Calibration

[shay_ohayon]

Hi All,

 

I have inaccuracy with metering light. Both my cameras (EOS3,300D)

are 2/3 stop above what my Minolta VF light meter says (in

reflective mode). Who should I believe ? Is there a way to verify

who has the better accuracy at home without going to a repair shop ?

 

Thanks,

 

Shay

[jeff medkeff anchorage, a]

My opinion is that you should believe whichever meter gives you the best exposure most frequently. Run some subjective tests, and decide.

 

The film calibration process I'm familiar with starts out with shooting test cards with your favorite film (one for which the curve is documented), processing, getting the lab to measure spot densities (which ever fewer labs seem to be willing or able to do), find out where the densities fall on the curve, and see if that ends up being consistent with either the camera or the Minolta. It probably won't align with either, so you will have to adjust something; and the process is really a pain in the neck anyway. The processor's ability to repeat development conditions, and differences between film batches, can also influence results.

 

Assuming the 300D converts to gamma 2.2 and that you use a custom white balance (or perfect preset) with unchanging lighting, you might do a sort of lightweight version of a digital calibration. Take a shot of a standard gray card. The big peak in the histogram should be dead center, IOW the average 8-bit pixel value should be 128, if the conditions mentioned above can be met. There really is more to it than this, but the software needed to do a good calibration from a 300D image, and knowledge of how to use it, is fairly obscure.

[lucas_wheat]

a grey card is actually about 110, 110, 110 in RGB format. not 128.

 

take a shot with the 300D with its meter settings and then with the minolta VF.

 

shoot with the same WB if you don't shoot raw and same settings, preferably manual so you can keep them the same.

[jeff medkeff anchorage, a]

<p><i>a gray card is actually about 110, 110, 110 in RGB format. not 128.</i>

 

<p>What exactly is "RGB format?"

 

<p>What RGB numbers you "should" end up with after a calibrated exposure depends on the reflectivity of the card; the gamma; and the image profile. I've bought standard gray cards of 12% and 18% reflectivity and obviously they shouldn't have the same pixel value in a digital image.

 

<p>The reason to use the gray card in the test I proposed doesn't have anything to do with its reflectivity, though; it is its neutrality and lack of detail that is desirable. The 300D TTL meter averages to middle gray, and the camera applies a gamma 2.2 curve and a white balance to the linear readout for display on the camera's LCD. In gamma 2.2, middle gray is, by definition, 128 in an 8-bit image. (There is some discussion of this <a href="http://www.aim-dtp.net/aim/calibration/middle_gray/">here</a> and <a href="http://www.normankoren.com/makingfineprints1B.html">here</a>; both sources agree 128 in gamma 2.2 is middle gray, but the first allows that at gamma 1.25 107 would be more representative of middle gray as the eye sees it.) The uncontrolled variable that we are testing for in this experiment is whether the meter averages to middle gray, or not - if it doesn't, something isn't right and one might then favor the VF.

 

<p>If he shoots raw, middle gray will be 2,048 (12 bit or 16 bit uncorrected linear); if he demosaics to 16 bit sRGB IEC61966-1, middle gray will be 32,768 - values nowhere near 110. Photoshop truncates one bit for purposes of pixel value display, though. Shooting raw complicates the test and depends heavily on taking the proper post-exposure steps, including converting the image from a device-specific profile to a working space with known neutral crossing points.

 

<p>How the VF is calibrated I don't know, and I still think it is a good idea to decide which one is "better" according to subjective tests, unless very accurate tonal reproduction is required. And if that <i>is</i> required, we can't be content to talk about RGB without also discussing profiles and gamma.