Opticallimits

Full Version: So finally ... the new Leica DG 100-400mm f/4-6.3 OIS
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
Pages: 1 2 3 4 5 6

Guest

They stick with 12 mp camera so that the test are comparable to previous tests. I can understand this line of thinking. Yea 16mp is more pixel but not really that much more.

De entering doesn't decrease overall lens performance in tests it gives typical characteristics very easy to spot by a qualified reviewer.

And as discussed earlier for most pro lenses decentering is easy to readjust.

Something looks fishy here.
Shall we start a third thread for the 12/1.4 since the second one has been - just as I predicted - locked after having devolved into chaos? Smile (actually, on one of the forums I used to hang out, there was an euphemism for locking the threads: *CLICK*, as in a moderator pushing the "Lock" button. Big Grin)

 

But speaking of the lens at hand, if I had been into the long teles AND mirrorless, I might've looked at another 100-400 instead: the Fuji one. The results seem a lot more harmonious there (at least from the Lenstip tests).

Quote: 

 

But speaking of the lens at hand, if I had been into the long teles AND mirrorless, I might've looked at another 100-400 instead: the Fuji one. The results seem a lot more harmonious there (at least from the Lenstip tests).
True, but you do have a reach difference due to the crop factor. 

With the "Leica" you get the same reach at 300mm as with the Fuji at 400mm. Then the comparison shifts a bit more to the positive for the "Leica"?
The Fuji is quiet a bit larger. Also the 24Mpix sensor of the fuji somewhat makes up for the difference in crop factor.


It would be interesting to know how much effective resolution the diffraction limit robs. The new 20Mpix mft sensors surely will suffer more in this regard.
Quote:The Fuji is quiet a bit larger. Also the 24Mpix sensor of the fuji somewhat makes up for the difference in crop factor.


It would be interesting to know how much effective resolution the diffraction limit robs. The new 20Mpix mft sensors surely will suffer more in this regard.
 

Hi Chris,

 

That is a misconception, really. Resolution of sensors do not even come close to diffraction limits, in any way, while good lenses at their optimum apertures do. And as effective or system resolution always is a combination of both sensor and lens resolution, and the total system resolution is always determined by both with the final system resolution by definition always being lower than the lowest of the two. Only until you get to the diffraction limit for a sensor with the diffraction limit of a lens, the sensor will "outresolve" a lens.

 

Considering we are talking 400 lp/mm and more for F/4 and F/5.6 lenses here, and a little less for smaller apertures, we still have a long way to go, if we even ever reach this point, as the small number of photons collected for sensors with that high of a resolution will generate too much noise to resolve anything from a practical POV.

 

I can set up a little table if you like, to show the effects, and include a 16MP and 20MP MFT sensor besides a FF and/or APS-C sensor. Just say so, and I'll gladly create such a table Smile.

 

Kind regards, Wim
So does the lens act as a low pass filter?
That is not quite true, Wim. Sensors very clearly get hit by diffraction. Any test shows that, as does the underlying physics predict that. Luminous Landscape used to have a very good and thorough article about the impact of diffraction and the maximum MP one could get for different formats with different aperture sizes, but it seems to have disappeared for non-members.

 

It (diffraction softening) is easily seen in the PZ tests for resolution, and even more clear in the lens tip tests (because lens tip does not sharpen results.

Example from PZ 16mp MFT:

http://www.opticallimits.com/m43/830-oly75f18?start=1

From f4 on, we see the resolution figures gradually drop. This is due to diffraction alone.

 

Another example, on the 12mp E-PL1, from lenstip:

http://www.lenstip.com/477.4-Lens_review...ution.html

[Image: 162810_roz.png]

We clearly see a drop in measured resolution from f4 onwards. Due to diffraction.

 

A quick test shows this clearly in images too, Macro lens wide open, at f8 and f16 at macro distances on EOS 6D:[ATTACHMENT NOT FOUND]

 

For e-line (a green) wavelength (546.1mm), fro f4 the max resolution is 375 lines/mm. For f5.6, down to 268. For f8, 188 (source: Nikon industrial lenses). That is for a single shortish wavelength. Longer wave lengths (which are part of normal visible light) produce much lower resolutions due to diffraction.

 

http://physics.stackexchange.com/questio...iffraction

 

 

Quote:So does the lens act as a low pass filter?
 

No, but but system resolution can in simple form be calculated with the old formula for this:

 

1/(system resolution) = 1/(lens resultion at specific f-stop) + 1/(sensor resolution)

 

For ease of use we are assuming that we are talking base iso for a sensor, and in that case sensor resolution is in principle equal to the nyquist frequency, plus or minus 5 to 15 % depending on de AA-filter stack being used.although fro ease of use we can ignore this.

 

Generally speaking, in the olden days, we worked all of this back to lp/mm, and for a 16 MP MFT sensor you then end up at approximately 133 lp/mm, for a 20 MP MFT sensor at base iso that is 150 lp/mm. If we assume a 10% reduction due to the AA-fiter, this then becomes approximtely 120 lp/mm and 135 lp/mm.

 

For diffraction based maximum lens resolution, we used to use the so-called Rayleigh criterion, and that amounts to using a wavelength of 512 Angstrom and an MTF of approximately 9%, which for normal shooting purposes is average. Note that MTF-50s are therefore typically a lot lower. Anyway, at F/1 the Rayleigh limit is about 1600 lp/mm, based on the Airy disk size, the diffraction circle size of a point, one gets at these parameters.

 

Because of diffraction, we then roughly get the following diffraction limits for perfect lenses at different f-stops, all based on the Rayleigh criterion:

 

F/1: 1600 lp/mm

F/1.4: 1143 lp/mm

F/2: 800 lp/mm

F/2.8: 571 lp/mm

F/4: 400 lp/mm

F/5.6: 286 lp/mm

F/8: 200 lp/mm

F/11: 145 lp/mm

F/16: 100 lp/mm

etc.

 

Now, there is no such thing as a perfect lens, and off-axis performance of a lens deteriorates as well, and at large apertures a lens is generally not at its best. However, many of the high quality lenses do rather well when stopped down a few stops, and some even get close if not equal diffraction limits at those f-stops, at least in the optical centre.

 

Because lenses tend to be best when stopped down a bit, generally speaking, and because diffraction gets worse with stopping down, what often happens when you measure lens resolution, is that the resolution starts at a specific point, gets higher, and when diffraction becomes teh limiting factor, lens resolution goes down again. You get a parabolic or elliptical type curve, with the top lying often at the f-stops a few stops away from maximum aperture.

 

The so-called diffraction limit of sensors, essentially is the aperture at which point a sensor cannot resolve beyond teh diffraction limit, purely caused by the fact that lights bend. It is a bit more complex than that, because sensor wells are indeed wells, and create shadows depending on the angle of the incoming light, but roughly this is correct. For MFT sensors at 16 MP and 20 MP these sensor diffraction limits are F/12 and F/11 respectively, but to get the most out of this you generally need to stop down  a full aperture less, so F/9 and F/8. it doesn't mean you can't take a picture, or can't get a usable image, but this is about making optimal use of what is available; the resolution of the sensor does not alter regardless.

 

As many testers use MFT-50 for their resolution findings, I have also calculated a set of MTF-50 diffraction limits. Here it is;

 

F/1: 760 lp/mm

F/1.4: 543 lp/mm

F/2: 380 lp/mm

F/2.8: 271 lp/mm

F/4: 190 lp/mm

F/5.6: 136 lp/mm

F/8: 95 lp/mm

F/11: 69 lp/mm

F/16: 48 lp/mm

etc.

 

I have created a few tables in which I also calculated the system resolution for different sensors, and perfect lenses, based on teh two diffraction limits I mentioned here (rayleigh criterion and MTF-50).

 

First, Panasonic Lumix GF2, a 12 MP MFT camera:

[ATTACHMENT NOT FOUND]

 

Olympus O-MD E-M10 Mark II (16 MP, same as O-MD E-M5, etc.):

[ATTACHMENT NOT FOUND]

 

Olympus Pen F, 20 MP (same as Panasonix GX8):

[ATTACHMENT NOT FOUND]

 

In short, as you can see, at F/4, 12 MP and MTF-50 maximum resolution is 72 lp/mm, at F/4, 16 MP and MTF-50 it is 78 lp/mm and at F/4, 20 MP and MTF-50 it is 84 lp/mm. Similarly, at F/5.6 it is 62 lp/mm, 67 lp/mm, and 71 lp/mm respectively. At apertures larger than F/4 lenses tend to have a lot of residual errors and resolve less, and as you can see at F/5.6 diffraction already starts to become signifcant (for any lens and any system).

 

Even so, resolution is well above what we used to have with film, as I mentioned before, in other posts.

 

HTH, kind regards, Wim

Quote:For MFT sensors at 16 MP and 20 MP these sensor diffraction limits are F/12 and F/11 respectively, but to get the most out of this you generally need to stop down  a full aperture less, so F/9 and F/11.
 

This is incorrect.

As we can see from the charts here at PZ and pretty much everywhere else, diffraction hits well before these aperture values.

A good example is the best MFT lens tested here in the zone, the Oly 75mm f1.8: http://www.opticallimits.com/m43/830-oly75f18?start=1

From f5.6 onwards, the resolution is hit by diffraction already.

Keep in mind that the higher the sensor resolution, the sooner it will be hit by diffraction.
Pages: 1 2 3 4 5 6