-
Notifications
You must be signed in to change notification settings - Fork 10
Description
Lines 1368 to 1369 in bf4f496
| sigmaeff = sqrtf(sumw + f1f2o2*(1.0/infields[o1]->fgain | |
| + 1.0/infields[o2]->fgain)), |
Line 1411 in bf4f496
| float sigmaeff = sqrtf(*(pixwstack+i)+amu/infields[o]->fgain); |
The doc says that setting the gain to 0 means infinity. However, this seems not properly implemented in the clipping, where a zero gain leads to infinite sigma and thus no clipping.
My use case is to provide SWARP with absolute weight maps computed as inverse variance and to turn off RESCALE_WEIGHTS. The variance already includes both background noise and Poisson noise of the sources. To avoid double counting, I need the second term to be zero and the gain to be infinitely large. I can manually set the gain to a very large value in the header keywords of resampled images before coadding. But it would be nice to see that SWarp can handle this.