Closed mzechmeister closed 5 years ago
This error can arise in coadding: spt.header['HIERARCH SERVAL COADD SN%03i' % o] = (float("%.3f" % sn), 'signal-to-noise estimate')
spt.header['HIERARCH SERVAL COADD SN%03i' % o] = (float("%.3f" % sn), 'signal-to-noise estimate')
Here a nan has entered the coadding. But using -skippre, the error doesn't occur.
nan
-skippre
The main problem is that the section attribute seems not to provide a copy. So when the data are loaded:
section
https://github.com/mzechmeister/serval/blob/f48e7cb6fc8af3a1fd789cd54ad7d5cc77cbe86c/src/inst_CARM_NIR.py#L132 or https://github.com/mzechmeister/serval/blob/f48e7cb6fc8af3a1fd789cd54ad7d5cc77cbe86c/src/inst_CARM_NIR.py#L155
and modified inplace by https://github.com/mzechmeister/serval/blob/f48e7cb6fc8af3a1fd789cd54ad7d5cc77cbe86c/src/inst_CARM_NIR.py#L173-L174 or https://github.com/mzechmeister/serval/blob/f48e7cb6fc8af3a1fd789cd54ad7d5cc77cbe86c/src/inst_CARM_NIR.py#L173-L174
the data are cumulatively multiplied with 10000 (related to #22). Moreover, the bad pixels are flagged and interpolated inplace (to avoid interpolation problems for the highest SNR spectrum): https://github.com/mzechmeister/serval/blob/f48e7cb6fc8af3a1fd789cd54ad7d5cc77cbe86c/src/inst_CARM_NIR.py#L150
Everything is fine when loaded the first time, but the second time the interpolated values are remembered, while its error is still nan and enters calculation, since only nan flux is flagged.
This error can arise in coadding:
spt.header['HIERARCH SERVAL COADD SN%03i' % o] = (float("%.3f" % sn), 'signal-to-noise estimate')
Here a
nan
has entered the coadding. But using-skippre
, the error doesn't occur.The main problem is that the
section
attribute seems not to provide a copy. So when the data are loaded:https://github.com/mzechmeister/serval/blob/f48e7cb6fc8af3a1fd789cd54ad7d5cc77cbe86c/src/inst_CARM_NIR.py#L132 or https://github.com/mzechmeister/serval/blob/f48e7cb6fc8af3a1fd789cd54ad7d5cc77cbe86c/src/inst_CARM_NIR.py#L155
and modified inplace by https://github.com/mzechmeister/serval/blob/f48e7cb6fc8af3a1fd789cd54ad7d5cc77cbe86c/src/inst_CARM_NIR.py#L173-L174 or https://github.com/mzechmeister/serval/blob/f48e7cb6fc8af3a1fd789cd54ad7d5cc77cbe86c/src/inst_CARM_NIR.py#L173-L174
the data are cumulatively multiplied with 10000 (related to #22). Moreover, the bad pixels are flagged and interpolated inplace (to avoid interpolation problems for the highest SNR spectrum): https://github.com/mzechmeister/serval/blob/f48e7cb6fc8af3a1fd789cd54ad7d5cc77cbe86c/src/inst_CARM_NIR.py#L150
Everything is fine when loaded the first time, but the second time the interpolated values are remembered, while its error is still
nan
and enters calculation, since onlynan
flux is flagged.