Hi,
I have clearly stated it was 3.8 meter and Bart probably feels that I used to be totally honest. No lady involved in that picture in any way, unless the feather is coming from a female specimen of that bird.
I can give some background to that image. It begins with Harry Nyquist, a scientist who formulated the sampling theorem called after him. It says that details below the half the sampling rate can be reproduced correctly. If a signal contains significant amplitude above that frequency lower frequency images will result.
A very good demonstration of that is to shoot Bart van der
Wolf's test target. The result will be something like this:
As you see the lines are converging to a certain limit, passing that they bend. The limit between straight and bending lines is the foretold Nyquist limit. If a lens throws high frequency detail on the sensor that sensor will show that high frequency as a lower frequency alias. This alias is usually colourful on Bayer sensor. So ideally, a lens would resolve at the Nyquist limit of the sensor, but pixel level contrast should be zero above the Nyquist limit.
The two images below illustrate this pretty well, I think. The upper one is shot with the Sony Alpha SLT 77. That camera has 3.8 micron pixels so Nyquist limit is 1000/3.8/2 131 lp/mm. The lens used here has around 3% MTF at 131 lp/mm. We still see some aliasing, and that is related to the Bayer filter as sampling frequencies for red and green are half of Nyquist.
The lower image is shot on a Hasselblad with a P45+ back. It has 6.8 micron pixels, which correspond to 73 lp/mm. At this frequency the Planar 80 lens used in this test transfers about 10% of contrast. Intense aliasing results.
Now, these cameras are different. The Sony is used with a Zeiss 16-80/3.5-4.5 lens and has in all probability an aliasing filter. The P45+ was used in this later test with a Planar 80/2.8 lens at f/5.6. The Planar is very sharp in the centre. The P45+ lacks an OLP-filter (AKA Anti-Aliasing filter). That AA filter is needed to supress MTF beyond the Nyquist limit. In general, most DSLRs have been using AA-filters while most MFD-s do not. The Mamiya ZD had an optional AA-filter costing around 3000 $US last time I checked.
On the P45+, aliasing is very obvious at f/11 but entirely gone at f/16. But, using f/16 reduces sharpness significantly, possibly to the same level I get from the 24 MP SLT at f/8.
Anyway, the way to avoid aliasing is to use small pixels or adding an OLP filter. For some reason unknown to me, MFD normally doesn't employ OLP filters.
On the other hand, we also can clearly see that sharp lenses transferring a lot of contrast are massive producers of aliasing.
The feather images were intended to investigate this phenomena, why:
- My suggestion of aliasing as a result of combining excellent lenses with large pixels was contested by a well known photographer of birds.
- I suspected that artificial detail could give the impression of fake resolution
- The strains of the feather gives a clear clue of what is true or fake detail
I used three camera bodies in this test, those were what I had available to me. Of the three cameras the SLT99 is closest to the Leica S, same pixel size and using gapless microlenses.
My conclusion of this test were:
- Small pixels are needed to avoid aliasing artefacts
- OLP filtering and gapless microlenses are not enough to suppress monochrome aliasing
- Small pixels and proper downsizing of the image give better reproduction than larger pixels
I would say that these findings are in line with basic image sampling theory and there is no magic changing it!
So, why don't photographers see this?
- They don't care about real or fake detail, as long it looks good
- The artefacts show up only when the equipment are optimally used. Small apertures, bad focus and camera shake can act as low pass filters eliminating aliasing.
Just to make it clear, I don't say that small sensor cameras give better reproduction than large sensor cameras. What I am saying that small pixels are needed for correct reproduction, especially when used in combination with inherently sharp lenses.
Very clearly, I don't think we have seen real world images from the Leica S. The micro-lenses may be helpful in reducing aliasing, that have been seen on the IQ-250 images published by Doug Peterson of DT. But those micro lenses will not cause a miracle to happen. I am 100% sure that smaller pixels are needed to make those Leica S lenses justice. At least if they are as good as they are said to be.
Best regards
Erik
Bart, was that shot from 3.8m away with a P45+? haha.
Small sections in isolation will never give the complete story, Eriks feather may be a tiny crop from a photograph of a stunningly beautiful woman in an incredible location and the feather or the detail has no baring on the image or it could be a picture of a feather in which case it's the wrong equipment and setup. Even after Eriks reply, I am still non the wiser as to how it relates to the camera in question.
