| RXTE GOF |
PCA Digest: Recent Developments | RXTE FAQ |
|---|
Background subtraction and PCA response are two areas in which constant progress is being made by the PCA team. Check back here for summaries of recent results and links to more information.
Table of Contents
- What's New: September 17, 2007 SAA History and Background Problems
- PCA Instrument Status
- Important Downloads and Links for background estimation and timing
- Ready Reference: Other Useful Calibration Info
- Discussion of Epoch 5 PCU0 background models
- Latest Developments in PCA Background Models
- Latest Developments in PCA Response
- Other Issues
- Number of PCUs On as a Fraction of Total Good Time
- High Frequency Timing Signatures
- Works in Progress, including Cross-Calibration with HEXTE
- References to other Web Pages with PCA Information
September 18, 2007
IMPORTANT: Problems and Corrections to RXTE PCA Background
Summary:
Problems were discovered in the SAA history file that result in errors in background rate estimation. A new file that corrects these problems is available.
In addition, a bug was discovered in the FTOOL PCABACKEST for faint models only, in Epoch 5c. The parameter maxmodels=2000 (instead of the default value maxmodels=600) must be set in order to return a complete faint background model.
Details can be found on the PCA Team SAA History and Backgrounds Problem Report Page
January 28, 2007
On 2006 Dec. 25 at 12:56:08 UT, PCU 1 (in the 0-4 numbering scheme) developed a pinhole leak in its propane layer, consistent with a micrometeorite hit. The PCA team recommends that observers should not use PCU1 for light curves and spectra until new response matrices and background models have been developed. Work is beginning on those issues, and expected to take several months. High frequency timing with PCU1 should still be possible after the propane loss.
August 6, 2006
New background models are available for Epoch 5 (observations taken in May 2000 and later). See below for downloads and discussion.
The overall PCA instrument is operating nominally. However, the five individual PCU detectors which make up the PCA have their own individual details, which are listed in the table below.
| PCU 0 | PCU 1 | PCU 2 | PCU 3 | PCU 4 | |
|---|---|---|---|---|---|
| Status | Operating | Operating | Operating | Operating | Operating | Propane Layer? | No (as of 12 May 2000) |
No (as of 25 Dec 2006) |
Yes | Yes | Yes |
| On-time (% of good-time) |
~85% | ~15% | ~100% | ~35% | ~15% |
In the table below, you will find quick links to the important
files you need. Start here if you're an experienced RXTE user, and
you just want to update your files with a few clicks.
| Description | File Name | Released / Updated |
|---|---|---|
| Bright background model (net count rate > 40 ct/sec/PCU) |
pca_bkgd_cmbrightvle_eMv20051128.mdl pca_saa_history.gz |
August 6, 2006 |
| Faint background model (net count rate < 40 ct/sec/PCU) |
pca_bkgd_cmfaintl7_eMv20051128.mdl pca_saa_history.gz |
August 6, 2006 |
| SAA History file (use for all background models) |
pca_saa_history.gz | ~Daily (few day lag) |
| SAA Prediction file (use for quick-look when History file is not available) |
pca_saa_predict.gz (also see README) |
~Daily (30-day forecast) |
| Clock file (use with faxbary for better than 70 μs timing) |
tdc.dat | ~Monthly |
| Ready Reference: | Energy to Channel Table |
Cycle 8 Simulation Data |
PCU On-Off History |
PCA Background Epoch Boundaries and Models |
|---|
Discussion of Epoch 5 PCU0
Background Models
Want to try the latest background models for Epoch 5 PCU0 data? With
the release of FTOOLs v5.2, users can apply the Epoch 5 models to
PCU0 by including a few new columns in the apidlist for XTEFILT.
Type fhelp xtefilt to get the full list of apids needed by
the tool. More details on PCU0 analysis
and discussion of what to expect, can be found on
Craig Markwardt's 2002 model documentation page.
Note that Craig's page was
written before the release of the FTOOLS 5.2 and contains
a few work-arounds that are no longer necessary.
August 6, 2006
New PCA background models have been developed that remove long term trends that have become apparent over the years. Only models for "Epoch 5" -- May 2000 to the present -- were changed. The original "Epoch 5" has been divided up into three new intervals, "5A", "5B" and "5C" to account for the trends and improve the background model. New faint and bright source models are available from the links in the table above. The new models have been tested extensively against the background data themselves, as well as by select observers on AGN data. There is marked improvement for faint sources in the energy range 25-30 keV, where a background feature caused artifacts in the previous models.
A more complete description of the new models can be found in this report by Craig Markwardt.
April 14, 2004
Last updated April 13, 2004 with links to new background model files that correct for a slow linear trend in the background rate. Available for Epochs 3a, 3b, 4 and 5 as individual models files, as well as in an updated Mission-long Faint Model File as of 2003-11-23. (now obsolete)
New faint model files are available through the link above and through the model table below. As the orbit of XTE degrades, the PCA background is gradually decreasing. Earlier, a new "linear drift" temporal model was incorporated in the "CM" faint background models which attempted to subtract this drift out. However due to an errant keyword in the background models, the computed term was CONSTANT and only correct at the START of the gain epoch. Eventually, this has led to an oversubtraction of background rates late in any Epoch. The performance degrades with time, and is worse at the present day, for epoch 5).
New models are being released that correct this problem, and eliminate these trends. One model is mission-long, the others contain the same data, but separated out into individual gain Epochs. The details of the faint model are the same in both representations; the mission long model is made available to simplify the analysis tasks for users.
The new models have been tested, and work well, on a variety of PCA data.
A detailed discussion of the differences between the old models and the new is available from the following PCA Team 2004 report.
The page includes several informative graphs of the cumulative error versus time, as well as pulse height dependences.
March, 2003
Download one complete mission-long background file for bright or faint
sources, and let the FTOOLs work hard so you don't have to! The
background files available here were created in March 2003, and contain
substantially more Epoch 5 data than previous versions of the model
files, which only included data through mid-2001.
Mission-long Faint Model File as of 2003-11-23 (now obsolete)
Mission-long Bright Model File as of 2003-03-30 (now obsolete)
...and don't leave without the latest
pca_saa_history file!
July 9, 2002
Information about the latest pcabackest and pcarsp in FTOOLs 5.2 (released June 25, 2002).
FTOOLS 5.2 includes software for improved response matrix generation as well as background subtraction.
The latest FTOOLS release, v5.2 (June 25, 2002) includes new versions of PCARMF, XPCAARF, and the wrapper script PCARSP as well as a new energy-to-channel FITS file pca_e2c_e05v02.fits. PCARMF v8.0 has the following features:
- Calculation of sigma in the response is greatly improved.
- Energy channels in the matrix are now spaced logarithmically.
- Each detector can now have a separate value for the thickness of the "dead layer" between xenon volumes (via new parameters in pcarmf.par).
- There is now an opportunity to specify a second a+bx law (and date when it starts) for the amount of xenon in the propane layer. At the moment this is only used for PCU0 (to set the amount to zero after the loss of that layer).
Derived fluxes from PCU0 still show a slight discontinuity around the time period of the loss of the propane layer, but much less so than before.
The only change to XPCAARF is to the default values (in xpcaarf.par) for the geometric areas of four of the five PCUs (PCU 2 was left unchanged). These were chosen to make the flux of the Crab similar when fit to the five detectors (epoch 3/4 chosen for the comparison). Note that this will likely result in a step function in monitoring fluxes unless the whole data set is reanalyzed with a constant set of XPCAARF coefficients.
PCARSP (v8.0) has been updated to ensure that the default value for the partial charge fraction matches the value in PCARMF (previously 0.0, now 0.02). A bug was also fixed in the previous version which caused the "-n" flag to fail to properly rename the final output response file in single-detector cases.
PCABACKEST has been updated to support the new Combined Models (CM) and to include more information in the output file headers documenting which models were used. In addition, XTEFILT (ie, FCOLLECT and XTEDERIVE) has been modified to add a new derived quantity, L6CNTPCU0, to the filter file. This can be used to filter out times when large flares appear in the background model due to the loss of the propane layer in PCU0. Six new quantities must be added to the input AppID list to allow L6CNTPCU0 to be computed (the full AppID list is given in the fhelp for XTEFILT). The new quantities are:
74 X2LX2RCntPcu0 74 X3LX3RCntPcu0 74 X1LX2LCntPcu0 74 X1RX2RCntPcu0 74 X2LX3LCntPcu0 74 X2RX3RCntPcu0
February 26, 2002
New and improved models are now available for Epoch 3
to the present (both faint and bright). Briefly, the new models
feature:
- sensitivity to X-ray variations improved by factor of ~2 over previous models;
- applicability to Epoch 5 PCU0 (needs new FTOOLS to fully exploit);
- new model packaging means fewer files to deal with;
- gain Epoch 3 is now divided into Epochs 3a and 3b for faint sources;
A quick reference table now appears at the top of the Digest so users may easily download new models by epoch and bright/faint criteria. The table includes new combined model files for all calibration/gain epochs. A pointer is included to Craig Markwardt's documentation on the differences between the new, updated models and the old, how to run them, and what to expect. The Digest has been reorganized for clarity.
June 13, 2001
Background models for the PCA detectors are created from observations of the "blank" sky (i.e. the recorded signals are the sum of the Cosmic X-ray Background and instrumental background). We observe that the background is time variable, so we parameterize the background as a function of spacecraft and instrument conditions. In particular, all useful models to date parameterize the background as a function of one or more of the coincidence rates reported in the Standard 2 data with an additional time dependent term to account for an activation component due to passage through the radiation belts in the South Atlantic anomaly.
The faint models use a rate called "L7", which is the sum of all 2-fold coincidences between adjacent xenon signal chains (L1+R1, L2+R2, L3+R3, L1+L2, L2+L3, R1+R2, and R2+R3). Even without the propane layer, this quantity is unambiguously defined, and our first attempt was to examine whether the L7 rate still tracks apparent X-ray rate for PCU 0. Summarized in a phrase, the answer is "not well enough". There are at least two contributing factors. Without the propane layer, there is additional noise in the apparent X-ray rate due to electrons which deposit energy in the first layer (and which previously would have stopped in the propane layer or the propane+first xenon layer). Additionally, without the proane layer, our electron screening criteria is undefined. The scatter is sufficiently large that we have abandoned the idea that we can model the PCU 0 background as a function of PCU 0 parameters.
However, we know that the behaviour adjacent and nominally identical detectors is highly correlated, and PCU 2 is always on. Over the course of summer 2001, we expect to explore a model for PCU 0 background that is based on the rates in PCU 2. The electron screening criteria can be applied based on PCU 2 data; we will explore whether additional screeing criteria are required based on combinations of data from PCU 0 and PCU 2. If successful, we will release a new background model file, and (if necessary) a slightly updated pcabackest. Users will be only slightly affected by the (potential) need to apply different screening criteria to PCU 0 than to PCU 2, and may therefore find that PCU 0 does not have an acceptable model for all the times that it operates. The GOF will work to update the REX script to accomodate this.
Keith Jahoda
February 26, 2002
New and improved faint models are now available. For the best possible background subtractions throughout the very long Epoch 3, a new boundary has been established, breaking it into Epoch 3a and Epoch 3b. The new Epoch boundaries are listed below.
Epoch "3a" 1996Apr15 at 00:24:05TT - 1998Feb09 at 00:01:00TT
Epoch "3b" 1998Feb09 at 00:01:00TT - 1999Mar22 at 00:18:38TT
The faint model components are:
- "cml7" - the part of the PCA rate accounted for by the L7 rate
- "cml7act240" - the part of the PCA rate accounted for by 240 minute radioactive decay
- "cml7tlin" - long term secular drift of background rate
New and improved bright background models are also now available for Epoch 3 and later. The new bright background models require the SAA_HISTORY file. The table at the top of this document organizes the models and extra files needed by Epoch and source count rate. For all epochs, the various model components have been combined into one file for ease of use. No changes have been made to the Epoch 1 and 2 models other than the merging of the separate components.
There are again three components, but not the same components as the faint model:
- cmvle - depends solely on VLE rate
- cmvleact24 - activation component - 24 minute decay time
- cmvleact240 - activation component - 240 minute decay time
All new model files can be used as-is with the existing PCABACKEST for Epochs 1-4 and for PCUs 1,2,3,4 in all Epochs. For spectral analysis of data from Epoch 5 PCU0, FTOOLS v5.2 (released June 25, 2002) or later is required.
Creating Filter Files and GTI Files for Use with Faint Models
Before running XTEFILT, you will need an updated apidlist so that some necessary new parameters (such as electron_n) can be calculated. Make sure you get the new list from the "fhelp xtefilt", save it as a text file, and use it as the appidfile when calling XTEFILT to create XTE Filter files.
Using MAKETIME on your XTE Filter files, create .gti files adding the additional constraints to your usual expression:
For Epochs 1 and 2:
- time_since_saa .gt. 30.0 .or. time_since_saa .lt. 0.0
- electron2 .lt. 0.1
For Epoch 3 and later:
- electron2 .lt. 0.1
These GTIs should be entered into the extractor at the GTIANDFILE prompt. See our pcabackest recipe for more details.
NOTE The CM models for Epoch 3 and later, unlike previous models, are based on blank sky observations with no screening on the 'time since saa' parameter. Tests show that the CM models work well in all parts of the orbit including shortly after SAA passages. Consequently, we no longer recommend screening on the time_since_saa parameter. The use of scripts that include this screening (i.e. REX as distributed in ftools 5.1) will eliminate a small amount of useful data, but will not affect systematics limited investigations.
NOTE We now recommend filtering on the ELECTRON2 level as
opposed to ELECTRON0. Since the loss of the propane layer in PCU0,
the value of ELECTRON0 is zero for all times.
Latest Developments in PCA Response Matrix
The PCA Response Matrix Generator has been updated in FTOOLS v5.2.
See the What's New section of this Digest
for details.
Other Issues
Number of PCUs on as a Fraction of Total Good Time
In the past, when proposers have prepared their technical feasibility calculations they have assumed that all 5 PCU will be turned on. However, the average number of PCUs on per observation is 3.3. At times there are now only 2 PCUs available. As a percentage of total good time, the number of PCUs on is:
|
Number of PCUs On 2 3 4 5 |
Percentage of Total Good Time 30 33 24 13 |
|---|
High Frequency Time Signatures
Works in Progress
- Database of slews spanning the entire mission, to increase the number of background observations in the models,
- Development of a single background model, appropriate for bright sources as well as faint sources.
- Cross-calibration of PCA and HEXTE
Learn More about PCA Background and Response
To find out more about PCA background and response issues, visit one of these experts' pages:- Craig Markwardt has collected some practical advice for users of the new PCA background models at Craig Markwardt's 2002 PCA Background Model Documentation Page
- Keith Jahoda maintains a list of useful information, including details of various background model components and model performance, at the PCA Instrument Page
- Many detailed notes from David A. Smith, including discussion of daily variation in the faint model 30-70 keV background subtractions, and systematic errors in the faint models (not the newly released versions) can be found on the PCA Background Issues Page
If you have a question about RXTE, please send email to one of our help desks.