Listening to Victor Ha from FUJIFILM in our interview recorded right here, there was one thing talked about that basically struck me – there’s this new sensor and picture processor within the X-H2S that permits studying the sensor information at 14-bit for as much as 30 frames per second. A brand new F-Log2 profile is added to accommodate the brand new sensor capabilities. This can be a distinctive function for a client digital camera that may actually enable a lot greater dynamic vary values. Therefore, I used to be fairly curious to run the brand new FUJIFILM X-H2S by our lab check procedures. Curious too? Then learn on…
To date, all the patron digital camera sensors that I’m conscious of enable the upper body fee recordings required for shifting photos solely at 12 bit sensor readout modes. 14 bit sensor readout modes had been reserved for high-end cinema cameras just like the ARRI ALEXA, for instance. Now, FUJIFILM brings this function right into a client hybrid stills / video digital camera, the X-H2S. Plus, inside 10 bit 4:2:2 ProRes HQ recording. Wow.
And, FUJIFILM declare as much as 4x greater sensor readout speeds. Time to torture check this new digital camera in our lab!
Utilizing our strobe mild, which creates the sequence of black and white bars (associated to the readout nature of CMOS sensors), we’re getting the primary end in UHD 25 frames per second: 9.7ms (much less is healthier).
This can be a good outcome, contemplating the 14-bit readout of the sensor! For 3:2 6.2K open gate mode, we’re getting 11.5ms. Fairly curiously, and in keeping with FUJIFILM’s statements that for greater frames per second the sensor readout mode modifications to 12bit, we’re getting 5.3ms at UHD 60fps. In UHD 120fps, the readout mode modifications once more, a slight crop is utilized (1.29x) and we get 3.9ms.
Now it will get attention-grabbing. In case you are not conscious of how we check the dynamic vary, please take a look right here.
Let me shortly summarize why we now have 3 methods to guage the dynamic vary of a digital camera:
- The waveform plot of the Xyla 21 chart at native sensor decision on a timeline with that decision: this offers a visible indication of what number of stops might be recognized above the noise flooring. Additionally, it exhibits the code worth distribution of the stops. Fairly often, the decrease stops are very shut collectively by way of code worth, therefore in case you underexpose and lift the shadows later in put up (i.e. increasing the shadow stops), you’ll not have sufficient code values between the stops and the result’s ugly banding (lack of advantageous shade transitions between the stops).
- IMATEST: IMATEST will calculate the sign to noise ratio for every cease. That may be a purely mathematical calculation and is useful to establish, how “clear” every of the stops is. Cameras that use quite a lot of inside noise discount naturally fare higher than others with much less noise discount. There isn’t any method to account for that in a significant approach, because the noise footprint of each digital camera / sensor is completely different. Therefore, there’s additionally no “customary” noise discount you could apply in put up to match cameras. That’s why we at CineD at all times flip off the noise discount – as beneficial by IMATEST.
- Latitude: publicity latitude is the aptitude of a digital camera to retain colours and element when over- or underexposed. Our CineD studio scene is the actual world check (in a managed setting) of how far a digital camera might be pushed. The great thing about this check is that it clearly reveals what number of stops are usable in our rigorously composed customary scene, because the play area will get equalized – and, it’s revealed if a digital camera is “dishonest” by utilizing extreme inside noise discount. Simply to say it: another actual world scene with a distinct distribution of lights and shadows would present a distinct outcome. Regardless of how a lot inside / or put up noise discount is used, cameras that present a stable 12 stops at a signal-to-noise ratio of two usually have 8 stops of latitude in our scene. Cameras, just like the ARRI ALEXA, that present 2 stops extra at SNR = 2 even have two stops extra latitude. Cameras that attempt to obtain one thing near 12 stops at SNR = 2 by utilizing heavy inside noise discount solely present 7 stops of latitude.
So, this trinity of exams could be very revealing and helps to establish if the mixture of sensor readout, sign processing and codec permits to push publicity round over a wide range.
FUJIFILM X-H2S at ISO1250:
Utilizing ProRes HQ, the brand new F-Log2 profile with a local ISO of 1250 and the native sensor decision of 6240×4160 open gate 3:2 mode, turning “noise discount” to -4 and interframe NR “OFF”, we get the next waveform plot:
13 stops might be recognized above the noise flooring. The noise flooring appears somewhat clear, which hints at extra in-camera noise discount even if we tried to show it down so far as doable.
There’s a stable code worth distinction between the eleventh and twelfth cease, in addition to from the twelfth to the thirteenth cease in F-Log2. This implies: in case you underexpose your pictures and you could push up these decrease stops (and successfully increase them), you might be much less prone to run in to issues like banding and lack of shade transitions.
IMATEST calculates the next:
IMATEST exhibits 11.9 stops at a signal-to-noise ratio (SNR) of two and 13.4 stops at SNR = 1 (on a 6.2K timeline). That may be a superb outcome for a client APS-C sensor (The ARRI ALEXA traditional confirmed 13.8 at SNR = 2 and 14.8 at SNR =1, to present you our present benchmark)!
Additionally, in case you have a look at the center graph above the blue “13.4” line, there are nonetheless about 2 extra stops recognized. That’s really a bit higher than the BMPCC 6K, not a lot for SNR = 2, however the SNR = 1 worth is half a cease higher (additionally measured in 6K, check right here). In comparison with that, the Canon C70 (lab check right here) trumps each with 12.8 stops at SNR = 2 and 14.2 stops at SNR = 1 for XF-AVC Intra C-Log2 (12.3 at SNR = 2 and 14 at SNR = 1 for CRAW).
Simply to present you this extra information level: scaled to 3840×2160 UHD, I get 12.2 stops at SNR = 2 and 13.6 stops at SNR = 1.
As described earlier, latitude is the aptitude of a digital camera to retain colours and particulars when over / underexposed and pushed again to a base reference stage.
For our CineD studio exams, the bottom publicity stage refers to a waveform worth of round 60% on the brow of my expensive colleague Johnnie. The digital camera was once more in 6.2K ProRes HQ F-Log2 open gate mode at ISO1250 (noise discount at -4 and interframe NR “OFF”):
For base publicity, we uncovered a bit beneath 60% to permit for 4 stops of overexposure with out clipping the purple channel on Johnnie’s brow, then introduced again to base publicity:
Now, lets successively underexpose the scene by closing the iris of the lens to T8.0, after which halfen the shutter worth.
At 3 stops underneath base publicity, pushed-back noise begins to kick in:
The noise could be very advantageous, very natural trying. No want for noise discount, in my view. Bear in mind, we’re already at 7 stops of publicity latitude. Different cameras survive this level solely by making use of noise discount in put up.
The 4 stops underexposed, pushed again picture appears like this:
Noise is now very dominant inside the picture, however nonetheless very advantageous, natural trying. We’re at 8 stops of publicity latitude! Noise discount in DaVinci Resolve simply removes any extreme noise:
We’re already at 8 stops of publicity latitude. Now, the massive query is, can we push it one cease extra, to 9 stops of publicity latitude?
Let’s take a look:
Heavy chroma and luma noise is now current within the picture, not so advantageous any extra. When utilizing noise discount in DaVinci Resolve, we get this:
This nonetheless appears surprisingly good, really! Nevertheless, the background shade modified to gray as a substitute of olive, and the shadow facet of Johnnie’s face is rendered pinkish as a substitute of orange – which I don’t like. However, there are not any massive blotches of chroma noise, or horizontal or vertical traces. Even the shifting picture nonetheless appears fairly okay.
Decide for your self, however that is near acceptable, and we’re at 9 stops of publicity latitude. Please take a look at different cameras just like the Canon C70 (check right here), full-frame cameras just like the Sony A1 (check right here), or the latest Sony VENICE 2 (check right here) – the FUJIFILM X-H2S undoubtedly trumps the C70 and even the Sony A1, whereas the 5 stops underexposed picture appears nearly nearly as good as with the VENICE 2 in 4K ProRes HQ – and we’re speaking a couple of full-size cinema digital camera right here.
If we examine the 5 stops underexposed picture with the chief of the pack, the ARRI ALEXA Mini LF (lab check right here) it turns into clear, nonetheless, that it can’t compete – the 5 stops underneath picture of the Mini LF undoubtedly appears a lot better, really even the 6 stops underexposed picture does. And, the Mini LF has one cease extra within the highlights as nicely, resulting in 10 stops of publicity latitude.
So, a stable 8 stops with extra wiggle room in direction of 9 stops – that’s fairly spectacular, and is definitely higher than a lot of the latest full body client cameras. Good job, FUJIFILM!
The FUJIFILM X-H2S fares very well for a client digital camera in our lab check, particularly when contemplating that it’s an APS-C digital camera, not full body. The rolling shutter efficiency is superb, the dynamic vary outcomes are spectacular, as is the latitude check. It turns into clear from these outcomes that the 14-bit sensor readout undoubtedly is a part of this good efficiency.
So, you possibly can really say it’s a BMPCC 6K on steroids with a ton of extra options like IBIS and all the different benefits of a real hybrid digital camera, together with a smaller kind issue.
Have you ever used the FUJIFILM X-H2S but? Please tell us what you assume within the feedback beneath