Hi Peter,
On Tue, Jun 17, 2008 at 04:17:33PM +0100, Peter W. Draper wrote:
> On Tue, 17 Jun 2008, Horst Meyerdierks wrote:
>
> >On Tue, Jun 17, 2008 at 03:29:30PM +0100, Peter W. Draper wrote:
> >>On Mon, 16 Jun 2008, Matthijs H.D. van der Wiel wrote:
> >>
> >>No, seems there's a long standing bug in the FWHM reported (took me a
> >>little while to appreciate that since I didn't write the code). The
> >>current value is actually the Gaussian sigma, so to get FWHM you need to
> >>scale by a factor 1.665. I've corrected PHOTOM for the next release.
> >
> >No, that can't be right. sigma is the half width at approx 60% of max
> >if you really have a Gaussian. This factor must be larger than 2, because
> >at 50% of max the function is wider than at 60%. I seemed to remember the
> >factor was sqrt(8 ln(2)), ex-radioastronomers find this hard to forget.
> >And Wikipedia confirms it. 2.355.
>
> Hi Horst,
>
> depends how you define your Gaussian, seems this code must use:
>
> exp(-(x**2)/(sigma**2))
>
> so that factor is sqrt(4*ln(2)), not sqrt(8*ln(2)). It initially confused
> me as well.
OK, if sigma is not the standard deviation, then one needs to
know the source code to figure out what sigma is.
Cheers,
Horst
--
Horst Meyerdierks Royal Observatory Edinburgh
Un*x/Network Manager [log in to unmask]
http://www.roe.ac.uk/~hme/ +44-131-6688-309
|