the field of view by triangulation (Positron Emission Particle Tracking, PEPT) [1]. Fig. 1 illustrates these concepts; a number of LORs from a point source are shown interacting within a cylindrical detector array. LORs are then described by a signal comprising the transaxial crystal number (0–511) and the axial crystal number (0–15) for both end points of the LOR. At the University of Birmingham Positron Imaging Centre variants of the ECAT series of scanner produced by Siemens/CTI are used for the study of industrial and engineering apparatus. These scanners consist of 128 individual detector blocks, each containing an array of 8 8 Bismuth Germanate (BGO) scintillator crystals viewed by four photomultiplier tubes (Fig. 2, part A). Due to their modular construction; where the detectors are arranged together in a number of identical small sub-systems
(called buckets, Fig. 2, part B) it has proved possible to remove the detectors from the gantry frame and arrange them in different geometries, thus creating a modular positron camera suitable for PEPT studies [2]. Custom support gantries have been constructed to hold the modular detector elements around the apparatus under study. An example geometry is given in Fig. 3 showing 16 separate modules each containing four detector blocks. Some of the allowed lines of response are shown to illustrate the field of view. The sensitivity varies as a function of position, this is shown as the density of the lines. A small ring camera was developed using components from a Siemens ECAT-951 full body PET camera. Studies were performed on this system in order to ascertain the operating characteristics before it was used for a PEPT study of annular shear flow. Parker et al. report experience using the modular camera and give a full account of the performance of the initial system [3]. In summary of these results a maximum event rate of around 700 k s1 (prompt þ delayed) was seen for activities exceeding 80 MBq. Many of these events were random coincidence noise. The maximum true coincidence rate occurred with a tracer activity of 43 MBq; the recorded data rate at this point, 305 k s1 , was limited primarily by the data transfer speed using the original data acquisition system. Optimum tracer activity lies between 20 and 50 MBq for this initial trial system.
2. Positron emission particle tracking Corresponding author.
E-mail address: [email protected] (T.W. Leadbeater). 0168-9002/$ - see front matter & 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2009.01.184
The PEPT method uses an iterative approach to locate a single tracer particle within the field of view by triangulation over a
ARTICLE IN PRESS 356
T.W. Leadbeater, D.J. Parker / Nuclear Instruments and Methods in Physics Research A 604 (2009) 355–358
small number of LORs. Disregarding events corrupted due to scattering, a sequential set of LORs should cross each other at the location of the tracer particle to within the spatial resolution of the camera. It has been shown [1] that for a set of N detected events, with a fraction f used to locate the tracer, the precision D of locating a stationary particle is given approximately by w
D pffiffiffiffiffi
(1)
fN
Fig. 1. LORs from a point source within a positron camera.
where w is the spatial resolution of the camera. It can be seen that the value of D can be set to the desired precision by the appropriate choice of N. The value of N is proportional to the activity of the source, thus over a long enough time scale (increasing N) a stationary particle can be located arbitrarily well. Using a high activity particle with the correspondingly high data rate allows accurate and frequent location. For example; acquiring event data at a rate of 100 k s1 , with f equal to 10%, a stationary tracer can be located to within about 0.5 mm every 10 ms (N of 1000) using the modular camera (w of around 5 mm). During this time, however, a tracer particle moving at 2 m s1 will move 20 mm with the LORs spread along its trajectory. To deal with fast moving tracers it is necessary to locate more frequently with the corresponding loss of precision (reduced N). Typical values for N lie between 50 and 300 events. In the trial study described above only 16 out of a maximum 32 modules were used. Higher data rates are expected with larger
Fig. 3. Example modular camera geometry.
Fig. 2. Detector block (A) and detector bucket (B).
ARTICLE IN PRESS T.W. Leadbeater, D.J. Parker / Nuclear Instruments and Methods in Physics Research A 604 (2009) 355–358
systems using more detector modules. Recent trials with the initial system [4] have reported location rates of approximately 1.5 kHz with a measured standard deviation in three dimensions for a stationary particle of 2 mm. A higher location rate of 7 kHz resulted in a standard deviation of 10 mm for a stationary source.
3. Event data In a complete system (of 32 detector buckets) there are 8192 discrete detection elements allowing over 1:5 107 possible coincident combinations (LORs). These data are presented at the output of the coincidence processor in the form of a 32 bit parallel data word. There are 6 bits describing the pair of detector modules containing the event. Then, for both detectors, 2 bits describe the detector block in the bucket, 3 bits describe the row and 3 bits describe the column for the excited crystal segment in each block. Two bits identify the coincidence processor board which processed the event, and 2 event related bits flag multiple and delayed events. The remaining 6 bits in the word are not used by the coincidence processor. In the new data acquisition system these bits can be used to give additional functionality, for example defining an event time and flagging external event triggers. A relationship between crystal segment number and crystal location relative to a fixed origin is derived after measurement of the detector module position. When running the PEPT algorithm LOR data are converted into cartesian (x1 , y1 , z1 ; x2 , y2 , z2 ; t) coordinates giving the spatial position of the two interactions and the time the event occurred. This conversion is customised to the camera geometry used for the specific application.
4. Data acquisition electronics Following the initiating gamma-ray event and electronic handshake the maximum theoretical data rate from each coincidence processor (of four) is 4 MHz, resulting in a maximum of 16 MBytes per second requiring processing. The original ECAT system was custom designed using off-the-shelf components in a compromise between speed and cost. The design [5] allowed a maximum data rate of 1 MHz transmission to the storage and processing system. The primary limitations were due to memory access times, however, storage to disk in list-mode was significantly slower due to large disk access times. Data storage rates could be improved by using two acquisition systems operated in parallel but with increased costs and added complexity. The demand for high speed list mode data storage required for PEPT experiments has therefore motivated the development of an improved data acquisition system. This system is based around the PCI-7300A Digital I/O Card manufactured by Adlink technologies inc. It features 80 MBytes per second, 32 bit data transfer via direct memory access (DMA) along the PCI bus, with a further 8 bits auxiliary I/O and simple handshaking/trigger process capability. In order to interface with the existing coincidence processor system a small number of dedicated support circuits have been built. These use TTL technology and deal with the handshaking requirements, timing tag word insertion and remote trigger handling. Conventional PET imaging does not require high resolution temporal information to be stored, however, this is essential for PEPT experiments as locations are made on a frequent basis. Time information for each event is determined by the time-of-arrival of the event in the processing unit. A dedicated board adds a timestamp to individual events with a temporal resolution of 0.5 ms. This is sufficiently frequent for PEPT as the location time is
357
calculated from the average time of the trajectories used in the final solution. This can cover a number of milliseconds. The frequency of time events can easily be increased if the data rate from the new system is high enough that multiple locations occur between time events (within 0.5 ms). This would improve the temporal resolution of the PEPT study allowing for the accurate tracking of high speed particles. Other instruments with a wide range of possible applications (e.g. tachometer (industrial), cardiac gating (medical)) can be used to provide additional data for any study. With this in mind a remote trigger system has been included in the hardware to allow up to two extra devices to insert data into the list mode data stream. A bit is set when the remote input is triggered, and the data word inserted into the list-mode stream with a time stamp. Fig. 4 shows a block diagram for the acquisition system electronics. The coincidence processor is driven by a handshake driver and event data words are placed on the output FIFO (First In First Out memory stack) following a successful handshake and trigger. Time information and remote trigger information is inserted into the FIFO by triggering the data capture as events occur. The new data acquisition hardware has been designed to interface with the ECAT-931 coincidence processor made by Siemens/CTI. Recently it has been used to capture and process data from ECAT-951 series detectors and is able to process data from similar systems with little or no modification. It is also possible to connect the new data acquisition system to multiple coincidence processor units allowing a further 32 detectors to be used, however, there has not been a need to experiment with a system this large. Control of the coincidence processor and individual detectors is made possible by high level software running on the same computer. In this a serial RS-232 link is used to interface with the systems, the software converts user level commands into machine code used to communicate with the camera. In this manner detector calibration is performed and system checks and event data can be accessed. The entire system is illustrated in the block diagram given in Fig. 5. Here the detectors and coincidence processor are controlled by serial commands issued from the computers USB port. Data from the individual detector modules undergo coincidence processing and the resultant coincidence data word is captured following successful handshaking. Data is transferred as a block from the local FIFO to a circular buffer sitting in system memory via DMA along the PCI and system busses. The PC resources can be
Fig. 4. Acquisition system electronics.
ARTICLE IN PRESS 358
T.W. Leadbeater, D.J. Parker / Nuclear Instruments and Methods in Physics Research A 604 (2009) 355–358
bursts of above 1.7 MHz have been seen during periods where the tracer particle was located in areas of high sensitivity. This corresponds to writing between 1 and 15 MBytes per second to disk. The increased data rate has allowed more trajectories (increased value of N) to be used for each PEPT location without the corresponding loss of accuracy due to tracer motion. In these experiments accurate locations were made every 11.3 ms on average using 2000 events per location (N). For this application a laser tachometer was used to correlate tracer motion with that of the extrusion shaft by inserting data using the remote trigger. While the new system has been seen to operate with performance exceeding the maximum performance of the initial system, it has so far not been tested to determine its full limits; these are expected to be much higher than reported here. Fig. 5. System block diagram.
used for processing the data, performing data analysis, and displaying real-time LOR data. The circular buffer consists of eight buffers of 32 kbits (10k array of 32 bits) wide. Data are transferred asynchronously from system memory to disk as each buffer is filled. This is a potential bottleneck as disk write speed is again the slowest transfer rate in the system. Buffers which are dropped due to slow disk access have been flagged, in tests this has not occurred. If this is deemed to be a problem in the future data can be transferred from the circular memory buffer into conventional system memory for temporary storage. It can then be transferred to disk post acquisition or during periods where data acquisition rates are low. Raw event data can be saved to disk or removable storage media (CD/DVD) in binary format. These raw data files can then be processed using the PEPT algorithm to determine the particle locations. This allows different processing techniques to be used to optimise the location data. The output from the PEPT algorithm is an ASCII file containing the average time of the location, three Cartesian coordinates describing the particle position, and a value which determines the quality of the location by calculating the standard deviations of the trajectories used to locate each event. The new data acquisition system has been used with a small modular camera setup consisting of eight detector buckets. These have all been configured to operate in coincidence and the data rate is consequentially high. In trial PEPT runs on a plastic extrusion process using this system sustained data acquisition rates of approximately 120 kHz have been demonstrated. Short
5. Summary The use of positron imaging techniques applied to industrial subjects has been introduced. Recent developments with the detector systems have resulted in a modular camera with flexible geometry allowing an increase in potential applications. Additionally the modular camera is transportable and has allowed PEPT studies to be performed at a number of sites across the University and off campus. Improvements in computer power and reductions in cost have allowed the development of a new data acquisition system constructed to interface with the modular camera. It can be used for data acquisition on a wide range of positron imaging devices as it offers high speed data transfer with handshaking capability. While its initial performance on small scale systems has been reported to be in excess of 1.7 MHz data transfer rate, the ultimate limit is expected to be much higher. The system has proved be reliable and operates simply and effectively. It will be used for many future PEPT studies. References [1] D.J. Parker, C.J. Broadbent, P. Fowles, M.R. Hawkesworth, P. McNeil, Nucl. Instr. and Meth. A 326 (1993) 592. [2] A. Sadrmomtaz, D.J. Parker, L.G. Byars, Nucl. Instr. and Meth. A 573 (2007) 91. [3] D.J. Parker, T.W Leadbeater, X. Fan, A. Ingram, Z. Yang, this issue. [4] D.J. Parker, T.W. Leadbeater, X. Fan, M.N. Hausard, A. Ingram, Z. Yang, Meas. Sci. Technol. A 326 (2008) 592. [5] W.F Jones, M.E. Casey, L.G. Byars, S.G. Burgiss, in: IEEE Nuclear Science Symposium, 1985.