Fujifilm GFX-100 actual *user* experience, tips, setup, please?

[32BT]

+4 stops

[32BT]

+6 stops

[32BT]

+8 stops

[Jim Kasson]

Spectra and histogram-equalized images of each raw plane from frame-averaged dark field GFX 100 exposures:

https://blog.kasson.com/gfx-100/fuji-gfx-100-dark-field-pattern-noise/

Jim

[Paul2660]

So far in the raw files I have been able to access and download at base ISO of 100 the effect of the banding seems very minimal at best.  Using Capture One 12.1.  2.5 stops of shadow push and you just start to see it. Simple enough to remove with many of the excellent noise removal software on the market.
Plenty of details and sharpness.


I doubt Fuji can do much with a firmware fix to remove it but they do try to improve on issues.

Paul C

[Jim Kasson]

So far in the raw files I have been able to access and download at base ISO of 100 the effect of the banding seems very minimal at best.  Using Capture One 12.1.  2.5 stops of shadow push and you just start to see it. Simple enough to remove with many of the excellent noise removal software on the market.
Plenty of details and sharpness.


I doubt Fuji can do much with a firmware fix to remove it but they do try to improve on issues.

Paul C

You can argue that you don't need all the DR of the GFX 50S and 50R, but this banding means that the effective DR of the GFX 100 is less than those cameras.

I think the banding is the result of a firmware modification of the raw data, as in the Nikon Z7. You don't see this kind of banding in the Sony a7RIII. You do see PDAF striping occasionally in the presence of lens flare, but that's another issue.

Jim

[Paul2660]

Hi Jim

Not trying to argue, I appreciate your findings as always.  I just don't it with a 2.5 push of shadows in any of the shots I have looked at raw, and used C1.  Can't speak to LR/ACR.  The Z7 banding for me is also not visable in any of the shots I have pushed 2.5 to even 3 stops.  The only issue I have seen with the Z7 (sorry off topic I know) is a slight red tint to highlights, but even that is rare.  I do see very clean shadow push, possibly cleaner than the Phase One 3100 at base of 50 pushing that much 2.5 stops. 

The raw files I have used are the ones from dpreview, from their various galleries.  I have looked at most of the images that have deep shadows. 

As always good to be aware of this, and hopefully Fuji can address it with a firmware update in the future.

Paul C

[Rand47]

I hope my question doesn’t sound snarky or anti-technology.  I appreciate and stand in awe of the knowledge and expertise demonstrated by this discussion of the PDAF banding “issue” (and the other analyses I see on Jim’s web site).  My question is, what is the actual impact on an optimally exposed image file?  I’m “just a photographer” who is also interested in making fine prints with the current state of the art printers and papers, combined with a competent color managed workflow.

I own a GFX 100 and so far have only been very surprised with how good it is.  How much “better” it is than my GFX 50s, in a lot of respects, based on what I see in prints on paper.  And, perhaps most of all how game changing it is in terms of versatility of use - especially in being able to shoot more “off hand” hand held - bringing stunning image quality (in prints) to a format that one doesn’t typically consider as portable and facilitating a more extemporaneous shooting style / experience.

Jim, I can sense that you’re in an “analyzing the machine” phase of your experience with the GFX 100, but do you have a more “lay person’s view” of it as an artist’s tool yet?  Say, as compared to the GFX 50s or one of the FF cameras you own and use?  I love the work you’ve done and posted on DP’s medium format page, by the way.

Rand

[SrMi]

You can argue that you don't need all the DR of the GFX 50S and 50R, but this banding means that the effective DR of the GFX 100 is less than those cameras.

I think the banding is the result of a firmware modification of the raw data, as in the Nikon Z7. You don't see this kind of banding in the Sony a7RIII. You do see PDAF striping occasionally in the presence of lens flare, but that's another issue.

Jim

Chamber Lloyds reports banding issues in certain shades of blue (e.g., sky), both with Sony A9 and GFX 100. Have you been able to confirm that issue?

[SrMi]

I hope my question doesn’t sound snarky or anti-technology.  I appreciate and stand in awe of the knowledge and expertise demonstrated by this discussion of the PDAF banding “issue” (and the other analyses I see on Jim’s web site).  My question is, what is the actual impact on an optimally exposed image file?  I’m “just a photographer” who is also interested in making fine prints with the current state of the art printers and papers, combined with a competent color managed workflow.

I own a GFX 100 and so far have only been very surprised with how good it is.  How much “better” it is than my GFX 50s, in a lot of respects, based on what I see in prints on paper.  And, perhaps most of all how game changing it is in terms of versatility of use - especially in being able to shoot more “off hand” hand held - bringing stunning image quality (in prints) to a format that one doesn’t typically consider as portable and facilitating a more extemporaneous shooting style / experience.

Jim, I can sense that you’re in an “analyzing the machine” phase of your experience with the GFX 100, but do you have a more “lay person’s view” of it as an artist’s tool yet?  Say, as compared to the GFX 50s or one of the FF cameras you own and use?  I love the work you’ve done and posted on DP’s medium format page, by the way.

Rand

After the Z 7 banding discussions on DPR, it must seem like a Déjà vu to Jim ;-).

I think Jim summarizes it best with the last sentence in one of his posts:
"How will this affect real-world photography? It's too soon to tell, but now we know what to look for."

I use Z 7 often, a camera that has been shown to have banding in some instances. I have yet to encounter any issues in my images, but I keep an eye out for banding in certain situations.

[Jim Kasson]

Hi Jim

Not trying to argue, I appreciate your findings as always.  I just don't it with a 2.5 push of shadows in any of the shots I have looked at raw, and used C1.  Can't speak to LR/ACR.  The Z7 banding for me is also not visable in any of the shots I have pushed 2.5 to even 3 stops.  The only issue I have seen with the Z7 (sorry off topic I know) is a slight red tint to highlights, but even that is rare.  I do see very clean shadow push, possibly cleaner than the Phase One 3100 at base of 50 pushing that much 2.5 stops. 

The raw files I have used are the ones from dpreview, from their various galleries.  I have looked at most of the images that have deep shadows. 

As always good to be aware of this, and hopefully Fuji can address it with a firmware update in the future.

Paul C

Then you don't need a lot of DR. This particular foible of the GFX 100 is a non-issue for you.

Jim

[Paul2660]

Jim, How do you define DR?

For me pushing shadows 2.5 to 3 stops from a single ISO or pulling down highlights all from a single file is use of DR. And the ability to do this means I want a lot of DR.  I assume DR means Dynamic Range?

How much more DR are you looking for from a single file as for me 2.5 to 3 stops is needing a lot DR.  That always has been my goal since the D800e.

I may have a different definition or understanding or need.

Paul C

[Jim Kasson]

I hope my question doesn’t sound snarky or anti-technology.  I appreciate and stand in awe of the knowledge and expertise demonstrated by this discussion of the PDAF banding “issue” (and the other analyses I see on Jim’s web site).  My question is, what is the actual impact on an optimally exposed image file?  I’m “just a photographer” who is also interested in making fine prints with the current state of the art printers and papers, combined with a competent color managed workflow.

I own a GFX 100 and so far have only been very surprised with how good it is.  How much “better” it is than my GFX 50s, in a lot of respects, based on what I see in prints on paper.  And, perhaps most of all how game changing it is in terms of versatility of use - especially in being able to shoot more “off hand” hand held - bringing stunning image quality (in prints) to a format that one doesn’t typically consider as portable and facilitating a more extemporaneous shooting style / experience.

Jim, I can sense that you’re in an “analyzing the machine” phase of your experience with the GFX 100, but do you have a more “lay person’s view” of it as an artist’s tool yet?  Say, as compared to the GFX 50s or one of the FF cameras you own and use?  I love the work you’ve done and posted on DP’s medium format page, by the way.

Rand

Thanks for the kind words. My intention is to use the GFX 100 mostly in the studio, and mostly for closeup work: say from 1:10 to 2:1 (those ratios imply a wider FOV on a 33x44 mm sensor than a FF one). I see no reason why it should not be excellent for that use.  It will replace the GFX 50S in that application, and I will probably eventually sell the GFX 50S, since the UI on the GFX 100 is sufficiently different that going back and forth is difficult for me.

WRT the technical tests I do, my intention when testing any new camera is to detect, quantify, explain any significant departures from canonical behavior, and to provide work-arounds for any issues that I find that have such workarounds. Case in point: I have already supplied RAW GFX 100 files to a Raw Therapee developer, and have hopes that there will be a fix soon in RT. No camera is perfect. I want to provide sufficient information to prospective users so that they can make informed purchase decisions, and to current users so they can get the most out of their cameras.

Jim

[Jim Kasson]

Chamber Lloyds reports banding issues in certain shades of blue (e.g., sky), both with Sony A9 and GFX 100. Have you been able to confirm that issue?

I have seen light-field anolmolies in the GFX 100 that are likely the same thing:

https://blog.kasson.com/gfx-100/fuji-gfx-100-light-field-pattern-noise/

I also observed similar effects with the a9, but give that camera's likely use spectrum, I think they are probably unimportant.

Jim

[Jim Kasson]

Jim, How do you define DR?

For me pushing shadows 2.5 to 3 stops from a single ISO or pulling down highlights all from a single file is use of DR. And the ability to do this means I want a lot of DR.  I assume DR means Dynamic Range?

How much more DR are you looking for from a single file as for me 2.5 to 3 stops is needing a lot DR.  That always has been my goal since the D800e.

I may have a different definition or understanding or need.

Unnormalized dynamic range is full scale over the mean value that provides the lowest acceptable image quality.

For many people, and for most uses, at base ISO, any current FF or larger camera has plenty of DR.

In the studio with artificial lighting, even a Hassy H2D-39 has plenty of DR.

Jim

[Jim Kasson]

After the Z 7 banding discussions on DPR, it must seem like a Déjà vu to Jim ;-)

Yep. Having done all the work with the Z7 meant that most of the programs to analyze this issue were all written. All I have to do is remember how to use them -- gotta work on commenting more.

Jim

[ErikKaffehr]

Jim, How do you define DR?

For me pushing shadows 2.5 to 3 stops from a single ISO or pulling down highlights all from a single file is use of DR. And the ability to do this means I want a lot of DR.  I assume DR means Dynamic Range?

How much more DR are you looking for from a single file as for me 2.5 to 3 stops is needing a lot DR.  That always has been my goal since the D800e.

I may have a different definition or understanding or need.

Paul C

Hi Paul,

I would say that the definition of DR is a ratio of signal strength at saturation over some acceptable noise ratio. Engineering DR uses SNR=1.

What I think Jim says that the banding is caused by the camera firmware compensating for the reduced sensitivity of the DR pixels. On that, I think that Jim is right. My understanding is that Jim wants his raw images raw and feels that almost any post processing is better done in the computer than in camera firmware.

The reason you probably don't see that banding is that your darkest parts of the image are kept near black, where they belong.

Any way, darks in images will always be noisy. Good sensors can keep the noise coming from the sensor low, but they can do nothing about shot noise.

My take on the issue is that Fuji should allow the user to select the level of in camera signal processing. Jim's earlier research on Nikon indicates that the cure is worse than the disease and he finds that Fujifilm has done something similar.

If it doesn't affect your shooting, just be happy!

I wouldn't say that you don't need DR, BTW, just that it may be that you employ DR in a way that the issue Jim is discussing is not visible.

Best regards
Erik

[eronald]

Yep. Having done all the work with the Z7 meant that most of the programs to analyze this issue were all written. All I have to do is remember how to use them -- gotta work on commenting more.

Jim

Maybe you need some new programs: given the effects of on-sensor focus grids, we need a new definition of photographic DR in the presence of such background pattern grids.

It would seem that the GFX100 is effectively something like a 12 bit camera, not bad but still images eg @6400 ISO are going to be visibly impacted eg. in the skies, and photographers workflow needs changing. Sshooting at dusk etc means strongly unbalanced channels and DR becomes an issue if no filterbis used.

Edmund

[Jim Kasson]

Maybe you need some new programs: given the effects of on-sensor focus grids, we need a new definition of photographic DR in the presence of such background pattern grids.

It would seem that the GFX100 is effectively something like a 12 bit camera, not bad but still images eg @6400 ISO are going to be visibly impacted eg. in the skies, and photographers workflow needs changing. Sshooting at dusk etc means strongly unbalanced channels and DR becomes an issue if no filterbis used.

Agree about another definition of DR to add to the ones we already have. Based on the theory that low frequency noise is more objectionable than HF noise, I once tried to derive a metric based upon the frequency content, but couldn't even get close to turning it into a scalar. Based on the theory that periodic noise is more objectionable than wide-spectrum noise, I have long been publishing spectra of image pattern noise. Unfortunately, that's not a scalar either.

I have hopes for a RT fix soon.

Jim

[ErikKaffehr]

Hi,

This article by Jim Kasson is worthwhile reading...

Best regards
Erik