As soon as we got confirmation from DJI that the recent firmware for the Ronin 4D is final in terms of image quality, we ran the DJI Ronin 4D through our lab test. Curious to hear how it fares? Then read on…

The new DJI Ronin 4D is quite a unique offering in the camera market, featuring the latest Zenmuse X9 gimbal (with an added stabilized Z-axis) combined with a main camera body that acts as the recording unit in one camera system. What makes this even more interesting is the fact that we are talking about a 6K dual native ISO (800 / 5000) full-frame sensor on the X9 gimbal, with internal ND filters, various exchangeable mounts (DL, M, E) plus the fact that it allows recording internally in ProRes RAW HQ and ProRes HQ codecs, along with an h.264 4:2:0 10bit option as well.


The DJI Ronin 4D in the CineD studio. Image credit: CineD

My colleague Nino made an extensive review on the Ronin 4D here, and please head over here for the technical info and specifications.

So, without further ado, let’s dive into our DJI Ronin 4D Lab Test cycle.

Rolling Shutter


As this camera system is intended to be used by its very design on a lot of fast-moving/action scenes, a low rolling shutter value will be very important.

Unfortunately, the DJI Ronin 4D disappoints on this end: I measured a rather high rolling shutter of 19.6ms in 6K full-frame mode at 25 frames per second, in 17:9 (DCI) mode (indicated by the black bars on the top and bottom):


DJI Ronin 4D rolling shutter of 19.6ms at 25fps in 17:9 6K full-frame mode. Image credit: CineD

Other recent full-frame cameras fare better in the rolling shutter department – the Sony A1 for example has 16.6ms, the Nikon Z9 has 14.5ms and the Sony a7S III shows only 8.7ms (less is better).

The rather high rolling shutter of the DJI Ronin 4D also explains why in full-frame mode only 48 frames per second are possible (higher fps are possible by changing the aspect ratio to 1:2.4).

In Super 35 (crop mode), the rolling shutter results improve to 14.5ms (in 4K at 25fps), and 13.4ms at 4K 60 fps.

Dynamic Range at ISO 800 and ISO 5000


If you are not familiar with how we test dynamic range, please head over here.

As mentioned before, the Zenmuse X9 gimbal features a dual native ISO sensor with a base ISO of 800 and 5000.

Starting off with 6K full-frame ProRes RAW HQ (highest quality settings) at ISO 800, developed in Premiere Pro V22.0 using D-Gamut/D-Log as the gamma curve, we can identify about 13 stops above the noise floor:


DJI Ronin 4D 6K ProRes RAW HQ at ISO800, D-Log. Image Credit: CineD

Even a 14th and a faint 15th stop are visible – wow! IMATEST confirms this result:


IMATEST dynamic range calculation for ISO800 on the DJI Ronin 4D. Image credit: CineD

12.4 stops at a signal to noise ratio (SNR) of 2 and 13.7 stops at SNR = 1 are calculated. In the middle graph, above the blue (13.7) curve about 2 additional stops can be identified. Those are very good results, considering the fact that with ProRes RAW we get a mostly unprocessed signal directly from the sensor!

At ISO 5000, the results stay almost identical, looking at the waveform plot below. The 14th and 15th stops are a bit more smeared out because of the noisier image though:


DJI Ronin 4D 6K ProRes RAW HQ at ISO5000, D-Log. Image Credit: CineD

IMATEST dynamic range calculation for ISO5000 on the DJI Ronin 4D. Image credit: CineD

For ISO 5000 using ProRes RAW HQ in 6K full-frame mode, IMATEST calculates 12 stops at SNR = 2 and 13.4 stops at SNR = 1. Hence, about 0.4 stops lower results than for ISO 800, but again, really good!

I also tested the dynamic range for ProRes HQ at both ISO’s: here, the camera offers a noise reduction setting from +2 to -2 in the menu, the default value is NR = 0. Using NR = 0 and ISO 800 I am getting 12.1 / 13.3 stops at SNR = 2 and 1 respectively. For NR = -2 and ISO 800, I am getting 11.8 / 13 stops at SNR = 2 / 1, and at ISO 5000 quite surprisingly the same results.

Hence, we are getting the best dynamic range results out of the box by using ProRes RAW HQ on the DJI Ronin 4D. This conclusion is somewhat surprising, as usual (on other cameras) better values at the SNR = 2 and 1 thresholds are achieved by using some of the internal, noise processed ProRes HQ files.

We contacted DJI about this and they replied “We read about that you found the dynamic range results under ProRes RAW HQ is better than under ProRes HQ. This is actually one of our image tuning directions for ProRes HQ that we hope the footage could restore more details to leave more space for post production, so we choose to minimise the artefacts caused by noise reduction process by keeping part of the noise that could be reduced easily in post to retain the original texture of footage. Hope you could understand this decision. Thanks!”

Latitude Test in 6K ProRes RAW HQ at ISO 800


As written earlier, latitude is the capability of a camera to retain colors and detail when over-or underexposed from a base exposure. This test pushes every camera (sensor) to its absolute limits and is very revealing.

Using ProRes RAW for the latitude test is very convenient because the exposure of the over-or underexposed images can be very easily adjusted by the exposure slider in Premiere Pro V22.0 (using “maximum bit depth” for the Premiere timeline settings). As for the dynamic range results above, I developed the files using D-Gamut / D-Log in Premiere and added a DJI_ZENMUSE_X9_DLog_To_Rec709 LUT (available on the DJI website).

Hence, here is the base exposure studio scene with my dear friend Nino as the model. As mentioned earlier, our CineD studio base exposure is achieved by having a luma value of 60% on the waveform plot at the subject’s forehead:


CineD studio base exposure scene with the DJI Ronin 4D. Image credit: CineD

One thing I realized is the fact that I could not get the framing right with the standard 85mm T1.5 Zeiss compact prime that I always use for full-frame cameras. I had to use the 50mm T1.5 instead (and get a bit closer), hinting at a slightly smaller than full-frame sensor on the X9 Zenmuse gimbal.

Now, let’s have a look at 4 stops of overexposure, by opening up the iris of the lens to f1.4 and having a shutter angle of 360°:


4 stops over, pushed back. Image credit: CineD

4 stops over, pushed back is no problem. Nothing clips, colors do get a bit cooler but all is fine at first glance.

At a second glance, some slight pink / greenish banding can be seen especially towards the right side where stronger gradients are present – I have increased the gradient and added massive saturation such that it is easier to see (of course, “max. bit depth” setting was used for the sequence settings):


Banding artifacts are visible when developing the files in Premiere Pro V22.

We contacted DJI about this issue, and they were able to reproduce it in Premiere Pro but not in Final Cut Pro. Hence we also developed the ProRes Raw files in Final Cut Pro and yes, there is no banding visible:


The ProRes RAW file developed on Final Cut Pro – no banding!

Obviously, this problem is related to the Premiere Pro 3D LUT calculation process in Lumetri Color. DJI told us they “will look deeper into the issue and keep optimising the official 3D LUT to provide better platform compatibility in the future”.

Now let’s move to underexposure, with 3 stops under, pushed back to base. Noise starts to show up, but it is a very fine noise (plus, the noise helps to hide the banding described above):


3 stops under, pushed back. Image credit: CineD

At 4 stops under, pushed back heavy luma and chroma noise starts to degrade the image:


4 stops under, pushed back. Image credit: CineD

The luma and chroma noise is still finely dispersed, and the image can be saved by using noise reduction in DaVinci Resolve (17.4.3) – however, rather pronounced horizontal lines become visible especially in the upper half of the image:


4 stop under, pushed back and using noise reduction. Image credit: CineD

Noise reduction settings in DaVinci Resolve for 4 stops under, pushed back. Image credit: CineD

As you can see above, quite heavy temporal and spatial noise reduction is needed to save the image. To bring the ProRes RAW files in Resolve, I exported a ProRes XQ 4444 file from Premiere.

Now, moving on to the 5 stops underexposed image, the limits are reached, see below:


5 stops under, pushed back. Image credit: CineD

Noise becomes excessive, and even very very strong noise reduction cannot save the image (plus the static horizontal lines that cannot be removed by noise reduction):


5 stops under, pushed back and noise reduction added. Image credit: CineD

DaVinci Resolve noise reduction settings for 5 stops under, pushed back. Image credit: CineD

That brings us to the conclusion, that the DJI Ronin 4D offers 8 stops of exposure latitude (4 over and 4 under) – a very good result, in line with the best consumer full-frame cameras available for now. The Panasonic S5, S1, and S1H cameras offer 8 stops, as well as the Sony A1 for example.

Just for your reference, the ARRI Alexa Mini LF that we tested (here) shows 10 stops of exposure latitude – 5 over to 5 under.

Summary


The DJI Ronin 4D is a very unique offering on the market, and it does not disappoint in the lab test. With the exception of a rather high rolling shutter (especially for a camera system that is designed for moving/action shots), it offers really good image quality with a very good dynamic range with internal ProRes RAW recording and equally good latitude results.

Regarding the banding artifacts found when developing the files in Premiere Pro V22, as mentioned above DJI is aware of it and working on it: “We’ll look into it deeper and keep optimising the official 3D LUT to provide better platform compatibility in the future.”

And I have to mention this, DJI were super helpful and very responsive in answering all our questions regarding our Ronin 4D Lab Test- so I am very confident this issue will be solved soon.

Guys, what do you think of this Ronin 4D lab test? Have you shot with the DJI Ronin 4D yet? Tell us your experience in the comments section below …

The post DJI Ronin 4D 6K Lab Test: Rolling Shutter, Dynamic Range and Latitude appeared first on CineD.