Circles of Confusion and Resolving Power

[David Regan]

I was just talking to a guy the other day who explained a system he had used for determining what circles of confusion to use for a particular lens set. He explained he was shooting 16mm, and had a particularily old set of lenses of poor quality, they were not very sharp at all. So he tested and determined the resolving power of the lens in line pairs per millimeter (lp/mm)

 

What he then realized was that he could use this to determing the minimum CoC needed for that particular lens. Say the resolving power was 40lp/mm, then 1mm/40 = .025, his circle of confusion for that lens. He based this on figuring, if the lens can't even resolve something of finer sharpness than .025mm, acceptable focus was not as criticle.

 

I thought about it, and to me it seems quite clever. Obviously it would require you to test each lens and could have different results for each lens in your set, but it just seemed like an interesting and smart idea, especially with today's palm programs, where you can dial in your CoC and calculate DoF from there.

 

Does anyone see any problem with this being a theoretical way to determine necessary CoC? (Or perhaps it's standard and I've just been oblivious, which is quite possible.)

[Chris Keth]

There's nothing wrong with it as a theoretical method. It's certainly accurate enough. The problem lies in it being a pain in the ass. I don't want to test every lens in my set and have a separate CoC for each lens. Making things even more complicated, you would want a Coc for each lens and each stop on that lens since most lenses change somewhat as they are stopped down. I don't want to do all that.

 

In reality, I am fine with the standard CoCs for a given format. If you have slightly more depth of field due to a soft lens that's fine. You will rarely if ever have less DoF and that will help keep you from screwing a shot (like a split in a car, for instance) up because of miscalculation.

Edited June 6, 2008 by Chris Keth

[Doug Brantner]

Dave, you should really take Salvaggio's M&P class, you'd love it. And I know Chris will agree.

[Chris Keth]

Dave, you should really take Salvaggio's M&P class, you'd love it. And I know Chris will agree.

 

Absolutely! At least try to take the first quarter of Photo M&P. You can get it counted as an elective. The other quarters are really good as well. I suggest them all but especially the first quarter. It is primarily optics, and the mechanics of how and why we see the way we do. I think there's a little color stuff, too.

[John Waterman]

interesting concept. Doesn't MTF come into play to determine the "correct" circle of confusion? Great to see so many RIT names in here.

[Mitch Gross]

You want to use the medium's CoC, no that of the optics placed upon it.

 

How d you calculate for filters, for changes in resolving power at different focal lengths of a zoom? Different T-stops on a gien lens will have different resolving power. Do you measure line pairs in the center of the frame or the corners?

 

The issue is the format you are shooting on and the presentation format. These are absolutes that will set your CoC, not the parameters that can change with ever shot, sometimes even within a shot.

[Chris Keth]

interesting concept. Doesn't MTF come into play to determine the "correct" circle of confusion? Great to see so many RIT names in here.

 

Not at all. Circle of confusion is derived from a formula containing the smallest angular details the human eye can percieve (1/60th of a degree, I believe), the viewing distance from the final exhibition (which varies in a theater, of course), and the enlargement the image will undergo.

[Simon Wyss]

If I may add something - there's a little thinking error (do smile about my English). A lens's resolution is maximum with the iris full open. In theory the resolving power increases with the opening. To compute circles of confusion from numbers of line pairs per given length in the image plane has no other value than changing units. Let's not forget that the concept of an objective with an iris diaphragm is already a compromise. A lens manufacturer gives us a product with a, say, 1:1.9 ratio. The lens will overall perform best at, say, 1:5.6 (geometric). In a certain sense it is silly to use a speedy lens only to stop it down for nice pictures when smaller diameters would behave more favourably. We could as well put on a 1:4 objective and use it less closed.

 

I have seen pictures of 1:2.8 triplets that outdid 1:1.4 six-element systems only because of the fact that smaller diameter elements can be produced in bigger numbers and then selected. They assemble better single lenses to systems at comparable prices. Unfortunately, this practice is gone. Today's lenses are too complicated, too heavy and too expensive. Just this before I close: There is an 1930 Hugo Meyer Euryplan in my hand, three elements cemented into one block for the front and the rear part of the system. I have never seen finer detail than out of that non-coated old-timer. It opens 1:6.3. I admit, this is for stills, not movies. If you want to capture what your lens projects try once Gigabitfilm. It has double the resolving power of the best lenses, so the image character is solely defined by the lens.

[Chris Keth]

A lens's resolution is maximum with the iris full open. In theory the resolving power increases with the opening.

 

Where did you learn this? I have never used a lens that does not improve somewhat when it is stopped down. Many lenses made for large format stills are made to perform best in the 16-22 range. It is true that there is a limit to this imposed by diffraction (which shouldn't be a problem until f32 or f45), but most lenses improve in the middle range of stops.

[Simon Wyss]

In theory resolving power decreases with stopping down. You are right when you say that a lens will improve in performance in general but who knows all the details of a given system. The physical laws do not dissappear in practice. On the contrary, they integrate all together. A recent computer calculated system can behave very different to an oldtimer.

 

The stop numbers you mention are geometrical ratios. The effective lens diameter left open at, say, 1:16 is not the same in 4" X 5" photography and in Academy aperture cinematography, comparing same focal lengths because you have quite different systems. Only with thin single lenses and three or four elements the values come close to each other. Cine lenses are optimized for use at full aperture and one to two stops closed: Cooke, Zeiss, Canon, Bausch & Lomb, Panavision.

 

What I know exactly is the behavior of the Kern Switar 1:1.9 or 1:1.4, a six element system. It shows best resolution-colours-sharpness-coma balance at 1:4. The Zeiss Tessar is also a rather-open diaphragm system. Closed to 1:22 you have diffraction of the worse kind. Modern Cooke lenses are best at 1:5.6 to 1:8 because of their warmer character. The red light parts are better aligned when under smaller angles which helps improve sharpness. To be able to compare resolving power values we need to ask the manufacturers for the curves. Some MTF curves have been published. In the end I'd like to know about the combination flange distance-film. Colour stocks have more but thinner layers for capturing the light. Silver stocks most often have only one layer but thicker. By closing the iris you link the film (depth of field). That produces less flange distance errors but the resolving power does not increase.