4k Lenses?

[Luke Lenoir]

Just read that canon is releasing a new line of 4k lenses to "compliment" the 4k cameras that are going to be coming out. Why do specific lenses have to be made for a 4k digital image. Can a standard 35mm film lens not resolve the amount of detail in a 4k image? I don't get it. Can anyone explain?

[Adrian Sierkowski]

It's about the ability of the lens to resolve 4K, and to be optimized to project onto the lensletts inside of a bayer chip. It gets kinda complicated, but sufficient to say, it's about getting as much sharpness across as possible (with a bit of marketing mixed in). You can of course use any lens on the camera which fits and often you will be begging for a lens with a bit less sharpness.

[K Borowski]

Calling a lens a "4K lens" is kind of silly if you understand the way an imaging *system* works.

 

 

Each part of the system (filters, each lens element, film/sensor - assuming everything is in focus taking into account DOF, COC) resolves a certain number of line pairs per millimeter (this is usually measured at contrasts of 1000:1 and I think 1.6:1).

 

A lense's resolving power isn't like a hole in the bottom of a bucket where 4K either "fits" through the hole or not.

 

 

 

To determine the resolving power of the system, you multiply the inverse of each part of the system in LP/mm and then the product of this equation is the resolution, in LP/mm, of the system.

 

Saying a lens is 4K is treating a consumer as being ignorant and thinking that the sharpest part of the system is going to be the system resolution, like adding the numbers together or taking the largest number of the system and applying it to the whole system.

 

 

It's like saying the F/11 is the next smallest stop from F/5.6, not F/8.

 

 

 

 

 

Another thing to be considered with system resolving power is the aperture of the lens, which is affected by two factors that both cut down on resolving power either being bounced around in the lens when shot wide open or fringe off the aperture blades when stopped down all the way.

 

If a lens is not used at 2-1/2 to 3 stops (depending on lens design) down from "wide open" its resolution in LP/mm is further reduced.

 

 

 

 

 

Anyway, digital is generally a lot pickier than 35mm film when it comes to light being perfectly perpendicular to the chip when it hits it. That doesn't mean film-optimized glass is "worse" merely that the sensor behaves differently.

 

I'd say calling a lens a "4K" lens is a blatant marketing move, taking advantage of customer ignorance as well as hype surrounding the number "4,096." Optimizing a lens for the peculiarities of a CMOS or CCD sensor is one thing. Optimizing for a "4K" chip is bogus. The SYSTEM is a product of each element; either film or digital can benefit from a sharper lens, regardless of resolution all the way down to a VHS-C camera.

[Keith Walters]

Just read that canon is releasing a new line of 4k lenses to "compliment" the 4k cameras that are going to be coming out. Why do specific lenses have to be made for a 4k digital image. Can a standard 35mm film lens not resolve the amount of detail in a 4k image? I don't get it. Can anyone explain?

There's an interview with Canon's Laurence J Thorpe which covers this subject.

The interview is fairly long, but quite interesting.

If you want to cut to the chase, go Ctrl/F (for "Find") and search for the word "Aha" (without the quotes) which marks the start of the bit about 4K lenses.

Basically Thorpe says a "4K" lens simply means one made with tighter tolerances than would be allowable for a lesser resolution camera. (Chromatic abberration for example). They were talking specifically about the lenses Canon would be supplying to suit this camera; there was no suggestion it wouldn't work equally well with standard cine lenses. In fact Thorpe points out that that is one of the advantages of using a PL mount.

Also, don't forget lenses designed for cine use are often of more rugged construction than equivalent units designed for stills work, so they would be distinct form Canon's usual lines.

[Dom Jaeger]

Just read that canon is releasing a new line of 4k lenses to "compliment" the 4k cameras that are going to be coming out. Why do specific lenses have to be made for a 4k digital image. Can a standard 35mm film lens not resolve the amount of detail in a 4k image? I don't get it. Can anyone explain?

 

There are various ways of testing lens resolution, the most informative for me as a lens tech is projecting a graticule (like a slide of a test chart) back through the lens onto a wall or screen.

 

The standard chart uses 200 line pairs per millimeter as the maximum resolution needed in the centre, with 50 line pair patterns at the edges. Virtually every cine lens I test, even ones 50 or 60 years old (if in good condition) will resolve this much at 2 stops down, modern lenses even wide open. For a S35mm cine frame of roughly 25mm width, this represents a minimum of about (50 x 25) 1250 line pairs of resolution to the edge. I'm certainly not an expert on things digital, but my understanding is that it takes at least 3 pixels to even begin to resolve a line pair (one black line next to one white line), which means virtually any decent 35mm cine lens should resolve at the very least the equivalent of 3.75K. Modern high end lenses could probably resolve 12K or more.

 

But resolution is a complex thing - contrast, chromatic abberation (colour fringing), field curvature, flaring, etc all affect the perceived sharpness of an image, and will vary lens to lens. As Karl mentioned, either end of the aperture scale will introduce abberations that also reduce the perceived sharpness. Still, I'd say any good cine lens from the last 30 years or so would easily out-resolve a 4K sensor.

 

With the advent of lower cost digital cameras entering the professional realm many lens manufacturers have sought to provide lenses that match the price/performance of these cameras. Often the trade-off (like the cameras) is less stringent tolerances and a reduction in build quality. The upside to lenses designed for sensors is that they are telecentric, meaning the light from the rear element emerges parallel, and hits the sensor perpendicularly. That was never a concern with film lenses. Most of the time it's not a problem, but occasionally a particular film lens with a certain digital camera combination does weird things.

 

I imagine the Canon 4K lens announcement is mainly advertising hype, to justify the cost of having to tighten tolerances back to where professional cine lenses used to be, and actually make proper lenses for this industry.

[Luke Lenoir]

Yes I had assumed this was primarily a marketing ploy because my 7D shoots 5184 x 3456 resolution stills and I have very sharp prime lenses to compliment the camera. I also have many F-series lenses that resolve every bit of image information available in slower film stocks. I don't see why I would need to purchase a 4k lens if I ever buy a 4k camera in the future.

[K Borowski]

Two things, Dom: The film doesn't "outresolve" a lens, nor does a sensor. You take the number of line pairs per mm inverted (say 1/250 for the center of the chart in your example) then multiply it by one over the resolving power of the film.

 

Then, by putting the two together you get the SYSTEM resolution. It's not the smaller of the two numbers, it's a combination of the two. It's not an arithmetic progression (adding, subtracting), I think this is, what, geometric? Sorry, been a long time since higher-level math.

 

 

As for line pairs, it should be two pixels (assuming monochrome, I don't even want to waste my breath going into all the silly different systems of stacking pixels, different colored pixels, different distributions, etc. etc.

 

 

 

One thing from this type of lens equation math, the film is a HUGE part of the equation, the size of it far outweighs the sharpness of the glass. Then all things being equal the lens becomes important. A larger film format, for instance, will have a far greater impact than the sharpness of the glass (or a 1"x1-1/2" sensor as opposed to a 2/3" chip or APS-C chip on a DSLR).

 

 

One final thing, I know that I saw some article saying a 35mm VV lens (8-perf., it was a still photography test but same difference) tops out at I think it was 22MP, or the equivalent in LP/mm. You'll never go above 22MP with film digital microfilm, anything going through that lens.

 

So a lot of these silly discussions are totally moot unless you're contact-printing gratings onto film or chips without a lens. Same thing if you're shooting wide open or at T 22.

 

Assuming your 7D is using that lens that can resolve 250lp/mm at the center:

 

 

1/72lp/mm (7D dividing by two to get PAIRS) + 1/250lp/mm = 1/56lp/mm Your system is only getting 4032 by 2688 (less when you consider that only the center of this lens can resolve 250 lp/mm So 10.8MP at the center probably translates to only 6 or 8 MP of actual resolution of a test chart going edge to edge, stopped down 2-1/2 stops from wide open.

 

Another thing to consider: These tests are at 1000:1 contrast. In the real world, things aren't black and white lines, they're far lower in contrast and therefore will resolve far fewer line pairs per millimeter, maybe 60% of these numbers I've given you here.

[georg lamshöft]

Then, by putting the two together you get the SYSTEM resolution. It's not the smaller of the two numbers, it's a combination of the two. It's not an arithmetic progression (adding, subtracting), I think this is, what, geometric? Sorry, been a long time since higher-level math.

 

It's simple multiplication but actually, it's not so much about extinct resolution of high-contrast test patterns but resolution-over-contrast (MTF). You have 50% camera contrast (film/sensor) at 4k and 50% lens contrast - you end up with 50% x 50% (or 50% of 50%) = 25%! The higher the frequency of the detail, the lower the contrast. So what is a "4k-lens" then? I have to agree, it's marketing. For S35 it means that it can handle 80 linepairs/mm with reasonable contrast. But what is reasonable? Will it be overshadowed by aberrations? Most "modern" (50 years?) 35mm-lenses do show some detail at 80lp/mm - not at every aperture, not with high-contrast, not with any image height - but they fulfill the very basic criteria for being "4k". It's the same stupid argument as with camera resolution...

[K Borowski]

Since you're using the inverse in lp/mm, it's not simple multiplication, either, at least with working with resolving power of the system at a given contrast. I'm not as familiar with modulation transfer functions. . .

 

 

 

You're right that it's better to use 50% contrast. I think a B&W gradient chart is 90%? The real world averages out to be more of the former. The only place that the high contrast situations show up in real life are edge effects.

 

Is this the same as 1.6:1 as opposed to 1,000:1 contrast? Those are the numbers I'm used to seeing when talking about test charts.