:::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.1 :::::::::::::: Minutes from PCA Background meeting 4/17/98 Those present: John Cannizzo, Tess Jaffe, Keith Jahoda, Paul Nandra, Mike Stark, Tod Strohmeyer, Jean Swank, Nick White & Will Zhang. The purpose to these meetings is to find a solution to the problems with the RXTE background model, so that the AGN community can satisfactorily analyze and publish their data. A secondary objective to resolve other related issues to do with the ease of analysis of AGN data and to take into account the differences that weak source analysis requires relative to bright sources. The meetings will be held weekly, on Fridays from 10.30-12 noon (except for the 1st Friday of the month when the X-ray brnach meeting takes place), until the problems are satisfactorily addressed to the communities satisfaction. The goal is a PCA background model that matches that achieved by Ginga of 0.01 ct/s/channel << 10keV and 0.02 ct/s/chan > 10 keV. The following decisions and action items came out of the meeting: 1) Before background models are publically released in future, they must be tested against blank sky, earth and slew data to ensure they meet the necessary quality criteria. Test data sets of blank sky and slew data will be assembled and kept permanently available on hard disk to enable rapid testing of the new background models (Tess, Keith, Paul). These tests will be done by somebody other than the person/people creating the background model. 2) A series of tests and criteria will be defined by Paul Nandra that will ensure that the background model is of sufficient quality to satisfy the AGN communities requirements. 3) Jean Swank will actively pursue hiring a new scientist who works on AGN and is also able/capable to work on background models. 4) Mike will make an internal release in the next week of the new VLE background model, that is based on the blank sky points made between late 96 and early 98. This will be tested against the blank sky points to identify the long term trends and to help with the strategy to model them (Tess, and the GOF). 5) The long term cycles and trends in the background data will continue to be studied and models devised to be included into the VLE background model. There will be an investigation of the various physical processes that are responsible for these variations (Keith, Will). 6) Background monitoring observations lasting 1000s, made every 12 hours will be added to the RXTE program, until further notice. These will provide an additional baseline of background data to test the stability of the background model (Jean). 7) It should be easier to analyse AGN data, i.e. it needs to be made less labor intensive. The GOF will look into this and will: - add all the relevant data selection parameters to the filter file - look into setting up a script or scripts to make it easier to collect together data from AGN observations where multiple pointings are made over long intervals (Tess, Alan) 8) A database of slew data that can be used to test the background model will be constructed (Mike, Tod). 9) A well defined method to work out the systematic errors in the background subtraction must be written (TBD) 10) Once a background model is devised, and passes the testing on background data, then a large cross section of data from the AGN community will be analyzed in house by the GOF and PCA teams to see how well the background model is working for a diverse set of data sets. The resulting data products will be sent to the PI for their feedback, and hopefully their publication (GOF and PCA teams). Next meeting, Friday April 24, conf room W20 Nick White :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.2 :::::::::::::: Minutes from PCA Background meeting 4/24/98 Those present: John Cannizzo, Tess Jaffe, Keith Jahoda, Jim Lochner, Frank Marshall, Paul Nandra, Gail Rohrbach, Mike Stark, Tod Strohmeyer, Jean Swank, Nick White & Will Zhang. Action items from previous meeting: 1) Before background models are publically released in future, they must be tested against blank sky, earth and slew data to ensure they meet the necessary quality criteria. Test data sets of blank sky and slew data will be assembled and kept permanently available on hard disk to enable rapid testing of the new background models (Tess, Keith, Paul). These tests will be done by somebody other than the person/people creating the background model. Tess has moved the Edelson bgnd, the AO3 and AO2 data to the PCA team disks. Tess will send an email around describing where the files are for the uninitiated. 2) A series of tests and criteria will be defined by Paul Nandra that will ensure that the background model is of sufficient quality to satisfy the AGN communities requirements. Paul circulated a note yesterday outlining a strategy, this will be discussed at the next meeting. There was some concern that the background sensitivity level to be achieved, as described in the previous minutes was not correctly defined. It was decided to quote this per KeV, not per channel. And also to quantify any dependency of this goal on exposure. Paul Nandra will redefine these objectives in terms of cts/s/keV and will circulate them to the group. 3) Jean Swank will actively pursue hiring a new scientist who works on AGN and is also able/capable to work on background models. Three candidates are under active consideration. 4) Mike will make an internal release in the next week of the new VLE background model, that is based on the blank sky points made between late 96 and early 98. This will be tested against the blank sky points to identify the long term trends and to help with the strategy to model them (Tess, and the GOF). The new model was released and testing is underway by Mike, Will Zhang and others. A discrepency in the results from pcabackest and the IDL programs was reported by Will and will be investigated by Will and Mike. It was agreed that any tests of the model should concentrate on: 2-10 keV band Layer 1 in 256s bins 10-20 keV band Layer 1+2 in 256s bins i.e. characteristic of those used for AGN analysis. Keith Jahoda will circulate the channel boundaries corresponding to the 2-10 and 10-20 keV to be used for the background model testing. 5) The long term cycles and trends in the background data will continue to be studied and models devised to be included into the VLE background model. There will be an investigation of the various physical processes that are responsible for these variations (Keith, Will). Keith Jahoda reported on promising results from his fitting three time constants of 22, 220, & 6500 minutes giving a reasonable chi2. This appears to take into account some, if not all, of the long term effects. The exception seems to be for data taken within 30 min of the peak of the SAA on any particular SAA orbit. Since the detector is off for 15 min of this time, its a relatively small loss. Further testing is required to see how this new model works. Since the model is probably made of of several discrete radioactive lines decay, the next step will be to reconstruct the coeficients for different spectral resolution elements. There was opinion expressed that this may be a significant breakthrough. Mike Stark will modify pcabackest to include this new model, with Keith Jahoda providing the components to be input. The new version ready for testing within 1 week, or shortly thereafter. Mike Stark will revise the current background model to include long term trends, in particular the 30 day cycle. This will provide a 2nd background model to compare to the 3 time constant model. 6) Background monitoring observations lasting 1000s, made every 12 hours will be added to the RXTE program, until further notice. These will provide an additional baseline of background data to test the stability of the background model (Jean). These observations have begun. There was much debate as to whether additional observations are required to: - gain sufficient data to determine the background model parameters in each spectral resolution bin. - establish if there are any other activation decay timescales that are being missed by the current sampling The conclusion was that a 4 day background observation would be beneficial on both counts and this will be scheduled as soon as possible, but to minimize disruption of coordinated observations. 7) It should be easier to analyse AGN data, i.e. it needs to be made less labor intensive. The GOF will look into this and will: - add all the relevant data selection parameters to the filter file - look into setting up a script or scripts to make it easier to collect together data from AGN observations where multiple pointings are made over long intervals (Tess, Alan) This is ongoing. A significant breakthrough is that Keith Arnaud has managed to read standard mode data into xselect v1. He is working on adding filtering for different layers, pcu, etc... This will hopefully enable rapid analysis and reanalysis of the data, and open up the accessiblity of the RXTE data to people more used to ASCA and ROSAT analysis (e.g. Richard Mushotzky). Tess and Gail will discuss with Paul Nandra changes to the analysis scripts that would make it quicker and more efficient to analyze the data. Keith Jahoda will send to Tess and Gail a list of PCA science data parameters that should be included in the filter file. 8) A database of slew data that can be used to test the background model will be constructed (Mike, Tod). Mike Stark has constructed the slew database, and will circulate an email to tell people how to access it. He is also going to construct a new database of dark earth data, to be used to search for other long term variations in the data. This will complement the sky background data, and because it is much deeper, can be used to refine the background model parameters. Before this can be done, differences between the sky and earth occult data need to be resolved and this will be the first use of this new database. 9) A well defined method to work out the systematic errors in the background subtraction must be written (TBD) No progress. 10) Once a background model is devised, and passes the testing on background data, then a large cross section of data from the AGN community will be analyzed in house by the GOF and PCA teams to see how well the background model is working for a diverse set of data sets. The resulting data products will be sent to the PI for their feedback, and hopefully their publication (GOF and PCA teams). The GOF will plan to support manual processing of critical AGN datasets to ensure that once the acceptable background model is ready, it can be quickly turned around to generate lots of high quality instantly publishable results. Next meeting, Friday May 1, conf room W20 at 9am. Nick White :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.3 :::::::::::::: Minutes from PCA Background meeting 5/1/98 Those present: John Cannizzo, Tess Jaffe, Keith Jahoda, Jim Lochner, Frank Marshall, Paul Nandra, Gail Rohrbach, Alan Smale, Mike Stark, Tod Strohmeyer, Nick White & Will Zhang. The meeting dealt mostly with a discussion of a new background model being worked on by Will Zhang that correlates the L3 rates against the L1. This seems to give a very good correlation and tight residuals. Further work is underway to compare this against the pcabackest model. Action items from previous meeting: 1) Test data sets of blank sky and slew data will be assembled and kept permanently available on hard disk to enable rapid testing of the new background models (Tess, Keith, Paul). This is essentially complete. 2) A series of tests and criteria will be defined by Paul Nandra that will ensure that the background model is of sufficient quality to satisfy the AGN communities requirements. Paul Nandra has recirculated the numbers and these will be posted on the web site. The feeling was that a better reference number was to get the background to within 30% of the diffuse background number. Beyond that source confusion would dominate the overall errors. 3) Jean Swank will actively pursue hiring a new scientist who works on AGN and is also able/capable to work on background models. Two candidates have been interviewed. A third candidate will be seen next week. 4) The VLE background model. Comparison with the new L3 model is underway. This needs to be done for the energy range specified above. Keith Jahoda will circulate the relative count rates in the three layers for an AGN with a 1.7 power law. 5) The long term cycles and trends in the background data will continue to be studied and models devised to be included into the VLE background model. Incorporating the new 3 time constant model in pcabackest is proceeding well. It should be available for testing by the next meeting. 6) Background monitoring observations: The conclusion was that a 4 day background observation would be beneficial on both counts and this will be scheduled as soon as possible, but to minimize disruption of coordinated observations. The 4 day observation will take place May 12. 7) It should be easier to analyse AGN data: Tess is working on a Perl script that incorporates all of various AGN peoples inputs as to how to best streamline the analysis. This will be a three part script, that will do the data collection, the background subtraction and the quality assessment. Alan Smale will look into getting good xenon into xselect and tidying up Keith Arnauds pioneering work on getting standard mode data working. 8) A database of slew data that can be used to test the background model will be constructed (Mike, Tod). No progress 9) A well defined method to work out the systematic errors in the background subtraction must be written (TBD) No progress. 10) Once a background model is devised, and passes the testing on background data, then a large cross section of data from the AGN community will be analyzed in house by the GOF and PCA teams to see how well the background model is working for a diverse set of data sets. The GOF is gearing up to support this, and sees no problem in being able to do so. Next meeting, Friday May 8, conf room W20 at 10.30am. Nick White :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.4 :::::::::::::: Minutes from PCA Background meeting 5/1/98 Those present: John Cannizzo, Tess Jaffe, Keith Jahoda, Jim Lochner, Frank Marshall, Paul Nandra, Alan Smale, David Smith, Mike Stark, Tod Strohmeyer, Nick White & Will Zhang. The meeting continued the discussion of the new background model being worked on by Will Zhang that correlates the L3 rates against the L1, comparing it against the 3 time constants model. Both models seem to work well, although there is a residual 1 hr variation in the L3 model that may be due to collimator activation. The following points were made: i) We need to compare the new background models with the currently released pcabackest, to demonstrate the improvement. ii) The 3 time constants model seems to show occasional spikes in the earth occult data. These might be due to solar illumination of the bright earth. Screening on bright/dark earth needs to be implemented iii) We need to test different electron screening criteria to see if there is a natural cut that will improve the background models iv) Putting the 3 time constant model into pcabackest is progressing but is not yet working. It should be straightforward to add the L3 model, when its ready. v) Standard plots were agreed to compare all the different methods being tried vi) We should consider better sampling the 30 day cycle. Keith wil investigate which phase the May 12 observation will take place. vii) Tod and Mike are working on collecting together the slew data viii) The GOF is working on collecting all the vital selection parameters into the filter file. Time since SAA is in, but has -ve numbers because of the formulae used, which can be confusing. This will be fixed. We need to also include a dark/bright earth flag. ix) Test scripts that make it easier to analyze lots of AGN observations are available for testing and Paul Nandra + the GOF will test them x) There was some discussion about the ease of use of XDF - the GOF will investigate and the PCA team will try it out to see if its improved from their initial impressions just after launch. xi) putting xte into xselect is progressing well. Next meeting Friday May 15, at 10.30am in W20..... :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.5 :::::::::::::: Minutes from PCA Background meeting 5/15/98 Those present: Tess Jaffe, Keith Jahoda, Frank Marshall, Paul Nandra, Alan Smale, Mike Stark, Tod Strohmeyer, Jean Swank, Nick White & Will Zhang. The "L9" background model (named after the number of background monitoring rates included) being worked on by Will Zhang is making extremely good progress. It seems to give a very accurate subtraction in the 2-10 keV band down to the sky fluctuation limit. This model only uses rates from the L2 & L3 layers and from other parts of the detector that do not show the bimodal rate distribution. There was discussion about whether L2 and L3 should be included, but since for the 2-10 keV band limit used for these rates there is effectively no source background for the sources of interest it was decided to leave these in to gain the maximum statistics. This correlation is very tight and appears to be universal over the mission lifetime. Will Zhang thinks the activation is due to electrons from beta decay and so may be relatively energy independent. Mike Stark and Keith Jahoda will give high priority to putting this into pcabackest so that Paul Nandra and the GOF staff can test drive it on datasets of interest. Getting the three time constants model working in pcabackest is still having problems. Keith and Mike are working on it. But they will give priority to getting the L9 mode in first. The next steps with the L9 model are: 1) try it with the 10-20 keV band 2) slice spectra based on L9 rate 3) to try it on the Edelson background monitoring data Progress with the L9 model is so good that it is probable results from it can be presented at the AAS meeting in San Diego. The GOF staff will assemble on disk critical data that will be presented at the meeting, and will be ready to process it through for distribution to the speakers and PIs. Assembly of the slew database is making good progress. Tod showed an impressive sky coverage map that shows PCA has covered a large fraction of the sky, with some areas covered to a large depth. Tess, and the GOF staff are working hard to support PCA team needs for test data. They are also making good progress on the scripts to make data processing easier. Alan Smale sent the following report about the progress with xselect: "Keith_A and I have been bouncing things back and forth, fixing problems and making progress. MAKE OBSCAT now works, and will find all the Std2 and PCABACKEST files in the directory treee beneath the pathname you give it. Thus the 'usual' XSELECT methodology can be used, and XDF (or Unix competence) is no longer required. Some GTI complexities remain, including the rather crucial one that we can't yet select source data based on values in the filter file. We're hoping to fix this within the next couple of weeks, and also figure out more details about how to handle the background analysis within XSELECT. I'll then write up a quick recipe (one or two of the steps in analysis aren't obvious to the neophyte) and we'll let some other local people loose on it, e.g. Tess "Codebreaker" Jaffe, Paul, and even (if we're feeling especially brave) Mushotzky. One encouraging sign: Keith gave Kim [Weaver] a demo of the new XTE-friendly XSELECT while they were both here for the ASCA meeting, and her only comment was to ask when she could get her hands on it." Next meeting May 22, at 10.30am in W20 :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.6 :::::::::::::: Minutes from PCA Background meeting 5/22/98 Those present: Padi Boyd, John Cannizzo, Tess Jaffe, Keith Jahoda, Frank Marshall, Gail Rohrbach, Mike Stark, Tod Strohmeyer, Jean Swank, Nick White & Will Zhang. The "L9" background model is now available in pcabackest. It appears to provide an excellent solution across the entire PCA energy range (2-60 keV). Its use on different blank sky points leaves good residuals, with only a sky fluctuation spectral feature left between 2 and 10 keV that varies from pointing to pointing (0.5 ct/s variations overall). When the L9 model is used on bright sources, contamination of the L2 and L3 layers by the source is causing an over prediction of the background. So a new L7 model that does not include the L2 and L3 data is being built and will be available in pcabackest later today. It is expected that this will provide the final solution. There was optimism that the final result will give a major gain over the sensitivity of Ginga above 20 keV. As soon as the L7 pcabackest model is available later today the GOF will start to run through the critical observations that are already on disk. The Edelson background monitoring observations will be a first priority. The GOF and PCA team will meet to prioritize other observations to be run through the new procedure over the coming week. The objective is to make these new data available to the PIs by next Friday (if not before), in time for their AAS presentations. If there is time, and it seems the model is working fine, consideration will be given to releasing binaries of the new pcabackest (this will be discussed further at the next meeting). Before the GOF starts to run through these data the PCA team will agree upon the best filtering criteria to use and then send them to the GOF. In particular the correct propane layer cut needs to be agreed. As the GOF processes data, the results will be shown to the PCA team for verification and validation, before it is sent on to the relevant PI. GOF and PCA staff were requested to forsake part of the Memorial day weekend so as to ensure results could be prepared in time for AAS presentations. A few technical issues will be looked into: - the new model works well up to 20 keV. But to make it work well over the entire band will require the long term drifts in the detector gain and the overall background spectrum to be removed. This will probably not be possible by next week, but should be available relatively quickly. - procedures for estimating the systematics errors in the background need to be given. The biggest is to allow the user to see how much background fluctuations will effect the spectral fits. Another is to simply include any other systematic changes due to the RMS variations from observation to observation. The next meeting will be May 29, at 10.30am in room W20. :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.7 :::::::::::::: Minutes from PCA Background meeting 5/29/98 Those present: Alan Smale, Padi Boyd, John Cannizzo, Tess Jaffe, Keith Jahoda, Frank Marshall, Gail Rohrbach, Mike Stark, Tod Strohmeyer, Jean Swank, Nick White & Will Zhang. Will Zhang showed that the residuals from the L7 model are correlated with the good event rate in layer 1. Keith Jahoda then showed that this correlation can be explained by activation during passages through the SAA. Keith has produced a new background model using L7 and activation in the SAA that decays with a 240-minute time constant that produces a much better fit to the blank-sky data than does the L7 model. Keith will make these results available on a Web page along with calculations of the rms residuals for the L7 and L7_240 models. He will also send email to each of the AAS presenters informing them of improvments in the background modelling that were found for our blank-sky pointings. Paul Nandra showed results of applying the old L7 model produced a good improvement over the old for his earth occult data. The remaining residuals should be removed by the new model. The new L7_240 model is now available in pcabackest, and the GOF will process Edelson's and Nandra's data using the new model. There will be a short meeting late Monday to consider the results of the runs made over the weekend. Keith Jahoda will contact the AAS presenters to help them understand the implications of these improvements on their results. The next meeting will be June 5, at 9.00 am in room W20, NOTE THE EARLIER TIME! :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.8 :::::::::::::: PCA Background meeting, June 19, 1998 Present: Padi Boyd, John Cannizzo, Keith Jahoda, Jim Lochner, Gail Rohrbach, Mike Stark, Tod Strohmayer, Jean Swank, Concerning the June 26 patch: - a new SAA history file, valid for all times through May 1998 has been created. Periodic updates to this file will be required. The current method of generating this file involves getting Duane Gruber at UCSD to process pcaket data one week at a time. Although straight forward, there was consensus that it is desirable to move this routine processing to the SOF if possible. Keith/Gail to explore the possibilities here - a new model file for the 240 minute component is expected by early next week. The new file will include spectra for PCU 3 and 4 (currently not included). Keith/Mike to deliver - we will clarify the status of warning messages in pcabackest when the SAA history file is not present. A version that produces a warning has been delivered though not universally installed locally. The version with the warning is to be in the patch - the names of the model files will be changed from those in use locally. Two model files, to be used together, will be distributed, one containing the "L7" model and one containing the 240 minute component model and the associated constant. (Mike) - the release notes documentation will include note on how to install (Mike Tripico/James Peachy), what is in the new pcabackest (Keith), what is in the new xtefilt and new fxbary (Gail/Jim). Drafts to be circulated if possible on Monday June 22. - a banner announcement will be added to the XTE GOF page, and a notification will be sent to the XTE_NEWS list (Gail/Alan) - recipes that concern the use of pcabackest will be updated or modified as appropriate (Padi/Jim/Gail/Tess) Concerning next steps: - Keith and Mike are working on a model with Vx and three time constants, hoped to be useful for brighter sources (and potentially for the fainter sources as well). - Keith/Mike will estimate a delivery date for an L7 model that can be applied to epoch 1/2 data. next meeting: June 26 :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.9 :::::::::::::: Minutes of the June 26, 1998 meeting present: Padi Boyd, John Cannizzo, Tess Jaffe, Keith Jahoda, Gail Rohrbach, Alan Smale, Mike Stark, Tod Strohmayer, Mike Tripico, Nick White The spectra for 3C273, obtained July 96, show an offset between the peaks in the data near 25 and 30 keV and the corresponding peaks in the model. For this data set, the shift turns out to be 0.7%. This is one of the larger observed shifts, probably due to the bulk of the data that went into the model coming from 11/96 through 3/98 and the presence of a slight gain drift. Correcting this problem was identified as the highest priority for immediate efforts. Calibrating the shift as a function of time is therefore the most urgent matter; algorithms for shifting the predicted background exist, and will need to be tied into pcabackest or a follow-on tool (TBD) Second priority: extending the model to brighter sources. Third priority: extending the model to epochs 1 and 2. There was lots of discussion about the release of the patch, which in fact occurred June 26 although the announcement was circulated on June 29. :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.10 :::::::::::::: PCA background meeting July 17,1998 present: Keith Jahoda, Tess Jaffe, Jim Lochner, Paul Nandra, Dave Smith, Tod Strohmayer, Nick White Personnel: Dave Smith (formerly of Leicester University) has arrived this week, and will participate in our on-going efforts to characterize and model the PCA background. Gain shifts: The gain shift has been parameterized by measuring the background peaks at ~25 keV and ~30 keV. PCABACKEST is expected to incorporate this parameterization by middle to late next week. Analysis scripts: XSELECT is close to working on RXTE standard data, although it has not been released to the community. Paul and Tess will work next week on a previously unanalyzed data set with (a) scripts developed by Tess and Gail, and (b) a script developed by Keith. User friendliness and usefullness of the two scripts will be compared with the idea of possible clean up and release to the user community. Global data sets: Dave may start working on the mission occulted data set; we have collected enough data at 12 hour intervals that it is now appropriate to apply the L7 model to this data (>~ 2 months) and to the 4 day background observation performed recently. Tess and Dave will work together to get this data on disk. next meeting: July 24 :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.11 :::::::::::::: PCA background meeting, August 14, 1998 Present: Glenn Allen, Padi Boyd, Tess Jaffe, Keith Jahoda, Paul Nandra, Gail Rohrbach, Alan Smale, Dave Smith, Mike Stark, Jean Swank - newest PCABACKEST is successful at correcting for the gain shift. The parameters which describe the shift enter the model file as keywords. (note added later: we/users should ensure that when several models are in use such as L7 plus the 240 minute time constant, that the gain shift parameters are consistent) - Mike observed, in the course of using pcabackest fullspec=yes, that the current "interpolation" scheme can cause problems. The current scheme simply divides counts equally among channels, which is fine where the background is flat, but inadequate in the presence of lines. A better interpolation scheme may improve our ability to translate epoch 3 background models to epoch 1 (to be determined via testing) - the ftools freeze is Sept. 1. We will aim to have the L7 models (for all 3 epochs) supported by this time, as well as all code needed to support the Vx + 3 time constant models (which should be better fro brighter sources) delivered. The Vx suite of models will not be ready by the Sept. 1 date, though the aim is to deliver them before the freeze on associated files. - Analysis of the 4 day background observation (D Smith). - a small daily term remains in the 2-10 keV band - pcabackest performance is poor in the 25-35 keV region - only 3 of the 4 days seem to be available; Gail will check on the processing - Links to the standard recipes from the PCA page would be useful (Keith) minutes prepared 8-28-98/KJ :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.12 :::::::::::::: PCA Background meeting, August 24, 1998 Present: Bish, Padi Boyd, Tess Jaffe, Keith Jahoda, Jim Lochner, Paul Nandra, Gail Rohrbach, Alan Smale, Dave Smith, Jean Swank, Nick White - as previously reported, pcabackest can account for the gain shift. However, the amplitude of the lines, particularly at 30 keV, is not quite right. Several toy models can explain why the line amplitude might not be tracked by L7, particularly since part or all of these lines consists of unflagged Am-241 calibration events. - Dave examined several bands of L7, but without strong conclusions on whether the peak was correlated with or independent of the continuum. The first studies used PCU 0, where the lines are not as strong as in PCU 1 or 2. - Mike (on annual leave) has made fixing the interpolation in pcabackest his next priority - Keith has made definition of the Vx model (anticipated to work for larger source rates) his next priority. - Questions about the partial charge collection in the response matrix came up; the Inoue et al. parameterization is not currently implemented, but the recent improvements around the Xenon L edge, by including time dependent effects, warrant another attempt at this. - Paul and Tess have discussed "standard AGN analysis" scripts. Tess is working on these, but the scripts are unlikely to be included in the September 1 FTOOLS freeze. - next meeting: back to Friday, 8/28 prepared by KJ/8-28-98 :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.13 :::::::::::::: PCA background meeting, August 28, 1998 Present: Glenn Allen, Padi Boyd, Tess Jaffe, Keith Jahoda, Donna Pierce, Gail Rohrbach, Alan Smale, Dave Smith, Mike Stark, Jean Swank, Tod Strohmayer, Nick White - deliveries for the freeze: PCABACKEST will do gain shifting, improved interpolation, and contain all the code necessary for using the L7/240 models, the Vx + 3 time constant models (which don't yet exist), and all previous models the default for syserr will be set to no the code is known to fail ocassionally on the alphas if syserr=yes. It was agreed that we could address this as a bug fix. Mike will deliver the pcabackest code to the GOF on Monday - Keith demonstrated that the Vx model generation is underway, though no results to report yet - next meeting (and all meetings this fall) will be on Thursday (9/3) at 11:00 am After the meeting a vigorous discussion broke out (Tod, Mike, Jean, Keith, Dave, Donna) about how statistics are handled by PCABACKEST/SAEXTRCT/XSPEC, and how they outht to be handled. This subcommittee will meet at 4:00 pm Tuesday (9/1) to present short and long term solutions to this problem, with notes posted on this page. minutes prepared 8-28-98/KJ :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.14 :::::::::::::: PCA background meeting 9-3-98 present: Warren Focke, Tess Jaffe, Keith Jahoda, Jim Lochner, Gail Rohrbach, Alan Smale, Dave Smith, Mike Stark - PCABACKEST v2.1a was delivered on September 1, 1998. new features include: - gain shifted background estimates - improved interpolation in going from 129 to 256 channels (affects gain shift, allows better operation of fullspec=yes option) - Discussion of statistics. PCABAACKEST/SAEXTRCT writes a bkgd.pha file with STAT_ERR = sqrt (number of counts), i.e. an error related to the length of the observation. The error should reflect the length of blank sky observation, and likely some systematic error as well. - Next steps: - determine the typical error (from the distribution of residuals of data that went into the model) and create a tool that will adjust the STAT_ERR column appropriately. - define what quantity should be summed as the extraction goes along (i.e. the syserr=yes option on PCABACKEST) and determine whether it is better to add functionality to SAEXTRCT so as to get data and stat_err in a single pass, or whether two passes should be used. Open issues, not discussed: moving the L7_240 model to epoch 1 and 2 the Vx plus time constants model - next meeting: September 10, at 3:30. We will try Thursdays at 3:30 as a regularly scheduled meeting time for the fall. notes prepared by KJ, September 3, 1998 :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.980910 :::::::::::::: PCA background meeting. September 10, 1998 present: Glenn Allen, John Cannizzo, Tess Jaffe, Keith Jahoda, Gail Rohrbach, Alan Smale, Dave Smith, Mike Stark - for the 20801-05 series, 1% of the data is screened out by a criteria based on Layer1 - (Layer2+Layer3). No other screening parameter has been found. This is unfeasible for source data, probably represents an unknown "source" outside the collimator. - daily variation of +/- 0.1 cts/s/3 PCU is observed. 5000 seconds bins were required to see this, but it is present in both NGC 4051 in an extremely low state (Phil Uttley) and the 4 day background obs of 30801 in May 1998 - Mike and Keith are "measuring" the errors in the L7_240 models. Every interval that went into the models gets the model removed; the resulting distribution of residuals (by pcu, by channel, by anode) can be used to estimate the error on the mean (i.e. pcabackest output) - pca_bs . This tool is adequately useful that we will keep it in the next release. The "Rots memo" needs wider circulation if this is to be useful for people defining their own channel boundaries. - automating the production of the SAA history files remains desirable. next target update: 9/22/98 minutes prepared by K Jahoda, 9-28-98 :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.980917 :::::::::::::: PCA background meeting. September 17, 1998 present: Padi Boyd, Keith Jahoda, Jim Lochner, Gail Rohrbach, Alan Smale, Dave Smith, Mike Stark, Tod Strohmayer - the variation of residuals for the data that went into the L7_240 models has been constructed, and used to derive the error on the mean for the pcabackest of the mean for each channel of each anode. The results are stored in pca_bkgd_sigma.fits available (as of 9-28-98) from keith.jahoda@gsfc.nasa.gov, mike.stark@gsfc.nasa.gov, or xtehelp@athena.gsfc.nasa.gov - under the assumption that these errors, appropriately combined, rather than Poisson statistics, represent the appropriate error to be associated with pcabackest output, it is possible to imagine the following process. PCABACKEST --> SAEXTRCT --> FIX_ERR where fix_err is a new tool that changes the STAT_ERR column in the .pha file output from saextrct. A script which does this, for 3 detector first layer only data, is attached as the appendix to these minutes. This script needs to be modified each time a different combination of detectors or layers is chosen. - For the L7_240 models, the error estimate of fix_err is smaller than the Poisson estimate for 3 PCU, first layer, for exposures up to ~240,000 secs (i.e. the value of chi-square is increased although the best fit parameters are unchanged for several examples investigated) - The GOF has accepted the job of modifiying fix_err to make it a little more user friendly. - An estimate of an appropriate sys_err is still needed. Appendix: fix_err suitable for 3 PCU, layer 1 (author Jahoda) #!/bin/tcsh -f set bkgd_file = $1 cp /home/keith/pcabackest/bkgd_err/pca_bkgd_sigma.fits temp.fits fcalc infile=temp.fits outfile=temp.fits clname=ssss \ expr="(X1LSIGMAPCU0 * X1LSIGMAPCU0 + X1RSIGMAPCU0 * X1RSIGMAPCU0 + X1LSIGMAPCU1 * X1LSIGMAPCU1 + X1RSIGMAPCU1 * X 1RSIGMAPCU1 + X1LSIGMAPCU2 * X1LSIGMAPCU2 + X1RSIGMAPCU2 * X1RSIGMAPCU2) ** 0.5" \ clobber=yes fdelcol infile=$bkgd_file+1 colname=STAT_ERR confirm=NO proceed=YES mode=h faddcol infile=$bkgd_file colfile=temp.fits colname=ssss history=yes mode=h fcalc infile=$bkgd_file outfile=$bkgd_file clname=STAT_ERR expr="ssss * #EXPOSURE" clobber=yes fdelcol infile=$bkgd_file+1 colname=ssss confirm=NO proceed=YES mode=h cat << EOD > modhead.txt TUNIT3 = 'count' / unit for column Comment / stat_err column modified by the fix_err script EOD fmodhead $bkgd_file tmpfil=modhead.txt exit :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.980924 :::::::::::::: PCA background meeting. September 24, 1998 present: Glenn Allen, Padi Boyd, John Cannizzo, Tess Jaffe, Keith Jahoda, Jim Lochner, Craig Markwardt, Gail Rohrbach, Alan Smale, Dave Smith, Mike Stark, Tod Strohmayer, Jean Swank - discussion of what models should be in CALDB. since we advise different models in different circumstances, CALDB can't know what is wanted. Keith proposed: - deliver existing models is $LHEA_DATA - set default for "modelfiles" to "-" - cause graceful exit if modelfiels aren't supplied. - update PCABACKEST page to give clearer advice, and ripple this advice through PCA page and recipes pages A disadvantage is that the ftools group is trying to eliminate $LHEA_DATA - next delivery of PCABACKEST will make syserr a hidden parameter with the default value set to "no" - residuals, and error estimates, are needed for other models beyond the L7_240 - Jim has joined the fix_err testing team. GOF is developing a feel for what the user interface should look like. - L7 rate vs 60-100 keV rate is extremely linear; this generates hope that if failures of L7 for bright sources are due to contamination of the L7 rates by source counts, that we can account for the contamination and enable the use of these models to higher rates. Keith is investigating. - Mike is examining the failure of pcabackest on certain files. This is apparently related to the gain correction code. - Keith noted that the Gain Correction is not implemented for the Q6 or VLE models. After noting "blank looks all around" he agreed to look into this. Given that the gain correction information is contained in the keywords, it should not be too complicated to implement, once the reference day for the models is implemented. minutes prepared 9-28-98 by K Jahoda :::::::::::::: /home/lhea1/keith/wwwkj_20dec2002/pcabackgd_mins//minutes.981022 :::::::::::::: PCA Background meeting - October 22, 1998 present: Padi Boyd, Tess Jaffe, Keith Jahoda, Jim Lochner, Craig Markwardt, Sandor Molnar, Gail Rohrbach, Dave Smith, Mike Stark, Tod Strohmayer, Jean Swank QUALITY OF THE CURRENT MODEL Keith presented his estimate of the systematic error in the L7_240 model, derived from the 20801 and 30801 data that went into the model. The systematic error is approximately 0.15 cts/s/3pcu/2-10 keV, and is invariant to the time resolution at which the data were binned. This is much smaller than the 1 sigma CXB fluctuation is approximately 0.7 cts/s/3pcu/2-10 keV, as determined from an extrapolation of the HEAO-A1/Ginga results. From an analysis of the background pointing position #6 dataset, which includes Rick Edelson's data, the 4 day background observations, and data acquired at approximately 12 hour periods, Dave independently determined a systematic error of approximately 0.15 cts/s/3pcu/2-10 keV, using the same selection criteria as Keith. However, the data-model residuals show a larger excess during the 12 hour pointings in the first layer only. This is consistent with a cosmic source, but Dave will look at the background pointing position #1 data acquired during the same period. Rick Edelson passed on kudos to Keith for the work done so far, and promptly pointed out where he thought further work was needed. Dave observed, during the course of his analysis of the 30801 observations, that the L7_240 model appears to "bottom out" during periods of low count rate. Dave will plot the L7 rate as a function time/time bin for the 30801 background data, to determine whether the periods of anomalous count rate correspond to values of L7 rate outside the current model. If so, Mike will make a new model file, incorporating the new background data, and/or alter the pcabackest code to accept a larger range of L7 values. Mike will look into making the L7_240 model available for epoch 1 and 2. To include the 30801 background data into the L7_240 model, the gain shift will have to be re-calibrated and the pcabackest code altered accordingly. Also, the long-term trends in the background will either have to be modeled, or the data split into two/three periods, and models made for each period. STATUS of the FTOOLS 4.2 RELEASE Release notes and any corrections to the fhelp for the lastest PCABACKEST need to be made for the FTOOLS v. 4.2 software release. Decision was made to start from the XTE Patch Release Notes and then include any further changes new in this release. Things that are new since the patch include: - gain shifting - interpolation to 256 channels improved - no longer crashes on Alpha (OSF) platform - syserr parameter now hidden (but still available) - CALDB no longer the default for model files - location and discussion of new model files (allsky,faint,saa history) [ask GOF for location] Note that the allsky model was never in caldb, and may not have been "formally" released before now The release notes from the patch will be e-mailed to Mike Stark with cc to Keith Jahoda, so they can begin from that text. Keith will be unavailable to work on this until after the RXTE User's Group meeting; Mike thought it conceivable that he could. The GOF will take care of changes to the fhelp for PCABACKEST and RUNPCABACKEST (Gail, et al). The model files will not be packaged in the $LHEA_DATA FTOOLS area for this release. This is primarily due to the size and number of model files involved, which affects the overall size of the FTOOLS delivery, in combination with the fact that these models are likely to continue to evolve in coming months. The model files will be offered in the XTE anonymous-FTP area on legacy. Keith (for the PCA team) agreed with this approach, and hoped the issue could be revisited for the next release, when conditions presumably could be different. minutes prepared by Gail Rohrbach, Dave Smith 10-23-98