The Sony FX2 was already introduced on the finish of Could this 12 months, which we additionally lined with a information article and a evaluation of the digital camera. What has been lacking to date (and what you have got been asking for!) is to run it by our checks and offer you our customary Lab Check. to listen to how this digital camera fares in our checks? Then learn on…

In case you missed the detailed specs, see our article right here. Additionally, take a look at Johnnie’s evaluation & mini-documentary right here. In essence, we’re speaking in regards to the following key specs:

• Sensor and Decision: 33 Megapixel sensor able to oversampled 4K (7K seize at as much as 30P) and 4K 60P (S35 crop).
• Recording Choices: 4:2:2 10-bit recording with S-Log3 and S-Cinetone choices, and exterior 16-bit uncooked video choice by way of HDMI.
• ISO and Dynamic Vary: Twin-gain structure with Base ISO settings of ISO800 and ISO4000 for S-Log3 taking pictures

Oversampled 4K from a 7K seize sounds fairly good, therefore we should always be capable of see clear, detailed 4K video with good noise traits. However one factor after the opposite, let’s look first on the rolling shutter. And as per standard, a giant thanks to my colleague Florian, who helped me shoot and analyze this Lab Check.

Rolling Shutter of the Sony FX2

Utilizing our 300Hz strobe mild, we get the next consequence:

27.5ms (much less is healthier) for a full body 4K seize – ouch! This can be a dangerous consequence and places the FX2 on the very backside of our check rankings. In APS-C crop mode, outcomes get higher, and we see 12.8ms.

Dynamic Vary of the Sony FX2 at ISO800 and 4000

Let’s first have a look at a waveform plot of our Xyla21 chart (please head over right here first in case you are not conscious of how we check dynamic vary), taking pictures full body 4K SGamut3.Cine / Slog3 at ISO800:

We will determine 13 stops above the noise flooring, and a 14th contained in the noise flooring. IMATEST reveals the next consequence:

We get 12.4 stops at a signal-to-noise ratio (SNR) of two, and 13.5 stops at an SNR = 1. These are good however not distinctive outcomes for a full-frame sensor, helped by the 7K downsampling. Nevertheless, there’s extra noise discount at play, which can’t be turned off, as could be seen within the “Noise spectrum” graph on the decrease right-hand facet. Amplitudes at larger frequencies rapidly drop to values beneath 0.2. This will also be seen within the waveform graph above; the noise flooring appears to be like actually clear.

It is very important point out the center graph within the IMATEST consequence above: there are extra stops obtainable within the noise flooring (above the blue “13.5” line). The latitude check will reveal whether or not these could be made “usable” by extra post-processing or not.

At ISO4000, the dynamic vary drops to 12.1 / 13.2 stops at SNR = 2 / 1. Right here, the power of a twin native ISO sensor is displaying up – growing ISO values from 800 to 4000 does probably not penalize dynamic vary (actually, lower than half a cease). Superb!

Publicity Latitude Check of the Sony FX2 at ISO800

As talked about earlier, latitude is the digital camera’s capability to retain element and colours when over- or underexposed and pushed again to a base publicity. This check could be very revealing, because it pushes each digital camera to its absolute limits – not simply within the highlights however largely within the shadows.

Moreover charts and IMATEST, which offer absolute (goal) mathematically obtained values, this latitude check is our “actual world” experiment on how usable the pictures are utilizing a rigorously crafted studio scene.

All latitude pictures have been accomplished at ISO800 with 4K 10-bit 4:2:2 XAVC compressed codec utilizing SGamut3.Cine / SLog3. Utilizing a colour area remodel (CST) on the primary node to DaVinci Huge Gamut, then an adjustment node, and on the finish one other CST from DaVinci Huge Gamut to Rec.709, the information have been reworked into the Rec.709 area.

Our studio base publicity is (arbitrarily) chosen as having an (ungraded) luma worth of 60% on the brow of our topic on the waveform monitor. On this case, my pricey colleague Johnnie:

From right here, we will overexpose 4 stops. The pushed again to base picture appears to be like like this:

Johnnie’s brow remains to be high-quality, however as standard, on the colour checker, some patches are already clipped.

Now, we begin to underexpose by closing down the iris of our trusty ZEISS 85mmT1.5 CP2 lens, and from T8 onwards, we’re doubling the shutter pace.

At 3 stops beneath, pushed again to base, we get the next picture (we’re at 7 stops of publicity latitude):

Noise ranges look good, no want for added noise discount in put up.

At 4 stops beneath (8 stops of latitude), noise ranges are elevated, however noise discount simply cleans this up:

Properly, all is seemingly high-quality, the (transferring) photos are wanting fairly good! Nevertheless, there’s a downside that we haven’t seen so removed from Sony, and that’s some kind of line splitting the picture, seen as a horizontal line from the underside of the white paper on the left, going throughout the picture to the fitting. We solely noticed this with different high-resolution digital camera sensors, just like the medium format FUJIFILM GFX100II (Lab Check right here).

Now, this phenomenon can occur if the pixel density on a sensor could be very excessive and {an electrical} cost impacts neighboring pixels (on this case, as a result of harsh horizontal transition from the gray background to the white piece of paper – thus the horizontal line).

Therefore, though the 4 stops beneath the picture look actually good from the noise/restoration perspective, the horizontal subject on the decrease finish of the image can’t be ignored. That limits the publicity latitude of the Sony FX2 to 7 stops solely.

That’s a disgrace, as in any other case we will push this picture even additional, to 9 stops of publicity latitude, with out and with noise discount beneath:

The noise could be cleaned up once more, with the picture nonetheless wanting comparatively good! If it weren’t for the sensor’s horizontal line making the picture non-recoverable.

For the report, right here is the ten stops latitude picture, therefore 6 stops of underexposure, pushed again, in addition to the identical picture utilizing noise discount in DaVinci Resolve 20.2.2:

I normally like to have a look at the shadow facet of our topic’s face, if pores and skin colour remains to be intact. Properly, at 10 stops of latitude, they aren’t anymore.

Therefore, if it weren’t for the sensor’s horizontal line subject, we might attribute 9 stops of publicity latitude functionality for the Sony FX2 – utilizing the interior, compressed 10-bit codec! I might think about {that a} RAW 12-bit codec would carry out even higher right here, however that’s not obtainable (internally) for the Sony Alpha or FX line of cameras.

This consequence (ignoring the unacceptable sensor horizontal line subject) can be on par with a few of the not too long ago examined full-frame client cameras, just like the Sony A9III (Lab Check right here), or the Canon R1 (Lab Check right here). The clear chief of the pack is presently the Panasonic LUMIX S1II (Lab Check right here), which exhibited a stable 10 stops of publicity latitude, even with some room in the direction of 11 (utilizing inside 12bit ProRes RAW, with DR Increase mode “on”) – on the LUMIX S1II this comes on the expense of a rise of rolling shutter to 27.5ms – the identical worth we’re seeing for the Sony FX2. The not too long ago examined Nikon ZR (Lab Check right here) confirmed 8 stops of publicity latitude.

Therefore, the Panasonic S1II has a stable 3 stops higher publicity latitude with none horizontal traces or different artefacts when in comparison with the Sony FX2 on the identical rolling shutter!

Abstract

Our lab check outcomes present mediocre rolling shutter efficiency however fairly good dynamic vary outcomes. The latitude outcomes of 7 stops the place sensor horizontal splitting line remains to be not displaying up and 9 stops ignoring the sensor subject, exhibits that the picture pipelines of latest Sony client cameras are getting significantly better not too long ago – the Sony A1 and A9III already exhibited 9 stops with their inside 10-bit codecs whereas the Sony a7IV (Lab Check right here) was tormented by giant blotches of chroma noise, limiting the latitude to usable 7 stops.

Nevertheless, the sensor subject can’t be ignored as it’s ruining the picture, therefore Sony wants to repair this asap!

Are you planning to shoot with the Sony FX2, or have you ever shot already with it? What are your experiences? Tell us within the feedback beneath!