HXDS SPECTRUM FITTING DEVELOPMENT TASK DESCRIPTIONS AND ASSIGNMENTS

This site under construction.

$Id: spectral.html,v 1.5 1998/03/18 14:59:15 emk Exp $

[M. McDermott, E. Kellogg] -- JMKMOD currently contains a synchrotron white light model (used for BESSY calibration).
This needs to be tested. The part of jmkmod that compares observed pulse height spectra with the calculated input spectrum from the BESSY synchrotron using the Schwinger model needs to be validated and applied to both the fpc and ssd data.
identified by M. McDermott, 4 Feb 1998.
We are developing a Mathcad model that uses the equations derived by E. Tsiang for jmkmod. This will permit independent testing of the model. The equations for calculating the flux through a circular aperture on the white beam line have been coded and run. There is a double integral for the general case when the aperture is not centered on the ring plane. It reduces to a single integral when the aperture is centered. This allows a test of the calculation by comparing the two versions. As of today, we are seeing a +8,-5 percent difference between the two. Mar. 17, 1998 [E. Kellogg]
[P. Barnes, S. Serej] -- JMKMOD doesn't do the pileup correctly:
It wants to put the pileup of a spectral feature at twice the channel number, not twice the pulse height (= energy). If there is an energy or a channel offset in the fit (which is ~always true), this produces bad fits. The current workaround is to turn pileup off, and for lines use an extra component constrained to be twice the energy, in the usual XSPEC way. (The down side here is that the pileup is off!) This should be fixed asap.
P. Barnes, 18 November 1997
A developmental version of jmkmod with the pileup correctly located at twice the energy has been tested, [S. Serej, Feb 26, 1998]
FPC and SSD deadtime correction Memo
[B. Wargelin]
Completed writing memo on deadtime corrections; emailed to axaf_cat on Mar 17(Happy St. Patrick's Day!!), in process of being posted [P. Barnes] (http://hea-www.harvard.edu/MST/hxds/hxds-home.html). [Mar 17, B. Wargelin]
Falling tail in jmkmod
We fixed the falling tail in the JMKmod. It used to start at zero energy. Now it starts at the energy of the line, as is needed to give better fits. This corrected feature is now included in the beta version of XSPEC_V10. Mar 12, 1998 [Shaun Serej]
Bug in Kramers continuum in jmkmod.
Working on a new bug that was reported by Dick Edgar. The cutoff energy of the Kramer continuum for some, yet unknown, reason is drastically affected by the offset energy. (Very small offsets on the order of 0.003 KeV change the cutoff by as much as one keV.) So far, I have not been able to find out why this occurs, but it seems that the problem lies in the integration routine for the continuum. I base this on the fact that the integration is wildly affected by the choice of number of intervals in the integration (nInt parameter in the code). More work on this is in progress. Mar 12, 1998. [Shaun Serej]
SSD Pileup
[M. McDermott, S. Serej, P. Barnes] -- Interpeak Events
Typically between the first and second order peaks in lower-energy (<~2keV) SSD spectra, the flat shelf below the 2nd order line appears to be higher than is accounted for by JMKMOD. Because this elevated "interpeak shelf" appears to be countrate-dependent, and not dependent on (eg) the crystal being used in the BESSY KMC monochromator, it is attributed to pileup but not included in JMKMOD. These channels then need to be omitted when running a fit, which distorts the solution for the various parameters and makes the overall fit less reliable. Modelling this effect would be very helpful.
P. Barnes, 18 November 1997
This effect has now been modeled successfully, by convolving the shaping amplifier triangular waveform with itself, simulating two overlapping pulses when the pileup rejection circuit is not functioning, as we now understand it doesn't for energies less than ~2 keV. The output pulse height spectrum shows promising qualitative agreement with observations [Brad Wargelin, Feb 27, 1998].

[M. McDermott]--Pileup in SSD spectra for low energy pulses ( < 2keV )
When an event produces a pulse in the detector that is very close to tne noise threshold of the Pile Up Rejection (PUR) Circuitry in the Shaping amplifier, the pulse does not activate the PUR circuit and the pulse is allowed to pass throught the shaping amp which can pile up with any preceding or subsequent pulse that comes along.
Mike McDermott, 20 November 1997
[B. Wargelin] -- Gain nonuniformity:
This applies for all the BND-H FPCs, except fpc_hn when it used its 36-mm aperture instead of the standard BND-H rectangular aperture. We have nice Pipe data showing that the gain curve is extremely repeatable--- 3 FPCs were checked, one of them at 3 widely varying energies, and the curves are all the same. We should incorporate a choice of gain curve (flat for fpc_5, x1, x2, and hn with 36-mm ap; curve for fpc_hb, hs, ht, and hn with full ap) into the mgf.

The curve is for the original 125-mm-long aps, but you can just chop off 13 mm on each end for the 99-mm-long aps we actually used at XRCF. We will also do more gain maps on a few FPCs on the Pipe in October to confirm that this is OK.
B. Wargelin, 4 September 1997 [edited]
[P. Barnes]-- Investigate full solution in JMKMOD ???
[nobody assigned??] -- Line complexes and broad lines
This is actually a rather messy problem, but we can make a good start by using Bearden's compilation of line energies, and fix the relative intensities of lines within a complex, preferably by some kind of editable lookup table. We have such tables for many elements in /data/axaff/bradw/lines---at least to start with. Ultimately we need to coordinate with other groups, using relative line intensities determined from grating data, etc. I think it's important to specify the relative line intensities as emitted from the unfiltered source---is that how XSPEC does it?

Once that's done it'll be easier to treat escape peaks. Just use the same fixed relative line intensities (fluoresence yields don't change that much with energy, and the escape peaks are only a fraction of the parent lines' intensities anyway) with a fitted (but limited to near 3 keV) offset.

Broad lines (B-K through Zn-L) should be decomposed into multiple lines so they can be treated in the same way as the line complexes.
B. Wargelin, 4 September 1997
"Ion Beam Handbook for Material Analysis", J. W. Mayer, E. Rimini eds., Academic Press, 1977, sec. 5, p. 311-484.
Ed Kellogg, 21 November 1997
[nobody assigned??] -- Check if XSPEC v10 has faster optimization routine
e.g., fit lines with Gaussian (don't use JMKMOD) with v9 and V10, and compare results.
[nobody assigned??] -- Profile XSPEC/JMKMOD and see where it's spending its time.
[P. Barnes] -- Determine feasability of determining analytical derivatives. Detailed information is available here [P. Barnes]
[nobody assigned??] -- Investigate improvement of optimization routines to make use of analytical derivatives. (Make backwards compatible with current XSPEC models.)
[nobody assigned??] -- Investigate new optimization routines. See The Optimization Technology Center.
[P. Barnes] -- Determine if rebinning of spectra (via --> -- grppha) speeds things up.
[P. Barnes] -- See if specifying fewer channels to fit affects fitting time.