Fujifilm hint at 44x33 large format "GFX-H100" filmmaker's camera / open gate 4K3K

[Andrew Reid]

With Canon debuting the C500 II today, demand for a full frame in a C-series body is clearly there but Fujifilm is looking further into to the future. Their X-Trans IV sensor technology has already debuted in the GFX 100 and X-T3 but I can reveal a leaked slide-show shows Fujifilm might be taking the technology much further than we believed...

A large format 44x33 camera optimised for open gate 4K3K RGB.

Read the full article

[Jonesy Jones]

Fake rumor. Thread locked....

Kidding.

Philip Bloom just recently posted how much he loves the GFX 100. He mentioned nothing about rolling shutter which seems to be a caveat with many of these large sensor cams. But you mentioned it could be pretty good so time will tell.

As far as hybrids go, the medium/large format are the ones that attract me most. If I were a serious photog that’s what I’d shoot for sure. 

Pricey though.

[frontfocus]

1 hour ago, Andrew Reid said:

Their X-Trans IV sensor technology has already debuted in the GFX 100 and X-T3

no X-Trans IV in the GF100. 
X-Trans is just the name of Fujifilms proprietary color filter array. Your pictures shows it. 
IV is the generation, the X-T3 has the 4th generation of those color filters. 

The GFX100, like the GFX50s and 50r, uses a Bayer color layout. 

[Nikkor]

This plus https://www.metabones.com/article/of/HasselbladVtoFujiGmountSpeedBooster 70mm look. Anyone wants to buy a kidney?

 

[DBounce]

This is interesting, indeed. I'm starting to feel like it's time for a refresh and "Large Format" would definite be an interesting addition. I hope they go the Sigma route to allow for cooling, and ditch the SD cards for CFExpress. Give us a bigger monitor. Three inches simply doesn't cut it. How about 4" or 5"? Make it tilty flip like Panasonic's S1H. And give it 15 stops of usable DR. Lastly, enough with the dual ISO... can we include real HDR? where two exposures are captured at once. The way Sony originally designed the sensor to work?

[Thomas Hill]

39 minutes ago, DBounce said:

This is interesting, indeed. I'm starting to feel like it's time for a refresh and "Large Format" would definite be an interesting addition. I hope they go the Sigma route to allow for cooling, and ditch the SD cards for CFExpress. Give us a bigger monitor. Three inches simply doesn't cut it. How about 4" or 5"? Make it tilty flip like Panasonic's S1H. And give it 15 stops of usable DR. Lastly, enough with the dual ISO... can we include real HDR? where two exposures are captured at once. The way Sony originally designed the sensor to work?

Isn't Sigma promising real HDR in the fp?

[androidlad]

38 minutes ago, DBounce said:

This is interesting, indeed. I'm starting to feel like it's time for a refresh and "Large Format" would definite be an interesting addition. I hope they go the Sigma route to allow for cooling, and ditch the SD cards for CFExpress. Give us a bigger monitor. Three inches simply doesn't cut it. How about 4" or 5"? Make it tilty flip like Panasonic's S1H. And give it 15 stops of usable DR. Lastly, enough with the dual ISO... can we include real HDR? where two exposures are captured at once. The way Sony originally designed the sensor to work?

15 stops DR? With current Sony sensors, 12bit ADC readout maxes out at 12 stops of DR.

With the Dual Gain HDR, it's a Digital Overlap Drive with double rolling shutter and half the framerates. Not very practical for cinematic work. But Fujifilm has already kind of implemented this DOL drive in X-T3/30 and GFX100, with the shutter speed slower than frame rate feature.

[Sid]

Hi Andrew, I'm probably speaking too soon, but you do those wonderfully funny spoof videos about Canon releases and specs, but when they finally release the C500Mk2 and give a date and a price, everyone covers it except for you who chooses only to mention en passing in a speculative article about Fuji.  C'mon, is that entirely fair, give Canon their due...

 ?

[tupp]

3 hours ago, androidlad said:

15 stops DR? With current Sony sensors, 12bit ADC readout maxes out at 12 stops of DR.

Dynamic range and bit depth are two different and independent properties.

 

You can have a 30-stop dynamic range mapped to 8-bit.  Likewise, you can have a 3-stop dynamic range mapped to 32-bit.

[androidlad]

11 minutes ago, tupp said:

Dynamic range and bit depth are two different and independent properties.

 

You can have a 30-stop dynamic range mapped to 8-bit.  Likewise, you can have a 3-stop dynamic range mapped to 32-bit.

Ha, I've explained many times, current commercial CMOS sensors behave linearly, and linear 12bit = 12EV = 12 stops of DR.

Yes you can store unlimited DR in 8bit, that's called logarithmic encoding.

[Video Hummus]

So glad @androidlad posts on this forum.

[tupp]

1 hour ago, androidlad said:

Ha, I've explained many times, current commercial CMOS sensors behave linearly, and linear 12bit = 12EV = 12 stops of DR.

Ha, ha!  Likewise, I've explained many times that dynamic range and bit depth are two different and independent properties.  I have also given practical, existing examples of cameras that offer variable bit depth while maintaining the same dynamic range -- the bit depth varies independently from the dynamic range.

 

In addition, there exist cameras in which one can change the effective dynamic range while maintaining the same bit depth.

 

It is a misguided notion that CMOS sensors (or any other types of digital sensors) have some sort of absolute linear relationship between dynamic range and bit depth.  12-bit ≠12 EV   and 12-bit ≠ 12 stops DR.

 

 

1 hour ago, androidlad said:

Yes you can store unlimited DR in 8bit, that's called logarithmic encoding.

The mapping of bit depth increments is independent from the bit depth and also independent from the DR.  You can map 8-bit logarithmically, linearly, rec-whatever or any other way -- regardless of the DR.

 

 

41 minutes ago, Video Hummus said:

So glad @androidlad posts on this forum.

Me too!

[Mako Sports]

3 hours ago, Sid said:

Hi Andrew, I'm probably speaking too soon, but you do those wonderfully funny spoof videos about Canon releases and specs, but when they finally release the C500Mk2 and give a date and a price, everyone covers it except for you who chooses only to mention en passing in a speculative article about Fuji.  C'mon, is that entirely fair, give Canon their due...

 ?

I thought he usually doesn't cover cameras in that price bracket ($10k+)

[androidlad]

5 minutes ago, tupp said:

Ha, ha!  Likewise, I've explained many times that dynamic range and bit depth are two different and independent properties.  I have also given practical, existing examples of cameras that offer variable bit depth while maintaining the same dynamic range -- the bit depth varies independently from the dynamic range.

In addition, there exist cameras in which one can change the effective dynamic range while maintaining the same bit depth.

It is a misguided notion that CMOS sensors (or any other types of digital sensors) have some sort of absolute linear relationship between dynamic range and bit depth.  12 bit ≠12 EV ≠ 12 stops DR.

The mapping of bit depth increments is independent from the bit depth and also independent from the DR.  You can map 8-bit logarithmically, linearly or any other way -- regardless of the DR

Me too!

You are talking about encoding bitdepth, 8bit H.264, 10bit ProRes etc.

Sensor ADC readout precision dictates the upper limit of the total DR the sensor can output. Do you know what ADC is? It's absolutely linear.

12bit ADC theoretically offers 12 stops maximum DR, and BMPC6K does 11.8 in lab test.

10bit ADC = max 10 stops, as demonstrated by the noisy shadows in ZCAM E2 4K 120p footage.

Alexa uses dual 14bit ADCs, it's one of the reasons why it's capable of 14 stops of DR (then stored logarithmically in 10bit ProRes)

Most stills cameras use 14bit ADC for stills and that's why A7 III achieves 13.9 stops DR in stills mode.

Of course the raw stream can be denoised to extend the DR below neutral grey, but in its raw state, ADC precision dictates the max DR a sensor is capable of.

 

[Jerome Chiu]

8 hours ago, Nikkor said:

This plus https://www.metabones.com/article/of/HasselbladVtoFujiGmountSpeedBooster 70mm look. Anyone wants to buy a kidney?

 

This is exciting.

0.79x0.71 gives us roughly a 0.56x crop , very close to the entire image circle of a 6x6 lens.

The 16x9 crop of this sensor is around 0.92x (on the current GFX 100 - a 16x9 rectangle cropped from open gate would yield a slightly larger crop ), i.e. most FF lenses should be able to cover it. A 645 lens with a speedbooster (0.92x0.71 => roughly 0.65) would cover it, too.

[Eugenia]

My husband's GFX100 arrived yesterday for his studio. In a very short test I did, I found it to be sharper than my BMPCC 4k (its native Fuji lenses are very sharp, they resolve a lot of detail). I honestly don't think I need even sharper or whatever "better" Fuji thinks it can give us (better color/midtone detail sure, but even sharper details heck no). I had to soften the image in post by a lot to make it look anywhere near cinematic. There was a reason why Alexa was sitting at 2.8k resolution for the longest time before Netflix forced them to make a 4k camera. For true traditional cine stuff, 4k is enough. I personally have no plans on updating on 6k/8k any time soon, not even when 8k TVs might be commonplace. I end up softening my 4k clips in post by either Panasonic or BMD, but the GFX100 required a lot more (f-log). If you're after super crisp, commercial style look, that's your camera.

[DBounce]

22 minutes ago, Eugenia said:

My husband's GFX100 arrived yesterday for his studio. In a very short test I did, I found it to be sharper than my BMPCC 4k (its native Fuji lenses are very sharp, they resolve a lot of detail). I honestly don't think I need even sharper or whatever "better" Fuji thinks it can give us (better color/midtone detail sure, but even sharper details heck no). I had to soften the image in post by a lot to make it look anywhere near cinematic. There was a reason why Alexa was sitting at 2.8k resolution for the longest time before Netflix forced them to make a 4k camera. For true traditional cine stuff, 4k is enough. I personally have no plans on updating on 6k/8k any time soon, not even when 8k TVs might be commonplace. I end up softening my 4k clips in post by either Panasonic or BMD, but the GFX100 required a lot more (f-log). If you're after super crisp, commercial style look, that's your camera.

It's rolling shutter that kills the GFX100 for video. If the GFX-H100 has decent rolling shutter then it's the better option for video of the two. 

[Shell64]

44 minutes ago, Eugenia said:

My husband's GFX100 arrived yesterday for his studio. In a very short test I did, I found it to be sharper than my BMPCC 4k (its native Fuji lenses are very sharp, they resolve a lot of detail). I honestly don't think I need even sharper or whatever "better" Fuji thinks it can give us (better color/midtone detail sure, but even sharper details heck no). I had to soften the image in post by a lot to make it look anywhere near cinematic. There was a reason why Alexa was sitting at 2.8k resolution for the longest time before Netflix forced them to make a 4k camera. For true traditional cine stuff, 4k is enough. I personally have no plans on updating on 6k/8k any time soon, not even when 8k TVs might be commonplace. I end up softening my 4k clips in post by either Panasonic or BMD, but the GFX100 required a lot more (f-log). If you're after super crisp, commercial style look, that's your camera.

I’m curios. How do you soften the clips in post?

[Eugenia]

Quote

I’m curios. How do you soften the clips in post?

On Davinci Resolve Studio (not the free version) there are various blurring algorithms/plugins. I use "sharpen and soften" plugin,  adjustment node, then Filmconvert itself blurs, and if it's still too sharp, I use any of these plugins or the default blurring found in one of the tabs on bottom half of the color page. At the end, I add one more adjustment node, and contrast pop (another Studio plugin). Modern digital cameras are too good in sharp details but with an anemic look, while film was traditionally softer but with much more acutance and midtone detail (particularly older film, modern film is closer to digital). Basically, details are still there, but "thicker". There's "texture" rather than outright "sharpness" (difficult to explain). These adjustments make the image look more like a moving painting, rather than video (at least to my eyes, who is after a vintage film look). Example below, using Panasonic's CineLike D and a rather soft, low contrast lens: https://imgur.com/a/FZoTc8j (open image in its own tab to look at full 4k size).

[kye]

1 hour ago, Eugenia said:

My husband's GFX100 arrived yesterday for his studio. In a very short test I did, I found it to be sharper than my BMPCC 4k (its native Fuji lenses are very sharp, they resolve a lot of detail). I honestly don't think I need even sharper or whatever "better" Fuji thinks it can give us (better color/midtone detail sure, but even sharper details heck no). I had to soften the image in post by a lot to make it look anywhere near cinematic. There was a reason why Alexa was sitting at 2.8k resolution for the longest time before Netflix forced them to make a 4k camera. For true traditional cine stuff, 4k is enough. I personally have no plans on updating on 6k/8k any time soon, not even when 8k TVs might be commonplace. I end up softening my 4k clips in post by either Panasonic or BMD, but the GFX100 required a lot more (f-log). If you're after super crisp, commercial style look, that's your camera.

You may be overlooking the difference between a 4K sensor and a 4K output file.  The 4K sensor has ~12 million photo sites but each is only one colour (R, G, or B).  A 4K monitor has ~12 million pixels, but each pixel is made up of 3 lights, one Red, one Green and one Blue.  In this way if you want a full 4K image you must have an 8K ~48MP sensor.  Of course, modern cameras are smart with how they process the sensor data and do a pretty good job from a 6K sensor to output a 4K file, but if you want true 4K then a 4K camera doesn't cut it.

For most of us it doesn't matter that much, and at standard viewing distances (ie, in the SMPTE and THX specifications) the difference between a 4K image and a 1080p image is barely perceptible, so we kind of talk about capture resolution and delivery resolution in the same context, but they're really not the same thing.

Just now, Eugenia said:

On Davinci Resolve Studio (not the free version) there are various blurring algorithms/plugins. I use "sharpen and soften" plugin,  adjustment node, then Filmconvert itself blurs, and if it's still too sharp, I use any of these plugins or the default blurring found in one of the tabs on bottom half of the color page. At the end, I add one more adjustment node, and contrast pop (another Studio plugin). Modern digital cameras are too good in sharp details but with an anemic look, while film was traditionally softer but with much more acutance and midtone detail. Basically, details are still there, but "thicker". These adjustments make the image look more like a moving painting, rather than video (at least to my eyes, who is after a 1970s 16mm look). Example below, using Panasonic's CineLike D and a rather soft, low contrast lens: https://imgur.com/a/FZoTc8j

I'm a big fan of blur and I used a ton of it in the recent $200 camera challenge.  I also think that resolution and sharpness is one of the reasons people are so fond of the older 1080 BM cameras (amongst other factors).