In my previous post, I have discussed what looks like an over-exposure effect on images of spectra taken with the CCDSPEC spectrometer.
See two other posts on this issue:
This made me think – does this effect the spectra themselves? Is it important not to over-expose the images with regards to identifying the elements/molecule present in stars and nebulae?
One way to look at this in a controlled way is to see what effect over-exposure has on the spectra of daylight (point out of window at the clouds) and compact fluorescent bulb (nicely defined lines) during daylight hours – easy to get the spectra and to control variables such as intensity and ensure spectrometer points constantly at same point.
Below is a 3 second exposure – drastically over-exposed with interlacing artefact appearing in conjunction with this.
FITS Liberator shows the white clipping in the above image:
Spectrum generated in MySpectra software from the above image:
The following is 0.5 second exposure where clipping is not evident:
Spectrum from 0.5 second image generated in MySpectra software:
Particularly between 450 pixels and 550 pixels, there are a significantly larger number of lines evident in the graph on the shorter exposure than the longer exposure, with no obvious missing lines in this shorter exposed spectrum even in areas of the image where exposure intensity is lower on both images.
Image below is calibrated for wavelength on x-axis and shows that the above indicates significant loss of data in the over-exposed image for 500-600nm in particular.
The following image is at the other extreme – here the exposure is only 0.002 seconds:
Spectrum of above 0.002 second image generated with MySpectra software:
Although the graph is not as smooth as fewer data points on the y-axis, there is little obvious evidence of loss of major lines from under-exposure.
The above suggests that it is better to slightly under-expose than over-expose an image of a spectrum from the CCDSPEC spectrometer (and probably other makes too as I can’t see how make of spectrometer would matter to this).
Compact fluorescent bulb:
Exposure 1.5 seconds (over-exposed):
Exposure 0.4 second (not over-exposed):
Although most lines are evident in both graphs, the lower exposure leads to narrower peaks and consequently measurements of those peak frequencies should be more accurate. This is especially important for research purposes.
In conclusion, not a great idea to over-expose spectra but nevertheless they are quite robust and a lot of interesting data can be seen even in under- or over-exposed images, so for outreach purposes it probably does not matter too much as long as the over- or under-exposure is not too extreme.