Laboratory Characterisation of a Commercial RGB CMOS Camera for Measuring Night Sky Brightness
Round 1
Reviewer 1 Report
Great article covering an important field of research.
However, the fact that an ad-hoc fitting was needed seems to imply that the calibration procedure could be improved.
In the attached document, yellow highlights are comments and suggestion on the content, green highlights notes on the form, red strikethroughs suggestions to remove and blue strikethroughs suggestions to edit.
Comments for author File: Comments.pdf
The language needs revision. Some sentence constructions are a bit heavy. The comments in the review document are not exhaustive on this matter.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
Dear authors,
The purpose of your paper seems simple: how to emulate the SQM spectral and spatial responses using standard RGB values from a digital camera. To that end you propose to create a source combining 9 narrow-band LEDs; and compare the camera and the SQM output values. Then you’ll try to define some weighting matrix to combine the numerical values of the RGB pixels into some scalar which should approximate the SQM result.
That approach is sound and safe.
Yet, as you enter the practical details, a lot of issues arise:
1) the NI USB 6008 digital IO are controlled digital sources, with a typical output current of 8.5 mA. That’s OK to drive a signal LED; but that is not adequate to control a reference source. LEDs should be current-driven, should be allowed ample time to reach thermal equilibrium, and if you operate a few of them from the same PCB there will be thermal feedback between them. Meaning that if you operate LED A, then LED B, then LED A together with LED B, the last flux will not be the sum of the two previous ones.
2) between lines 228 and 250, you define a few approximations, the latter being based on refinement of the former. Except your approach is loosely described. This part must be more detailled. Another scientist working in the field would not be able to reproduce your approach
3) between lines 251 and 299, I’m lost. Do you use your Minolta CS1000 in tri-stimulus or in spectral mode? How are the curves in Fig. 2 constructed? Why uses this approach? If you want some RGB to XYZ transformation, the CIE recommends the use of the 3x3 matrix based upon the white point CCT. In the present case, there is no white source.
4) around line 335, the comment ‘errors present significant fluctuation’: yes but … y(.) and z(.) are zero outside some range. This point must be reviewed.
5) Fig.3, Fig.4 and their introduction are even more cryptic to me. Typically, I would have expected a weighting matrix between the 9 LEDs output values and the Minolta CS1000 X, Y, Z reading. Then use the GUM uncertainty propagation method to deduce errors (as standard deviation) on x and y. Those two sections give the impression that you’re searching some way of modeling the error, instead of using establish methodologies.
6) for Sec. 3, there is a relationship between ISO, f number, exposure time, and luminance. Why not use it as starting point?
7) Sec 4.2 is about approximating the SQM transfer function by a linear combination of the RGB channels. Did you check your SQM sensitivity w.r.t. its theoretical value? Once you’ve established the spectral approximation error, how do you translate that to errors on the SQM readings obtained by combining the RGB values? Basically, the main concern of the whole study is that you use a few times linear regression approaches to say « I compare this to that which is supposed exact ». Except that for each comparison, you do not take into account the intrinsic noise of the « reference ». THAT noise is transformed into bias, as the zero-mean white noise on the reference is squared by the regression computation. This is what you observe in Fig. 14a: there is a constant offset between the curves.
Conclusion :
The subject of your work is interesting, but the methods used have too many measurement biases that lead to significant uncertainties in your results.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
This paper conducted the analysis of the spectral responsivities of the SQM and a specific RGB camera. The peeformances are compared when the two instruments are excited by real night sky spectral radiances. Here are the comments:
1. The Introduction mainly clarifies the digital RGB cameras and its principle to measure the night sky brightness. However, literature review about previous studies and the gap about current study are not mentioned. Please elaborate in details.
2. Please elaborate the contribution or innovation of this method.
The quality of English language is ok. Minor English editing can be just required.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
NO additional coments. Very good article. No authors corection needed.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf