What's New in OSPEX

thick_warm - Warm Thick Target Bremsstrahlung
multi_therm_linear_gauss - Multithermal, Gaussian in linear T DEM distribution
multi_therm_pow_exp - Multithermal, Pow Temp * Exp Temp
thin2sm - Thin Target Bremsstrahlung Version 2 for a Smoothly Broken Power-Law
line_asym - Asymmetric Gaussian
line_nodrm - Gaussian, does not go through DRM
line_asym_nodrm - Asymmetric Gaussian, does not go through DRM
1pow_nodrm - Single Power Law, does not go through DRM
3pow_nodrm - Triple Power Law, does not go through DRM
exp2_nodrm - Exponential with log normalization, does not go through DRM

1. New button in the Fit Component Setup Widget for each thick or thin component.
2. New item in the plot 'in Energy Bins'. column in the View Fit Results Widget
3. New plot_edf method in the spex_fit object:
   fit_obj = o->get(/obj, class='spex_fit')
   fit_obj->plot_edf...
4. Standalone function called calc_electron_distribution.
5. Standalone routine called plot_electron_distribution.

1. New item in the View Fit Results Widget when you expand the vth or vth_abun plot items.
2. Standalone function called therm_energy_plasma

1. New button in the Fit Component Setup Widget for each multi_therm_xxx component.
2. New plot_dem method in the spex_fit object:
   fit_obj = o->get(/obj, class='spex_fit')
   fit_obj->plot_dem...
3. Standalone routine called spex_get_dem.

spex_auto_erange now sets the upper limit of the energy range to fit by finding the first energy bin where the counts/bin becomes less than a threshold (defined by new parameter spex_fit_auto_emax_thresh). Previously used  count rate  > .01 test, and rounded the energy up to the next multiple of 10.

Added option to set lower limit (spex_fit_auto_emin) for RHESSI based on attenuator state.

Now each energy band for defining background can use any of these options:
0poly,1poly, 2poly, 3poly, exp, High E Profile, or This E Profile. 'High E Profile' is like the old bk_ratio method, except now it can be chosen for any specific energy band (previously it was global). 'This E Profile' is new- it's the same idea as ratio to high-energy band profile except the profile is created using this eband. For all of these methods, the error is the square root of the counts in the time interval(s) used to calculate the ratio and when combining time bins, errors are averaged.

2vth - sum of two vth functions, with a shared abundance value
2vth_abun - sum of two vth functions, with shared abundance values for ~10 elements
drm_mod2 - RHESSI pseudo-function for fine-tuning RHESSI DRM parameters.  Like drm_mod, but added center thickness ratio parameter.

1. Previously: Error on high-band profile was wrong - just used error that came from the bin_data command.
   Now: Treat high-band profile error the same way we're treating flux profile - use selected time intervals, smooth, interpolate, smooth, etc.
    
2. Previously:  To create high energy band profile, interpolated between profile in selected time intervals, replaced first and last two points with average of first/last 11 points (if selected intervals didn't include first and last points of data), then smoothed. This had two problems: interpolation was based on single points at end/start of time intervals - if those points were high or low, threw off interpolated points. Also, when taking average of first/last 11 points, didn't ensure that they were all in first/last time intervals. Then convolve with savgol filter.
   Now:  Find groups of contiguous points (i.e. the points in each specified bk time interval) and smooth them before doing the interpolation between groups - use boxcar smoothing with width=10 (or less if fewer points in group). Also when replacing first and last two points with average use up to 10 points, but fewer if there are <10 points in first/last group. After interpolation on smooothed data replace non-interpolated points with original data, then convolve with savgol filter.
    
3. Changed wording on widget - smoothing width is really half the number of points used to smooth.
    
4. If half width to smooth is 1 (means 3 points: 1left +1right+1point) set degree in savgol to 2 so doesn't reject. And if half width to smooth is 0, skip final smoothing.
    
5. Added button to display a plot of the high band profile used in bk_ratio method.
    
6. Added buttons to include errors in time and spectrum plots in background widget.

When you select the template function, there are new broad, narrow, and broad+narrow line nuclear templates, as well as pion decay templates for five spectral indices. See http://hesperia.gsfc.nasa.gov/ssw/packages/spex/idl/object_spex/fit_model_components.txt - scroll down to the template component.

Function returns the photon spectrum at the Earth, d(Flux)(eph)/dt, from a differential emission measure distribution which has a Gaussian dependence on the logarithm of the temperature.

Please see http://hesperia.gsfc.nasa.gov/fermi_solar/analyzing_fermi_gbm.htm for more information.

Previously the albedo and pileup_mod corrections were applied to the model (for counts units) or the data (for photon units), but not shown explicitly. Now they are calculated (by taking difference of spectrum with and without correction), and returned by calc_func_components and displayed on plots showing the separate function components. 
Also, previously the albedo correction was applied to the separate function components as well as the composite spectrum (in plots, or results from calc_func_component) - now it is applied only to the composite spectrum.
Neither of these changes affects fitting, only the display or retrieval of the separate function components.

1.New option to remove highlighted intervals
2. New option to highlight or display intervals that have been fit
3. The 'Select -> Intervals with filter n' option didn't include filter=-99 (unknown filter state).  Now it does, so you can select the intervals with unknown filter state, and then use the new option to remove highlighted intervals.

RSP files for the b0 detector are now written with 200 input energy bins rather than 140 bins, still covering the range 100 keV to 200 MeV.  All existing b0 RSP files were regenerated. See PS plot of changed bins

Nonthermal bremsstrahlung X-ray spectrum from a thick target with step-function or linear-function ionization profiles. Relativistic energy loss and full cross section are included. Based on THICK2. 

The poshist file contains information about the directions of the NaI detectors.  Now when you download a data file for a specified time, the corresponding poshist file is downloaded too.  That way, if you go offline, you have the file and can plot the detector angles if desired.

  Error introduced in  Feb 2011:
  For data before December 2008, times were late by one second
  For data after December 2008, times were late by two seconds.

Thick2_vnorm is similar to the thick2 component, but has a variable normalization.  A seventh parameter was added - the reference energy at which the function is normalized.  This allows a[0], electron flux at ref energy,  to be independent of a[4], the low energy cutoff.  In thick2 a[0], the total integrated electron flux, and a[4], the low energy cutoff, are closely coupled.

The template function component supports photon functions of any shape over a specified energy domain given by enucl in the save file. The function is interpolated onto the energy bins used with the detector response so the count response of the detector can be computed. Previously, the function was normalized to 1 photon/cm2/sec over the range of the response energy bins. Now it is normalized over the domain of the input template, enucl. This makes a difference when the energy domain spanned by the input side of detector response matrix does not fully include the range of enucl. We noticed that this was causing problems understanding the normalization of a Pion template for LAT data where the low end of the LAT range was 70 MeV but the template enucl started at 200 keV. For the gamma-ray line spectral templates this would also cause a problem when looking at the low energy counts from a nuclear line spectrum where the detector response input does not extend to 8 MeV. Now, with the fix, the normalization parameter in the template function, A[0], returns the same photon spectrum regardless of the span of the energies on the detector response input.
 

When spex_error_use_expected is set to 0 (the default is 1), the weights used in fitting are determined from the error on the observed data combined with the error on the background.  Previously the error on the observed data was computed by sqrt(observed counts).  Now we use the error in observed counts that was set when the input data was initialized.  These are often the same, but may not be (e.g. RHESSI data, user input data through SPEX_USER_DATA or replacedata).

When Fermi GBM input is selected, a new button appears on the 'Select Input' widget to either plot the GBM detector angles to the Sun as a function of time, or to print the angles at the peak of whatever flares are in the selected time interval.  From the command line (for 12-jun-2010 for example):
   print,gbm_get_det_cos('12-jun-2010 00:56')
   gbm_plot_det_cos,['12-jun-2010 00:00', '12-jun-2010 02:00']

A thin-target bremsstrahlung spectrum for electron n-distribution function component called thin_ndistr is now available (from J. Kasparova).

Lat data can now be analyzed in OSPEX.  Spectrum and response files must be requested from LAT team.

When spex_use_expected is set, the weights used in fitting depend on the model, and therefore the current model parameter values.  Chisquare mapping should be done with fixed weights.  Now the weights matching the best-fit parameters are used throughout the chisquare mapping process.

1.  The weights used in fitting  are 1./errors^2, where
errors = sqrt (sqrt(c)^2 + eb^2)
c = total number of counts (either model+background, or total observed, depending on value of use_expected)
eb = systematic error in the background. 
Previously the background counts were not included in c.  Now they are.

2. Previously, the error in the background was the sum of two terms, the statistical error in the counts used in the fit, and the error from the fit.  The second term is no longer included.
(This change only applies to background computed by fitting a function to the data in user-defined background intervals.)

For in-orbit background, these two changes will change the fitting errors in opposite directions, so it's not obvious whether your new chisquare will be larger or smaller.  For inserted background, the second change has no effect, so the fitting errors will be larger, and your chisquare will be smaller.

Chisquare mapping analysis consists of varying one parameter through a range of values surrounding the best-fit value, fitting the other parameters, and then examining the chisqr vs parameter value curve to determine 1- and 2-sigma error estimates.  More details are in the full OSPEX documentation.

In GUI, Browse button now has two options - the normal browse on your computer, and browse across the network by time.  Downloads necessary files.  Use spex_find_data (see header in .pro file for arguments) to find and download files from command line.

Now you can fit on the anisotropy value in the albedo correction by adding the 'albedo' function component to your fit function.

Previously if you enabled albedo correction, but also used the RHESSI drm_mod function component, the albedo correction was undone.  This has been corrected.

Monte Carlo analysis consists of doing many fits to Poisson random deviates of the model computed from the best-fit parameters, and determining the spread of the resulting fit parameters.  More details are in the full OSPEX documentation.

Please see http://hesperia.gsfc.nasa.gov/fermi_solar/ for more documentation.

In some regimes (e.g. high-energy rear detector RHESSI data analysis), you may want OSPEX to ignore the filter states.  There's a new button on the Input widget called 'Ignore filters',  From the command line:  o->set, spex_ignore_filters=1.

bpow_ep is the same as bpow, but allows the user to select the pivot energy easily (via an additional parameter).

Previously if you created a spectrum from multiple ROIs, any overlapping parts of the ROIs were counted twice. 

Background estimated from pre- and post-flare intervals is only as good as the fit to those intervals.  The percentage uncertainty should not change when binning over multiple time bins.  Previously, combined time bins by summing the error in quadrature, as though the background were based on statistics - this resulted in an incorrectly small error for larger time integrations.

A new parameter, spex_bk_poisson_error (default is 0) can be set to 1 if your background is in fact uncoupled (e.g. if you inserted background from a previous orbit).

A new version of mcurvefit was onlined that provides more reliable, stable sigmas on fit parameters.  A problem with the old version allowed sigmas to become unreasonably small in some cases.

After selecting a YOHKOH data file (wda...):
In the GUI, a popup widget will let you choose which data type you want from the file - GRS1, GRS2, or HXS
At the command line, select the data type via the spex_data_sel parameter, e.g.:
o -> set, spex_data_sel = 'GRS2'       ; or 'GRS1' or 'HXS'

o->template_select, ['a.sav','b.sav'], [3,4]  ; sets template file for template components 3 and 4

Also made it possible to enter user template file names in the fit components widget.

d = o -> getdata (class='spex_fitint') now returns the observed data and background data (and errors for both) binned into the fit analysis time intervals (previously only returned bk-subtracted data).  Adding spex_units='rate' or 'flux' returns those units.

Visit the RHESSI wiki site for a detailed guide to using OSPEX to analyze RHESSI data.  Direct link is here.

On fit widget, new button called 'Init Fit Parameters Only' allows you to keep values of min, max, free, erange, uncertainty, #iter and only transfer the fit parameter values depending on your selections for first and subsequent interval initialization method.  Parameter name is spex_fit_init_params_only.

Button on fit and fit components widgets called Auto-set Max to select automatically selecting upper limit for energy range to fit based on background-subtracted flux levels.  New parameter name is spex_fit_auto_erange.

Returns the separate or combined fit function components as a function of energy in counts, rate, or flux, in photon or count space.  See Methods section of OSPEX document for more details.

SPEX_SPECFILE parameter is now an array.  For MESSENGER data, concatenates data from files (any number, don't have to be consecutive).  This can be done for other data types if needed, but currently only done for MESSENGER.

The new option, 'current', enters the fitting process without changing the fit parameters from their current value for the first interval.  For subsequent intervals, the 'previous interval' method is used. (Note that this was how the 'previous interval' method used to work.)

In the 'Previous Interval' method, the first interval selected for fitting now gets its starting values from the previous interval (in the full list of intervals).  Note that the meaning of 'previous' depends on whether you are looping forward or backward through intervals.

Plots residuals for selected fit intervals, showing chi-square.

If this parameter is set to 1, then you can use fit results whose energies are different (in value but not number) from the current data set.  Useful if you use RHESSI native energy bins which are slightly different for each detector.

Previously if you changed the fit time intervals (spex_fit_time_interval), the spex_summ arrays would be reinitialized and you would lose any fits you had already done.  Now you can modify your fit time intervals and the spex_summ arrays are adjusted accordingly.

There is now a 'Reset All Comp.' button in the fit component widget that lets you reset the parameter values, minima, maxima, and/or free masks for all components at once to default values, previous interval, previous iteration, or original (values when you entered the widget).  It is similar to the Reset button that applies to each fit component separately.

Also a bug was fixed so that now if you select Reset to previous interval, and your loop direction is backwards, then it resets to the interval+1 values (previous in the reverse direction).

Previously you had to include all fit function components you might possibly need for every interval right from the start - otherwise if you changed them, the spex_summ arrays would be reinitialized.  Now you can change the fit function after you've started saving fits - the spex_summ arrays are adjusted accordingly.

In the fit widget, the Select button now has some options to help you choose intervals based on filter state.
Also the filter state is now displayed (in parentheses next to the time interval) in the Fit Widget if all the intervals don't have the same filter state

New option to use the time profile of the background in the highest energy band to define the shape of the background for all energy bands. The amplitude is adjusted by the ratio of the data in the lower energies to the high-energy profile.  Refer to the 'Background' section in the OSPEX documentation for details.

There are new options for selecting contours in the ROI selection tool.  There's also a new HELP button on the widget to explain the region selection options.
 

Relative abundance of Fe, Ni is now a fit parameter.
Chianti is now default, previously was Mewe.
vth_noline function is obsolete, use vth with continuum option.
chianti/mewe and full/continuum/lines are now options.
New fit_comp* and spex_summ* parameters to handle these features.

Previously had a choice of genx (save) or FITS format for fit results output file.  Now FITS is the only choice.  Existing genx files can still be read into OSPEX.

o->list_function displays the details of the function components you've selected and their parameters
o->make_func_obj makes a fit_function object from current energy edges and function component parameters or from a specific interval that you've already fit

Start OSPEX main widget or any sub-widget with gui_label keyword to put a label on the widget title bars:
o=ospex(gui_label='your label')
o->gui, gui_label='your label'
o->xfit, gui_label='your label'

A new parameter, spex_image_full_srm, was added (in the GUI, it's a check box on the Select Input widget).

When spex_image_full_srm is not set (the default), the image cube is in units of photons/cm^2/sec/asec^2.  The conversion from counts to photons was made using the diagonal SRM in the RHESSI image object.  No SRM is used in OSPEX in this case.  You shouldn't analyze images at low energies in this mode since K-escape is a significant factor, and the off-diagonal elements of the SRM have not been taken into account.

When spex_image_full_srm is set, the spectrum computed from your selected image regions is converted back to counts by applying the diagonal SRM to reverse the original conversion.  The full SRM is calculated and used to convert to photons.

Previously in the GUI, when you created a new plot with the same name as an existing panel (e.g. SPEX HESSI Count Flux vs Energy), the old panel was replaced by the new panel in order to limit the number of panels stored.  Since panels showing different information may have the same name, all panels are now stored.  You can delete selected panels via the Multi-Panel Options button under Window_Control.
 

When you change to a new input source file by changing the spex_specfile parameter, now all of the input-dependent parameters (but not parameters to control configuration options) are reinitialized to program defaults.

You can always initialize all parameters via o->init_params (or via the Reset... button under File in the GUI).

To see which parameters are reinitialized with a new input file, type print, o->get_param_names(/new_input)

In the OSPEX Select Input widget, when you select a RHESSI Image Cube input file, the Region # button becomes sensitized.  After you select all ROIs (up to 4) on all images, then you can select which ROI (0,1, 2, 3 or All) to make spectra from.  From the command line, e.g.:
o -> set, spex_roi_use = 2
o -> set, spex_roi_integrate=1

In the OSPEX Select Input widget, when you select a RHESSI Image Cube input file, the Select Regions button becomes sensitized.  This button has two options  to configure the region selection parameters, and to make the region selections on the images.  Right or left click in the image panel display to get a list of options for defining ROIs or displaying spectra and time profiles.

To use the Region selection tools from the command line, after selecting an image cube input file, type
o -> roi_config
o -> roi

Full explanations of these features are found in Imaging Spectroscopy with OSPEX.

There's a new environment variable, OSPEX_MODELS_DIR, used to find the file fit_model_components.txt which lists the available functions.  Users can copy that file to a local directory, add functions, and define OSPEX_MODELS_DIR to point to that directory.

In the GUI, there is now a button under File to set parameters from a script procedure file (either initializing all parameters first, or not).  To do the same from the command line, type:

o->runscript   or  
o->runscript,/init   or  
o->runscript, init, file='ospex_script.pro'   or  
ospex_script, obj=o

The writescript method now writes a procedure instead of a main program.  Previously we ran it with the .RUN executive command.  Now we run it by calling the procedure, with obj as a keyword argument.  If obj is an existing OSPEX object, the script will set parameters in it, otherwise a new OSPEX object will be created.  The procedure name will be the same as the file name you selected.  For example, if you write the script in ospex_script.pro, then call it as follows:

ospex_script, obj=o

The spex_summ... parameters now include spex_summ_bk_rate and spex_summ_bk_error.  These are written to the save or FITS output file along with the other spex_summ.. parameters.   The calc_summ method has two new allowed values for 'item': 'bk_ct_flux' and 'bk_ph_flux'.

In the "Fit Options" widget, there is now a button to "Show Error" bars when plotting spectra.  From the command line, use the /show_err keyword when calling plot_spectrum.  To show errors on plots while looping through fit intervals, set spex_autoplot_show_err to 1.

There is an option to correct for albedo now.  In the GUI, the controls are in the "Select Input" widget.  From the command line, there are four new parameters: spex_albedo_correct, spex_source_angle, spex_source_xy, and spex_anisotropy.  (Note: currently DRM edges must be on integer boundaries.)

Previously fit results were stored in IDL save file (.geny file).  Now have additional option to store them in FITS file.  FITS files are more standard and can be read without SSW or IDL, whereas save file require SSW IDL to read them.  The spex_summ... parameters are stored in both, but the FITS files contain additional general information in standard FITS keywords, as well as an extension storing the values of all OSPEX control parameters.

In the OSPEX GUI, under File, there is a new button called 'Set Params Manually'.  From the command line, use the setparams method (o -> setparams).

Either of these brings up a widget interface with all OSPEX control parameters.  You can change them through this interface and click Commit to change them in the OSPEX object.  Be careful - it's easy to mess things up unless you're sure of what you're doing.

f_ion is now included in the fit function choices in OSPEX.  f_vth_ion is not included because you can use the vth component combined with the ion component to form f_vth_ion.

RHESSI SRM files written after this date will contain the response matrices for all attenuator states that occurred during the time interval of the spectrum file.  In OSPEX, you set spex_drmfile to that one SRM file, and OSPEX will automatically retrieve the correct response matrix for each fit time interval.  The 'Plot Time Profile' button in the Input Widget will display the filter states on the plot.  Also in the Fit Options Widget, the 'Show Filter States' button shows the filter states on a time plot.  In the 'Fit Options Widget', the 'Remove Bad Intervals' button will remove fit time intervals that include a change in filter state.

The progress bar now always (almost) responds when you click the 'Cancel' button.  Previously, it usually ignored 'Cancel' clicks.

The spex_interval_range control parameter was replaced by the spex_intervals_tofit parameter.  Previously you specified the range of intervals to fit, and all intervals in that range were fit.  Now you specify an array of the interval numbers to fit.

f_3pow is now included in the fit function choices in OSPEX.

You now have the option to initialize OSPEX with a FITS file, and then replace the data or background array in the objects with your own data arrays.  Refer to the 'Input Data from Command Line' section in the OSPEX Documentation.

Implemented the option to select different background time intervals and fitting polynomial order for separate energy bands.

Now you can set data directly into OSPEX without a FITS file.  You need to supply, at the minimum, a spectrum and an array of energy edges.  Refer to the 'User Data' section in the OSPEX documentation for details.

Previously you were limited to specifying a single energy range to fit over (spex_erange parameter).  Now you can select multiple energy ranges.  This will allow you to exclude artifacts in the spectrum.