Affiliation
Variable Stars South (VSS)
Thu, 10/04/2012 - 23:17
Hi All,
I recently purchased a Star Analyser diffraction grating with the aim of measuring spectral response curves (is this the right terminology?) of my DSLR cameras. What I want to determine is the individual red, green and blue channel curves and compare these with standard BVRc filters. Does anyone have a step by step tutorial for doing this?
From these filter response curves it should be possible to calculate coefficients for transforming instrumental magnitudes to standard BVRc magnitudes. I've determined emprical transformation coefficients from observations of standard stars but it would be an interesting exercise to calculate them from filter response curves. Cheers,
Mark
Hi Mark,
You have a good idea, but I think a transmission grating isn't the right answer. These units don't give a very good absolute response - there is different efficiency depending on the wavelength, and the DSLR will have vignetting, etc. so that the spread of the spectrum across the sensor will include these aspects.
The best way IMHO (and I've never seen it done - it would be an excellent paper) is to illuminate the DSLR with a monochromator. These are sorta inverse spectrographs - they use a blackbody source (like a tungsten lamp), feed the light through a prism, and then use a slit to isolate one wavelength of light to exit through the system. If you used that to illuminate the DSLR, and then took lots of images, one for each "step" of the monochromator, you would see the response vs. wavelength for the red,green,blue pixels, including any red/blue leak that their filters might have.
Arne
Actually I've seen it done well with the Star Analyser. You do need to do a spectral calibration of the system first. This is done by using a tungsen lamp to show the spectral response of the system. This can then be used with a filter to show the filter response across the spectrum. While a DSLR camera can be used, a monochrome camera would be much better.
Jeff
Hi Jeff,
it is the DSLR's inbuilt Bayer matrix filters I'm trying to measure. I was thinking of recording a spectrum of, say, an A0 star, calibrating wavelength axis using the zero order spectrum and a known absorption feature then dividing by a library spectrum of the same (or similar?) star. Does this sound feasible? Cheers,
Mark
Hi Mark,
You can practice on Vega, an excellent star to experiment with. Wavelength calibrating the line profile should be very easy. I've attached a calibrated line profile of Vega taken with a Star Analyser.
To make a instrument response calibration curve is more detailed than just dividing line profiles. I discuss this in detail in my book. Basicly you need to use both the library standard star and your line profile. The trick is to determine smooth curves for each line profile. This is done by deleteing all features from the profiles and then using a spline function to fit a smooth curve the the resulting profiles. These splinned curves are then used to produce the calibration profile. It can be confusing, but once you have gone through this a few times it is easier.
Jeff
Hi Mark,
Here is an attachement showing the original Vega spectrum line profile, the wavelength calibrated line profile and finally the instrument response corrected line profile.
Jeff
Hi Jeff,
Vega unfortunately is not an option for me. I'm at about 34 S latitude and Vega culminates at less than 20 degrees altitude - and behind buildings as viewed from my observatory. I'd have to use much fainter A0 stars, but thats ok since my setup can accommodate unguided exposures up to 2 minutes. If necessary multiple images could be stacked. Cheers,
Mark
Hi Arne,
some months ago Ed Budding and I discussed using a monochromatic light source to measure DSLR response curves. Ed made some enquiries with contacts in New Zealand (a unversity I believe) however they weren't interested.
I read a paper recently by Munari and Moretti about characterising photometric filters in the ANS collaboration program. The UV-Vis spectrometer they used would be ideal. I might make a few enquiries with Australian labs.
The Star Analyser approach is definitely a poor mans option but might prove useful. Cheers,
Mark
Hi Mark,
Yes, you can do it easily with the star analyser. I do it and I have submitted a paper to the JAAVSO about a simulation based on the response curves of the DSLR and using the Pickles spectra library. I have no news about that paper, I have no idea when it would be published. You can contact me through the "contact" flap of my profile.
The setup to make the pass-band response (the combination of all filters and the sensor) is simple. I put the SA100 in front of a 85 mm lens (or 200 also) with a slit about 1 meter away at the end of a carton tube, black inside and adding some baffles. The light source could be the sun reflected by a white paper target (never in direct). The spectrum of the sun is published at proper air-mass by the ASTM and other sources. The lines will enable you to calibrate the wavelength. Then you divide your DSLR spectrum by the sun spectrum (needs to be filtered at the same resolution). You shall also convert the sun spectrum to photon-count instead energy (our CMOS sensors are photon-count sensitive, not energy). Some correction of the efficiency curve of the SA100 should be applied too (The SA100 is a very specific case and its efficiency varies few between 400 and 700 nm). In the involved SA100 configuration vignetting is no problem. The results are in very good agreement with other sources for similar DSLR.
Another way is to use an halogen lamp with a known temperature (2800~3000 K) and the resulting Plank's spectrum (photon-count corrected). The wavelenght calibration could be made using a CFL lamp (to see http://www.astrosurf.com/buil/calibration/lamp1.htm , spectra of CFL is at the end of the paper.) The two techniques have provided exactly the same results.
For this simple work you can use the simple tools of IRIS (http://www.astrosurf.com/buil/index.htm) or ISIS, or VSpec... You could find tutorials and number of papers on the Buil's site and the ARAS site (http://www.astrosurf.com/aras/)
Clear Skies !
Roger
Hi Roger,
thanks for your detailed reply. Sounds like a nice experiment but I need to digest this info. Cheers,
Mark
Hi Mark,
Is this what you are looking for ?
http://www.astrosurf.com/buil/50d/test.htm
By Christian Buil comparing with a previously spectrophotometrically calibrated camera.
Note he used a LISA slit spectrograph. If you plan to use the Star Analyser, you need to arrange for the light source to be a point (or slit) and as Arne pointed out you should (as for any spectrum you are going to make quantitative measurements on) apply a flat field correction to the spectrum.
EDIT :
(Except in this case you can't use a flat as a spectroscopic flat contains the very thing you are trying to measure and therefore removes it - tricky)
Cheers
Robin
Hi Robin, Mark,
In fact the vignetting is not an issue with the SA100. In any case it's ok to make a classical DSLR flat like we do for photometry (not a spectro flat as denoted by Robin ! ).
In addition the vignetting should be negligible given the F# and the low dispersion of this grating. With my Nikkor 85mm F1.8 used with the SA100 in front, we get only F4 and the corner/center vignetting is <10%. Then the 400-700 nm spectrum extends on about 700 pixels in a image width of 4290 by the way a quadratic vignetting is < 0.2% at the spectrum level ! No issue...
The point that needs more consideration is the efficiency curve of the grating. The SA100 is a transmissive grating at low dispersion, such grating doesn't have significant s-p polarization problems (reflective have). Then it is well blazed for a DSLR pass-band. The efficiency of such grating is very high and the curve smooth. We do not have (confirmed by Robin) a curve measured in lab, but the modelization and measures made on similar gratings show the max transmission at 500~550 nm is ~80% , ~60% at 400 nm, ~70% at 700 nm. Those points can be interpolated at second order, fitting well the curve. I made an indirect check by comparing the R,G and B integrated output of the DSLR with and without the SA100 and this confirmed the assumption. Anyhow we don't need much accuracy around 400 and 700 nm, the transmission of the DSLR being very low at those wavelengths.
Checking further the efficiency curve of the SA100 should be very interesting, we need only a couple of points, I think 5 is enough given the smooth curve.
Cheers,
Roger
Although I'm a bit late to this thread, I thought the link below might be helpful. It shows the steps to do instrument response calibration.
Although this video is specific to the RSpec software (of which I'm the author) it should be useful to anyone interested in the topic, since the concepts and general work-flow would be the same in other software too.
http://www.rspec-astro.com/more-videos
Select video 15, entitled "Adjustment for Instrument Response"
Tom