Open kbarbary opened 8 years ago
Please clarify the issue: all these exposures are obviously references, and ES is not supposed to run on them.
This is about the cubefit-determined position, not ExtractStar. Since these are reference images, this likely means that something is going wrong with the alignment between the reference and exposures with the SN.
Sorry, I thought it was an issue which was assigned to me.
For PTF11dzm R band, cubefit is not getting the SN right at all:
You can see here that the SN does not appear in the scene model in the first three epochs, but is clearly visible in the residual. It appears that the minimizer converged to a local minimum, where the SN is positioned on top of the galaxy.
I think the initial position for the SN is pretty far off the true position, and this is causing us to miss the true minimum. Expanding the bounds for the fit doesn't help. To fix this, we need a robust solution for finding a global minimum. This could take the form of:
(x_data, y_data, x_sn, y_sn)
for each epoch (too many parameters!)We just need to get within a spaxel or so and then we can use the normal minimizer.
Could you just have a look at the residuals, see if there's anything structured there and use it as a new starting point?
Mickael reports a few SN with positions that look inconsistent between the B and R channel by up to 6 spaxels. PTF11dzm, at least, is a case where the galaxy is obvious, and the SN position with respect to the galaxy is obviously inconsistent. Epochs:
15_081_051_003_4/2
,13_101_072_003_4/2
(PTF11dzm)15_081_055_003_2/4
,15_080_059_003_4/2
(PTF12ena)15_110_042_003_4/2
,15_109_046_003_2/4
(SN2011bl)