DSLR Linearity

Hi Roger Whats the meaning of the green, blue, and red linesper ISO setting in your Canon EOS 450D linearity plot ? CS Wolfgang

Hi Wolfgang, It's simply the red, green and blue channels outputs of the DSLR (the post is probably not fully clear about that, sorry). The green is one of the two green channels, I have not interpolated them, normally I use the average of both G ofeach Bayer cell. I didcheck the red and blue as I use them in my specific process of color indexcorrection and blue extinction correction ( I plan to explain thisin a coming paper ) The green curveis the pure raw values. The blue isscalled to (B-V) ~0.4 The red scale is increased by a factor 1.5formaking the graphreadable. This rescallingwas needed due to the fact the spectrum of the"white" LED being used was far from anythingstellar! The red and blue outputs of a DSLR at T~6000 K are much below the green due to the poor transmission of the red and blue filters, by the way the saturation occurs first on the green in most cases. The saturation could occurs on blue first only for very blue stars at (B-V) ~ -0.18 like eta AURat zenith. The redremains below the green for any star even the very red ones. Yours truly, Roger

Hi Roger, This is the exact sort of thing we need to have for DSLR cameras. Is your illuminator small enough that it could be shipped (or cheap enough that we could make a few) so that we may characterize more cameras? I have been thinking of a simple test for spectral response too as not too many manufacturers release that information. The combonation of spectral response and linearity for a variety of cameras would make a nice article in JAAVSO. Brian

Hi Roger, Seemeasurement of the spectral response of an unmodifiedCanon 350D compared with a sony CCD based camera using a standard star and an objectivediffraction grating on my website here http://www.threehillsobservatory.co.uk/astro/SA_DSLR_instresp.gif Note these curvesalso include the efficiency of the grating(and a small contribution from the lens and the atmosphere). Measuring the response of the camera independentlyis tricky but can be done forexample by comparsion with a sensor with a known (eg published) response. Buil has done thisfor the40D and 50D here http://astrosurf.com/buil/50d/test.htm Robin

Hi Brian, The illuminator is a simple LED torch available at superstors for a couple of $. This is the only simple solution I have found to achieve a stable illumination using a battery in good condition. But white LED have a poor spectrum and their pattern is very non-uniform reason why two diffusers are needed. One is near the imager and the illumination is measured by the photometer at this level. The illumination is set by adjusting the distance of the torch. The key item is in fact the photometer. You can probably find one at labs of your university. If someone is interested I know a solution to build one having a good linearity in a wide range even if the output will be in arbitrary units, cost is low(this is a solution we have been using a lot in my labs in the past) Someone could also use the exposure time of the camera to vary the illumination. In CMOSit issetat theelectronics leveland there is no issue with the shutter accuracy (it is here just to protect the chip during the unloading operation) BUT... the displayed times arethe traditional valueswell knownof thephotographers instead the true valuesdefined by the electronics !!! Those values are in general just in geometric progression instead. It's easy to readjust that series to thenearesttraditional values. It's not myfeeling we should test any CMOSDSLR. The imager responseshould bewell linear up to nearthe saturation point (seen only at the lowest ISO) and that saturation point is anyhow set near the top of the range ofthe ADC. At next ISOthe response curve is clipped by the ADC rangenear the middle of the imager response, by the way well linear. In fact I suspect the circuits that read theelectron charge INEACHPIXEL in CMOS imagersare not perfectly linear but the processor or any LUT makes a linearity correction. The fact thereare codes systematicallymissingin the data output is typical from suchnon-linear recoding.Under that circumstance it's probably notvery useful tomake a lot of tests... CMOS imagers are a true revolution in imaging technology, CCD are near over !! And apparently astronomersare not considering it... This fact has been debated in therecent astro conference"Rencontres du Ciel et del'Espace" in Paris. CMOS is a very different world than the small specialized CCD business, on many aspect: principles, technology, manufacturing technology, mass manufacturing constraint,business actors...Beside the fact it's a challenge for companiesserving the astronomy needs we have also torevise notions that have been developped with the CCD. Yours truly, Roger

When Idid some linearity checks of a long exposure modified webcam, Iput it in a light tight box and used a constant total exposure time, switching the lamp on for avariable time during the exposure. This avoids the problems of needing a photometer or relying on the program exposure times which are notoriously inaccurate in webcams. It also means you only need one set of darks. Unevenness of Illuminationproblems can beminimised by taking the mean value in a small area of the field (noteonly onecolour channel should be used of course) Robin

Hi Robin, Spectral response: We have been discussing this a year ago if you remember. I use a smallhome made spectrometer. The light source could also be an halogen lamp corrected from the Planck law or the sunny daylight then corrected from the ASTM standard. But this is true the continum is probably not very accurate in particular due to the blase pattern. I think it's not very critical forwhat we are interested ininphotometry. Theobjective here is essentially to understand the differencebetween the GG channels andthe V Johnson filter. Idid show it for a G9 last year and I will soon post somethingon that subject. Light source for linearity: Yes, this is also an option but you need an accurate timer to drive the lamp if you are looking for an accurate characterization of the linearity.Another issueI got is the color variation of the lamp, if you switch a thermal lamp the rise and fall time could affect the exposure and the color during such phases. This could be minimized using aLED but even the LED have some drift depending the junction temperature, this is probably ok in case ofexposure long enough that should minimize the transient contribution. But, ok, today all DSLRare CMOSbased and asexposed in my answer to Brian I do not see much reason to make extensive test of linearity of such devices. Yours truly, Roger

Hi Roger and Robin, Good points in your posts above. I agree with you that linearity tests probably aren't required, but I think it is worth going through the motions of such a test in case one of our participants were to branch into CCD or PEP. Perhaps we can find a faster solution. Mass-produced cameras will have very similar spectral response and linearity characteristics. Perhaps we could make a series of pages on which the characteristics of a "typical" camera of a product line (i.e. Cannon D350) is tested. This will serve as a first-guess for the camera's characteristics. If the participant wishes to do their own linearity / spectral response test we can have instructions for doing so. Do you think that is a reasonable solution? I can get some true wiki pages setup for this so I needn't be involved in the entire process. Brian

Hi Brian, I agree this isa good approach. Getting into brand / type comparative test is not the way I would go. A typical unit would be the 450D I think. I have one it's easier for me !! But, ok, from what I have seenmany of us use it and this type has been sold a lot even if it is now near out of commercialisation. New 500 / 600with so many(smaller... ) pixels aremaybe questionnable for photometry (needs more analysis, sure), but thevery new low end 1100D is in factnot farthe 450D (evenbetter !) and I think many are gone to buy it. Next we could use an old CCD model(without brand mention) as a comparison. This could be the restart of Camera Characterization (Maybe under anothername? like"DSLRand Photometry" ? ) One pointthat made things difficultfor me in thepast was the editing, at a certain level of complexitythe documentwasvery difficult toeditproperly ?Could we havea bettertool ? Yours truly, Roger

Roger, I too am frustrated with the editor here on the CSky website. I've asked for it to be changed, but the web developer is bogged down with other tasks. I could move the tutorials over to Google Docs and provide people with editing privilages. What do you think would that help? Brian

Brian, I have no experience with Google Docs but if you think it's better why not ? Roger