Hi,
Thank you very much, John and Doug, for the comments, which helped a great
deal. Let me conclude this discussion in the following:
1) The weights are indeed NEARLY independent of the instruments for a same
band and a same spectral model.
2) The thread should definitely be revised because people are confused and are
getting wrong results. I'd suggest that the spectral fit part should not be
included in the thread. The thread should only tell how to generate the
weights, assuming a spectrum.
3) An alternative way, in addition to Sherpa, should be described. Not
everybody uses Sherpa. For example, I faked an ARF file (making specresp=1 for
all channels) and then used XSPEC to generate the weights for whatever bands
and models. This approach actually accounts for the RMF effect(!), which is
however slightly instrument-dependent. Do you think that this is a more
transparent, straight-forward, and more accurate way to do the job?
For your reference, I get the weights for the 2-4 keV band as
2.25 0.372433
2.75 0.264904
3.25 0.202489
3.75 0.160174
assuming a power law with an energy index =0.7 and an absorption =0.03E22
as well as resp=acisi_aimpoint_rmf.fits is used.
Cheers,
Daniel
> In chandra-users, you wrote:
> >This is about the creation of a spectrally-weighted instrument map in an
> >energy band. I was trying to understand the weights produced by a member of
> >our group for various chips, using Sherpa. The weights are drastically
> >different from one chip to another, which does not make sense to me. Then I
>
>
> Generally speaking, if different energy bands are chosen for
> different chips (to reflect instrument specific effective area
> curves) one would have different sets of weights for each chip.
> But in this case, the FI and BI chips behave similarly enough
> that one could use the same energy bands for both, and the
> weights should be the same for each chip.
>
> Aside from that, the weights should normally be derived from
> the same spectrum model and are instrument-independent.
>
> [snip]
> >
> >The thread requires the use of both RMF and ARF files, in addition to a
> >spectral data file. I am not exactly sure how these files are incorporated
> >in the calculation of the weights, since I don't use Sherpa. It seems to me
> >that none of these files should be needed to calculate the weights, if they are
> >instrument-independent. However, the files might be required in order to
> >get the model spectrum, although the results don't depend on the files at all.
> >If this is the case, the thread should be designed in a much simpler way and
> >should explain clearly that the results are independent of the input files.
>
>
> I suspect that the response files are used only to provide a
> grid for Sherpa to compute the model spectrum.
>
> The thread should make it clear that the weights are mostly
> independent of the instrument response.
>
>
> > Then, I still don't understand why the weights should vary from one
> >chip to another. Am I missing something here? I'd very much appreciate
> >comments from experts on the issue.
> >
>
> You're right -- apart from some dependence on the specific
> bands selected, the weights should not vary from chip to chip.
> They should all be computed using the same model spectrum.
>
> John Houck
> On Thursday 23 May 2002 08:23 pm, wqd@gcs.astro.umass.edu wrote:
> > Hi,
> >
> > This is about the creation of a spectrally-weighted instrument map in an
> > energy band. I was trying to understand the weights produced by a member of
> > our group for various chips, using Sherpa. The weights are drastically
> > different from one chip to another, which does not make sense to me. Then I
> > was reading the online thread
> > (http://asc.harvard.edu/ciao/threads/spectral_weights/), which is
> > complicated and seems to problematic. However, I do understand the help on
> > the routine mkinstmap (ahelp mkinstmap) and the quoted article by John
> > Houck. Let me explain my problems in the following:
>
> Daniel,
>
> If you're having trouble with the threads, the best place for help is the CXC
> helpdesk (http://cxc.harvard.edu/helpdesk/ - which is linked to from the
> bottom of each page).
>
> > The thread requires the use of both RMF and ARF files, in addition to a
> > spectral data file. I am not exactly sure how these files are incorporated
> > in the calculation of the weights, since I don't use Sherpa. It seems to me
> > that none of these files should be needed to calculate the weights, if they
> > are instrument-independent. However, the files might be required in order
> > to get the model spectrum, although the results don't depend on the files
> > at all. If this is the case, the thread should be designed in a much
> > simpler way and should explain clearly that the results are independent of
> > the input files.
>
> The thread is designed for the case of a user fitting a model to a spectrum
> so as to determine what parameters to use. The spectrum.sl script used to
> calculate the weights file used by mkinstmap is just a list of energies (kev)
> and fluxes (photon/cm^2/s) [ie is independent of chip as you said], so if you
> already know the model parameters you want to use you can create this
> directly in sherpa (or whatever fitting package you use).
>
> Doug
>
> --------------------------------------------------------------
> Doug Burke | Email: dburke@cfa.harvard.edu
> Harvard-Smithsonian | Phone: (617) 496 7853
> Center for Astrophysics | Fax: (617) 495 7356
> 60 Garden Street MS-83 | Office: P-138
> Cambridge, MA 02138 |
> --------------------------------------------------------------
-- ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Address : Daniel Q. Wang + Astronomy Department, B-524 LGRT + University of Massachusetts, Amherst, MA 01003 + E-mail : wqd@astro.umass.edu + Phone : 413-545-2131; Fax: 413-545-4223 + URL : http://www.astro.umass.edu/~wqd/ + ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
This archive was generated by hypermail 2b29 : Thu Dec 13 2012 - 01:00:07 EST