Thomas Fitzgerald tested it and has written about it on his blog here. His conclusion:
The changes are very, very subtle to the naked eye.
The update fixes some issues that can occur when there is very fine repetitive detail such as tree branches, leaves, foliage and things like that.
It’s as if they have improved the anti-aliasing algorithms ever so slightly.
It’s something most people probably won’t notice, and they didn’t have to do this, but it pretty much eliminates most rendering artefacts from X-Trans files.
the black line in the center part has been manually added by Bill from Photons to Photos
Fujifilm GFX100
For years, mirrorless cameras lagged behing DSLRs in terms of autofocus speed.
In order to catch up, companies started to incorporate phase detection pixels on their sensors, and modern cameras have phase detection pixels spread all over the sensors, very much to the delight of photographers, who enjoy fast autofoucs, eye autofocus all over the frame and reliable subject tracking.
But no technology is perfect, and so also phase detection has its downside.
When pushed to the extreme (meaning extreme shadow recovery for example), sensors with phase detection pixels can show some banding.
This has been documented with Nikon, Sony and so forth, and of course Fujifilm is no exception. They all use the same Sony sensor at the end of the day :).
It looks like also the Fujifilm GFX100 is (unsurprisingly) showing the same banding issue, when its RAW files are pushed to the extreme.
In fact, the Fujifilm GFX100 sensor has
a total of 3.78 million phase detection pixels
7,776 PDAF pixels every 18 lines
The more phase detection pixels a sensor has, the more you can use phase detection also in lower light.
Bill Claff from Photons to Photos has published a Fujifilm GFX100 sensor heatmap (via dpreview), showing a short black line every 18 rows (see image above).
Should we panic?
I guess not. Or we could just throw any modern mirrorless camera into the garbage that uses phase detection pixels (unless it’s X-Trans ;) ).
It’s, as always, a tradeoff.
Do you want faster autofocus? Or do you want RAW files that even when pushed to its limits and beyond don’t show banding?
The Solution
First off: Fujifilm is fine tuning the firmware for the Fujifilm GFX100, and of course they are aware of banding. They are working to optimize sensor readout and the final firmware will show, how much banding the camera will really have.
But in any case, there is partially a solution to that, even without optimized and final firmware.
As I told you already months ago, Fujifilm is working to bring pixel shift multishot into the Fujifilm GFX100.
The original goal was to have it ready for GFX100 launch, but it needs a bit more time of development.
But pixel shift mulitshot will come, and as we have seen from other phase detection mirrorless cameras offering this feature, pixel shift reduces or even eliminates banding completely.
So, as long as you are shooting static subjects on a tripod and use pixel shift multishot, you won’t have any issues with banding.
The Straits Times posted a short video, with clips from inside the Fujifilm factory in Sendai.
Interestingly, at some point they display a rendering of the Fujifilm GFX100S, basically a Fuji GFX50S marked as 100S. I don’t know if they were confused and just made an incredible error, or if they just wanted to hint something to us :)
Anyway, here are some facts we hear in the video:
227,000 sqm
2,000 workers
18,000 cameras manufactured each month
30,000 to 50,000 lenses manufactured each month
the first room they visit is a room, where the “image sensor is produced“, as well as some of the lenses assembled
the IBIS of the GFX100 consists of 110 pieces. It takes surprisingly little space around the image sensor
the camera is assembled from close ot 900 pieces [to be precise, 860 pieces, as we reported here]
When he says about “image sensor being produced” he probably means “assembled on the GFX“, since it is produced by Sony.
Max Yuryev compares the Sony vs Fujifilm color science.
It’s a blind test, so you have to go through 10 frames, note on paper if you prefer camera A and B. Of course I did it, and it turns out I prefered Fujifilm in 80% of the cases.
The video frames from 1 to 4 are taken with standard picture profile, whereas from 5 to 10 he compares Eterna to a custom made picture profile.
Some general conclusions Max made, before we watch at the frames one by one:
in harsh daylight, with light bouncing, the Sony got some heavy tint, but Fuji’s smart color science got rid of it
Sony is usually more accurate, but you have to spend time to get rid of that tint
Fujifilm got it figured out, and you don’t have to worry about color tint
If you just want to shoot, and not spent time color grading, Fujifilm is great
We remind you that, as reported here, Fujifilm makes a distinction between metric color and memory color. Metric color is more realistic, whereas memory color is interpretation, it’s how your brain sees and remembers colors in a certain scene.
Certain Fujifilm film simulations are not ment to be accurate representation of real colors, but interpretations closer to how you remember those colors.
Other film simulations instead just have to give a unique vibe to the image, think at the vintage touch of the less saturated Classic Chrome for example, great for documentary photography, thanks to the “art of omission” that puts the story/content in the center, instead of colors.
Colors science is a crucial factor at Fujifilm, and we told you already that the engineer, who developed the original Velvia film, has also developed the digital Velvia film simulation.
