L-Mount alliance: cameras with different sensor glass thickness

[Jon Warwick]

Advertisement (gone after registration)

8 hours ago, hdmesa said:

Not really, they're about the same in practice, though lens choice makes a difference. I never tested them side-by-side when I owned them, but I also never thought the M11M punched above the GFX 100S, though I did think it was more or less its equal.

That sounds about right .....I've compared the GFX100S, SL3, M11, M11M very closely together (back-to-back images of the same subject, tripod mounted, f5.6 etc).

To my eyes, and assuming very large prints (say, 50" wide) ....

- M11 + 50mm APO Lanthar and SL3 + 50mm SL APO Summicron look identical in the middle of the frame, but the sides of the frame are notably better with the SL APO Summicron lens (ie, the latter records detail pretty much perfectly from far edge to far edge, whereas the APO Lanthar drifts out of focus a bit at the edges at f5.6). I assume any M mount lens, however, excellent (and the APO Lanthars are brilliant), is a compromise vs the more larger telecentric SL APO lens for edge performance.

- GFX100S and M11M are more similar than not, but one can see small signs of aliasing in the finest of detail with the GFX, whereas it doesn't exist with the M11M; the M11M is exceptional in naturally recording fine detail due to lack of color filter array. I also prefer the tonality and acutance (no micro blurring from a color filter array) with the Monochrom sensor. Still, the edges of the frame with most GF lenses are better than what I get off the APO Lanthar, which again drifts out of perfect focus at the sides of the frame compared to either the GF lenses or the "perfect" SL APO Summicron primes;

- personally, i often prefer the rendering of images that I've taken with the M11 and SL3 over the GFX100S, given (to my eyes) they sometimes seem more "alive" (more "natural"?) and with more "bite" in their look, even if the 60mp full frame Leicas are giving up a tiny bit of sensor resolution and can generate slightly more aliasing in the finest of details (the latter can be partly controlled by better DNG processors such as "Enhance" in ACR);

- if the SL3 ever had a high resolution mode, i'd be delighted for those times when i want to obliterate as much aliasing as possible. I see aliasing in landscapes too (ie, distant moraine next to glaciers), so it would be a really useful firmware update for me for those times when nature is completely still and i have the camera on a heavy tripod.

 

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[gbpost]

vor 15 Stunden schrieb Jon Warwick:

Yes, I often see aliasing on the color M11 (and SL3 that I tested), but it’s a bit less common on the GFX100 (not sure why - the GF lenses are sharp, but maybe less sharp than some of the best of the best Leica and Voigtlander APOs?). “Enhance” in ACR seems to help to clean up some of the artefacts though.

In comparison, the M11 Monochrom is magnificent in its purity of capture (ie, no aliasing) when head to head with all those color filter array cameras, and due to the lack of aliasing in particular, it seems to record more fine detail as a result (remarkably, to my eyes, I see more fine detail with the M11M than I get off the 100mp medium format).

It’s aliasing and getting rid of it at capture is why I’d like to see the SL3 with a multi shot mode. Will Leica ever produce pixel shift for the SL3 in a firmware update?!

Though it is a little off topic, that is an interesting question to me: Which camera shows more aliasing, the Leica M11 Monochrom or the Fuji GFX 100? I do not own any of these two cameras but I understand something of sampling theory and so I try to look at it from a theoretical perspective.

tl;dr: It depends, but for most structures the Leica M11 Monochrom will show less alias.

First, we need to understand the concept of spatial frequency. This is the number of sinusoidal light/dark structures per unit length. In image processing we assume that the image sensor is perfectly aligned with the horizontal and vertical direction. Then we have horizontal and vertical spatial frequencies and in combination spatial frequencies in every other direction too. We can imagine spatial frequencies in a two-dimensional spatial frequency plane.

Following the sampling theorem all frequencies below half of the sampling frequency can be obtained without aliasing. For the two-dimensional spatial frequencies this is a little more complex and we need to know both the horizontal and vertical sampling frequencies assuming an orthogonal sampling grid, which we have on both cameras. The frequencies without alias are an area in the two-dimensional frequency plane, then.

Both cameras have the same pixel pitch of about 3.77 micrometer in horizontal and vertical direction, which we call s in the following. So, we have fs = 1/s as the sampling frequency in both vertical and horizontal directions.

The Leica M11 Monochrom’s alias free frequency area will look like a square in the two-dimensional frequency plane:

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

The frequency plane includes both positive and negative frequencies because of the mathematical background. In our comparison, only the positive frequencies are relevant.

For the Leica M11 Monochrom all structures with horizonal and vertical frequencies below ½ fs will be alias free. The maximum alias free frequency in 45° direction is square root of 2 greater than the maximum alias free frequency in horizontal or vertical direction.

With the Bayer Pattern of the GFX100 it is a little bit more complex. For a sensor with a Bayer Pattern the alias free frequency area depends on the color and the frequency direction. For the green channel the alias free area is a square rotated 45° in the frequency plane and for the red and blue channels the alias free area is a smaller square and includes only spatial frequencies below ¼ fs.

Why is this? The complex answer is, that we can calculate it. The simple but not fully correct answer is, that the size of the alias free area is proportional to the number of pixels for each color channel.

The different behavior in the color channels is the reason for the funny ‘false’ colors when you have aliasing in the red and blue channels but not in the green channel. Depending on the direction of the original structure the colors are orange/blue or magenta/green. If you look into the color channels in Photoshop, you will notice that only the blue and the red channel show aliasing and the green channel is alias free in that case.

The GFX100 has a normal Bayer Pattern and its alias free frequency areas will look like this in the two-dimensional frequency plane, depending on the color channel:

For high frequencies the alias will occur only in the red and blue channels, not in the green channel. Only for higher frequencies the green channel will also produce alias. Only for diagonal structures are the maximum alias free frequencies identical in the three channels. Green structures with high horizontal frequencies and a vertical frequency of 0 are as alias free as on the Leica M11 Monochrom. The same holds for green structures with high vertical frequencies and a horizontal frequency of 0.

Before we can compare the areas in the frequency plane, we have to take into consideration that the sensor in the GFX100 is physically larger than the sensor in the M11 Monochrom. A structure that is captured with the full sensor height of the M11 Monochrom is 24 mm long on the sensor. The same structure on the sensor of the GFX100 has a length of 33 mm using its full sensor height. Therefore, all details on the sensor of the GFX100 are 33/24 larger and the spatial frequencies are 24/33 smaller.

When we map this on the chart of the M11 Monochrom, the corresponding alias free areas of the GFX100’s sensor have to be adjusted and increased by the factor of 33/24.

The comparison of the alias free frequency areas of the two cameras will look like this in the two-dimensional frequency plane:

What does this mean?

Only green structures with high horizontal and low vertical frequencies or with high vertical and low horizontal frequencies do not show alias on the GFX100 while they show alias on the Leica M11 Monochrom.

For red or blue structures or diagonal structures of any color the M11 Monochrom is the clear winner as its alias free frequency range is greater than the alias free frequency range of the GFX100.

Your practical observations are in line with this theoretical approach.

I apologize for the long text. I tried to keep it as short as possible.