Daniel Browning
Well-known member
Should similarly built lenses, a FF35 f2.8 lens covering a 6K sensor & a S35 f1.9 lens covering a 4K sensor, have the same MTF across their image circles?
The short version is: yes.
Smaller sensors need MTF at higher spatial frequencies in order to match the image of larger sensors. Basically, in order for two formats to have the same image, the "crop factor" needs to apply to MTF. If it doesn't, then the image quality wont be the same.
The crop factor between RED ONE 4K recording area and FF35 is 1.6X (36/22mm). So in order for a S35 f/1.8 lens to have the same resolution and contrast as a FF35 f/2.8 lens, the MTF must be the same at 1.6X higher frequencies. That is, if the f/2.8 FF35 lens has MTF80 at 19 lp/mm and 14mm image height, then a S35 lens must have MTF80 at 30 lp/mm and 9mm image height.
And not just one MTF value, but the entire range of MTF values at all frequencies relative to the viewer's CoC (or Nyquist, if both sensors have the same resolution) and relative to image height.
The question is: does MTF actually tend to scale in this way in real life? The answer to that is above my level of expertise, but what I know leads me to believe that it does scale, but is not directly proportionate; that large sensors are favored.
I think the reason larger sensors tend to have superior MTF is at least threefold:
Faster f-numbers are more difficult.
My understanding is that S35 f/1.8 is more difficult than FF35 f/2.8, even though they have about the same size and open aperture. Similarly, I understand that f/4 on Medium Format is easier than f/2.8 on FF35.
Miniaturization is difficult.
The manufacturing tolerances are tighter for smaller formats. A 1mm centration error is a bigger problem on 2/3" than medium format, so the tolerances must scale with the size, and I would think that it makes smaller sizes more difficult. However, there are certainly counter points to this: f/2.8 lenses for digicams and mobile phones prove that highly miniaturized multi-element plastic lenses can be constructed within amazingly small tolerances, at low cost, and amazingly high MTF (over 400 lp/mm). I don't know how well that translates to the high quality lenses we want for 2/3", S35, and other sizes, if at all, but it's interesting none the less.
Optical viewfinders do not allow back focus distances to vary
Optical viewfinders that were necessary for film forced lens manufacturers to have a certain distance between the exit pupil (real of the lens) and the film itself. This distance is based on whatever the manufacturer chooses, not format size. So a a FF35 lens design cannot always be scaled down for S35 if it would mean the back focus shrinks too much to fit onto the mount without hitting the mirror. So then the designed must be changed for retrofocus with more rear elements, perhaps reducing the quality.
Rangefinder lenses, for example, can have rear elements just 4mm from the surface of the sensor, because there is no mirror to get in the way. That's the opposite of telecentricity, and while it can save cost, improve speed, and reduce aberrations, it increases falloff greatly: at least because of cos^4 falloff, and perhaps also due to sensor angle of response.
So if my understanding is correct, these factors combine to favor large formats such as FF35. Not that it matters for cine purposes, as there is not many cine lenses for FF35.
The golden route
Good points all around, Hans, especially about how DOF affects sharpness.
The golden route would be a S35mm DMSC camera that resolves 5K+ with the DR of the upcoming FF35 cameras.
I think FF35 will always be superior (when used with more light), but at some point S35 will be good enough that the difference will not be important in many circumstances.