Idaho Airships, Inc.

Do the Canon 1DsMII and MIII exceed the Canon lenses available to them?

The production of an image actually begins far beyond the front element of a lens and ends long after the image has been demosaiced from the RAW format. It involves mechanical issues such as autofocus accuracy, camera movement, and the physical quality of lens elements. It involves the performance of light itself as it enters an optical element from the air medium, as it responds to lens coatings, as it disperses from the lens’ node, and as it enters the photosite. It involves the RAW conversion’s treatment of resolution extinction, color temperature and any in-camera (God Forbid!) processing. It involves manipulations such as rotations and Transforms, Sharpening, Global and Local Contrast adjustments, and gamut conversions in Photoshop or your RIP. It involves ambient air quality/atmospherics, the direction of the sun, and even ambient temperature!

Image production involves esoteric issues such as Modulation Transfer Function (”MTF”), Refraction, Diffraction, nanometers and microns, Dawes Limiting, Nyquist Frequencies, Rayleigh Scattering, Seidel Aberrations, Spurious Resolution, the Circle of Confusion, Transverse and Lateral Chromatic Aberrations and Bayer Array Demosaicing, et. al. Some of these matters are less manifest in our photographic application (e.g. Dawes) than in other optical systems.

It is an extremely complicated and convoluted process. It far exceeds my ability to comprehend it all, encapsulate it, or describe it effectively. For some reason it wasn’t covered in my Contract, Statistics, Antitrust or Industrial Psychology curricula. Go figure.

Unfortunately, though the efforts are valuable to us on a fundamental level, it appears that our conventional investigation of the matter of whether or not sensors are beginning to exceed lenses has been distilled down to individual considerations such as MTF or Nyquist.

On the other hand…and this is a very positive preadolescent development, to my way of thinking…the conventional discussion is at least turning more toward these matters and we’re developing a widespread sensitivity to issues like MTF and Nyquist that simply didn’t exist 10 years ago.

Any conclusive opinion on the matter of sensor resolution vs. lens performance, however, must include all relevant and substantial performance by all involved elements, including those external to the camera body, because different cameras respond to those distal influences differentially (e.g. smaller sensor sites will be more sensitive to motion blur).

I’ve reached some conclusions based upon rudimentary understanding of the matters listed above and my own investigations, and although those conclusions are valuable to me in my own little universe, they may be of little or no value to anyone else who doesn’t model their business after mine. If I were to simply list my findings they would most certainly be devalued to fodder for deprecatory reviews of how little I really know…which, frankly…is true. About all I do know is how little I really know. The point is that the opinion of some clown in Boise is really of little consequence to the rest of the world.

We need a real expert to look into the matter and come to a definitive conclusion-a conclusion that incorporates all of the optical, processing, and contextual issues that effect image production…if that is even possible.

If you wish to study optics and processing on your own, I do respectfully encourage you to do so. The commercial rewards, should you be able to bring a rarefied expertise to bear on your market, are substantial. Expertise in your profession is also a potent last line of defense in any argument about whether or not you are a complete idiot. Be prepared to spend 2-4 years “in the books,” though-what you are entering is very, very deep and (at least in my experience) very difficult to grasp in toto.

If you look in my BlogRoll in the left margin of this page, you’ll find a couple of links to sites that I consider excellent resources for most photographers who are looking to improve their understanding of the science beneath their craft. Of particular value on this question are the Cambridge in Colour site, Norman Koren’s site, and this document on the ImaTest site Norman authored. Books are useful as well.

With that said, I’ll present my conclusions, which I pray you’ll consider with the understanding that while I am not an optical physicist, were constructed with a fairly comprehensive consideration of the science tempered with my own empirical observations:

• If you are shooting under excellent or controlled conditions (e.g. a studio, with good lighting), on a tripod, with a select few Canon L primes such as the 50/1.8, 135/2.0L, or 200/1.8L, you can harness the spatial resolution of a 1DsMII or 1DsMIII.
• If you are shooting handheld, even with IS lenses shorter than 200-300mm, using ISO’s that exceed 400 and in existing light, you will probably not exceed the spatial resolution of a Canon 5D or Nikon D3. Depending on your techniques and processing knowledge, you may not exceed the Canon 1DMIII. For rough service and some advanced features, the D3, 1DsMII and 1DMIII are superior to the 5D. The D3 and 1DMIII may be superior to the 5D for low light shooting (e.g. dusk aerials), but that remains to be proven.
• If you are just starting your business and are not an accomplished photographer you should be looking at the Nikon D300 or Canon 5D. By the time you learn to use them or earn the money to buy the lensing armementarium a professional aerial shooter really needs, they’ll (the bodies) be competitively obsolete.
• I can’t speak to the quality of Nikon glass-however-Canon I do know. If you are starting out, I recommend the 24-70L and the 70-200 2.8L, each on their own body. There are very few things quite so cool as shooting double-fisted…and then there’s the more functional argument that if you’re paying $900+ an hour for a helicopter or on a time-sensitive $4,000 shoot and one camera goes into syncope, a spare $2,500 body pays for itself quickly. Don’t laugh. Stuff happens.
• It is my opinion that in the field, in a preponderance of handheld aerial photographic situations, the 5D produces superior images equal in bona fide spatial resolution to the 1DsMII. I have struggled with this conclusion for months, but it keeps distilling down to photosite dimension (the advantage of the slightly larger sites in the 5D). I maintain the conclusion even printing to 60+ inches, which I do regularly. The post processing is absolutely critical on such enlargements.
• The single greatest potential for overturning my second conclusion above (the 5D’s image superiority) lies in the superiority of the 1DsMII’s autofocusing.
• A professional grade 8MP 1.3x SLR produces very close to the spatial equivalent of scanned 645 negative film.
• A prosumer/professional grade 12MP FF SLR produces very close to the spatial equivalent of a scanned 67 negative film.
• It will be necessary to capture ~45MP with ~6+ micron Bayer Array sites to replicate the spatial performance of a tripod-mounted, accurately focused 4×5 film camera and Imacon/Tango scan of the properly exposed chrome. This equivalence presumes antecedent perspective correction (not performed during post processing) when necessary.
• In any of the three comparisons immediately above there are so many variables (exposure, focus, lens quality, processing quality, et. al.) that it is not possible to provide a concise, generally accurate conclusion.
• The “next step” is to capture all three primaries in a single photosite of approximately 8-9 microns. Perhaps someday we’ll have a hemispherical sensor, but that’s getting ahead of myself…

There are two points I’d like to submit as a matter of prudence:

• You cannot simply look at an enlarged presentation of regular photograph on your monitor, notice that the adjacent pixels are different, and conclude that the difference represents pure resolution!
• When I come up for air I’ll run some exercises on ImaTest to determine whether or not I’m out of my tree. The testing battery will be tightly controlled so that anecdotal influences are minimized. Only then will I be secure enough to publish any opinions without all the qualifiers.

L