How do we solve this problem? A network exists in which analysis activites are related to one or more of the objectives. Drawing this network allows us to determine what intermediate steps must exist. The network also shows whether a given task is essential or peripheral. There may be some overlap in the outputs of some of the activities, in which case comparison of their results will help validate the analysis process. The presently understood data analysis network is shown in Figure 1.
We identified the main sources of Basic Data, including data derived from the flight instrument, and data already available in the record of laboratory work. We then listed the information that could be found from these data sources, which we called the Derived Data. We then started to determine how the intermediate products could be synthesised into Final Data sets. Data sets and Activities are referred to by Letter and Number, as shown in Table 1.
Category Identifier Objectives, Goals G Basic Data sets B Derived (intermediate) data sets D Final data sets F Analysis or data gathering tasks A Models M
Table 1. Data or Activity category and identifiers.
Goal or Objective Identifier Describe useful future detector life G1 Maximise life G1.1 Efficiency vs. Observation condition G2 Performance optimisation G3 GSET s G3.1 Procedure G3.2 Processing G3.3
Table 2. Goals of the calibration programme
The relative importance of these goals will be a matter for further discussion.
In the table of Goals, Efficiency should be regarded as a quantity that represesents the Detective Quantum Efficiency as a function of position (on the detector). Observation condition stands for any adjustable parameter of the detector (GSET number, HV, slit number, MCP age, etc.) or condition of the source (spectrum,.flux).
Subject to a compromise analysis (A15), the existence of G2 allows G3 - Performance optimisation, to be achieved. The output of this, whilst not known at this stage, might be a recommendation to acquire data in a particular way, together with GSET parameters to be used, a procedure for data taking, and a description of any necessary post-processing steps that might be required.
It should also be understood that there may be a conflict between Goal 1.1 "Maximise Life" and Goal G3 "Optimise Performance". If this is the case then a compromise should be sought.
Data Category Identifier PHD vs. Position and Time F1
Table 3. Final data category
At present, there is only one data set in this category. It will be a record of the Pulse Height Distribution at all positions on the detector surface, consistent with all known observations in the past, and capable of extrapolation to future conditions.
An important input to the preparation of this data set will be a model of the Pulse Height Spectrum (M1).
Using data set F1, it should be straighforward to predict the quantum efficiency at any given condition of observation (G2). Goals G1 and G1.1 can then also be stated.
Data Category Identifier Science Data SPECT1 study data B1 GIMCP study data B2 other study data B3 Raw Data All raw dumps B4 Other Telemetry Engineering telemetry B5 Special Calibrations Flat field vs. LLD - obtained 18-19 B6 May (GC011) Lines vs. LLD B7 Lines vs. HV B8 Laboratory, Design and Theory Line Positions B9 Gain vs. HV B10 Gain vs. Charge B11 slit B11.1 flat field B11.2 Gain vs. Rate B12 position-dependance (aka B12.1 "long-range") Electronics Description B13
Table 4. Basic Data categories
Analysis [A1/GC022] consists of determining the area of a (pre-defined) selection of lines, and plotting these against time. The plots will also be examined for evidence of "ghosting", etc.
Analysis [A2] consists of defining a metric [GC023] which is sensitive to such differences, and plotting this metric against time [GC024].
A possible drawback is that there may exist periods in which the detectors were active but no study was ongoing.
Some rudimentary PHD as a function of position can be derived from the raw data. This is because the MCP/SPAN system exhibits dependance of "R" on pulse height.
Analysis consists of taking slices of event density across the R-theta locus at the position for which pulse height information is desired.
Activity [A7] denotes extraction of TM data using emon. Actions [GC015, GC021, GC026, GC027, GC028] also relate.
The value of this data was discussed at previous meetings [GC011], and a plan of operations was drawn up by Eddie with Alice and Matt. At the time of writing, some of the results are in the bag, though maybe not all.
Justification: Ideally, the sensitivity (Quantum efficiency, QE) of the MCP detectors should be fairly uniform across the face of the detectors. If there are any variations then these must be qauntified.
The QE is partly determined by the location of the peak in the MCP pulse height distribution (PHD) in relation to the setting of the Lower Level Discriminator. If, as we believe may be the case, the PHD is a function of position-across-the detector - due to count-rate related ageing effects in the MCP - then adjusting the Discriminator and observing the data for a constant "flat" field (i.e. the filament) should allow us to determine where such variations in PHD exist.
If there is a gross change in the PHD at a single position, then one might see two peaks in the PHD. Acquiring PHD's at other voltages will allow us to see features in the PHD that would not have been seen in the above LLD scanning tests.
Further analysis [A9] will consist of determination of the differences between the flat fields, and plotting these against LLD setting. If there are many positions where this occurs then the differences should be plotted at each position.
The following items are laboratory/theoretical data and are all required.
Data set description Identifier Source(s) Line area vs. time D1 B1 "Metric" of Slit 1 vs. Slit 3 vs. time D2 B2 Total count vs. position D3 B3, D7 PHD vs. Time D4 B4 RMSIG D5 B4 PHD vs. HV D6 B4 Total, Processed & ULD counts D7 B5 HK parameters D8 B5 Efficiency vs. Position D10 B6 Efficiency vs. Line D11 B7, B8 Historical Record D12 logbooks
Table 5. Derived (intermediate) data sets
RMSIG is Root Mean Square Inverse Gain, a previously used metric of Pulse Height Distributions.
HK Parameters are the (TBD) set of telemetry parameters derived from telemetry files.
Efficiency is ideally Relative Detective Quantum Efficiency, although it might be found that some derived parameter (e.g. slope of count rate vs. LLD setting) is all that can be known directly.
Historical record is a list (file?) of important details in the detector history, such as dates of HV changes, turn-on or off, which can then be the basis of an estimate of Total Charge vs. Position.
Other parameters should need no further explanation.
Model Identifier Model of PHD vs. Any parameter M1 Adaptation of Simpha M2
Table 6. Models used in analysis tasks
As alluded to above, some of the observations or derived data sets may not be directly interpretable. To make sense of the data, it has been proposed to construct a model of the pulse height distribution of the MCPs, which can then be used to simulate the distribution at any given condition.
An existing model (simpha) of detector performance under conditions of varying pulse height distribution can be adapted [M2] to the present problem. Simpha was produced (by Matt) to help understand gain depression effects in the Yohkoh-BCS proportional counters. It was described in the meeting on 17 May. Model M1 will be a necessary component of the adapted scheme, which should be able to reproduce the effects of the observations.
M2 is not yet shown on the network, but its use will probably be a component of A14.
Activity Priority People Identifier / Action Analysis of SPECT1 data high Eva A1 / GC022 Analysis of GIMCP data high Alice, Eva A2 / GC023, GC024 Derive Total Count from Alice A3 Science observations PHD vs. Time from Raw Dumps Alice A4 "RMSIG" Metric from Raw Alice A5 Dumps PHD vs. HV from Raw Dumps A6 Event Counters from TM if Eva A7 archive needed PHD vs. Time from trickle A8 data Analysis of FlatField vs. Alice, Matt A9 / GC008, GC009 ULD data Analysis of Lines vs. LLD, Later? A10 / GC007 Lines vs. HV Define content of history "database" General Analysis Alice, Matt, A14 Alan Performance Tradeoff A15 Analysis
Table 7. Plan of work
A14 - General Analysis is a catch-all task representing all future activities that lead to understanding of the evolution of the PHD.
The meaning of other tasks will be apparent from their description and relationship with their input and output data sets.