ap--/python-seabreeze

correct_nonlinearity=True gives incorrect intensities

acpo opened this issue · 4 comments

acpo commented

spectrometer and system information

  • model: HR4000
  • operating system: Raspbian (Buster) freshly installed last week
  • python version: 2.7.16 and 3.7.3 (tested in both)
  • python-seabreeze version: current master
  • installed-via: python setup.py install; really via pip

current problem

When the correct_nonlinearity option is set true {spec.intensities(correct_nonlinearity = True) } the data returned is incorrect. The numerical values alternate between very small and very large, like 1e-5 to 1e8.
When set to false, the data is correct. So there is an easy short-term fix.
However, the version of python-seabreeze from about a year ago didn't have this issue.

steps to reproduce

With an operating spectrometer,
Try spec.intensities(correct_nonlinearity=True)
then look at spec.intensities(correct_nonlinearity=False)

ap-- commented

That's unexpected.

could you post the return value of

# nonlinearity coefficients
spec._nc
# and dark pixel indices
spec._dp

Best,
Andreas

acpo commented 
>>> import seabreeze.spectrometers as sb
>>> spec = sb.Spectrometer.from_serial_number()
>>> spec._nc
[0.939935, 4.48307e-05, -1.58006e-08, 3.23411e-12, -3.96006e-16, 2.8124e-20, -1.05914e-24, 1.61494e-29]

>>> spec._dp
[5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17]
>>>
ap-- commented

@apco
thanks! found the issue! Currently all spectrometers that support non_linearity correction should return broken spectra when non_linearity correction is enabled.

Current Code:

python-seabreeze/src/seabreeze/spectrometers.py

Line 67 in 59c9f6d

self._nc = nc_feature.get_nonlinearity_coefficients()

Should be:

python-seabreeze/seabreeze/spectrometers.py

Line 125 in 8b50ed5

self._nc = numpy.poly1d(sbnc[::-1])

workaround for now

You should be able to do the following to get nonlinearity correction working until I release a new version:

import numpy
from seabreeze.spectrometers import Spectrometer

spec = Spectrometer.from_first_available()
# do the following to correct the nonlinearity coefficients:
spec._nc = numpy.poly1d(spec._nc[::-1])

# now it should work correctly
spec.intensities(correct_nonlinearity=True)
acpo commented

Thanks! and I look forward to the new version.