Re: back_fun problems

On Sun, Oct 28, 2007 at 15:04 -0400, David P. Huenemoerder wrote:
> Hi -
>
> I'm trying to create a model background for HRC-S for simulation
> purposes. So I used tg_bkg (a ciao contrib script) and dmtype2split (a
> ciao tool) to create PHA Type I files of the summed _up, _down
> background from the default "bow-tie" LETGS extractions. These have a
> constant ratio of background width to source width of 10, so the
> BACKSCALE is 10, and the counts are scaled by 1/BACKSCALE when loaded.

Isis doesn't perform this scaling on input.  Does your script
do this explicitly?

>
> So I then load the background spectrum for -1 order, and fit a
> poly(1)+gauss(1) (and no responses assigned).  So the resulting model
> defines the counts/s/A/area.  (where "area" means spectral region
> width of 1 unit).  This model gives a reasonable reproduction of the
> LETG/HRC-S background.
>
> If I then try to use this model as a back_fun via:
>
>   back_fun( 1, "get_data_exposure(1)*bin_width(1)*( poly(1)+gauss(1) )" ) ;
>
> which is what I think I need according to the B(h) term in eq. 7.35
> of the ISIS manual:   s * A * (counts/s/A) = counts/bin.
>

I don't think the factor of bin_width() is needed.  The computed
model values already include the integration over the bin-width.

>
> To verify, I define a dummy model with zero normalization:
>
>   fit_fun("Powerlaw(1)");
>   set_par( "Powerlaw(1).norm", 0.0, 0, 0, 1);
>
> I should recover the background model when I eval_counts.
> I get the right shape, but it seems to be a factor of 8 too low.
> (the data_backscale is 1.0).
>
> This factor is supiciously 1 / bin_width / back_backscale ( 1./ 0.0125 / 10 ).
>
> Am I misinterpreting some scaling in fitting the background pha and/or
> in defining the background model?    I can renormalize this to derive
> an empirical background model, but I'd like to understand the
> discrepancy.

Isis doesn't rescale or otherwise modify the values computed by
the background function, so that function should include
whatever scaling is appropriate for the relevant background
exposure time and extraction region area.

At the moment, I'm not sure where the problem is occurring.
Could you send a script that reproduces the problem?

Thanks,
-John

----
You received this message because you are
subscribed to the isis-users list.
To unsubscribe, send a message to
isis-users-request_at_email.domain.hiddenwith the first line of the message as:
unsubscribe