The Nikon ZR is producing fairly a hype because it was launched shortly earlier than IBC 2025: it provides a compact, full-frame physique with inside 6K REDCODE RAW NE (Nikon EXPEED) recording and a joint RED / Nikon label screaming “cinema” on the again at a very enticing value! We had been wanting to run it via our CineD Lab Take a look at procedures to see if the picture high quality lives as much as the hype. to listen to extra? Then learn on …
I’ve to confess that I wasn’t fairly resistant to the hype, and when the Nikon ZR lastly arrived at our CineD headquarters right here in Vienna, I couldn’t wait to seize my colleague, Florian, and collectively run it via our customary CineD lab torture take a look at. Once more, an enormous “thanks” to Florian for serving to me shoot and analyze this take a look at. This time, it was actually three lab assessments in a single, as we wished to check the three completely different RAW codecs out there on this digicam: REDCODE RAW NE, N-RAW, and ProRes RAW.
Additionally, please stand by for Johnnie’s digicam evaluation, coming very quickly too.
The Nikon ZR in our CineD studio. Picture credit score: CineD
So, let’s begin as typical with rolling shutter.
Rolling Shutter of the Nikon ZR
Once more, we use our strobe mild at 300Hz to find out the rolling shutter through the sequence of black and white bars.
For full body REDCODE RAW NE, we get the next picture, 9.4ms rolling shutter (much less is best). For a full-frame 6K readout, this can be a actually good end result and precisely the identical because the shut sibling Nikon Z6 III, out there in our lab database right here.
In DX (or APS-C mode), we’re getting 6.3ms, thus enabling a 4K120 frames per second readout mode as nicely.
Nikon ZR – Dynamic Vary at ISO800 and ISO6400
In case you are not accustomed to how we carry out our dynamic vary testing, then please head over right here first.
Let’s begin with REDCODE RAW NE in full body 6K. These are our growth settings in DaVinci Resolve 20.2.1 within the digicam RAW tab (word that we now have disabled the default “Chroma Noise Discount” – we all the time take a look at with out noise discount):
Wanting on the waveform plot, we see a well-recognized phenomenon:
Nikon ZR utilizing REDCODE RAW NE at ISO 800
I’ve expanded the RGB curves in direction of 5600K utilizing the white stability slider to display a RED-specific phenomenon, which known as “spotlight restoration”, constructed as default into the IPP2 coloration science (REDWideGamutRGB, Log3G10) – seen for the second patch from the left, which isn’t clipped, however the pink and inexperienced channels are recovered and never seen. That is why we are able to solely begin to rely from the third patch (from the left), which isn’t clipped – please additionally take a look at our RED KOMODO X Lab Take a look at right here, or the RED V-RAPTOR [X] 8K VV Lab Take a look at right here. The third to the fourth patch is the primary cease of dynamic vary, the fourth to the fifth is the second cease, and so forth.
We are able to see 10 stops of dynamic vary above the noise ground. Operating this picture via IMATEST, we get 10.2 stops at a signal-to-noise ratio (SNR) of two, and 11.5 stops at SNR = 1.
IMATEST end result for R3D at ISO800. Picture credit score: CineD
Wanting on the “Noise spectrum” graph on the decrease right-hand facet, we see good amplitudes at excessive frequencies, hinting at a really detailed picture with out a lot inside noise discount.
Now, taking pictures the XYLA21 chart utilizing Nikon RAW (N-RAW), we get 9.7 / 10.9 stops at SNR = 2 / 1 (nearly precisely the identical because the Nikon Z6 III), and utilizing ProRes RAW, we get 10.2 / 11.2 stops at SNR = 2 / 1. In essence, the “spotlight restoration” characteristic constructed into RED’s IPP2 coloration pipeline appears to trick IMATEST into barely greater scores.
These values are all within the center to decrease floor for a compact hybrid shopper digicam. Lately, the Panasonic LUMIX S1 II confirmed what is feasible, particularly with their “dynamic vary increase” characteristic turned to “ON” – it reveals 2.5 stops extra dynamic vary than the Nikon ZR for ProRes RAW (when evaluating dynamic vary values between cameras, it is best to all the time examine apples to apples, therefore RAW codecs vs RAW codecs and inside compressed codecs vs inside compressed codecs).
Please take a look on the CineD database for numerous further outcomes, together with ISO6400 (spoiler: we’re shedding between 0.5 to 1 cease at ISO6400 relying on the codec). Let’s simply rapidly take a look on the H.265 N-Log ISO800 waveform:
H.265 ISO800 N-Log waveform for the Nikon ZR. Picture credit score: CineD
You’ll be able to see a large, huge quantity of inside noise discount at work (have a look at the skinny noise ground, and the way clear it’s). IMATEST reveals the next:
IMATEST end result for H.265 ISO800 N-Log. Picture credit score: CineD
Effectively, we are actually getting 12.5 / 13.4 stops at SNR = 2 / 1. However take a look on the “Noise spectrum” graph within the decrease proper nook: you may see that frequency amplitudes drop very quick, having values of solely round 0.1 on the greater frequencies – therefore, noise discount destroys all of the excessive decision particulars. REDCODE RAW NE nonetheless reveals amplitudes of round 0.3 on the similar excessive resolutions.
Once more, please take a look on the databases for all the opposite dynamic vary outcomes.
Nikon ZR – publicity latitude at ISO800
Latitude is the potential of a digicam to retain particulars and colours when over- or underexposed and pushed again to base publicity. A while in the past, we selected an arbitrary worth of 60% luma worth (within the waveform) for our topics’ faces (truly their brow) in our customary studio scene. This CineD base publicity ought to assist our readers get a reference level for all of the cameras examined, no matter how they distribute the code values and which LOG mode is used.
REDCODE RAW NE: we used 6K 25fps R3D NE at ISO800, our trusted Zeiss Compact Prime 85mm T1.5, and developed the recordsdata utilizing the digicam RAW tab once more in DaVinci Resolve 20.2.1. Utilizing the publicity and ISO, we normalized the recordsdata to base publicity. Then, we introduced every little thing into the REC709 coloration house through the use of a Coloration House Remodel (CST) from R3D to DaVinci intermediate/huge gamut, after which utilizing one other CST node to REC709.
Equally, we used the next settings as a place to begin for ProRes RAW and N-RAW in DaVinci Resolve’s RAW tab (lastly, we’re in a position to make use of ProRes RAW in DVR!!), after which once more we used coloration house transforms from the respective coloration areas to REC709:
Nikon ZR – comparability of ProRes RAW versus R3D at ISO800, near clipping highlights
First, we have to discuss a phenomenon we noticed after we began with ProRes RAW after which converted to R3D, utilizing the very same publicity settings (ISO, lens iris, and shutter pace + studio mild setting):
Exposing for ProRes RAW within the typical approach, therefore having the brow of Nino as shut as potential to clipping, we get the next waveform – leading to 4 stops above base publicity. You’ll be able to see, all the colour channels are intact, whereas the white paper on the left is partially clipped:
Waveform of 4 stops overexposure with ProRes RAW (ungraded)Remodeled to REC709 and pushed again to base publicity.
Now, leaving the publicity precisely the identical and switching the digicam to REDCODE RAW NE, we get the next waveform (ungraded):
Waveform of REDCODE RAW NE on the similar publicity as ProRes RAW – spotlight restoration is lively, unnecessarily merging the pink and inexperienced coloration channelsHighlight restoration in motion with R3D on the similar publicity as with ProRes RAW.
Therefore, as we’re near clipping on Nino’s brow, we appear to be already within the area of RED’s spotlight restoration, which mixes the pink and inexperienced channels and thus destroys the pores and skin tones (rendering them yellowish). Additionally, publicity values are mapped to decrease code ranges in Log3G10 and RED WideGamut RGB for compatibility with the RED workflow.
It is a little bit of a wierd implementation of the built-in spotlight restoration characteristic. It ought to solely change into lively if one of many coloration channels is already clipped! However as may be seen within the ProRes RAW implementation, on Nino’s pores and skin, all coloration channels needs to be intact at this publicity degree.
Therefore, this truth reduces the spotlight latitude of REDCODE RAW NE by about 1 cease when in comparison with ProRes RAW (and in addition Nikon N-RAW). We have to decrease the sensor publicity (studio mild ranges) to get out of the spotlight restoration area and maintain the pores and skin tones intact. It will positively penalize shadow restoration. Therefore, I don’t suppose we’re getting the total potential with this implementation of R3D within the Nikon ZR.
We reached out to Nikon, and so they responded within the following approach:
“What you’ve got confronted is as we now have designed. The reason being distinction of spotlight clipping factors between N-Log and R3D NE. Due to this fact, while you shoot utilizing R3D NE, you have to set the publicity in accordance with the clipping level of R3D NE. In addition to, to prioritize lowering noise in darkish areas, spotlight has been shifted one cease to low mild … The precise picture sensor sensitivity differs by one cease between R3D NE and different codecs.”
Therefore, please concentrate on this truth. Wanting on the base publicity of all 3 RAW codecs, we are able to see R3D and N-RAW look very comparable, whereas ProRes RAW is a little more greenish:
Base publicity of all three RAW codecs
In abstract, to forestall spotlight restoration in motion (utilizing the decrease mild publicity ranges) throughout the RED Log3G10 gamma curve, this can lead to 2 stops overexposure above base as the utmost for R3D, and 4 stops above base for ProRes RAW and N-RAW. Inside the R3D workflow, that is according to different cameras just like the KOMODO-X and V-RAPTOR [X] VV 8K.
Don’t confuse this with a lack of 2 stops spotlight latitude – code values are remapped to decrease ranges within the R3D workflow. We’re speaking a couple of lack of 1 cease within the highlights in comparison with ProRes RAW or N-RAW on the similar publicity degree to forestall spotlight restoration for correct pores and skin tones.
Now, shifting to three stops of underexposure (i.e., 7 stops of publicity latitude), we get the next photos (2 over to five beneath = 7 stops for R3d, 4 over to three beneath = 7 stops for ProRes RAW). We are actually using the default “chroma noise discount” within the digicam RAW tab of DaVinci for R3D:
In each photos, we’re already seeing heavy noise, and the R3D picture will get barely pinkish, whereas the ProRes RAW picture will get greenish. Let’s apply noise discount:
They each clear up properly. Now let’s have a look at 8 stops of publicity latitude, particularly 4 stops beneath for PRR and 6 stops beneath for R3D:
Once more, however extra pronounced: R3D retains the colours general in a greater form, however shadows change into extra pinkish, whereas ProRes RAW general loses coloration accuracy and shadows change into greenish. Let’s see if we are able to clear this up with NR:
Now, it turns into troublesome to use the correct quantity of noise discount with out shedding an excessive amount of decision and getting that plasticky look. R3D nonetheless retains the colours considerably intact (though shadows have this cyan/pinkish coloration solid), whereas ProRes RAW loses all colours, and pores and skin tones change into orangey/greenish.
N-RAW seems like this, making use of noise discount – coloration illustration is a bit higher than ProRes RAW, however the picture additionally turns greenish:
We’re at 8 stops of publicity latitude, and issues are positively very borderline right here. The shifting photos of all three RAW codecs are literally probably not usable anymore, as bigger blotches of chroma noise are hovering over the picture.
Only for the sake of completion, listed below are R3D and PRR at 9 stops of publicity latitude:
This can’t be saved in publish, take a look in any respect three RAW flavors right here, utilizing heavy noise discount:
All photos are damaged past restore, however R3D and N-RAW behave higher than ProRes RAW. The place the one turns into pinkish, the opposite turns into greenish. Therefore, 8 stops of publicity latitude is the restrict! I feel there would positively be further potential for the R3D implementation, as a result of thus far, it penalizes spotlight stops as described above.
Placing this into context, the Panasonic LUMIX S1 II managed 10 stops of publicity latitude (Lab Take a look at right here) with DR Increase “ON” (9 with DR Increase “OFF”), albeit on the expense of a excessive rolling shutter. I actually urge you to take a look on the LUMIX S1 II outcomes once more. You’ll be able to clearly see how significantly better the LUMIX S1 II behaves within the publicity latitude self-discipline! The Sony A9 III (Lab Take a look at right here) with the worldwide shutter sensor exhibited 9 stops of publicity latitude, simply utilizing the 10-bit inside compressed codec! And naturally, the king when it comes to latitude remains to be the ARRI ALEXA 35, exhibiting 12 stops of publicity latitude.
Nikon ZR CineD Lab Take a look at – abstract
The Nikon ZR is a really well-rounded package deal at this value level from a picture high quality perspective. Superb (low) rolling shutter values, common however OK dynamic vary figures, and a mean 8 stops of publicity latitude put it within the center floor of full-frame shopper cameras. You do get a lovely-looking picture with REDCODE RAW NE inside recording and RED coloration science (though we see some potential for enchancment right here, as talked about above), and in complete, three inside RAW taste choices to select from, with the bonus of an in depth match with different RED cameras when utilizing the R3D workflow.
Are you interested by utilizing the Nikon ZR to your work? Have you ever shot with one already? Tell us within the feedback under.
