Leica 006 VS Fuji GFX 50

[ErikKaffehr]

Hi,

I will check out live view in that shop.

Personally, I seldom use peaking on the A7rII, except when I use Scheimpflug. In that case it is very convenient for finding the approximate tilt.

Best regards
Erik

It is certainly a very key feature, and that's why I was so surprised by how poorly implemented it is in the GFX, which has an EVF that on paper is superior to the EVF in the X1D. When you look through the GFX viewfinder with magnified live view and compare it to the X1D or the Sony A7rII, the difference is not subtle. It is dramatic. I remember trying to focus the GFX and struggling to see the optimum focus point. I do recall Jim Kasson stating that he "solved" the issue with live view in the GFX by turning on peaking at the same time. I haven't tried it myself, but that strikes me as a kludge that may not work well in many situations. Personally, I would not like having peaking on all the time in my viewfinder. (I assume you can't set it up so that it only comes on when you go into magnified live view; with the X1D, if you have peaking on, it goes OFF when you go into live view.) I find it compromises my ability/enjoyment of the composition process as I look through a viewfinder. Peaking also does not seem to work well in low light or low contrast conditions, and you would probably have to adjust the sensitivity in certain situations.
When you visit the Hasselblad dealer, take a look at the magnified live view with the X1D. I think it works very well and is quite comparable to the A7RII.

 

[Bernard]

Erik,

I understand your argument perfectly. I don't think it is very relevant to most photographic situations. Regularly repeating patterns that happen to be at or near the Nyquist frequency (for your imaging system) are few and far between in nature.
The example you show of printed text is the most typical, perhaps because it is easy to reproduce. My argument is that you should not be copying text at such a relatively low resolution. You need to allocate more pixels to each letter.
Other potential sources of aliasing are some construction materials and fabrics.

The examples of a mast or waves are different. The mast in your example is too far away to resolve cleanly. It could disappear entirely in an adjacent shot, if the pixels do not line-up right.
The waves are a natural example of a diffraction grating. You can see the same effect without a camera.

I have only rarely seen aliasing in my own photography, including tens of thousands of images taken with the S. Even then, the aliasing that I see at 1:1 on screen is not visible in prints. Monitors and printers dither images differently.
In short, I agree with your basic premise, but it's not something that is relevant to me. The kinds of non-random repeating patterns that create aliasing are rare in nature. I certainly would not give up the great local contrast and imaging quality of the S system for this reason.

By the way, there is an example of an Apollo Hasselblad with a 3.5/100 Planar in the upcoming WestLicht camera auction catalogue. That lens was built for a long time (as were most Hasselblad Zeiss lenses), but it's optical design is from the 1960s. I don't want to imply that it is a bad lens (I do love the Planar look and use it often), I simply pointed-out that the S system benefits from 50 years of opto-mechanical advances.

 

[Paul2660]

It is certainly a very key feature, and that's why I was so surprised by how poorly implemented it is in the GFX, which has an EVF that on paper is superior to the EVF in the X1D. When you look through the GFX viewfinder with magnified live view and compare it to the X1D or the Sony A7rII, the difference is not subtle. It is dramatic. I remember trying to focus the GFX and struggling to see the optimum focus point. I do recall Jim Kasson stating that he "solved" the issue with live view in the GFX by turning on peaking at the same time. I haven't tried it myself, but that strikes me as a kludge that may not work well in many situations. Personally, I would not like having peaking on all the time in my viewfinder. (I assume you can't set it up so that it only comes on when you go into magnified live view; with the X1D, if you have peaking on, it goes OFF when you go into live view.) I find it compromises my ability/enjoyment of the composition process as I look through a viewfinder. Peaking also does not seem to work well in low light or low contrast conditions, and you would probably have to adjust the sensitivity in certain situations.
When you visit the Hasselblad dealer, take a look at the magnified live view with the X1D. I think it works very well and is quite comparable to the A7RII.

The problem with the GFX and magnified Live View is that by default, the image is over zoomed, way way past a 100% view. You can easily shrink the view back down with a two finger pinch, much like you do on on iPad or iPhone. However as of the latest firmware, I don’t know of anyway to get it to only zoom to 100%. Of course if you are using the EVF, you can’t pinch the screen, and you have to manually dial the view back down, which on the GFX is an award process at least to me. Not an issue on the LCD which is what I tend to use.

This effect is just like taking the zoom on a Nikon D810 or D850 all the way up. 100% view is about 3 steps from the top.

The one click view that Fuji has, is I agree with Howard, very hard to use, eve with peaking on unless you are focusing on something very close. But it’s easy enough to just zoom it back down to a manageable level.

This issue is also present on other Fuji Cameras that I have used, like X-T2 and XT1, more of an issue on the 50MP camera however.

Although I agree with Jim K on his assesment of focus peaking, Fuji again just IMO doesn’t get it. Compare to Nikon or Sony Peaking the Peaking offered by Fuji is just different and harder to work with, at least for me and I tend to leave it off and rely on the LCD and or EVF resolution to determine sharp focus. The LCD is so easy to do this on I just tend to use it more.

As Fuji has done this implementation on all their cameras, I don’t see a change coming.

Sure would be nice to see Phase add peaking to the IQ100 camera, that would just be excellent. Really no real they can’t as I believe that Hasselblad has it on their 100MP back, and P1 could add it to the 50MP backs also as Pentax and Fuji both somehow figured it out. IMO peaking is a much better tool than a after the shot focus mask, pretty much worthless to me.

Paul Caldwell

 

[ErikKaffehr]

Hi Bernard,

If I go back to that place and reshoot that picture with the 40/4 another day, it will show the same kind of aliasing. Why I know? Because I have done it I guess four times. Stopping down to f/16 eliminates it. The same mast will show aliasing also on the Canon 16-35/4 with the A7rII, but of a lesser magnitude.

Here is an example of a sailboat rig:


If you print an aliased image at decent size, the print will show the aliases, but it may not be obvious if you don't look close enough. But, I can identify which of my prints were shot by the P45+ on aliasing alone.

If you don't see aliasing artifacts in your images shot with the Leica S, it may depend on one of these factors:

• You are not sensitive to aliasing. So, it's there but you don't see.
• There is some factor that causes that you don't get optimal sharpness on the sensor.It could be use of small apertures, for instance. In theory, you need to stop down to f/16 to get rid of aliasing on the S/S2, but Lloyd Chambers indicated that f/11 seems to reduce aliasing quite a bit.
• Other factors that could affect aliasing is focus and camera vibration.

Aliasing always occurs when there is signal strength at sampling frequency, but it may be less than obvious. Aliasing cannot be removed in post.

This is one of the reasons moviemakers use very expensive lenses, they will have a strong anti alias filter, because aliasing has much worse effects in film making than in stills. It destroys compression and makes havoc of chroma key, as far as I understand.

You can check on this great info from Panavision: https://www.youtube.com/watch?v=DrW-8u2ukbk

Best regards
Erik

Erik,

I understand your argument perfectly. I don't think it is very relevant to most photographic situations. Regularly repeating patterns that happen to be at or near the Nyquist frequency (for your imaging system) are few and far between in nature.
The example you show of printed text is the most typical, perhaps because it is easy to reproduce. My argument is that you should not be copying text at such a relatively low resolution. You need to allocate more pixels to each letter.
Other potential sources of aliasing are some construction materials and fabrics.

The examples of a mast or waves are different. The mast in your example is too far away to resolve cleanly. It could disappear entirely in an adjacent shot, if the pixels do not line-up right.
The waves are a natural example of a diffraction grating. You can see the same effect without a camera.

I have only rarely seen aliasing in my own photography, including tens of thousands of images taken with the S. Even then, the aliasing that I see at 1:1 on screen is not visible in prints. Monitors and printers dither images differently.
In short, I agree with your basic premise, but it's not something that is relevant to me. The kinds of non-random repeating patterns that create aliasing are rare in nature. I certainly would not give up the great local contrast and imaging quality of the S system for this reason.

By the way, there is an example of an Apollo Hasselblad with a 3.5/100 Planar in the upcoming WestLicht camera auction catalogue. That lens was built for a long time (as were most Hasselblad Zeiss lenses), but it's optical design is from the 1960s. I don't want to imply that it is a bad lens (I do love the Planar look and use it often), I simply pointed-out that the S system benefits from 50 years of opto-mechanical advances.

 

[Bernard]

Erik,

Do you really print 300 pixel crops from your 40MP+ files? You see aliasing because significant subject detail is at the limits of what your imaging system can handle. Stopping-down more won't help, it will only make this detail disappear entirely. The only real fix is to allocate more pixels to your image (or to re-frame your shot). This is true regardless of imaging format.

In other words, you need to ask: "Am I documenting this boat's rigging, or am I documenting the bay in which this boat is moored?" If it's the former, then you need to get closer to the boat.

Off topic: Different motion picture cameras use different OLPF strategies. For instance, Blackmagic Design uses none. Red (who build Panavision cameras, as well as their own) offer a choice of interchangeable OLPFs. It's all very interesting, but not so relevant to medium format imaging. Motion picture images are typically shown in the home, at low resolution (compared to a medium format print), on a large screen. The fact they they are "motion" images also makes aliasing more conspicuous: the aliasing tends to move around a little bit, and human eyes are very sensitive to that type of motion. This creates a distraction for viewers, taking away attention from larger image details.

 

[Bernard]

Erik,

Have you tried developing your P45 image with different raw processors? You may find one that renders the sailboat rigging in a more agreeable way.

 

[ErikKaffehr]

Hi Bernard,

Yes, I did. There is no significant difference. Capture One has some kind of automatic moiré supression but it would not kick in on such small detail. These stuff is a part of the raw data, so you can supress it to some extent but it is always there. Supressing it essentially means smearing colour information.

Capture One is tuned for MFD backs and does a decent job at keeping this kind of aliasing at bay, but Capture One only supports Team Phase One backs. You could use RawTherapee with LMSE, but it will still just mask the issue to some extent. Aliasing is theoretically impossible to get rid of, but it can sometimes be masked.

There are four solutions:

• Smaller pixels, that actually works. I don't think an IQ3100MP would show that issue with the Hasselblad lens.
• Stopping down to f/16 or f/22. This would reduce resolution on the lens to match the resolution of the P45+. But, it throws away most advantages of MFD.
• Adding an OLP filter. But, OLP filters are expensive and reduce apparent sharpness.
• Use a Foveon type sensor. That sensor will still alias, but aliasing will be monochrome. The downside is that Foveon type sesnsors are very expensive in manufacture and need low ISO for acceptable colour separation.

TMaking pixels smaller is the best option, why?

• Aside from marketing, it is a zero cost option.
• It increases system MTF while reducing aliasing.
• It will be able to deliver more detail.

But, we seem to stuck at:

• 150 MP on 54x40 mm. Sensors have been announced.
• 100 MP on 44x36 mm. Sensors have been announced.
• 50 MP on 24x36 mm.
• 133 MP on 43 mm diagonal has been implemented by Forza Semiconducter for Japanese broadcaster NHK.
• 24 MP on APS-C seems to be standard.

Jim Kasson has done a lot of "kitchen research" indicating that something like 2.5 microns would be optimal for the Otus line of lenses. That would be around 138 MP, but I got the impression that Jim feels that even 600 MP would yield benefits.

But, sensor design needs also adopt things like power management, readout speed and video capability.

The leading 24x36 mm devices are at 42MP (Sony A7rIII), 45MP (Nikon D850) and 50MP. Leica is stuck at 37.5 MP on 45x30 mm or 24 MP at 24x36 mm. So, Leica may claim to have the best lenses, but they don't have the sensor resolution to match.

I am not so sure that those Leica S-lenses are so great. I have not see any head to head comparison of Leica-S lenses to Fuji GFX, but I have seen quite a lot of evidence that they are not fully corrected for axial chroma. So what I know right now is:

• Fuji has great lenses for the GFX that are designed for 100MP resolution.
• The GFX has an EVF that allows for optimal focusing using magnified live view.
• The GFX has the best DR on any 120 crop sized sensor.
• The GFX lenses have been demonstrated to perform extremely well

The things listed above are kind of facts. No alternate facts but real ones.

So, would I have the need and money, I would go with the GFX. What about the Hasselblad X1D? I don't know. I have seen a lot of good data on the GFX, thanks to users like Jim Kasson. I would think that the X1D is as good as the GFX. But fact is that I have very little solid info on the X1D. My guess is that the X1D is good, but I have seen just a few usable test images from that system.

MFD makers may have a small issue, 24x36 mm with modern sensors is perfectly good enough for almost any things we do. 100+ MP digital offers some real advantages, if we print really large. But most of MFD is 37.5 to 50 MP, right now, and that is to little for to much.

Best regards
Erik

Erik,

Have you tried developing your P45 image with different raw processors? You may find one that renders the sailboat rigging in a more agreeable way.

 

[Bernard]

Erik,

Respectfully, there are other solutions. The simplest is to retouch the offending line in your image. A few dozen pixels are troubling you, out of 40 million in your image. No sense throwing out the baby with the bath water. It's no different from what photographers have done with spotting ink and a #00 sable-hair brush for centuries. Some photographers are philosophically adverse to doing this, but it is restoring contrast lost to Bayer array processing.

I understand that you have a bias against medium-format systems in the 37-50 MP range, based on your previous experience (and your preferred style of photography). I respect your opinion, but you should not extend that bias to the point where you believe that such systems are without merit for all photographers. Not everyone is interested in the tiniest detail that can only be resolved at close range, on a giant print. The benefits that you get from higher modulation transfer (micro contrast) outweigh this for some photographers. It lends subjects a depth and richness that can not be achieved by fiddling with sliders in post-processing.