androidlad

Fuji Kaizen update news:

1. X-H1 to receive firmware update to offer dual IS (IBIS + lens OIS), which adds 1.5-2 stops of additional stablisation.

2. X-T3 (and very likely X-H1/X-T2) to receive firmware update to enable file size larger than 4GB when using SD cards formatted as exFAT (64GB+).

Just now, CyclingBen said:

@jonpais we get it, you don’t like Fuji, and love Sony. I love Sony too, their broadcast cameras are awesome, if I still worked in broadcast I wouldn’t want anything else. To be fair, if I had your prior poor experience with Fuji I wouldn’t like them either moving forward. I feel the same way about Panasonic for hybrid cameras (really like their broadcast offerings though).

In response:

First, obviously look at both camera bodies. The specs are what I was talking about.

Second, no one thought Panasonic would go full frame, yet here we are. I agree, Fuji will stay with apsc until smartphones catch up and their medium format might always be a niche product, sorry for being hyperbolic.

Sensor wise, it’s different than any other Fuji sensor. So either:

1. Sony completely changed their design

2. Fuji redid their X-trans color array (although the GFX50 R and S don’t have X-trans and are almost the same as the X-H1, X-T2, etc when it comes to colors and expected performance for noise and ISO)

3. Samsung made the sensor and Fuji mirrored the specs of the NX1

4. BSI makes way more of a difference than we (me at least) thought

5. Canon made the sensor and accidentally sent them to Fuji instead of putting it in the R. (Note this is sarcasm)

I’ll leave these frame exports here, from the X-T3, NX1 and A73.

 

 

Great pictures!

You're right about #1, it's a completely new generation of sensor, 3.76um photosite architecture that's also used in the 100MP sensor in GFX100.

8 minutes ago, KnightsFan said:

Cropping for 120 doesn't mean it has less processing power then the NX1. NX1's 120fps is not a full sensor readout either; line skipping vs 1:1 cropping, same processing power (if the XT3 is oversampling at all in 120fps, then it is using more processing power. The NX1's 120fps looks like a 1920x1080 readout at BEST).

Edit: also, adjusted for inflation, the XT3 is cheaper than the NX1 was.

120P on X-T3 seems to be pixel-binned instead of line-skipped, it's a bit soft but aliasing artifact free.

6 hours ago, kye said:

What exactly does "super sampled" mean?

4K60 down sampled from 8K??

That would be absolutely stunning to see!

The source on Weibo was suggesting a supersampling algorithm similar to how C300 derives 2K 444 from 4K sensor - 4K 444 from 8K readout.

Just now, Ingerson said:

nonscientific DR comparison between Eterna simulation and F-Log. Mid exposure, then one and two stops up and down. Included waveform.

https://www.dropbox.com/sh/rw3ecm3tnft1a21/AADf4nRWTJ8WqSQK_TWKiKXUa?dl=0

No idea how to measure that into numbers, sorry.

Yeah I understand, just curious about the higher rolling shutter in 50/60P compared to 25P, 16ms/9ms almost 2x. Did you notice that? I though higher framerate usually result in lower rolling shutter.

Whici DR mode did you use for the comprison? DR mode is now available in video mode like X-H1, Fuji quotes 12 stops DR with DR400 and Eterna.

1 hour ago, Ingerson said:

Rolling shutter tests here:

4k 24p: https://vimeo.com/292103456/f01f36ff9b
4k 60p: https://vimeo.com/292103784/bae8f67cef
1080p24: https://vimeo.com/292104075/3cca0aa757
1080p60: https://vimeo.com/292104319/73eacbf42b
1080p120 (24fps playback): https://vimeo.com/292104581/7488279a51

Thanks! Fuji website says 16ms rolling shutter for 4K 60P, cinema5D says 9ms for 4K 25P. Do these numbers align with your own observation?

Some love for the almost forgotten A7S III:

https://***URL not allowed***/where-is-the-sony-a7s-iii-video-interview/

Leaks from Weibo:

Exmor RS sensor

4K 120P with crop

8bit/10bit X-OCN with grip

Latest leak from Weibo:

S1/S1R use Sony global shutter sensors. S1R has 150MP pixel shift mode for photos.

4K60p supersampled from 8192x4320 sensor area.

https://www.atomos.com/drives

You don't have to use the Angelbird mini SSD, full size drives work too but they will protrude to the left side a bit.

1 hour ago, CyclingBen said:

@androidlad yep, it’s all face tracking. Any other features I can review for you?

Thanks! I'd love to see a low light ISO test with noise reduction all the way down to -4. Judging from the JPEG performance, X-T3 could potentially retain excellent detail up to 6400.

13 minutes ago, CyclingBen said:

@Anaconda_ yes it is. I was impressed

Did you try face detection & tracking?

The newer f/2 primes and f/2.8 zooms are quad-linear motor so AF-C performance is much improved. On X-T3 older gen lenses got a bump in performance but still cannot compete with the newer ones.

4 minutes ago, Andrew Reid said:

It's full frame but 0.8x crop!

F1.2 = looks like F0.95 on full frame. But glass can be much cheaper (adapted). Canon FD 85mm F1.2L cost me 600 euro. It's a monster, no vignetting.

Bear in mind true medium format goes larger than the Fujifilm GFX 50S. My Hasselblad H3D-39II has a even larger sensor so adjust that 0.8x.

Medium format is also about resolution - both sensor and glass. You can still have a very shallow DOF stopped down at F4, with supreme sharpness compared to a fast lens wide open. On the sensor side, 100MP would be pretty noise on a full frame sensor, without some sort of technological breakthrough.

Regarding noise, it depends on how you look at the pixel count. On a 1:1 level, the 100MP sensor has identical noise performance as X-T3, it's the same 3.76um photosite architecture, the 100MP one is like four X-T3 sensor stitched together. If you normalise the images to standard 8MP (DxO "print" test method), the 100MP will have at least 3EV more DR in theory.

It's about FOV and perspective distortion. A 50mm lens on MF will have 50mm perspective distortion but equivalent of 35mm FOV, making things look "grand".

Of course it's easier to achieve shallow DOF with MF, but it's not the only thing.

Just now, Brian Williams said:

This was posted earlier in the thread, but if you look at his test of FLog converted to Eterna vs Eterna straight out of the camera, the FLog retained way more detail in the highlights, which makes sense. I don't know enough about DR mode to know if it wouldve made a difference here.

DR400 in Eterna would definitely make a difference, since they both share the same base ISO of 640. I'd think of it as Eterna LUT applied internally.

Just now, CyclingBen said:

@androidlad no, there really isn’t a reason to unless shooting F-Log.

What do you mean? Some of the shots have harsh highlight clipping, I think if you used DR400 with Eterna you would have better highlight roll-off. On X-T2 the DR mode isn't available in video mode.

Just now, CyclingBen said:

Here are some samples that I've been able to take in my limited amount of time at home the last few days.  All shot with the 18-55mm. Most of it is in 4k h.265 10 bit the rest is 120p 10 bit. I did do a touchscreen focus test as requested. I also made sure to put in some annoying stock music

 

 Thanks for the test, did you use DR400 with Eterna? You get about 12 stops of DR.

Again from the senior semicon enginner on Weibo:

X-T3 uses a Sony sensor, same 3.76um photosite architecture seen on medium format sensor IMX411/IMX461, with BSI, copper interconnect and 16bit ADC. Think of it as a crop from the medium format sensors.

According to this engineer, Samsung does have a APS-C sensor but it's only recently taped out, 50MP ISOCELL+HD which is equivalent to Sony's Exmor RS, but without the DRAM. 

Looks like 100P/120P on X-T3 went from line-skipping (X-T2/X-H1) to pixel-binning, sharp horizonal lines like the fence are clean unlike X-Trans III. Reasonable amount of detail but definitely not supersampled from the entire 1.29x sensor area.

9 hours ago, gethin said:

I hope it's towerjazz. Just because everything looks like sony these days

Yep, TowerJazz FF global shutter, according to a senior semicon engineer:

ETA mid Oct!

https://www.proav.co.uk/swit-5-5-inch-full-hd-4k-hdmi-lcd-monitor

I was told that TowerJazz, a CMOS sensor manufacturer, just signed off a large order with a big client for its FF global shutter sensor, the client is very likely to be Panasonic.