Pages: [1]   Go Down

Author Topic: Fuji GFX100 Dynamic Range Testing?  (Read 691 times)

Shrev94412

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 89
  • Phase One, Nikon and Leica Shooter
    • Robert Shreve Photography
Fuji GFX100 Dynamic Range Testing?
« on: June 20, 2019, 11:38:10 am »

Has anyone figured out the Dynamic Range of the Fuji GFX100?
Logged
Check out my work at www.robertshrevephotography.com

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2164
    • The Last Word
Re: Fuji GFX100 Dynamic Range Testing?
« Reply #1 on: June 20, 2019, 12:00:16 pm »

Has anyone figured out the Dynamic Range of the Fuji GFX100?

PDR here:

http://photonstophotos.net/Charts/PDR.htm#FujiFilm%20GFX%20100

What you need to get EDR here:

http://photonstophotos.net/Charts/RN_ADU.htm#FujiFilm%20GFX%20100_16(p)

Analysis here:

https://blog.kasson.com/gfx-100/bill-claffs-gfx-100-analyses/

PDAF striping/banding may affect usable DR. Too soon to know yet. I'll test when I get my GFX 100.

Jim

Doug Peterson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4066
    • http://www.doug-peterson.com
Re: Fuji GFX100 Dynamic Range Testing?
« Reply #2 on: June 20, 2019, 12:27:08 pm »

I'll just throw in my typical disclaimer to take, with a huge grain of salt, the single numerical value that most people try to distill Dynamic Range down into (including all camera makers and most testing websites).

Dynamic range to an engineer means (roughly speaking) "can I readily identify where one wedge ends and another begins on a target with progressively darker patches?" which bears only passing resemblance to photographers. Assuming you actually mean "Does anyone know how camera X handles the highlights and shadows of a high contrast scene?" then some very important attributes of a camera system in a high-contrast scene include:
- How fast, easy, and accurate is it to establish the perfect digital exposure? Is the camera providing you a histogram and blinky highlight warning based on the in-camera JPG (with curve/wb applied) or based on the underlying raw data. The former is basically useless.
- Where there is noise in the shadows, what are the aesthetics of that noise? Is it clumpy? Is it full of color? Or is it evenly and finely distributed in an organic way, free of meaningful color warbling?
- Are there any tools in the camera to reduce shadow noise or improve dynamic range (internal stacking, bracketing with useful metadata, exposure evaluation tools, slow-readout-low-noise sensor control modes)?
- Is there any dark frame technology available? If so, is the data it captures supported and leveraged in the raw processor you use? If so, is it considered by the websites doing the "reduce this down to a single number" testing?
- Is there any structured banding or break up in tonality deep in the shadows? What about in deep shadows of a subject that is colorful (e.g. the shadow of dark green tree foliage)?
- Is the color of a subject accurate from highlight to shadow, or does the hue/color twist around or go muted (less saturated) in the shadows?
- How does the camera handle being on for extended time? Does the sensor get unusually hot in a way that increases shadow noise? Or is there a generous sized heat sink to sap the heat away?
- How does the camera handle long exposures?
- How well does your preferred raw software handle the files when pulling up shadows and pulling down highlights? This is partly to do with the data in the raw file and partly to do with how extensively fine-tuned the raw processor is to that raw data?
- How good is the glass, and especially the coating of the glass, and lens hoods/shades to keep contrast high in the image projected in the scene. Flare and glare can quickly reduce the DR of any modern high quality camera system.

I have no idea what the answers are to the questions above in regards to the Fuji GFX100; we don't sell Fuji and none of the above should be considered a direct comparison between any particular cameras but rather a laundry list of things that can and should be considered for evaluating the DR of any camera. I suspect the answers in most cases are pretty favorable for the GFX as it's using a more modern sensor and Fuji makes nice cameras. But as the list of questions points out, the answer is more than just a number on a spec sheet somewhere, and should be asked and answered in a way relevant to the underlying true question ("will it handle this scene with grace?").

Martin Kristiansen

  • Sr. Member
  • ****
  • Online Online
  • Posts: 1111
    • Martin Kristiansen
Re: Fuji GFX100 Dynamic Range Testing?
« Reply #3 on: June 20, 2019, 12:32:37 pm »

No way Doug. You need a graph. ;)
Logged
Commercial photography is 10% inspiration and 90% moving furniture around.

Doug Peterson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4066
    • http://www.doug-peterson.com
Re: Fuji GFX100 Dynamic Range Testing?
« Reply #4 on: June 20, 2019, 12:37:53 pm »

No way Doug. You need a graph. ;)

Graphs and numbers are definitely helpful. But I find them more useful for comparing (given a specific test maker) between two models of the same camera.

That is, if Jim or DXO or whomever, test two camera generations (using the same test criterium) from Brand X then it's usually the case that the RELATIVE values of those two tests will be informative.

But as soon as you start trying to use an absolute value, or compare between brands, or worse between testers, then you really don't get that much confident information from the number/graph.

If someone is considering a P1 and asks me how many stops of DR it has I'll answer "P1 says 15 stops, but I'd rather you come and compare your current camera and this camera, and see how the highlights and shadows handle when shooting the same tricky scene. I suspect you'll be very pleasantly surprised, but the whole point is to find out if that's the case!" because this is answer the question they are really asking (or at least should be asking).

Jim Kasson

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2164
    • The Last Word
Re: Fuji GFX100 Dynamic Range Testing?
« Reply #5 on: June 20, 2019, 01:33:10 pm »


That is, if Jim or DXO or whomever, test two camera generations (using the same test criterium) from Brand X then it's usually the case that the RELATIVE values of those two tests will be informative.

But as soon as you start trying to use an absolute value, or compare between brands, or worse between testers, then you really don't get that much confident information from the number/graph.


In general, I am in agreement with this. But there's an asterisk.

There are many different DR testing protocols, and they test different things. If you just say "this is the DR", without specifying the kind of DR you're measuring and the protocol you're using, you've said nothing that can be used to someone else's saying "this is the DR" without specifying those things. Also, in general, DR varies with ISO setting, shutter speed, noise reduction settings, and temperature. Also, all noise -- the denominator in DR -- varies in its visual presentation. Photon noise and Gaussian read noise are fairly benign, but there are some kinds of pattern noise that are much more visually objectionable.

The threshold of what's acceptable noise also varies with application, subject, and photographer. That's why I prefer showing the shadow noise at various shadow densities like this:



Take those numbers with a grain of salt. They are modeled, and the GFX 100 line is based on a prototype camera that doesn't look as good as I expected it to look.

Jim

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11249
    • Echophoto
Re: Fuji GFX100 Dynamic Range Testing?
« Reply #6 on: July 07, 2019, 10:58:33 am »

Hi,

In a way, that is the reason that curves like Bill Claffs data are usable. Camera histograms, histograms in LR or Capture One distort the information. Just as an example:

  • Capture One's film curve shows images to bright. That tricks the photographer to expose lower which of course allows more recovery on highlights.
  • LR often applies some highlight recovery, so LR may encourage expose high enough to cause some clipping

Bill's data on the other hand shows the real world data. Another great tool is RawDigger, which presents the actual data on screen. RD is a great tool to understand how the camera exposes, but also to choose the best exposure from bracketed shots.

As far as I understand, the XF has some advanced functions actually showing raw histogram and selectable warning for clipped exposure.

We have two sets of data, engineering DR, which means DR with limit at SNR=1. DxO mark shows that when data are viewed in 'screen mode'. The same data is available on sensor spec sheets and when such data can be found, it is generally in good agreement with DxO mark.

The "print mode" display they have is normalized. That is a reasonable way to compare sensors with different number of pixels.

Bill Claffs numbers are also normalized. Bill thinks they are more accurate than the DxO mark data.

The good thing is that the data is available and that is a good guide to how the systems perform.

Let's show a real world example:

CameraYear introducedDxO-mark pixel modeDxO-mark, normalized
Phase One P45+200711.7512.9
Sony A900200811.512.3
Sony A7rII201512.6813.89

Now, what does those figures look like in a real world test? I made one and done it with some care.


For full size use this link: http://echophoto.dnsalias.net/ekr/Articles/TMP/Darkside.png

It is quite obvious that the A900 and the P45+ are quite close, with some advantage to the P45+, that is a very good match with the DxO results.

The 'Bright Side' of the same high luminance ratio image is here:
http://echophoto.dnsalias.net/ekr/Articles/TMP/BrightSide.png

Full size: http://echophoto.dnsalias.net/ekr/Articles/TMP/BrightSide.png

My take from this test is that the DxO mark data is quite relevant, at least for the three devices tested. Doing the test was not really easy, building a high luminance range image was not that easy, I also needed to adjust highlight exposures accurately. I would say that my accuracy here was about +/- 0.1 EV.

Best regards
Erik

Graphs and numbers are definitely helpful. But I find them more useful for comparing (given a specific test maker) between two models of the same camera.

That is, if Jim or DXO or whomever, test two camera generations (using the same test criterium) from Brand X then it's usually the case that the RELATIVE values of those two tests will be informative.

But as soon as you start trying to use an absolute value, or compare between brands, or worse between testers, then you really don't get that much confident information from the number/graph.

If someone is considering a P1 and asks me how many stops of DR it has I'll answer "P1 says 15 stops, but I'd rather you come and compare your current camera and this camera, and see how the highlights and shadows handle when shooting the same tricky scene. I suspect you'll be very pleasantly surprised, but the whole point is to find out if that's the case!" because this is answer the question they are really asking (or at least should be asking).
« Last Edit: July 07, 2019, 11:09:15 am by ErikKaffehr »
Logged
Erik Kaffehr
 
Pages: [1]   Go Up