Fuji GFX100S adapted lens testing - tilt/shift [Canon TS-E90] and lenses from the 70's.
Vložit
- čas přidán 10. 06. 2023
- Fuji GFX100S adapted lens testing - tilt/shift [Canon TS-E90] and lenses from the 70's. How do some older lenses perform with the 100MP medium format camera?
Many do rather well, not up the the very latest standards but at a fraction of the price and interesting to use.
My Tilt/Shift book
www.crowood.com/products/phot...
Available in all good book stores...
For my detailed TS-E90 review see
www.northlight-images.co.uk/t...
The new 90mm
www.northlight-images.co.uk/t...
Assorted old lenses tested
www.northlight-images.co.uk/u...
-----------------
If you'd like to make a small donation, I have a Kofi page:
"Buy me a coffee" ko-fi.com/keithcooper
-----------------
My articles and videos are always free to access.
Any help with running this channel is gratefully received.
-----------------
I also have some affiliate links which earn me a commission if used.
US Amazon photo/print gear: amzn.to/3l9vJC6
B&H Photo: www.bhphotovideo.com/?BI=2008...
Adorama: www.adorama.com/?... - Krátké a kreslené filmy
A good bit of glass is worth its weight in gold and will outlive many generations of camera bodies. Great video, love the channel and what you bring to us.
Thanks - glad it's of interest!
Do you have lens rental companies in UK? Be nice to see Nikon 19mm tilt shift on Fuji 100s without having to purchase it. We have 2 here that ship to your door. Great way to test the latest lenses. Brilliant videos. Purchased your book last week.
I'm sure we do, but I don't have an F->GFX adapter.
Also, I usually get test [loan] lenses direct from the makers [and sometimes dealers] - which has a great advantage in being free.
Sorry, but I don't really have much budget for buying/renting stuff just to test.
Thanks for getting the book - if you've any questions, feel free to email me at Northlight Images!
Kieth, FYI: All of my Zeiss Contax primes cover the GFX sensor with no noticeable vignette.
Thanks - good to know
I recalled that Olympus made a TS lens back in the day so I just looked it up: Olympus OM Zuiko - 24mm Tilt Shift f3.5. The price is bonkers.
I use the vintage lenses on my M4/3 which they suit well because you get the image circle from the centre of the lens to cover the sensor so they tend to be sharp across the frame. I use the Helios and have a Jupiter 11 135mm and a few others. They are great to use.
This was fun to watch. I am enjoying your journey with the new camera and I continue to learn.
Thanks - I did try to get a zuiko one for when I was writing the T/S book, but had to settle for some pics of it ;-)
This was very interesting. Can't wait to see more. By the way, I found out a Finnish online bookshop has your book on tilt and shift lenses available. Bought it, and now waiting for it to arrive.
Thanks - was going to go out and use the 8-15 today, but it's a bit too warm for my liking ;-)
Hope you enjoy the book!
Interesting, Keith. Your position or assumption that between different MP cameras you mention the detail resolution would not necessarily differ a lot, but that the larger file size of the sensor with more photosites gives more flexibility - totally agree to that.
If I need that kind of flexibility, or even large prints at high DPI, I use Topaz Gigapixel AI to upsample [1] a shot to 32,000 on the ling side or even the members of a panorama shot.
This app gives better detail recovery, or prevention of detail loss, than any app I have ever used to do this upsampling (Ps and LrC can go to 200% but do a much less good job).
As Adobe Camera Raw (ACR) now has been amended with better demosaicking [2] one would hope that detail would be better and the weird digital artefacts we get from sharpening would disappear because raw processing did better - but no, detail is not better in ACR (=LrC and Ps where ACR does the raw processing).
One thing that plays a role at the camera side in this reasoning is the presence or absence of fuzzy filter [3], as I call it. But not in a narrow sense, but rather in the sense that the algorithms in your raw processing software may not have been adapted to do a proper job on images shot without the fuzzy filter.
[1] I distinguish upsample from upscale. In the case of upscale there is a simple extrapolation of detail along mathematical lines. In the case of upsample, more AI is applied to extrapolate detail.
[2] As a raw file has no RGB pixels but monochrome data elements that are either red (R) only, or Green (G) only , or blue (B) only, the data element from each photosite needs to have the missing two other dolours being reproduced - through a process called deBayerisation - so the R data elements get G and B added and mutatis mutandis the G and B data elements all become RGB too. This process generates digital artefacts like Moiré that we easily recognise but also noise that we call luminance noise or colour noise and we blame the camera or its sensor for it.
The removal of digital artefacts like Moiré require relatively simple AI and as the general concept of digital artefacts is called mosaicking, the removal of the artefacts is called demosaicking. The problem with apps like ACR (does raw processing for both Lightroom Classic and Photoshop) is that they do the demosaicking on the basis of what their programmers deem good enough. The problem is that this causes lots of detail to be mutilated and leaves a lot of noise. But, the raw processing is fast. Well, with its CPU specialised in image processing your camera can do it much faster and hence it probably does a better job.
So, Adobe had an artificial limit - excuse - in trying to be fast, for doing a lousy job on demosaicking. One step they made to improve this is the release of the new processing option in ACR called AI Denoise.
It's faster and a lot better at demosaicking but the detail loss is still far below par. Topaz and DxO apps do much better.
[3] the fuzzy filter is generally called an anti-aliasing filter (AA filter) or low-pass filter. This was conceived to make raw processing easier at times when the required processing for raw processing was big, heavy, expensive, and needed loads of energy.
It helps raw processing in a couple ways: [a] reduces uncertainty in the guessing of missing colours, [b] helps make jagged edges fluent, [c] might help reduce hard banding a bit. While it was needed because where the computing hardware was and consequently software engineering still needed to develop more diverse and better algorithms that would be viable if more power could be used, at some point its downside started to exceed the benefits.
So about 10 years ago, Nikon Eliminated the fuzzy filter from the D800E version of their D800.
The fuzzy filter reduces contour sharpness, reduces colour space, reduces low light sensitivity, reduces contrast envelope (the dynamic range usable in a single shot/frame). And it increases vignetting.
These problems increase when MP increase. While C.anon seems to religiously want to stick to the fuzzy filter even at high MP, Nikon did away with it above the photosite density of a 24MP full frame sensor. Consequently it is also absent in a 20MP DX camera like the D500.
How does this relate to "resolution" in terms of detail and sharpness thereof? Well, the pros and cons are clear, but what we see from that depends actually on the raw processing app we use.
In the case of Adobe (ACR) we can now get better (less noise) images on that end, but at the detail loss it still is the same old, same old (really bad).
Agree there.
It's one of the key reasons I've told people to keep their original raw files since about 2004 ;-)
Every improvement lets me get a bit more out of my best Canon 1Ds files from 2004 ;-)
Look forward to the canon 8-15 fisheye on the GFX100s! I own that leans and it is superb! I am fishing around for my next camera - waiting to see the high megapixel canon vs the next GFX. Hoping the next GFX will be more compact with improved "other" features. More image stabilization, better video (which I don't do much of). But since I am waiting loooong for canon high megapixel one, must gove gfx another shot too!
Your reviews are awesome! You should get the first fuji tilt shift one 🙂
Yes, along wait for any high MP Canon
There was a time they were confident it would cross the 100MP mark...
The 90mm TS-E performed remarkably for being such and old design and having a small, 58mm filter size! I It's new news to me of how *focus peaking* on mirrorless can help know where and how much to tilt the lens. This plus your book (which I just ordered from Amazon USA) will make it possible for guys like me to successfully use tilt in the field!
I thought maybe you had bought the Laowa 15mm shift, seeing what looked like it sitting between the two Canon lenses on your table, but then you said it's the 17mm. Maybe the Canon 8-15mm fisheye defished can do some of what only the Laowa 15mm could do - to get wide enough when need be, unless the fisheye is too wide.
When shooting buildings, it seems that maybe shift isn't needed with fisheyes?, because the image is plenty big enough to tilt the camera up or down a bit until the buildings' vertical walls appear straight up and down near the center of the image.
___________
I'm still trying to figure out how both Canon fisheyes (8-15mm and the ancient 15mm) are stated to be 180° diagonally, corner to corner, but the image is even wider on the GFX's larger sensor.
You're saying that the 8-15 vignettes on the MF sensor, but not by a lot.
And I've seen a 15mm Canon image on the same sized MF sensor on a Hasselblad, and it looks like it's going pretty far into the corners too, but the 3:2 fixed petal lens hood blocks what it can really do.
I tested my 15mm Canon fisheye today, just by looking through it on my 5DsR, angled diagonally in our house to see if I can see objects directly to the left and right, and I can.
So it's *likely that Canon is telling the truth* that these lenses are 180° diagonally on a full frame camera, and that when they show even wider coverage on the medium format sensor, they're *probably going beyond 180° diagonally,* maybe 220°, or something like that - showing some of what's actually behind us to our right and left?
Could you maybe take a shot diagonally of a straight sidewalk, or something with the 8-15 at 15mm with the Fuji in 35mm FORMAT mode, and then take another diagonally with the full width of the sensor set to 3:2 LARGE IMAGE SIZE, and then compare how much further back you can see behind the sidewalk, curb, or whatever straight line you're shooting?
If the 100S is set to 3:2, that should be an apples-to-apples comparison, and would keep the same 45° diagonal angle. [I understand that RAW images aren't saved as 3:2 LARGE when shot in 3:2 LARGE, but in full 4:3, so it would have to be cropped to 3:2 in post.]
___________
I'm also interested in seeing what the image circle ends up being on the 100s' MF sensor.
I mentioned that the image circle of the Samyang 12mm f/2.8 shot on the GFX starts starts appearing on the left and right edges right where the 65:24 image would be. So 16:9 and 3:2 can't happen on the full GFX sensor.
The way you're talking, if the vignette is small enough, maybe 3:2 large can be achieved without vignetting?
Even if 3:2 LARGE does vignette, Fisheye Hemi software should be able to result in a full sized, uncroped image, I'm guessing, but would like to see if this is true.
__________
100S' 100MP ADVANTAGE WITH FISHEYE?
I've heard that edges can be soft when full frame fisheye shots are defished, which is because the straight lines are curved in, so when they're made straight with software there are less pixels per square mm, after they're stretched out.
But with a 100MP sensor, the defished images may be plenty sharp, because there still are plenty of pixels between the squished curved lines before defishing.
The 50MP sensor of the 5Ds may do well too though, since the image is 40% smaller and pixels are 10-15% less densely packed?
__________
I'm especially interested in seeing what the 15mm fisheye image looks like on the full GFX's sensor.
I haven't seen any examples of that online so far. You may be one of the first to do it in a video or post.
CHEERS TO CANON for not putting an unremovable, 3:2, petal shaped hood on it, like they did with the 15mm!
The hard image circle with the Canon FE lenses is the 180 FOV - the focal length just adjusts the diameter.
With the original 15mm the 180 diagonal id nicely shown in a shot of a beach I took with the shore running across the diagonal - see my written stuff for detail I'd never bother including in a video
www.northlight-images.co.uk/category/articles-and-reviews/fisheye/
Note the image circle diagram here
www.northlight-images.co.uk/canon-ef-8-15mm-f4-l-usm-review/
@@KeithCooper Thanks. I looked through your excellent fisheye articles. I see your image circle diagram at the top of your EF 8-15mm f4 L review. That's how it's supposed to work: the image circle almost touching the four corners of 35mm sensor at 15mm. And that's how it works for what I've seen for the Samyang 12mm f/2.8 fisheye. But with the Canon 15mm and 8-15mm it looks like the image circle is much bigger than that.
For example, in the fourth set of images from the top in your 8-15 review, "Sample Images" of the car lined street with brick buildings on both sides, the 8mm image circle extends significantly beyond the four corners of the rectangular, 3:2 full frame image at 15mm.
The top-left corner shows the chimney about two windows further back in the 8,10 and 12mm circular images compared to where it is in the 15mm image.
If the image circle is 180°, then maybe what's seen with the lens at 15mm on full frame diagonally is less than 180°, because a fair amount is cut off from what we see in the full circle, 8mm image.
So I suspect that if that same location is shot in exactly the same spot 13 years later with the GFX, both of those white windows to the left of the blue ones may also be in the GFX's *full sensor, 3:2, rectangular image;* though the tree may be too big to see those windows now.
In comparison, the Samyang's image circle doesn't seem to have any of this extra space, according to the Hasselblad MF image I saw. In 35mm FF, the 3:2 edges do seem to almost touch the image circle, so it probably won't present a wider view when shot on the GFX's full sensor for 3:2.
65:24 and 1:1 can use the full sensor though, so that's a big advantage, and 4:3 gets bigger too, because it can touch the image circle; doesn't have to be cropped from full frame 3:2.
@@KeithCooper Could it be this simple to test?: Is the biggest image (widest field of view) for the Canon 8-15mm fisheye is somewhere *between 14 and 15mm* in full frame? When in full frame mode or on a full frame camera, *dialing the 8-15mm back* from 15mm until the widest field of view is seen, until just before it vignettes, just before 3:2 touches the image circle is capturing everything the image circle can deliver?
It just dawned on me, having never shot a zoom fisheye, thinking about your 8,10, 12 and 15mm comparisons in your lens review.
Canon may have designed the 8-15 to go beyond the widest field of view at 15mm. Backing off towards 14mm should get to a place where the vignette is seen. That's the point with the widest field of view.
Does the 8-15mm reach the widest point between 14 and 15mm, and beyond that it' zooming in?
I wonder what that number is? Closer to 14mm than 15mm?
What do you think? This would be easy to check, even in the house.
The corners of the full frame are just outside the image circle edge, so on the GFX 44 x 33 there is a small bit of vignetting at 15mm
The circle just gets smaller as you reduce focal length until it fits inside the GFX frame to give a fully circular image.
The 8-15 works exactly the same on the gfx as it does on my 5Ds - just a bigger sensor
@@KeithCooper Is this the greatest GFX advantage?? A theory.
First, I'd love to what the 8-15mm looks like on the GFX sensor. How much vignette there is compared to what I've seen for the original Canon 15mm and the Samyang 12mm f/2.8. I can't find any photos of the 8-15mm on the GFX sensor online.
I did sign up at the getdpi forum, so I could open the thumbnail of what the Canon 15mm fisheye looks like on the Hasselblad X1Dii, which has the same size sensor as the GFX. The image circle can be seen enough even with the petal shaped lens hood blocking some of it, so it looks like it's a tight fit similar to the Samyang. So I doubt if much field of view will be gained at 3:2 on the bigger sensor, UNLESS DEFISHED, my new theory.
Your lens may have a wider field of view, but I'd love to see a photo to see how much it differs.
But IS THIS THE, or at least a MAIN GFX ADVANTAGE for fisheye, besides being able to shoot in 1:1, 33mm x 33mm, etc.?
Since Fisheye Hemi straightens the vertical lines, images defished that way can cover the entire medium format sensor in full 4:3?
The angled vertical lines at the top and bottom will be straightened by Fisheye Hemi, so the vignetting will be totally gone when defished, and the image will be huge, and will show a much wider field of view area?
And because we have 100MP, the top and bottom edges will be sharp enough even when the vertical lines are spread apart?
If the sensor were only 50MP, like in the 50s ii, then the upper and lower edges would be noticeably softer?
I've never used Fisheye Hemi, but plan to after I get the GFX, so I'm guessing based upon your reviews and others of the software.
Hi Keith, Perhaps an idea for one of your reviews or articles... What can you tells us about the Pixel Shift mode (if that's the right term) used to create 400MP images with the Fuji GFX100S?
Cheers,
Mike.
Thanks - it's on the list ;-) I did [briefly] look at it in my original review overview, but that was when I was first testing a [loan] GFX100S last year
@@KeithCooperMy apologies - I didn't search back that far. Cheers
Thanks - this is one reason I'm working on some proper index pages on the Northlight site for all my videos. From my own POV, CZcams really sucks in its ways of presenting an extended collection of related videos. I can see how/why they do it, but it's why YT simply does not figure in any search for info for myself beyond fixing plumbing etc ;-)
Yes, I am still slightly surprised that people watch my stuff... :-)
Thank you, that is helpful. It's indicate that I can get a 2nd hand 19mm PC instead of a Zeiss Milvus 15mmF2.8 for use on my D850, I just order a D850, and for in the future use of a Fuji GFX 100S/GFX50Sii. I checked on reviews on using old Mamiya 645 lenses and they were soft in the corners and some chromatic aberration on edges and prone to flare. I have Mamiya 7ii lenses that I like to use and also some HB lenses. I still need a 50mmF4 to my HB 503CW though. I still shoot film but more seldom each year...
The M645 55 and 80 both work well, if I needed 55mm and 80mm ;-)
@@KeithCooper That’s good to know. All my MF lenses will be short or long tele lenses on the GFX, the closes is 43mm and that will be normal on GFX. A HB 50mm will be as 60mm and that is acceptable. I have to get a new Zoom lens for the wide angle. So maybe I only need one new lens to the GFX system and two adapters,. I’m waiting on your review on the wide angle zoom 32-64mm? that you will get soon ..
The 32-64 is due 'some time' - it's a loan so always dependent on stock...
Great info Keith - I still have a stack of EF canon lenses - can I ask what converter you use? Any idea how the 180mm macro might perform on the GFX?
It'a a fotodiox one.
I've not used the 180 Macro at all so I don't know - there are lists 'out there' of lenses checked by people, but of course, you've no idea how competent the people doing the testing were ;-)
@@KeithCooper thanks Keith - shame we don’t live closer, could lend you some of my GF lenses - waiting patiently for the the new TSE lenses - have you tried the new Fuji app ?
Thanks ;-)
I tried it when I first set up the camera - yes, the GPS works, but needs the phone on and with me - neither of which I can guarantee ;-)
The remote control app looks more useful, but I'm just not sure how... ;-)
I am curious if the Fuji and Fotodiox adapter work together to auto set the aperture or SS, or if you have to shoot manually. I realize 70s lenses have manually adjusted apertures, but was curious if you are using a modern lens (no manual aperture ring) that can work with the camera and Fotodiox adapter in auto.
The EF lenses [purely electronic] work normally, for aperture/EXIF - every lens I've tried works OK from a functionality POV [ignoring image circle etc]
Shutter/Aperture priority seems to work fine - P mode? I don't know, I've not actually used it or the equivalent on any camera I've ever had. It seems to work, but perhaps I need to read that bit of the manual again ;-)
For the old manual lenses, they are just that. Aperture priority works - the camera just doesn't know/record the aperture
some lenses may not work simply because rget are not sharp enough for that amount of pixels
I've long questioned the idea of lenses 'not being sharp enough' - it all depends what you want to do with the images.
Some are less useful than others for sure, especially if the image circle isn't big enough, but the lens sharpness police get short shrift if they turn up here ;-)