Pages: [1]   Go Down

Author Topic: Test of how much difference sensor size and resolution makes...  (Read 2481 times)

Dan Wells

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1044

I've posted this on the medium format forum, but wanted to get input from folks who don't end up over there, too.

I'm soon to have an opportunity to look at three different sizes of very similar sensors (X-T4, A7r IV, GFX 100S) all on the same subjects in Yellowstone NP. The A7r IV is mine, and the GFX and (hopefully) X-T4 are on their way in for review. What makes them interesting is that they use the same (current) generation of Sony sensor, at exactly the same pixel pitch (3.76 µm). It's as close to a pure resolution/sensor size test as there is.  I won't have the fourth step in the ladder - I can't come up with the $50,000 Phase One IQ4 150 at the same moment.

The sensor toppings are NOT perfectly aligned - Fujifilm and Sony colors are a bit different, and the X-T4 is, of course, X-Trans. I'm going to minimize as much of the color difference as possible by shooting raw and processing to get as close as possible. Fortunately, none of the three has an anti-aliasing filter, so that variable is out. Matching field of view will, of course, mean different lenses, and, unfortunately, zooms. Even without the field of view issue, I can't think of a lens with high-quality adapters to all three mounts. The closest would be manual focus Nikon F mount, possibly an Otus (which I don't have lying around). Aspect ratio also differs, with the GFX being squarer...

My thought is to print big - I have a Canon Pro-2000, so 24x36" is the maximum easily achievable print, and look for the differences on the print. I may also try some LightJet prints to see if dot gain is messing with the results.

Fujifilm's technical experts are telling me that 100 MP medium format should be about as much of a difference from high-res FF as high-res FF is from APS-C. Mathematically, that seems very plausible - the ratios of sensor area and resolution are similar. While I've never done it in this controlled a manner, I have tried prints on this scale from APS-C and high-res FF, and the difference is very noticeable - the 24 MP APS-C print looks good until you put it next to a similar print from a 45-60 MP full-frame camera...

Anything else I should be aware of...



Logged

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 13983
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Re: Test of how much difference sensor size and resolution makes...
« Reply #1 on: April 30, 2021, 11:50:12 pm »

For the lens I would use a P1 Bluering lens such as the 45mm or 120mm macro. They are pretty much the best there is and can easily resolve that density of pixels even in the corners.

Being Mamiya mount if should be adaptable to all.

They lack an aperture ring, but if you have an XF body lying around I think it's possible to keep the aperture at a certain level if you shut down the camera in the middle of a long exposure. I haven't tried recently but I think it's how they behave.

Cheers,
Bernard
« Last Edit: April 30, 2021, 11:53:25 pm by BernardLanguillier »
Logged

mcbroomf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1534
    • Mike Broomfield
Re: Test of how much difference sensor size and resolution makes...
« Reply #2 on: May 01, 2021, 05:35:02 am »

One important point to be discussed and considered is how the files should be post processed.  You mention it briefly on the other thread.  My 2c;

Each files needs it's own conversion settings.  As they are different formats and have different CFA and type they need custom and optimized capture settings (sharpening, NR, WB and possibly other adjustments).  The "norm" in reviews is to set the software at default but I think this is lazy, or the only possibility given time constraints of a review.  How you get to "optimized" is a difficult question mind you.  Perhaps too subjective but worth considering and discussing.
Logged

Dan Wells

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1044
Re: Test of how much difference sensor size and resolution makes...
« Reply #3 on: May 02, 2021, 03:36:04 am »

I was planning on processing so they all looked similar and as good as possible to my experienced eye (no defaults, because nobody's going to use defaults to make big prints). I do want to use the same software on everything - and DxO, which is my usual go-to, won't handle X-Trans. C1? Adobe?
Logged

David Sutton

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1345
    • David Sutton Photography
Re: Test of how much difference sensor size and resolution makes...
« Reply #4 on: May 02, 2021, 06:25:58 pm »

I suppose there is an argument to be made that each camera will need a different optimum software for its files. I doubt you can please everyone on this. Probably better to keep it simple and use your judgement on what looks good to you.
C1 has better colour profiles for X-Trans than Adobe. Good enough that I've stopped making my own.
I can't comment the latest LR updates to demosaicing X-Trans. I gave up on them a few years ago as unusable for large prints. I went to Iridient's X-Transformer. C1 is at least as good.
For large prints I like a minimal amount of sharpening during the raw conversion and finish in Photoshop with Topaz AI Clear. It does however shift the colours so either I need to use an appropriate blend mode or a hue/saturation adjustment.
Referring to your post on the medium format forum, Fuji does use a different system for ISO. When I compared an X-T2 to a Canon 7D, Fuji's 200 was equivalent to Canon's 100.
Cheers
Logged

Dan Wells

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1044
Re: Test of how much difference sensor size and resolution makes...
« Reply #5 on: May 02, 2021, 11:35:12 pm »

Do you happen to know if the GFX also uses a different system for ISO? If it does, there's an argument for also throwing in the Nikon Z7. It has a different pixel pitch, but it does ISO 64 (probably very close to the GFX's unusual ISO 100), and its sensor is similar in design. If the GFX is going to be operating at an unusually low effective ISO, it's probably worthwhile to have another camera that can do that (and I have a personal Z7 I can include)...
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
My two cents...
« Reply #6 on: May 03, 2021, 12:29:26 am »

I've posted this on the medium format forum, but wanted to get input from folks who don't end up over there, too.

I'm soon to have an opportunity to look at three different sizes of very similar sensors (X-T4, A7r IV, GFX 100S) all on the same subjects in Yellowstone NP. The A7r IV is mine, and the GFX and (hopefully) X-T4 are on their way in for review. What makes them interesting is that they use the same (current) generation of Sony sensor, at exactly the same pixel pitch (3.76 µm). It's as close to a pure resolution/sensor size test as there is.  I won't have the fourth step in the ladder - I can't come up with the $50,000 Phase One IQ4 150 at the same moment.

The sensor toppings are NOT perfectly aligned - Fujifilm and Sony colors are a bit different, and the X-T4 is, of course, X-Trans. I'm going to minimize as much of the color difference as possible by shooting raw and processing to get as close as possible. Fortunately, none of the three has an anti-aliasing filter, so that variable is out. Matching field of view will, of course, mean different lenses, and, unfortunately, zooms. Even without the field of view issue, I can't think of a lens with high-quality adapters to all three mounts. The closest would be manual focus Nikon F mount, possibly an Otus (which I don't have lying around). Aspect ratio also differs, with the GFX being squarer...

My thought is to print big - I have a Canon Pro-2000, so 24x36" is the maximum easily achievable print, and look for the differences on the print. I may also try some LightJet prints to see if dot gain is messing with the results.

Fujifilm's technical experts are telling me that 100 MP medium format should be about as much of a difference from high-res FF as high-res FF is from APS-C. Mathematically, that seems very plausible - the ratios of sensor area and resolution are similar. While I've never done it in this controlled a manner, I have tried prints on this scale from APS-C and high-res FF, and the difference is very noticeable - the 24 MP APS-C print looks good until you put it next to a similar print from a 45-60 MP full-frame camera...

Anything else I should be aware of...

The way that I compare systems is:

  • Use comparable lenses
  • If feasible, include a color checker. That I often forget...
  • Use same software for comparing.
  • In general, I always use my own color profiles, generated by LumaRiver
  • I would adjust exposure to yield an L value (in Lab) around 52 on fourth neutral patch on the ColorChecker
  • Adjust WB on second neutral patch on the ColorChecker
  • I prepare two images, for some given print size say 30"x40" at 180 PPI, without sharpening
  • Sharpen both images using FocusMagic
Regarding the choice of size and PPI, I think looking at a 180 PPI image on a normal 24" 2K monitor at actual pixels is a pretty good representation of a large print.

For actual prints, I make small crops, note the equipment on the back side. I mix the prints until I don't know which is which.

A funny comparison I made was this one, comparing my Sony A7rII with my Phase One P45+:

http://echophoto.dnsalias.net/ekr/Articles/Shoots/Voigtlander65/V65_vs_Hassy/0001_HalfAndHalf.jpg

One half is P45 and one half is the Sony A7rII. Lenses: Planar 100/3.5 at f/8 and Voigtlander Apo Lanthar 65/2 at f/5.6

Best regards
Erik
« Last Edit: May 03, 2021, 12:35:30 am by ErikKaffehr »
Logged
Erik Kaffehr
 

David Sutton

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1345
    • David Sutton Photography
Re: Test of how much difference sensor size and resolution makes...
« Reply #7 on: May 03, 2021, 04:44:33 am »

Do you happen to know if the GFX also uses a different system for ISO? If it does, there's an argument for also throwing in the Nikon Z7. It has a different pixel pitch, but it does ISO 64 (probably very close to the GFX's unusual ISO 100), and its sensor is similar in design. If the GFX is going to be operating at an unusually low effective ISO, it's probably worthwhile to have another camera that can do that (and I have a personal Z7 I can include)...
You've got me there.
Apparently Fujifilm, Panasonic, Olympus and Pentax use Standard Output Sensitivity (SOS) which is supposed to be more accurate.
The rest use Recommended Exposure Index (REI) which is supposed to be more pleasing.
The GFX 100 lists its iso as Sensitivity [Standard Output] 
I think this will end up a distraction if you are comparing image quality at a camera's lowest native iso. You probably have enough to do as it is.
When comparing sensors the main consideration may be DoF differences with the advantage going to the smaller sensor, but again if the subject is far enough away that can be ignored.
Logged

rdonson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3263
Logged
Regards,
Ron

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2005
    • http://www.guillermoluijk.com
Re: Test of how much difference sensor size and resolution makes...
« Reply #9 on: May 04, 2021, 01:15:38 pm »

I have a Canon Pro-2000, so 24x36" is the maximum easily achievable print, and look for the differences on the print.

Hi Dan, you don't need to print huge to know how a huge print would look like. If the goal is just to compare, you can print a crop of the entire image (the most interesting part) to emulate any imaginable print size with your Canon Pro-2000. It's just a matter of camera Mpx vs print DPI ratio.

Regards

Dan Wells

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1044
Re: Test of how much difference sensor size and resolution makes...
« Reply #10 on: May 04, 2021, 04:47:14 pm »

I certainly can print crops (and will), but I want to look at real prints as much as possible to get the feel for "how does this look in a properly composed photograph on a wall"? I'll occasionally use some Canon ultra-glossy RC paper in a test (that should show extra detail at the cost of a lovely surface), but I'll do most of the printing on Canson Platine, because my real prints are generally on Platine...

A crop from a 60x90" print is only relevant to people with access to a 60" printer and the ability to display and market 60x90" prints. It can also be viewed closer, in a way that the real print cannot... There's a minimum viewing distance to take in the whole image, and that's pretty substantial (6-8 feet?) for a really huge print...

The photographer in the YouTube video admits that he rarely prints... Why bother with anything bigger than the (already excellent) X-T4 in that case? The extra dynamic range will be crushed by JPEG compression, the colors will be converted to lowest common denominator sRGB, and the resolution will be reduced to less than a phone by any photo sharing site. Instagram's image width is only 1080 pixels (some others are larger, but I'm not aware of anything that goes over 3000 or so).

There are good reasons to use an X-T4 or the like online (superb lenses, a wide range of focal lengths), but I can't see why to go bigger than that, since whatever you get will simply be crushed by the social media process... Am I missing something?

Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Test of how much difference sensor size and resolution makes...
« Reply #11 on: May 04, 2021, 06:03:50 pm »

There are good reasons to use an X-T4 or the like online (superb lenses, a wide range of focal lengths), but I can't see why to go bigger than that, since whatever you get will simply be crushed by the social media process... Am I missing something?

Hi Dan,

I agree. There's no reason to aim at posting/printing anything beyond (or close to) inspection/viewing output dimensions. An exception mighjt be subject matter that would benefit from accounting for "Vernier acuity".

Cheers,
Bart

Logged
== If you do what you did, you'll get what you got. ==

David Sutton

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1345
    • David Sutton Photography
Re: Test of how much difference sensor size and resolution makes...
« Reply #12 on: May 04, 2021, 07:05:48 pm »

Am I missing something?
I don't think so. A while back I shot the a scene using a 24 mp camera with a top lens and then with the 3mp camera and plastic lens on my phone. Projecting the photos at a camera club, no one could pick out which was which until I pointed out that the abominable camera phone blew the highlights.
Some people need to squeeze the last ounce of dynamic range and resolution from their files for the type of work they do and large format makes perfect sense. But for others its a question of how the camera, lenses, processing and printing combine for their needs, and I suspect in real world use for them resolution is not high on the list of things they need. For myself I use APS-C because I can carry and use it all day without exhausting myself and losing an opportunity. I like it that I don't have to upgrade my entire computer system to deal with the larger files. Then again, I like using much the same software and operating systems as I did 10 years ago because it forces me to be more creative to achieve what I see in my mind's eye. Meaning I prefer to limit my tool set but master what I have, rather than putting time into chasing the latest and greatest.
Last year I exhibited 24 by (up to) 40 inch prints from a 16 mp X-T1. I was focussing on content not detail and in fact needed to slightly blur parts of each image because they were too sharp and proved a distraction. Undoubtedly they would have fallen apart if I'd printed larger but the buyers here don't have or want the wall space to take a larger size image. Horses for courses.
Logged

smthopr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 612
    • Bruce Alan Greene Cinematography
Re: Test of how much difference sensor size and resolution makes...
« Reply #13 on: May 12, 2021, 06:43:05 pm »

When you conduct this test, set the aperture for each camera so that depth of field is similar. Otherwise you might feel the larger imager is “sharper” as the in focus parts are more separated from the background.
Logged
Bruce Alan Greene
www.brucealangreene.com

Dan Wells

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1044
Re: Test of how much difference sensor size and resolution makes...
« Reply #14 on: May 14, 2021, 02:23:28 am »

The only lenses I have that will give comparable fields of view on each camera are a trio of f4 zooms. They aren't ideal - ideal would be three Zeiss Otii of different focal lengths, with the mythical perfect adapters. Not only do I not have 100mm,85mm and 55mm Otii lying around, the right high-quality adapters would be very difficult to come by (Canon or Nikon to Sony, X and GFX all exist, but the only one that I know is available from a major brand is the Sigma adapter from Canon to Sony), and there is a question of whether the 100mm Otus would cover the 33x44mm sensor (some full-frame lenses do, and that one seems a reasonable candidate).

What I ended up with is the Sony 24-105mm f4, Fujinon XF 16-80mm f4 and Fujinon GFX 32-64mm f4. They were all available, and they have a significant overlapping focal length range. Would I prefer to use really good primes? Sure, if I had them at focal lengths that matched field of view - the zooms will make that much easier! There really aren't that many sets that would work - the only two I can think of where all three lenses more or less exist are (23mm, 35mm, 45mm - actually exists) and (55mm, 85mm, 100mm - almost exists, but the medium format lens is a slightly too long 110mm). Would I prefer to use zooms with less range on the smaller sensors (24-70 and 16-55, for example)? Again, that would be ideal, but I was limited to what I had on hand or could borrow. Fujifilm chose the 16-80 (as well as the 32-64 on the GFX), and they really like that lens. The 24-105 is the normal zoom I happen to own for the Sony.

Besides the zooms being good lenses, but not always great ones , the other problem is that adjusting for depth of field is hard. Either the APS-C lens ends up wide open OR the medium format lens is in diffraction territory. It needs 1 stop between APS-C and FF, then another 2/3 stop between FF and medium format f4, f5.6, f7.1 works, but the poor APS-C lens is wide open. Probably the best compromise is f5, f7.1, f9. The medium format lens will be diffracting a little bit, but not badly...

Dan
Logged

John Nollendorfs

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 623
Re: Test of how much difference sensor size and resolution makes...
« Reply #15 on: May 19, 2021, 06:17:13 pm »

Dan:
It will be interesting to hear about your impressions after the test. I predict that in viewing the files on screen at 1:1, you will notice a lot of differences in the files, but when you print out the sample swatches the the desired size, I think you  will see the differences a lot less.   That's what happens when you convert digital to analog. If you can use Qimage to print the files, you will get optimum results because of it's great output sharpening algorithm, and optimum printer uprezing features.

Looking forward to the results!
Logged

Dan Wells

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1044
Re: Test of how much difference sensor size and resolution makes...
« Reply #16 on: May 21, 2021, 09:31:24 am »

Still out on the road... About a week or a bit more, depending on how I decide to go, from returning to my printer. I printed one shot from the GFX at 30x40" (native 300dpi from that camera) on a Canon 8400 at a local FedEx Office, which was the only place I could find a big printer... Really impressive detail - the camera holds that big with absolutely NO sign of faltering, and quite a few signs of that look you get from a large-format print (the extra detail at the edge of vision that draws the viewer in). It's CLEARLY in a different class from APS-C - no head to head needed, my knowledge of large APS-C prints is more than enough.

How clear is it compared to the very best of full-frame (the 45 mp + class)? It's better - but I don't have a similar print with me from full-frame, and it's close enough to need a head to head comparison (unlike APS-C, where it's just night and day). Without a head to head, the maximum detail in the plane of focus seems higher (but I couldn't swear to that - really nice full frame is also good at that scale). The bigger difference might be the smoothness of the transition from in focus to out...

Another difference that might seem obvious, but I wasn't expecting, is just how nice the big Fujifilm's color is. Of course - it's a Fujifilm, and they always have good color. Color science varies significantly by manufacturer (I personally prefer Fujifilm and Nikon, but that's taste), and some, but not all of that comes out in raw processing.

Can't wait to make better prints, on paper I know, without FedEx's RIP messing with the process (I did my best to reduce the impact as much as I could, and did manage to avoid any resizing)...
Logged

EricV

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 270
Re: Test of how much difference sensor size and resolution makes...
« Reply #17 on: May 22, 2021, 01:49:46 am »

This sounds like a system comparison, not a sensor comparison.  A perfect lens at f/2.8 has resolution (Airy disk diameter) the same size as one pixel of your sensors.  If you stop down much more than that, the sensor size still matters, but not the individual pixels.  Your real life results are likely to be dominated by lens performance and sensor size.
Logged

Dan Wells

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1044
Re: Test of how much difference sensor size and resolution makes...
« Reply #18 on: May 22, 2021, 10:13:44 am »

I've seen different Airy disk results (Cambridge in Colour's calculator), which suggest that diffraction sets in around f7.1. That leaves plenty of room for a good lens (including good zooms) to have a number of near-optimal apertures. Yes, it's inevitably a system comparison, at least in large part, but it's a system comparison with the interesting and highly informative feature of very similar sensors...
Logged
Pages: [1]   Go Up