EOS calibration test

Some while ago, Chuck Westfall of Canon USA described a simple practical test suggested by Canon to give a general confirmation that the overall "calibration" of the AE system of EOS dSLRs was generally at the point intended by the design.

The test involved making an automatic exposure with the camera regarding a known reflectance patch on a certain test target and then examining the exposure result for that in the JPEG output file. The criterion was that, with Photoshop having "virtually converted" the image to the gray gamma 2.2 color space (for examination purposes only), the value of "K" (the only coordinate of that color space, which we can call "blackness") for the patch should be about 55%. (That corresponds to a relative luminance of about 17%.)

[Note that the property, examined by this test (which we can call "standard exposure result") results from the combination of the assessment of camera sensitivity ("ISO speed", for example) and the "calibration" of the AE system.]

However, since that era, there has been much discussion about the "standard exposure result" of various new EOS dSLRs being different than that for earlier models. (This is sometimes spoken of in terms of a difference in the manner of rating the "ISO speed".)

Do we know whether Canon endorses this same simple check for all EOS DSLR's, or only for certain ones, or maybe no longer at all?

How it looks here

Today I decided to do some tests on my EOS 20D with regard to the matter discussed at the head of this post. (I had done some of this a couple of years ago, but didn't really capture the detailed results!)

One thing I found is that the workings of the AE system do not follow the idealized relationships I often use as the basis of my lectures on metering principles. For example, the "exposure result" of a metered exposure of a frame-wide uniform-reflectance neutral target varies with the so-called "ISO setting" of the camera and (if we imagine Tv mode metering) of the setting of shutter speed. (The range of the variation is about 0.5 stop, not really a biggie.)

Some of this may be due to improper shutter speeds in my machine, although it is rare to encounter these in the modest shutter speed regime in which I tested today.

For now, I will just chalk this up the the well-known fact that there are a lot of little wheels whirring in the mind of an EOS camera, in ways we mortals can never hope to know.

That having been said, for an "ISO" setting of 200, and a shutter speed (in Tv mode) of 1/200 second, the exposure result for the (metered) test exposure is right at that prescribed by the Canon handy test: a value of the K coordinate (for a pro forma transformation of the image into gray gamma 2.2 color space) of 55%.

The RGB values for this image are 116/116/116. This "suggests" a focal plane photometric exposure of 17.5% of saturation (an implication that must be taken with a grain of salt given the various unknowns in the Canon image processing).

So it looks as if the Canon handy calibration test, originally enunciated with respect to the EOS D30, may be equally applicable to the EOS 20D.

So, you may ask, did you do this all day instead of taking pictures of cute girls?

No, I did it after taking pictures of a cute girl. And after taking pictures of some guys finishing the concrete on our new street paving job.

Best regards,

Doug

Doug_Kerr said:

So, you may ask, did you do this all day instead of taking pictures of cute girls?

No, I did it after taking pictures of a cute girl. And after taking pictures of some guys finishing the concrete on our new street paving job.

Click to expand...

LOL

Doug, there is no real need/use in explaining the motivation for shooting what one shoots (not for me anyway) . I e.g. have been photographing (amongst others) a plant disease on oak leaves, and other macro stuff of weeds (no girls in front of the lens today), but intend to shoot some wild water canoeing later this week.

The participants of OPF have a variety of backgrounds, some (AKA artists) will never grasp the pixel peeping aspect, and others (AKA pixel peepers) may have difficulty in grasping the poor fundamental understanding of our tools by others. In the end we all like to make beautiful images, some just with greater predictability than others.

IMHO it takes both types of photographers (and those somewhere in between) to color the fascinating spectrum of photography we all enjoy.

Bart

Hi, Bart,

Well said!

Best regards,

Doug

Hey Doug, what about posting a pic of the cute girl here?
It would give us a rest in the middle of a tech (but interesting, as usual) discussion!

PS Ok, Bart a little bit of plant disease too!

Hi, Nicolas,

Nicolas Claris said:

Hey Doug, what about posting a pic of the cute girl here?
It would give us a rest in the middle of a tech (but interesting, as usual) discussion!

Click to expand...

Well, if you insist.

Our granddaughter Tiffany (16), done for a school sports publication of some sort:



Sorry about the irregular catchlight situation - my rig wasn't set up properly at the time.

Best regards,

Doug

Thanks Doug for the fresh air… seems to be a lot of joy there!

Metering equations

I thought I would expound a little on my remark about the "classical" exposure metering equations on which much of my "idealistic" discussion of exposure metering, sensitivity ratings, and the like are predicated.

The classical reflected light metering equation for digital cameras is based on the underlying objective that the working of the metering system will always produce a constant value of Havg, the average photometric exposure on the sensor, as a fraction of Hsat, the saturation H for the imaging system. This will be true for any scene.

[Photometric exposure, H, is, basically, the product of illuminance on the sensor and the duration of the exposure.]

[Saturation H, Hsat, is the value of H that produces the largest distinct digital output.]

This can be expressed as:

Havg = D*Hsat [1]

where D is a constant that expresses our desire for what Havg will (always) be as a fraction of Hsat.

Hsat will vary with the sensitivity of the camera, and in fact one standard measure of sensitivity (the "saturation" flavor of ISO speed) is defined to directly reflect that:

Ssat = 78/Hsat [2]

where Ssat is the ISO speed, saturation flavor. If the camera has different sensitivity settings ("ISO settings"), then properly this should apply separately for each.

If we skip over a lot of the intermediate algebra, we get to one form of the classical reflected light metering equation:

t/N^2 = K/(Ls*Ssat) [3]

where t is the exposure time (shutter speed), in seconds; N is the f/number describing the aperture; Ls is the average luminance of the scene as measured by the meter; and Ssat is the sensitivity of the camera (as the saturation flavor ISO speed). K is the "reflected light metering constant", whose choice reflects the manufacturer's metering strategy (actually coming from the choice of the constant "D" I use above, describing the target value of Havg as a fraction of Hsat.)

If we do not wish to think of sensitivity in terms of the standardized definition of ISO speed, but rather on some more arbitrary (proprietary) basis (but still based on Hsat), we can rewrite this as:

t/N^2 = J/(Ls*Sx) [4]

where now Sx is the sensitivity in "proprietary terms" and J is a new constant, which differs from K by the amount that values of Sx differ from values of Ssat.

Now, if this equation were true, then regardless of the value of J, it will lead to the relationships we have come to expect in metered exposure, including:

-- If we cut the exposure time in half, the aperture will "enlarge" (f/number decrease) by "one stop".

-- If we double the sensitivity rating and keep the same aperture, the exposure time will be reduced to 1/2 its previous value.

-- If we double the sensitivity rating and keep the same exposure time, the aperture will "reduce" (f/number increase) by "one stop".

Now, the comment I made about my test of the Canon EOS 20D is essentially, "its exposure equation evidently is not as simplistic as equation 4, above".

This of course scrambles simplistic efforts to evaluate such things as Canon's "rating" of sensitivity.

Best regards,

Doug

Nicolas Claris said:

Hey Doug, what about posting a pic of the cute girl here?
It would give us a rest in the middle of a tech (but interesting, as usual) discussion!

PS Ok, Bart a little bit of plant disease too!

Click to expand...

How about this (not as cute) girl, admittedly with hairy legs and eyes that are too big for comfort ...



Bart

Hi Bart

LOL!

I've already seen your amazing image in the "close-up and macro forum", interesting technic! as for the girl, this one should have a shower and a shaver !-)

I have to recognize that I'm far too old for the first one but she's far more cute!
I wouldn't feel very comfortable with your girl, but obviously, if we share the same air, we are not in the same world. A world that you and other macro shooters do bring to our sight… Thanks!
This is one of the fair aspects of photography, nonetheless macro permits artistic point of view (your seeds in the already mentionned post), but also helps to discover and learn, that is good, it is a very old human character that when one knows one appreciates better (in all meanings) and when one appreciates better, one destroys less as one starts to like…
Obvious but sometimes good to remind… (eh Kombiz ;-)

Well, getting back to the original question (what a concept - on dpr you can get expelled for doing that), I was still hoping to hear whether that Canon handy test is deemed applicable to such cameras as the EOS 20D.

I've done some testing in that regard on my machine, but the first batch of data was useless since I had forgotten to cork the viewfinder eyepiece! Bart van der Wolf was kind enough to remind me about this after I had reported some anomalous results.

But preliminary analyis of the new (properly done) results for my machine seem to suggest that it exposes about 1/2 stop hotter than what would be expected (as I make it) from the Canon "handy test". So of course my question is, "does that mean my machine is out of calibration, or what?"

Best regards,

Doug

RAW camera calibration

Hi, Doug....

I don't know about calibrating the camera as you discuss....and I don't mean to digress, but I think this is associated with your question.

I only shoot RAW and use ACR to "develop". One of the key things I found was the need to "calibrate" the camera for proper color rendition...i.e. a color profile for hue & saturation.

I used http://www.rags-int-inc.com/PhotoTechStuff/AcrCalibration/ which works very well. Actually, the hardest thing was to get a good, properly exposed image of the color chart.

One side benefit of the calibration is a good starting point for the other sliders (brightness, contrast, etc.), which one would probably then "tweak" for the individual image.

This is probably not much different than what you are trying to do "in camera" for the 20D. If you had a 30D or 5D, you would have the add'l "levers" to consider of 'picture styles'. Me...I rather do it "out of camera".

John

Hi, John,

John_Schwaller said:

This is probably not much different than what you are trying to do "in camera" for the 20D. If you had a 30D or 5D, you would have the add'l "levers" to consider of 'picture styles'. Me...I rather do it "out of camera".

Click to expand...

No, that's quite a different matter altogether.

My issue relates to the confluence of two things:

1. How Canon rates the "ISO speed" of each of its sensitivity settings.

2. What is the calibration of the automatic exposure system.

The two conspire to give a simple "earmark": if we take a metered exposure of a uniform neutral test target, what will be the average RGB output over the image?

For example, if the answer to (1) above were, "As defined by ISO 12232", and the answer to (2) above is, "as prescribed by ISO 2721", and if the image processing doesn't really cause the JPEG output to diverge from the sRGB definition, then the answer can be expected to be RGB=100/100/100. This is the so-called "12.8% of saturation" value.

At least for older EOS dSLRs, Canon tells us to expect RGB=116/116/116. This is the so-called "17.3% of saturation" value. Of course, they don't describe it that way, but that's what it comes to. (They describe it as 55% black when the image is converted to the gray gamma 2.2 color space.)

If we for some reason (folklore, perhaps) believe that the result "should be" the "18% of saturation" level, then the indication would be RGB=118/118/118.

Best regards,

Doug