Error estimation in v2.8.6

[mlnashby]

I've processed a mosaic with SEv2.5.0 and v2.8.6 using identical configuration files and weight maps. I get different uncertainty estimates in the resulting aperture photometry catalog, in the sense that under v.2.8.6 they are larger. Can someone point me to v2.8.6 release notes somewhere that describe all the changes made since v2.5.0?

[mlnashby]

(10-19-2011 17:17)mlnashby Wrote:  I've processed a mosaic with SEv2.5.0 and v2.8.6 using identical configuration files and weight maps. I get different uncertainty estimates in the resulting aperture photometry catalog, in the sense that under v.2.8.6 they are larger. Can someone point me to v2.8.6 release notes somewhere that describe all the changes made since v2.5.0?

By way of followup, I'll just mention that the errors are quite a bit larger under v2.8.6 -- factors of a few. So either something is awry or the definition of the errors has changed, I figure.

[Emmanuel Bertin]

Hi,
the detailed history of changes can be found in the SExtractor repository Log. I am puzzled by your findings. Do you see a similar effect without weighting?
- Emmanuel.

[mlnashby]

Hi again,

Thanks for suggesting something I can try.

I re-ran the two versions of SE again on the same input Spitzer/IRAC mosaic but this time without weighting. I do see that the effect persists.

I hope this is diagnostic of some common issue...but even if not I'm happy to try any other tests you can suggest.

Cheers,

Matt

[Emmanuel Bertin]

Hi Matt,

Looking at the differences between both SExtractor versions, the first explanation that comes to my mind is a different way of managing imager "gains" (actually ADU to e- conversion factor). V2.8.6 and above read the value from the GAIN keyword in the image FITS header and use that by default, whereas V2.5.0 would only rely on the GAIN parameter in the SExtractor config file (infinite by default). This should not lead to a measurable difference for far-IR images though, because detector "gains", when stated, are generally extremely high.
- Emmanuel.

[mlnashby]

Hi Emmanuel,

That was the issue. My mosaics had a carry-through keyword GAIN parameter (the Spitzer detector gain), and it was very different of course than what SE was looking for in terms of a GAIN parameter. So, when I replaced that GAIN value with an appropriate (much larger) number and then re-ran SEv2.8.6, I got appropriate error estimates. Specifically, they were the same as what I got with SEv2.5.0 to several significant figures.

Many thanks for helping me track this down...it is much appreciated!

Matt