dante stella stories photographs technical guestbook

An open reply to Mark Dobovoy

Dear Mr. Dubovoy:

It is difficult to think of a method of communication more confrontational and unproductive than an "open letter to major camera manufacturers." Was your intent to get an answer from Canon, Nikon or Sony? Or was it to grandstand? Did you even send it to them? By the way, your piece is not even in the form of a letter, except the last sentence. It's more of a screed that invites an equal-opportunity rebuttal. And that rebuttal can be summarized in just two words.

Who cares?

You initially blow hard with the statement that DxO Labs" is the "premier site to fine [information about RAW performance]. Although DxO has an impressive site, you don't give anyone reason to believe that DxO has any better capability to test cameras than do the people who make them. And so you proceed from an assumption that is untested. Any holes in DxO's methodology (this is, as you will note, an organization that assigns single scores to sensors based on weighted combinations of factors) necessarily become holes in your thesis.

But to move on to your theory itself, it adds up to a lot of sound and fury, no practical application, and no class-action consumer lawsuit. In short, you appear to be angry that sensor gain goes up with certain wide-aperture lenses, you've somehow been cheated out of your chosen ISO setting, and it's time for manufacturers to come clean on this perfidious scheme. If I were a camera manufacturer, I would demurr.

First, I am glad that you caught onto the concept of T-stops many decades after filmmakers came up with the metric and several decades after anyone rejected the serious notion of using them for still photography. T-stops are important in 20-element zoom lenses; modern multicoating makes the transmission loss much less - in fact, fairly negligible even for complex lenses. The entire purpose of a T-stop is to match scene brightness with multiple lenses on narrow-latitude motion picture film. But the grand conspiracy loses its steam when you look at the T-stop scatterplot and see that yes, the cheap cameras have higher loss. The higher-end cameras all run at about 1/3 EV off. This is practically insignificant. The metering system of most cameras only reads in 1/2 or 1/3 stops, and 1/3 stop is considered accurate enough for shutters. It is natural that different sensor designs have different efficiency (what you call T-stop loss). The point is that manufacturers have to account for this - and it always happens somewhere, whether on the chip, in firmaware, or in a RAW developer.

Second, manufacturers fudge the f/stops on their lenses all the time - which itself is part of the reason that T-stops exist for motion picture films. If you have any doubt, check out the tests that have appeared in Popular Photography for the past 30 years. All of them have show measured apertures, which are often as much as 10% smaller than the stated size. The opposite, understating the real aperture is very, very rare. So is the "f/1.2" Canon lens in the first graph even an f/1.2 lens at all? How do you know that any ISO boosting isn't primarily designed cover up a mis-marked lens? Your f/1.2 lens might be f/1.4; your f/1.8 lens might actually be f/2.1. Faster lenses are still faster. It's just that virtually nothing is what it says it is.

Third, DxO's own data shows that manufacturers cook the ISO numbers to begin with (at least if DxO's own test results are accurate). Take a look at the various graphs of ISO sensitivity there and how the ISO points for any manufacturer fall above or below the line (Leica being one exception). "ISO" - a standard designed around film - has been a tricky concept for digital, and different manufacturers measure it in different ways.

Fourth, mechanical vignetting occurs to some degree with every lens. Cameras are already compensating for this by brightening the corners. Is this offensive too? The cosine-fourth-theta rule leads to a lot bigger brightness deltas than the scandalous 1/3 to 1/2 stop you are asserting is reason for manufacturers to "come clean."

Fifth, where is the comparison of 24x36mm and 18x24mm sensors? One of your initial graphs shows that bigger sensors have less sensitivity shortfall than small ones. So is the "ISO boost" happening with the big boys? Or is it just something that is used to make smaller sensors expose the same way bigger ones do? Or are small-sensor cameras simply not designed for critical use?

Sixth, what is the particular application justifying the statement "When I select a specific ISO, I have a reason to do so. I find it unacceptable for the camera to change it without my knowledge"? Is it equally unacceptable that the manual setting of shutter speed or aperture rounds to 1/3 stop increments at the finest? Would you would freak out about having the camera shooting at 1/125 instead of 1/100? Why? It's not as if the noise on any camera over a 1/3 to 1/2 stop range jumps radically. And there are few scenes where that would make a difference. I question whether modern matrix metering is even that repeatable.

Seventh, isn't Canon always fudging things? Andre Agassi as a spokesperson. Fake mullet. Enough said.

Somehow, it's not suprising that the majors declined to answer your piece any more than NASA answers letters inquiring about the faked moon landing. Increasing ISO subtly to even out exposures isn't offensive; it may even be desirable in a lot of circumstances to make things consistent - precisely what you would want to do with close control of ISO. I'm sorry that you feel that you have been deprived of critical information. Idea: get some sleep. Relax. Enjoy a sunny day and stop worrying about this stuff.