Sony FX5 Lab Test: Rolling Shutter, Dynamic Range and Exposure Latitude

Lab Test

by Gunther MachuToday

Subscribe to Lab Tests

Long anticipated, now a reality: a production version of the new Sony FX5 with final firmware has arrived at our CineD headquarters in Vienna. Time to run it through our standard CineD Lab Test! Curious to hear more? Then read on …

The new Sony FX5 created quite a hype when it was announced, mainly because of headline features like internal X-OCN LT RAW recording, 3 base ISOs, dual gain, and high frame rate shooting up to 4K240 frames per second (with a future firmware). We have covered the key specs here, and Johnnie made a review including a mini documentary here. Also, Johnnie and Nino sat down with Philipp Bloom for a special FX5 episode of the weekly podcast here.

Of course, we were also very interested to see how it would fare in our standard CineD Lab Test. And, I have to say, it performed exceptionally well. But first things first: again, a big shout out to my colleague Florian Milz for helping to shoot and analyze this test!

Similar to the recently tested Sony a7R VI, the new FX5 features a dual gain sensor read out mode. At a fixed ISO of 800 when shooting in S-Log3, this mode combines a low and high ISO gain in one exposure, resulting in expanded dynamic range in the shadows.

We have seen dual-gain implementations in various cameras before, such as the Canon C70, the LUMIX GH6 and GH7, and, more recently, the LUMIX S1II, which set a new benchmark for consumer full-frame cameras in our standardized CineD Lab Test. For all of these cameras, turning DG “ON” delivered about 1 to 1.5 stops more dynamic range and exposure latitude, but this mode also doubles the rolling shutter compared to turning it “OFF”. Hence, let’s see how the Sony FX5 behaves in that respect! Current firmware tested is V1.00.

Rolling shutter of the Sony FX5

As usual, we used our 300Hz strobe light to reveal the sequences of black and white bars, an artifact of the read-out nature of rolling shutter sensors.

For 5K 17:9 X-OCN LT with dual-gain readout “ON”, we get the following 13.3ms (less is better):

This is a tad better than the Sony a7R VI, which showed 15.5ms in DG “ON” mode (at 4K). Switching DG to “OFF” gives a strikingly good 5.6ms!

Quite interestingly, so far all the cameras using a dual-gain function showed about double the rolling shutter when DG was activated. The Sony FX5 actually shows more than that, about 2.4x the rolling shutter – but the absolute value (13.3ms) is still top-notch compared to the Panasonic LUMIX S1II, for example, which exhibited 27.5ms when dual gain was “ON”.

For all the other modes, please refer back to our camera databases.

Dynamic range of the Sony FX5

If you are not aware of how we test dynamic range, please have a look here first.

Again, let’s start with a waveform of X-OCN LT at 5K, ISO800 and DG “ON”, S-Gamut.Cine / S-Log3:

We can easily identify 13 stops above the noise floor, with a 14th still slightly above and even a 15th inside the noise floor. The noise floor itself looks very clean – a result of the dual-gain implementation!

Notice also that the clipping code values of S-Log3 don’t change whether DG is turned “ON” or “OFF”, in contrast to other cameras like the Panasonic Lumix S1II, where dual gain “ON” (DRBoost) for V-Log lifts the code values in the highlights (where clipping occurs). Hence, any existing S-Log3 workflow stays fully compatible with the dual gain implementation on the Sony FX5.

IMATEST reveals the following:

IMATEST shows 13.4 stops at a signal-to-noise ratio (SNR) of 2, and 14.4 stops at SNR = 1. For an internal RAW codec, this is a fantastic result! For comparison, the ARRI Alexa Mini LF (Lab Test here) showed very similar values: 13.4 / 14.5 stops at SNR = 2 / 1 in ARRIRAW!

And, compared to the Sony BURANO 8K (Lab Test here), which also allowed internal X-OCN LT recording (exhibiting 10.8 / 12.2 stops at SNR = 2 / 1, ISO800), the FX5 shows 2.6 stops more dynamic range.

Also, please note how the “Noise spectrum” graph on the lower right-hand side above shows very high amplitudes at high frequencies- the amplitude never drops below 0.4 – a clear sign of a very detailed 5K image.

Turning DG “OFF” in 5K X-OCN LT at ISO800 results in 10.6 / 12 stops at SNR = 2 / 1. This is remarkable, as we are getting about 2.5 stops more dynamic range with DG “ON”.

The Sony FX5 also offers 2 more base ISO’s with S-Log3, namely ISO4000 and ISO12800. For ISO4000 X-OCN LT SG3.C/SLog3, we get 10.1 / 11.6 stops at SNR 2 / 1, and for ISO12800 9.1 / 11.2 stops. Hence, dynamic range drops significantly towards the two higher base ISO’s.

In internal compressed 10bit 4K H264 4:2:2 XAVC S-I mode, we are getting the following waveform and the corresponding IMATEST result:

We are getting 14 stops at SNR = 2, and 14.4 at SNR = 1. These are really high figures, but of course some noise processing is involved here (as usual, we turned off any additional noise reduction in-camera), but some NR remains, as can be seen in the “Noise spectrum” graph on the lower right-hand side – amplitudes drop to below 0.2, but only at higher frequencies.

4K DG “OFF” reveals 12 / 13 stops at SNR = 2/ 1 and ISO800 – not bad either, considering that DG “OFF” enables a super-fast rolling shutter of 5.6ms! For the other ISO combinations, please visit our databases!

Exposure Latitude of the Sony FX5

Now here is where things should get really interesting – a 16-bit RAW mode combined with dual gain should yield impressive results.

As stated in previous articles, latitude is a camera’s ability to retain detail and color when over- or underexposed and then pushed back to a base exposure. This test is very revealing as it pushes the complete image pipeline of any camera to its absolute limits, not just in the highlights but mainly in the shadows. And typically, it’s in the shadows that dual-gain technology is a massive plus.

With the new DaVinci Resolve 21 version, native X-OCN LT support is available, and we mostly used the camera RAW controls to adjust exposure. We then brought the footage into Rec709 space using a colorspace transform (CST).

Our studio base exposure is (arbitrarily) chosen as having an (ungraded) luma value of around 60% on the forehead of our subject on the waveform monitor – in this case, my colleague Nino:

From here, in typical S-Log3 fashion, we can overexpose 4 stops until the red channel on Nino’s forehead is at the cusp of clipping (but still intact):

And now, we start to underexpose by closing down the iris of our standard compact prime 85mm T1.5 lens in one-stop increments until T8, and then the shutter angle is halved.

And quite amazingly, there is not much to report until we have reached 5 stops of underexposurer (9 stops of exposure latitude) – here, we get the first signs of increased noise.

If needed, noise reduction easily cleans this up. Hence, let’s move to 10 stops of exposure latitude, 6 stops underexposed, and brought back to base:

Color-wise, this image is fully intact, and we have nice, finely dispersed noise that looks very organic. Noise reduction in DaVinci Resolve 21 still cleans this up nicely:

Let’s pause for a moment here. For one, we are at an ARRI Alexa Mini LF level of performance here. It already showed in the IMATEST results, but the latitude test is real-world confirmation. Second, compared to the Sony BURANO 8K (Lab Test here), which also (but barely) managed 10 stops of exposure latitude with an OK’ish image, we have to state that the FX5 image looks much better at the same 10 stops.

Let’s push it to 11 stops, 7 stops underexposed:

Now, noise becomes very pronounced. Noise reduction yields the following:

As you can see, things become very borderline. Horizontal stripes are starting to appear (e.g., below the text caption in the image, more visible in the moving image), but the shadow side of Nino’s face can still be recovered color-wise. In general, colors stay more or less intact without big shifts towards green or pink – I will still count this image as valid.

At 8 stops underexposure, we get the following:

Noise reduction cannot clean this up any longer – game over:

Now, the color noise can still somehow be mitigated, but the luma noise cannot be controlled any longer without smearing out all image details. Large blotches of luma noise hover around the image (the dark species of noise). But there is no strong color cast, and overall the image is somewhat still intact!

Hence, we end up at 11 stops of exposure latitude on the Sony FX5! Now, this is a new benchmark for full-frame cameras, as it is a tad better than the ARRI Alexa Mini LF (Lab Test here) and the Panasonic LUMIX S1II (Lab Test here).

The king of the latitude test however, is still the ARRI Alexa 35, which exhibited 12 stops of exposure latitude (Lab Test here).

Now, one question remains, and it is one that frequently gets asked by our dear readers – how does the internal, compressed 10-bit All-I 4:2:2 (XAVC S-I) codec perform? There is always speculation that a dedicated noise-reduction algorithm from the camera manufacturer, fine-tuned to the specific camera, could potentially outperform DaVinci Resolve’s algorithms (which have improved as well, with the AI noise reduction, for example).

Our argument to test the RAW option (if available) so far was always the technical one – as the exposure latitude of a camera is mostly determined by the shadow performance (we are mostly pulling shadows from underexposures), a 12-bit (or 16-bit linear, as Sony claims) RAW codec would always beat a 10-bit compressed codec, as colors should be better preserved, including less potential for color banding.

Well, here it is – 10-bit 4:2:2 XAVC S-I (using SGamut3.Cine / S-Log3): as before with X-OCN, we use a color space transform to convert the S-Log3 footage into DaVinci Wide Gamut, then do our adjustments, and then another CST to transform into Rec709. Let’s start with 5 stops of underexposure, brought back to base:

As shown, at 5 stops of underexposure (9 stops of latitude), a pinkish color cast creeps into the image. In addition, the noise is blockier; it doesn’t have the fine, organic look of the X-OCN LT option.

At 6 stops of underexposure (i.e. 10 stops of exposure latitude), we have already reached the limit:

We have large blotches of pinkish chroma noise hovering around the image, and there are these very nasty individual red / pinkish pixels that cannot be removed by noise reduction:

This is not acceptable. Look at the shadow side of Nino’s face, for example.

Hence, we have to conclude that XAVC S-I reaches its limit at 9 stops of latitude. That is 2 stops below the X-OCN LT result and clearly demonstrates the power of RAW. To put things in perspective, 9 stops of exposure latitude was a superb result back in 2024 / 2025.

Summary

Sony’s FX5 camera sets a new benchmark in our standardized CineD Lab Test for full-frame cameras. Rolling shutter values are superb (with dual gain set to “OFF”), and still very manageable when DG is set to “ON”. Dynamic range is in a similar league to our top contender so far, the ARRI ALEXA Mini LF, and this observation is confirmed in the exposure latitude tests.

Part of the superb results is the new internal X-OCN LT RAW codec. All of that in a consumer price bracket! Wow – well done, Sony!

To check how the Lab Test results of the Sony FX5 stack up in our ranking of all tests, and how they compare with other cameras we tested, head over to the CineD Camera Database (Sony FX5 entry) to find out!

Have you already shot with the new Sony FX5? What are your experiences so far? Please let us know in the comments section below.