Why is Chromatic Aberration worse at higher f-stops?

[Jim Snow]

This is a question for lens experts. Why is Chromatic Aberration worse at higher f-stops? At first thought, it would seem that with less of the lens area used, the demands on the optical precision of the lens would be less. I realize there is obviously more to it than that and I'm curious what causes this.

[Peter Manojlovic]

I believe the answer has more to do with Focal length...

Wiki has a good explanation..

[Jeff Donald]

Hi Jim,

Here is a link to an article I wrote here some years back.

Camcorder Lens Defects Explained DV Info Net

It looks like the links to all the photos have been broken, but you can read the text. Neither chromatic aberration (longitudinal and transverse) is improved or gets worse by stopping a lens down. Diffraction gets worse as you stop a lens down and that may be what you're thinking of.

The most through discussion on the web about optical defects, causes, etc can be found here.

Chromatic aberrations

Paul van Walree is spot on and his writing is very clear, if somewhat technical at times. His photos and drawings help illustrate the concepts and causes of the major optical defects in lenses.

If you have any questions after reading the links please post back and I'll try to help explain and or clear up your confusion.

[Peter Moretti]

Also keep in mind that how the lens and camera adjust for CA play a role too.

I remember seeing a test comparing CA of Canon's XH-A1 to JVC's HD 110U. The JVC had more CA at longer focal lengths, the while Canon had more CA at the shorter focal lengths.

[Jeff Donald]

The differences between two lenses will have to do with the lens designs and choices of materials and lens coatings.

[Giroud Francois]

that is easy to understand if you suppose that a wide open lens lets the light go straight thru, so lens does not need to bend light too much. (considering that raylight coming from outside the lens are all parallel).
if you close the light (up the the point you just got the iris leaving only a small hole), you create a new light source (the hole) and the light has to expand from that hole/point. Then lens has to bend lights heavily to redirect (focusing) it in the right direction (in that case since the "light source" is considerably closer to the lens , the rays are heavily diverging) . the only rays that do not need to be redirected are the ones passing directly straight thru the hole and the center of the lens, you can imagine the heavy work is done at the border of the lens, usually where the CA occurs.

[Bill Ravens]

The truth is that whenever a lens designer sits down to optimize a lens, he/she generally does a "point design". What this means is that the design process involves a TON of variables. The design process is so much more simplified when a single operating condition (ie focal length and aperture) is considered. So, whatever singular design point is picked, that's where things are optimized. In the field, when practical operating choices are made, it may , and usually does, vary from the singular design point. When this happens, things like optimized suppression of CA go out the proverbial window. Now, one may ask why not design for a range of operating conditions? The answer is that so many compromises are made at multiple design points, that NONE of them work well, at all.

So, a lens designer picks an operating condition that seems very common, and everything else is a downhill slide from there. The more one deviates from the design point, the more aberrations are experienced.

Physicists don't understand that everything is a compromise. An engineer has to make it work, damn the torpedoes, theory to hell.

[Jim Snow]

I imagine there is another fly in the ointment as well - the prism that's splits RGB to the sensors in a 3 CCD or CMOS camera. In addition there has to be an alignment consideration as well. Is that an argument for a single, very large sensor? No doubt a single 1-inch sensor for example would be very expensive but is there any rational to this from a technical point of view? This gets rid of the RGB splitter prism as well as the sensor's alignment with respect to each other. Perhaps there are a number of factors that my question doesn't take into account but I'm curious what the issues are.

[Craig Parkes]

Originally Posted by Jim Snow

I imagine there is another fly in the ointment as well - the prism that's splits RGB to the sensors in a 3 CCD or CMOS camera. In addition there has to be an alignment consideration as well. Is that an argument for a single, very large sensor? No doubt a single 1-inch sensor for example would be very expensive but is there any rational to this from a technical point of view? This gets rid of the RGB splitter prism as well as the sensor's alignment with respect to each other. Perhaps there are a number of factors that my question doesn't take into account but I'm curious what the issues are.

None whatsoever. See RED / Index for confirmation of that.

[Dean Sensui]

As the aperture gets smaller, diffraction starts to become more prevalent and the image starts to get softer.

Here's an explanation with some samples: Understanding Lens Diffraction

Photographers who work with large format cameras have to wrestle with this a lot since large formats also mean very shallow depth of field, and smaller apertures to get greater depth of field sacrifices some sharpness. There are techniques where the front and rear standards can be tilted or swung to shift the focal plane but sometimes it's not practical to do that.

[Peter Moretti]

Originally Posted by Jim Snow

I imagine there is another fly in the ointment as well - the prism that's splits RGB to the sensors in a 3 CCD or CMOS camera. In addition there has to be an alignment consideration as well. Is that an argument for a single, very large sensor? No doubt a single 1-inch sensor for example would be very expensive but is there any rational to this from a technical point of view? This gets rid of the RGB splitter prism as well as the sensor's alignment with respect to each other. Perhaps there are a number of factors that my question doesn't take into account but I'm curious what the issues are.

The "problem" with single sensor designs is that they can only record one primary color per photosite (except for the Foveon sensor Sigma has used).

This means that each photosite can record only either red, green or blue. The Bayer pattern, originated by Kodak, groups photosites into fours (two rows of two): two green sensitive photosites are diagonally postioned from each other and one blue and red sensitive photosite fill the other diagonal.

So while there is information sampled at the each photosite, color information is incomplete at every photosite. So the adjacent photosites need to be used to reconstruct the complete color information. This is called demosaicing or de-Bayering.

A three chip camera with an RGB splitter does not have this issue. It samples complete color at every photosite, so recording accurate color is theoretically easier. Now like you said, sensor and splitter alignment can be problems in practice. Also the camera design is bigger and more power intensive b/c it has a prism and three sensors, which a one chip camera doesn't have.

And that's about all I know on the subject. ;)

[Jeff Donald]

Prisms (beam splitters) are really not an issue in regard to CA.

The concept of lens design is a little more complex and almost any issue can be solved if you through enough money at it. Consider, for example 35mm still lens from Nikon, Canon, Zeiss etc. You can find a 300mm zoom lens from them for under $300 USD (not Zeiss of course) or you can find a very well corrected prime 300mm lens for $3000 to $5000 USD. I've owned several long lenses over the years form Nikon, Canon, Olympus and Pentax. They all showed very little, if any CA. Well, for $4000 you would expect a little better performance than say a $400 lens.

In the mid, late '80's I owned an Olympus 250mm f/2.0 lens. NASA certified it for shuttle flight and at the time, they said it was the best, most well corrected lens they had ever tested.

[Bill Ravens]

What you say is true, Jeff, with one caveat. 35mm zoom lenses have a much smaller zoom ratio than cine lenses. And primes have no zoom feature at all, making the design point considerable easier to choose. As the zoom ratio goes up, the design constraints on no. of elements, coatings and correction element design goes up, astronomically. Sure, anything can be designed to work properly, the question is, can the market bear the development cost passed on to the prosumer. NASA is a good example of a customer whose optics design needs come with no cost limitations.

[Jeff Donald]

Hi Bill, good to chat with you again.

Yes, video lenses do have longer zoom ratios in most cases, so only one factor really changes. When I owned my production company in the '90's we had a Fujinon lens with a large zoom ratio (14 or 16 to 1 as I recall) and it had almost no CA. The downside was I paid over $17,000 USD for the lens. I had several adapters at the time to convert 35mm lenses to 2/3" mount (Century Precision Optics and JVC) and the adapters induced the CA. Hence the need for a $17,000 lens.

[Jim Snow]

Originally Posted by Jeff Donald

Prisms (beam splitters) are really not an issue in regard to CA.

The concept of lens design is a little more complex and almost any issue can be solved if you through enough money at it. Consider, for example 35mm still lens from Nikon, Canon, Zeiss etc. You can find a 300mm zoom lens from them for under $300 USD (not Zeiss of course) or you can find a very well corrected prime 300mm lens for $3000 to $5000 USD. I've owned several long lenses over the years form Nikon, Canon, Olympus and Pentax. They all showed very little, if any CA. Well, for $4000 you would expect a little better performance than say a $400 lens.

In the mid, late '80's I owned an Olympus 250mm f/2.0 lens. NASA certified it for shuttle flight and at the time, they said it was the best, most well corrected lens they had ever tested.

This helps explain why the Sony HVR-S270U is getting mediocre reviews with respect to its lens. I guess Sony set a price point ($8.25K at B&H) that didn't leave enough room for a lens that is appropriate for this class of camera. This is a HD ENG camera that is 1/3 the price of many of its SD predecessors. I guess if you factor in the cost of adding a higher quality lens it levels the field a bit.