Crop Factor, Low Light and Aperture with m4/3 lenses? Part 2

Started Jun 16, 2014 | Discussions thread
Re: Crop Factor, Low Light and Aperture with m4/3 lenses? Part 2

noirdesir wrote:

Anders W wrote:

noirdesir wrote:

The problem is that we cannot fully test any hypothesis because so far with the exception of the Nokia 40+ MP phone cam, larger sensors come with higher pixel counts and we cannot verify whether a m43 lens in front of a 80 MP sensor would deliver the same resolution per picture height as MF lenses in front of a 80 MP back. We can compare a m43 lens in front of a 16 MP sensor with a MF lens in front of a 16 MP sensor (Phase One P20) but while we know that the MF lens has a lot of reserve resolution as visible when used in front of a 80 MP sensor, we cannot test whether the m43 lens would do equally well in front of 80 MP sensor because there is no such sensor.

If two lenses do equally well when tested on sensors with 16 MP, then we can predict with pretty good accuracy that they will also do equally well on a sensors with 80 MP. The following approximation is known to be a pretty good one:

1/i = sqrt(1/l^2 + 1/s^2)

where i is image resolution, l is lens resolution, and s is sensor resolution.

How good that extrapolation is depends on the accuracy of the measurement.

Sure. So what?

And there other, indirect aspects, like the chief ray angle which smaller sensors (with the same resolution as larger sensors) are more picky about.

Based on what evidence?

The closest we can get is to compare 'equivalent' DX with FX lenses adapted to smaller-sensored (and thus higher pixel density) cameras. For example the Nikon 40 mm f/2.8 DX vs the Nikon 60 mm f/2.8 G FX or the Nikon 35 mm f/1.8 DX vs the Nikon 50 mm f/1.8 G FX tested on the 10 MP Nikon 1 V1 (as done by Photozone), the 10 MP correspond to a 32 MP DX sensor. We see that DX lenses have a somewhat higher resolution but nowhere 50% higher than they would need to deliver the same resolution when used on DX and FX with the same number of MP.

On a variety of grounds, that does not strike me as a particularly good comparison. What is known based on the laws of optics is that if we take a lens design for a certain format, say a 50/1.8 for FF, and scale it down in all relevant dimensions so as to make 25/1.8 for MFT, the two lenses will deliver the same resolution if we measure per image diagonal (and the MFT lens twice as much as the FF lens if we measure per mm) at the same f-stop as long as resolution is limited by lens aberrations only.

If you can scale down manufacturing tolerances by a factor 2

Why wouldn't manufacturing tolerances scale?

and if you can scale down the wavelength of light by a factor of 2.

There is no requirement that they'd be scaled down by a factor of 2 for what I said to hold.

Already on 10 MP Nikon V1 lenses aren't able to live up to the full sensor potential from f/5.6 onwards which means you couldn't scale down a f/5.6 FF lens and get the same performance.

Why not?

And all Nikon 1 zooms already have a maximum f-stop at the long end of f/5.6. If we look at the sensor side, while smaller sensors are built with smaller 'manufacturing tolerances' (process geometry), technically, the smaller process geometry could be applied to large sensors as well. And this indirectly affects optical performance because smaller sensors cannot be made with the same tolerance for chief ray angles.

Why not?

I think it is a general rule in any kind of manufacturing process that the improvements in precision and accuracy don't scale fully with size as one goes to smaller sizes.

A general rule based on what evidence?

There can be jumps sometimes which break that rule but overall I would say it holds. And there are physical limits like the wavelength of light.

The wavelengths of visible light simply are what they are. On what grounds would they matter?

So, on paper, in theory, you are right but in practice you aren't and that is what my two examples illustrate.