Jump to content
dreifish

Full Frame Upgrade for Hybrid Shooters - Sony A1 vs. Canon R5 C vs Nikon Z9 (AMA)

Recommended Posts

Posted (edited)

I'm upgrading from a GH5 camera to full frame. I shoot both video and stills, so I'm looking for a true hybrid camera option. 8k (and preferably at 60fps) is important to me for future-proofing stock footage. 

Through a series of incidents, I ended up in the possession of a Sony A1 (with EF 16-35F4 image-stabilized lens), Canon R5 C (with RF14-35F4 and RF15-35F2.8 image stabilized lenses) and Nikon Z9 (with Nikkor Z 14-30F4 lens), not image stabilized. I No underwater housings yet, but I figure I can put them through a series of top-side tests to figure out which is the most capable solution for an underwater shooter looking for a true hybrid camera. And since I'm doing the tests, why not share my results in a series of posts and invite you guys to suggest other tests as well. So let's begin!

IMG_2794.thumb.jpg.10d4983642dafd68284634278f24c400.jpg

Image Stabilization

I was worried that I'd be giving up something with the R5 C because it lacks IBIS. But for handheld shots and panning with the two lenses I have (both which come with lens stabilization), I didn't really see much difference as between the three cameras. Canon's lens stabilization seemed to make up for the lack of IBIS in the R5 C. The Z9's IBIS did just fine despite the 14-30 not having lens image stabilization. Overall, for non-moving handheld shots on land either trying to hold a particular framing or panning around, there was no meaningful difference in the results. All three were good. 

Video Codecs - File Size

With 8k resolution comes massive file sizes. So which camera offers the best compression options when it comes to RAW and h.265 codecs? For reference, the GH5 shoots 4k 4:2:0 60p at 150mbps. 

1. Raw Codecs - 8K60p

The Nikon Z9

  • N-RAW at 8.3k60P (High Quality) - 5780mbps or 12 minutes on 512gb card
  • N-RAW at 8.3k60P (Normal Quality) - 3740mbps or 20 minutes on 512gb card
  • N-RAW at 8.3k30P (High Quality) - 2890mbps or 24 minutes on 512gb card
  • N-RAW at 8.3k30P (Normal Quality) - 1740mbps or 40 minutes on 512gb card

The Canon R5 C

  • Canon RAW LT at 8k60P - 2570mbps or 27 minutes on 512gb card
  • Canon RAW ST at 8k30P - 1980mbps or 34 minutes on 512gb card
  • Canon RAW LT at 8k60P - 1290mbps or 53 minutes on 512gb card

Sony A1 doesn't offer any 8K raw options. So the winner is clearly the Canon R5 C, with RAW LT files that are 75% the size of the Nikon normal quality files. I'll update this with info on how easy the raw codecs are to edit tomorrow.

2. h.265 LongGOP Codecs - 8K30p

None of the cameras offer an 8k60p non-raw codec. The R5 C and Z9 should in theory be able to film at those frame rates, so the limitation appears to be the MP4 compression pipeline. The A1 can't film at 8k60, only 8k30. 

The Nikon Z9

  • H.265 422 10 bit at 8k30 - 400mbps or 171 minutes on 512gb card

The Canon R5 C

  • HEVC 422 10 bit - 8k30 (DCI) - 540mbps or 126 minutes on 512gb card - works with a V60 SD card. 
  • HEVC 420 10 bit - 8k30 (DCI) - 400mbps or 171 minutes on 512gb car

The Sony A1

  • XAVC HS 8K 422 10 bit - 8k30  - 520mbps (131 minutes on 512gb card)  OR 260mbps. For some reason, the 520mbps requires a CFEXPRESS TYPE B or V90 SD card. Won't work on a V60 card.
  • XAVC HS 8K 420 10 bit - 8k30  - 400mbps or (171 minutes on 512gb card) OR 200mbps  The 400mbps works with V60 card. 

The A1 has the most compression options, including a codec that's half the size of the R5C and Z9 codecs. But the A1's weakness is that the highest quality 540mbps codec  requires a V90 SD card, whereas the other two can record this format on cheaper (per GB) CFEXPRESS Type B cards.  Thus.. no clear winner, in my view. If your goal is to have the smallest 8k30 files though, the A1 wins with file sizes 50% the size of those coming off the Z9 and R5 C. 

3. h.265 LongGOP Codecs - 4K60p

The Nikon Z9

  • H.265 422 10 bit at 4k60 - 340mbps or 201 minutes on 512gb card

The Canon R5 C

  • HEVC 422 10 bit -4k60 (DCI) - 225mbps or 303 minutes on 512gb card
  • HEVC 420 10 bit - 4k60 (DCI) - 170mbps or 402 minutes on 512gb card
  • XF-AVC YCC 422 10 bit 4k60 (DCI)  - 260mbps or 232 minutes on 512gb card

The Sony A1

  • XAVC HS 4K 422 10 bit - 4k60  - 200mbps OR 100mbps (341 or 683 minutes on 512gb card)
  • XAVC HS 4K 420 10 bit - 4k60  - 150mbps OR 75mbps OR 45mbps 

 

Here, the Sony A1 has the most compression options, but comes last in terms of image quality because the 4k60 uses line-skipping, whereas both the Z9 and the R5 C downsample the full 8k sensor to 4k when filming at 60 fps. The Canon R5 C is the winner for most usable 4K60 LongGOP format, allowing for file sizes 66% the size of those from the Z9. 

In the next post, I plan to cover Subject Tracking in Video, Animal Eye Detect and whether it works with Fish, Manual White Balance and battery life. But please ask me anything -- happy to run other tests if I can. 

 

Edited by dreifish
  • Like 1

Share this post


Link to post
Share on other sites

Sounds like an interesting test.  The obvious question to me how about testing with macro lenses?  This is probably where IS and auto focus will be put to the test more so than wide angle lenses.

Share this post


Link to post
Share on other sites
1 hour ago, ChrisRoss said:

Sounds like an interesting test.  The obvious question to me how about testing with macro lenses?  This is probably where IS and auto focus will be put to the test more so than wide angle lenses.

=> Testing C-AF of a small, central, field with subsequent trekking for re-framing (in plenty of light and low light) would be a great test (I think the fancy algoritms for subject, face, animal etc. detection may not be so useful for UW)...

With establishes macro lenses (Nikon 105, Canon EF and RF 100, Sony 90, Sigma 105 (for Sony)). Comparison with GH5 and Zuiko 60mm or Pana 45mm will give a good reference for many of us...

I know, I am asking too much, but you asked for suggestions...

 

Wolfgang

 

Share this post


Link to post
Share on other sites
6 hours ago, ChrisRoss said:

Sounds like an interesting test.  The obvious question to me how about testing with macro lenses?  This is probably where IS and auto focus will be put to the test more so than wide angle lenses.

Unfortunately I only have the Sony 90mm macro. No Canon RF or Nikkor Z macro lenses, or I'd be happy to test. If anyone in South Florida is willing to lend me a Canon or Nikon macro, I'd be happy to team up to test them :)

But I do plan to test animal eye detect at 30-35mm with the wide angle lenses, and object tracking in video for tracking fish (using underwater videos on YouTube as 'targets', hah). The object tracking implementation appears quite different across the three systems, with Canon's cinema operating system in particular widely perceived as more limited than their photo interface (and the interface offered by Nikon and Sony). So this might be a meaningful point of differentiation between the different systems. 

Share this post


Link to post
Share on other sites

What happens if you find that one of the three is best for video, e.g., the Canon, another of the three is best for stills, e.g., the Nikon, while the third one is best for swimming  with (because it is the smallest and should have the smallest housing), e.g., the Sony? Be aware that Nikon just updated the FW for the Z9 yesterday to improve the eye AF. Not using FW 2.1 would invalidate your conclusions regarding AF.

I also suspect that using a video image to test AF, tracking, etc.will have issues due to nature of the electronic image. It might be better to go to a public aquarium that has a huge tank and shoot there especially if the room is darkened. Keep in mind you will be shooting through a flat port so best try to be perpendicular. It might need three trips doing one camera at a time unless you like carrying that much around with you.

  • Haha 1

Share this post


Link to post
Share on other sites
Posted (edited)
3 hours ago, Tom_Kline said:

What happens if you find that one of the three is best for video, e.g., the Canon, another of the three is best for stills, e.g., the Nikon, while the third one is best for swimming  with (because it is the smallest and should have the smallest housing), e.g., the Sony?

It would be quite the conundrum (for me). But at least we'll document that so others know the various strength and weaknesses of each system. Truthfully, the only reason the Sony A1/A7S III are in the discussion for me is because they have unique access to the WWL-1 and smaller housings (vis a vis the Z9 and R5 C.) They would make a more compelling package for travelling, but perhaps they give up too much in other areas. 

3 hours ago, Tom_Kline said:

 Be aware that Nikon just updated the FW for the Z9 yesterday to improve the eye AF. Not using FW 2.1 would invalidate your conclusions regarding AF.

Yep, Z9 has v 2.1 firmware installed as of yesterday and A1 has v. 1.3 firmware with the new 8k 4:2:2 codec. 

3 hours ago, Tom_Kline said:

I also suspect that using a video image to test AF, tracking, etc.will have issues due to nature of the electronic image. It might be better to go to a public aquarium that has a huge tank and shoot there especially if the room is darkened. Keep in mind you will be shooting through a flat port so best try to be perpendicular. It might need three trips doing one camera at a time unless you like carrying that much around with you.

This is why I'm not testing for focusing speed (no depth to video images anyway) but merely the algorithm's ability to recognize underwater subjects (when it comes to Animal Eye autofocus) or fish/fish eyes as a discrete subject (when it comes to general object tracking like Nikon's 3d tracking, Canon's 'tracking' and Sony's Real Time Tracking.) All these algorithms are doing is recognizing objects. An aquarium would be a better test indeed because it introduces the element of depth as an additional data point for the algorithm to use (if it does). Maybe I'll visit one on the weekend.

Edited by dreifish

Share this post


Link to post
Share on other sites
Posted (edited)

Good points. I would be concerned if there was aliasing between the AF and the raster image of the video. For example if there is a moire pattern, how would the eye detect algorithm deal with that?

There is a moire pattern when shooting images of my high rez (6K) computer monitor which I do when following instructions on a pdf manual and playing with the settings - great to do on rainy days.

Edited by Tom_Kline

Share this post


Link to post
Share on other sites
Posted (edited)

Subject Tracking Autofocus

For this test, I used the excellent "Mucky Secrets" YouTube documentary from Bubble Vision and tried to get the tracking algorithms to recognize and follow the various fish, critters and nudibranchs of the Lembeh straight. 

1. Sony A1

The A1 has real time tracking (in photo mode) which advertises the ability to identify animal and bird eyes. I tested this with various fish and critter targets. It doesn't work. In photo mode, it also has a general object tracking  (similar to Nikon's 3d tracking) which you can select from the Focus Mode by by picking "Tracking: Center Fix" or "Tracking: Spot M". You engage tracking by pressing the AF-ON button, so it doesn't require the touchscreen, which means it can be used in a housing. This worked quite well with entire fish or their eyes (if large enough in the frame). 

As far as I can see there is no animal eye AF in video mode. Apparently only the newer A7IV has it. But there is object tracking system, which works fine for selecting eyes or entire fish. However, the only way I could engage it was with the touch screen, which obviously doesn't work underwater. I wonder if there's a way to map activation to one of the custom buttons.

2. Nikon Z9

In photo mode, the Z9 has 3d tracking and animal + bird eye detect. Like the A1, the animal eye detect doesn't seem to distinguish fish. However, the 3d tracking can be activated and the subject selected by placing the rectangle over a fish or its eye and pressing the AF-ON button. Like with the A1, this works very well for tracking the fish or its eye.

No 3d tracking in video mode. Subjects could be selected and tracked just like on the A1 by touch focus. This worked well on fish eyes and entire fish. You can also set the AF-area mode to Subject Tracking AF. I this mode, you get a single central spot that you can place over a target and initiate tracking using the AF-ON button. You can then release the subject by pressing the OK button. Overall, seems to work the same as 3d tracking in stills mode and can reliably track fish eyes or fish. 

3. Canon R5 C

In Photo mode, the R5 C can set AF method to Face + Tracking and Subject to detect to Animals. With these settings, the R5 C sometimes would recognize the eye of fish even if the target was set elsewhere. It's the only one of the three that can recognize eyes of sub-aquatic animals. Moreover, you can place the target square over an entire fish or its eye and initiate subject tracking with AF-ON. This works as well as with the A1 or Z9. 

In Video mode, the interface is different. There is no automatic detection of animal eyes, but if you set the AF Frame to "Whole Area", you can initiate subject tracking by touching the object on the screen. Alternatively, you can configure a custom key to "Tracking". Pressing the custom key will then bring up a "+" target on screen which you can place over the fish or its eye. Subsequently pressing "Set" will start tracking whatever object the "+" is on top of. You can stop tracking by pressing cancel. This seems to work a pretty well, but is a bit more clunky to activate than the Nikon Z9's Subject Tracking AF in video mode.

Winner - Canon R5 C for photos since its Animal Eye Detect actually finds fish eyes. For video, Z9 wins by a hair over the R5 C because subject tracking is easier to initiate. But both are capable. The Sony A1 is the least capable with no way to activate subject tracking underwater AFAIK. 

Edited by dreifish
  • Like 2

Share this post


Link to post
Share on other sites
Posted (edited)

Z9 3D tracking mode works differently than having the camera find the subject within a designated area (this area is quite variable and is selected separately from detection mode). Doing vids, 3D might be preferred if you are able to see your subject at the start. I tend to have the camera find the subject for me using one of the rectangles (from FW 2.0 on)  but pick where in the frame to pick from. For example I picked a narrow horizon box so the AF picked a bird from the row of birds that I selected - keeping the frame over the row if I recomposed. Birds were entering and departing by flight from each row (resting shorebirds). Worked well for that.

Edited by Tom_Kline

Share this post


Link to post
Share on other sites

Thanks for doing these tests Dreifish!

Following...  :D

Share this post


Link to post
Share on other sites
Posted (edited)

Can you Edit The Video Codecs?

First, I tested this on my 2018 15" Macbook pro ( 2.2 GHz 6-Core Intel Core i7, 16gb ram, Radeon Pro 555X 4 GB). I have a 2022 M1 Max Macbook Pro on order, so I'll repeat the tests with the more powerful system later.

Nikon Z9

  • 8k N-RAW would play back at no more than 16-17 fps in Davinci Resolve, whether decoded in full, half, or quarter resolution
  • 8k h.265 8k30p 422 10-bit plays back at almost but not quite real time (30fps)
  • 4k h.265 4k60 422 10-bit plays back in real time most of the time (60fps) but stutters once in a while
  • Raw files allow for white balance to be pushed up to 50000k and +150 tint
  • Raw files with Custom White Balance set in camera start out at no more than 10000k, suggesting in-camera custom white balance maxes out at 10000k.

Same white wall lit up with a 5000k video light through a Lee 353 Lighter Blue gel, First, with CWB done in camera. Second with WB deliberately set wrong at 3000k in camera and adjusted to 10000k to match the in-camera CWB during raw decode.  No matter how I adjusted the color temperature and tint, despite the wider range than what's available on the Canon RAW files, I could not neutralize the Lee 353 gel in post. Whereas both the Canon and Nikon were able to neutralize the gel almost perfectly when setting CWB in camera.

1736990795_ScreenShot2022-07-07at9_04_09PM.thumb.png.fea6d7d68d5e4c1383417ce4668fd55b.png

1082167352_ScreenShot2022-07-07at9_05_17PM.thumb.png.7a33cae980700738cb773f057e4dc201.png

Canon R5 C

  • 8k Canon RAW LT & ST plays back at 17-18 fps in Davinci Resolve if decoded at Full Res (Resolve) or 2-3 fps if decoded at Full Res (Canon). Decoding at Half Res is still 17-18fps but going to Quarter Res gets you ~26fps. Still not editable (unless filming at 24p I guess)
  • 8k30p h.265 in 422 10-bit plays back at 13-14fps
  • 8k30p h.265 in 420 10-bit plays back at 24fps
  • 4k60 h.265 in 422 10bit plays back at ~20fps
  • 4k60 h.265 in 420 10bit plays back at ~29fps
  • 4k60 XF-AVC in 422 10 bit plays back at 29fps.
  • Raw files only allow for white balance in raw decoding to be pushed up to 15000k! and +150 tint. 
  • Raw files with Custom White Balance set in camera also seem to max out at 15000k and +150 tint. However, the results are not the same! If I set the Custom White Balance in camera and shoot in raw, vs shooting in raw and putting in the same settings again in post, the CWB in camera achieves better results with more neutral white balance. This suggests that with the Canon raw files, some of the white balance setting are nonetheless baked in! 

Same white wall lit up with a 5000k video light through a Lee 353 Lighter Blue gel, First, with CWB done in camera. Then with WB deliberately set wrong at 2700k in camera and adjusted to 15000k during raw decode. Judge the histograms for yourself.. 943993689_ScreenShot2022-07-07at8_51_57PM.thumb.png.6880e8bcc695d4eccac5f233212a74a1.png 

1905024456_ScreenShot2022-07-07at8_52_09PM.thumb.png.67af8fb1f2dba7c57bace3e5cfb88898.png

Sony A1

  • 8k30p h.265 in 422 10-bit plays back at 22 fps
  • 8k30p h.265 in 420 10-bit plays back at 32 fps
  • 4k60 h.265 in 422 10bit plays back at ~35 fps
  • 4k60 h.265 in 420 10bit plays back at 60fps -- only editable format!
  • 4k120 h.265 in 422 10 bit plays back at 29fps.
  • 4k120 h.265 in 422 10 bit plays back at ~27fps.
  • 4k120 h.265 in 42010 bit plays back at ~52fps.
  • CWB in camera was able to neutralize the Lee 353 Lighter Blue gel just fine:

1531851998_ScreenShot2022-07-07at9_35_18PM.thumb.png.33c16c62c3c4c3fe53825f34a42999d6.png

Conclusion -- wow, terrible. I would say none of the Z9 or R5 C codecs is really editable without proxies on an older machine. Of the A1 codecs, only the 4k60p in 4:2:0 10-bit was editable. Guess if you're getting any of these cameras, budget for a newer computer, too!

Next up -- a deeper exploration CWB's ability to neutralize various shades of green/blue introduced using Lee and Rosco gels to simulate underwater ambient light at depth. Will test both with CWB set in camera and set in post for the Canon R5 C and Z9 when shooting raw to see how much of the white balancing you can leave to post with these 'raw' video formats.  

 

 

Edited by dreifish
  • Like 1

Share this post


Link to post
Share on other sites

The current generation of hardware and graphic cards digest 4k60 very well in either prores raw (mac) or even all intra codec

In water as none of the RAW has auto white balance you will have to white balance anyway so personally I do not see a significant benefit of RAW as cameras operate differently in video than still

Also from the bitrates I see here only the Nikon seems to be nearing the bitrate of ProRes RAW I doubt the canon files can really be pushed

I am sticking to 4k for the foreseeable future. I can shoot 5.7K raw with the GH6 but actually I use all intra 800 mbps for 60fps and works very well

Share this post


Link to post
Share on other sites
7 hours ago, Interceptor121 said:

The current generation of hardware and graphic cards digest 4k60 very well in either prores raw (mac) or even all intra codec

Yes, and the newest Intel and Apple processors / NVIDEA/AMD graphics cards also have dedicated video decoders/encoders for most common formats, including h.265 & h.264 LongGOP. I expect editing both the 4k120 and 8k30 LongGOP files to be non-issues on newer hardware. But it's a challenge for this 4-year old Macbook. Nikon made a big deal about their N-RAW video format being easy to edit. Doesn't seem to be the case (or any different than the Canon RAW) in my testing so far. I expect Prores Raw may indeed be a bit more edit friendly, but only the Z9 allows for filming that internally, and only in 4k. It also according to other reports has quality issues in the downscaling. I will test that codec as well in my next batch of tests to see how it handles white balance adjustments in post.

8 hours ago, Interceptor121 said:

In water as none of the RAW has auto white balance you will have to white balance anyway so personally I do not see a significant benefit of RAW as cameras operate differently in video than still

All the cameras allow for auto white balance when shooting video. But I think your point is correct -- from everything I've seen so far, leave the white balance on auto and shooting in N-RAW or Canon RAW for video with the idea that you'll be able to correct it in post is not a good strategy. The "raw" video files don't have the same flexibility for WB correction in post as RAW stills do. You'll still achieve the best results by doing a CWB in camera whether shooting in a "raw" video format or an h.265 codec. I will test this more extensively. Quite disappointing if true, and it does eliminate the biggest advantage of a RAW video workflow in my view. 

8 hours ago, Interceptor121 said:

Also from the bitrates I see here only the Nikon seems to be nearing the bitrate of ProRes RAW I doubt the canon files can really be pushed

Why would you conclude there isn't sufficient data to push the files around? As you'll see below, the Canon RAW LT files still have 2.5x the information of the h.265 files, even not accounting for the inherent efficiencies of compressing RAW RGGB sensor data rather than debayered data. I haven't come across any reports suggesting the Canon RAW LT files are too compressed.

I actually think the smaller bit rates from the Canon RAW are an advantage.  ProRes RAW, like ProRes, may be optimized for requiring limited processing resources to edit, but that also likely means it doesn't have the most efficient compression algorithm. Just like the h.265 4.2.2 10-bit codecs end up being able to retain the same image quality as ProRes at much lower bit rates.

For what it's worth, Canon 8k60p RAW LT appears to use around 10:1 compression.

8K60p Uncompressed raw = 8192 x 4320 pixels x 12 bits per pixel x 60fps = 25,480mbps. So based on that rough math, the compression ratios are:

  • Nikon N-RAW High Quality = 4.4:1
  • Nikon N-RAW Standard Quality = 6.8:1
  • Canon Raw ST = 6.4:1
  • Canon Raw LT = 9.9:1
  • Prores Raw = ~3.8:1 
  • ProRes 4:2:2 IntraFrame = ~5.4:1
  • 8K30p h.265 LongGop (available on the Z9, A1 and RC5 ) = 24.5:1 or 50:1 when using the A1's 260mbps 8k30 codec. 
  • 4k60p h.265 LongGop (downsampled from the 8k full sensor readout on the Z9 and RC5) = 75:1 compression on Nikon Z9; 113:1 compression on the Canon R5 C

What does this tell us? The raw codecs are anywhere from 2.5x (Canon RAW LT) - 6x the size of the h.265 LongGOP formats. You have to budget for a hefty premium in storage needs if adopting a raw workflow. I think the 10:1 compression on the Canon RAW LT is the most usable of the lot (requires the least storage). The Prores Raw 4:1 and Nikon N-RAW High Quality 4.4:1 compression are too light to be useful because of the massive file sizes.

If you look at what other companies are doing, Blackmagic with the Ursa Mini 12k offers 5:1, 8:1, 12:1 and 18:1 compression in BRAW.  According to them:

  • 5:1 is perfect for effects heavy feature film and commercial work, where you need the highest possible quality.
  • Blackmagic RAW Q5 and 8:1 are extremely high quality making them great for episodic television and independent films.
  • Blackmagic RAW quality is so high, you can even use 12:1 and 18:1 settings for incredibly small files that are suitable for all kinds of uses such as broadcast news, live sport events, digital signage, A/V, corporate video, weddings and more!

 The RED Raptor offers slightly lower compression ratios with RedCode Raw in three flavors:

  • - HQ (~3.5:1 compression ratio) - Very detailed scenes, VFX, still workflows
  • - MQ (~4.3:1 compression ratio) - For Cinema and Higher End Series featuring not VFX heavy content
  • - LQ  (~7.5:1 compression ratio) Ideal for television, online, documentary, interviews, live events
9 hours ago, Interceptor121 said:

I am sticking to 4k for the foreseeable future. I can shoot 5.7K raw with the GH6 but actually I use all intra 800 mbps for 60fps and works very well

This is a sensible position given the data rates required to shoot in the RAW formats. If they indeed don't offer any advantages when it comes to adjusting white balance in post, then 4k60p downsampled from a 6k or 8k sensor starts to look much more practical in terms of storage needs and workflow. Filming in 4k60p on the R5 C gets you all that data from the 8k sensor compressed down ~113:1. So you get file sizes 1/10th the size of Canon RAW LT. 

Share this post


Link to post
Share on other sites
Posted (edited)

White Balance Flexibility for RAW Photos

I don't have any housings, so I tested this by illuminating some illustrative scenes with video lights shot through various filters to simulate gradually cooler ambient light at different depths. This is the setup:

IMG_2798.thumb.JPG.9baf928e7bb82e8eec5a604bbb97cbab.JPG

In the rows below, the order from top to bottom of the images is A1 -> R5C -> Z9. DM me if you want raw files to play around with.

With no filter (~5000k)

SonyA15000k.jpg.0c88e13f465615172d087514f1b57bfb.jpg

CanonR5C5000k.jpg.27b3b5fb130a35649c7aa3432ce3924b.jpg

 

NikonZ95000k.jpg.6e076790eed3704efd317af4326c4818.jpg

Lee 353 Filter (~20000k, +100 Magenta)

SonyA1Lee353.jpg.0f8e17e451a544668291591e54fed885.jpg

 

 

CanonR5CLee353.jpg.68ce0f5336b5e46ac9019f491f50dea5.jpgNikonZ9Lee353.jpg.b1d167afe429a5956caa6ab3f3a1e76b.jpg

Lee 353 Filter Stacked w/ Rosco 4330 Filter (~25000k, +120 Magenta)

SonyA1Lee353Rosco4330.jpg.25ef7f11a23676966d391571f2ed6e94.jpg

 

CanonR5CLee353Rosco4330.jpg.f5a1deb3b4d6f08f5d9ddd3367cf3a47.jpg

NikonZ9Lee353Rosco4330.jpg.89e83d59ad31c5366ef80dcdf67c3c80.jpg

Lee 724 Filter (~30000k, +80 Magenta)

SonyA1Lee724.jpg.77e83f774f1c572f4500ebd36f496911.jpg

 

CanonR5CLee724.jpg.a78ccf4b51c8e8a1e2bad32fdb7612df.jpg

NikonZ9Lee724.jpg.c03e7800b80528dc3fb012d8199d293d.jpg

Lee 724 Filter Stacked w/ Rosco 4330 Filter (43000k, +110 Magenta)

SonyA1Lee724Rosco4330.jpg.0c0e85cc7312ab45706b127ca8d12ab0.jpg

CanonR5CLee724Rosco4330.jpg.d70f6caa9c3912a2e705f24f357504c4.jpg

NikonZ9Lee724Rosco4330.jpg.7e266c9ffe4cc9c7ed3a206f35fcf1bc.jpg

Lee 724 Filter Stacked w/ Rosco 4360 Filter (50000k, +125 Magenta)

SonyA1Lee724Rosco4360.jpg.313f7f238d2964612539cbd2eb813ba8.jpg

CanonR5CLee724Rosco4360.jpg.6a9bbb752976e039b7573008856794c6.jpg

NikonZ9Lee724Rosco4360.jpg.cf5e3b319447f9baa7efa10f8fbcc0df.jpg

The results are remarkably similar through the first 6 tests, though the specific kelvin temperature and tint used to achieve them did vary a fair bit in Lightroom. In general, the Nikon needed the lowest kelvin temperature, and the Sony the highest. Canon sat somewhere in the middle.  The other thing I noticed is that the Canon seems to produce slightly more saturated blues. 

The last test was with a Lee 172 Filter (~50000k, +150 magenta).

 

SonyA1Lee172.jpg.3b1ce1a1e88b2b388165f600c514936c.jpg

CanonR5CLee172.jpg.2156b75291c7e55b37e87df883ff8f00.jpg

NikonZ9Lee172.jpg.a5ba86a4afebf631aca8e0910056e71d.jpg

None of the cameras did great in this very challenging light. But I'd say the Canon R5C was the best of the lot, retaining a blue water column, and the Nikon Z9 clearly the worst, turning the water column purple. (The Nikon Z9 could not set an in-camera CWB in this light, whereas both the Sony and Canon could. But more on that later when we get to WB in video).

SonyA1Lee724Rosco4360.jpg

Edited by dreifish
Removing extra photo.
  • Like 1

Share this post


Link to post
Share on other sites
Posted (edited)

Manual White Balance in Camera For Video

All three cameras allow setting CWB off a neutral target like an 18% grey card fairly easily. The Canon has a true one-touch white balance, but the downside of this is that it uses the entire frame for the white balance. The A1 and the Z9 allow you to set the white balance off a smaller portion of the screen but take a few more button presses to get there.

I tested the results using the same setup as for the raw stills, but doing a white balance in camera. On the Nikon Z9, I filmed in 8k N-Raw, and 4k Prores Raw and Prores 4:2:2 in the Flat picture profile. On the Sony A1, I filmed in 8k 4:2:2 10-bit in Picture Profile 11 (S-Cinetone). On the Canon R5C, XF-AVC 4k 4:2:2 in 709 Wide DR.

Results are presented Sony A1 -> Canon R5C -> Nikon Z9

Lee 724 Filter Stacked w/ Rosco 4360 Filter (50000k, +125 Magenta)

SonyA1Lee724Rosco4360Still_1.4.1.thumb.jpg.6bfc5a3527dc530ba9f8d153fd1cd475.jpg

CanonR5CLee724Rosco4360Still_1.3.1.thumb.jpg.9574992113c9170894b2c80f304e97cd.jpg

NikonZ9Lee724Rosco4360Still_1.6.2.thumb.jpg.71bfe1860e406a191dc713517a89679c.jpg

 Both the Sony and the Canon are still holding up quite well, but the Sony is more contrasty and saturated. I rather prefer the lighter blue rendition on the Sony, it seems more saturated. Thoughts? To me, the Nikon produces the worst result.

Lee 172 Filter ( ~50000k, +150 magenta).

 

SonyA1Lee172Still_1.4.1.thumb.jpg.2bbfaf03f73621aa046a6b1b456c832f.jpgCanonR5CLee172Still_1.1.1.thumb.jpg.8b1f82a521a869532d577a8c96328004.jpg

NikonZ9Lee172Still_1.5.1.thumb.jpg.b4da845be3c7650b0a1fb6ab6cd272af.jpg

None of these great, but the Sony A1 file clearly (and surprisingly) looks better, no? The Nikon z9 refused to set a custom white balance, it's outside its range in this challenging light.. 

Surprisingly, I'd say Sony A1 wins the video manual white balance battle. Canon comes 2nd. Nikon also is fairly capable, but breaks down at the edge cases and to me at least doesn't look as good as Sony or Canon. 

Edited by dreifish
  • Like 2

Share this post


Link to post
Share on other sites

Battery Life and Recording Times

For this test, I set all cameras to 4k60p and recorded until the battery run out. 

  • The Sony A1 lasted 1 hour and 45 minutes of recording.
  • The Canon R5C lasted 45 minutes of recording (with the Nauticam housing USB-C battery pack, it should be 4.5 hours total)
  • The Nikon Z9 with it's larger battery got to 2 hours and 35 minutes.
  • Like 2

Share this post


Link to post
Share on other sites
Battery Life and Recording Times
For this test, I set all cameras to 4k60p and recorded until the battery run out. 
  • The Sony A1 lasted 1 hour and 45 minutes of recording.
  • The Canon R5C lasted 45 minutes of recording (with the Nauticam housing USB-C battery pack, it should be 4.5 hours total)
  • The Nikon Z9 with it's larger battery got to 2 hours and 35 minutes.


There are many misconceptions on how a battery pack works
If a pack has 19200 mah it means it holds that amount however when you look at what it can deliver the efficiency drops depending on the output current
You can expect around 65-70% on average so your 4.5 hours will be more like 3 to 3h15m
This is still a very good time and you will know where the external pack is dead so you can decide to replace it if you have another one but of course you need to open the housing


Sent from my iPhone using Tapatalk

Share this post


Link to post
Share on other sites
21 hours ago, Interceptor121 said:

 


There are many misconceptions on how a battery pack works
If a pack has 19200 mah it means it holds that amount however when you look at what it can deliver the efficiency drops depending on the output current
You can expect around 65-70% on average so your 4.5 hours will be more like 3 to 3h15m
This is still a very good time and you will know where the external pack is dead so you can decide to replace it if you have another one but of course you need to open the housing


Sent from my iPhone using Tapatalk

 

Might be shorter in real life. I will receive the 19200 mah battery pack tomorrow and test it to see. 

Still, I think the Nikon and Canon batteries at least should last for a full day of liveaboard diving. The A1 battery might also, since you're not really going to be filming all the time. But maybe more dicey. 

Share this post


Link to post
Share on other sites
Might be shorter in real life. I will receive the 19200 mah battery pack tomorrow and test it to see. 
Still, I think the Nikon and Canon batteries at least should last for a full day of liveaboard diving. The A1 battery might also, since you're not really going to be filming all the time. But maybe more dicey. 

When you use the camera in real life start stop focus white balance the battery life shortens around 25-35%
Hence the idea of a battery pack in the housing is appropriate


Sent from my iPhone using Tapatalk

Share this post


Link to post
Share on other sites
Posted (edited)

How well do the codecs edit on a Macbook M1 Max?

My 16" Macbook Pro 2021 with M1 Max (32 core GPU) and 32gb of ram has arrived. So I redid the playback tests for the various codecs in Davinci Resolve.

Sony A1 -- all the video codecs play back in real time, no problems.

Canon R5 C - all the video codecs, including 8K RAW ST at 24fps and 30fps play back in real time. 8K RAW LT at 60fps also plays back in 60 fps, though I've seen it stutter now and again.

Nikon Z9 -- all the video codecs, including 8K Raw at 24fps play back in real time. I no longer have a 8k60 raw file to set unfortunately.

So no issues trying to edit the raw files on a new Macbook.

Edited by dreifish

Share this post


Link to post
Share on other sites
Posted (edited)

How much latitude do you get to set WB and ISO in post if shooting "raw" video?

I wanted to revisit the latitude of the raw video codecs when it comes to adjusting white balance in post-production rather than setting a 'correct' white balance in camera at time of acquisition. For this purpose, I shot the same scene illuminated with 5000k video lights modified with various filters to achieve 15000k, 30000k and 45000k. In each case, I filmed once with CWB set in camera to match the illumination, and once with CWB set in camera to 5000k to see whether the WB could be rescued in post-processing.

White Balance Adjustments

For the Canon R5C Cinema RAW files and the Nikon Z9 N-Raw files, white balance and tint can be adjusted in the Color page under the Camera Raw tab. Canon Cinema RAW files can also be adjusted in a similar fashion using Canon's own Cinema RAW development app.

  • Nikon N-Raw allows adjustment of color temperature from 2000 to 50000 kelvin, and tint from -150 to +150 in Davinci Resolve. (there is no other software that works with N-Raw at the moment.) However, this is deceptive, as dialing in white balance in post to 50000k barely shifts the colors, unlike with the Canon Cinema Raw. Refer to the results under 45000 kelvin illumination below.
  • Canon Cinema RAW only allows adjustment of color temperature from 2000 to 15000 kelvin, and tint from -150 to +150. This is the same whether you process the files in Davince Resolve or canon's own Cinema RAW development app.
  • For the Sony A1 files, which are h.265, I performed the white balance adjustments using the white balance picker under the Primaries - Color Wheels tab of the Davinci Resolve Color page to compare what WB adjustments can be achieved in the same scenario using 12-bit "RAW" formats.

ISO Adjustments

  • Nikon N-Raw (in Davinci Resolve) does not allow you to adjust the ISO in the Camera Raw tab.
  • Canon Cinema RAW allow for ISO to be adjusted if you use the Decode Quality - Full Res - Canon setting. They don't allow ISO adjustments if Decode Quality is set to Full Res -  Resolve 

For reference, here's what the Sony A1 and Canon R5 C shots looked like illuminated with the 5000 kelvin video lights:

SonyA15000_1.1.1.thumb.jpg.1bcd27ea509b85eed4c7733aaa149037.jpg

CanonR5C5000_1.1.1.thumb.jpg.150e38339d2f420342853bf41f7f00b1.jpg

Already we can tell there are some differences in color science between the Sony and the Canon files. Sony files are slightly more saturated but less contrasty. Which color science you prefer might be a matter of debate..

And the Nikon Z9 at 5000 kelvin:

NikonZ95000_1.3.1.thumb.jpg.2b91ab3aadf0f78317afb153f98cc3ca.jpg

15000 Kelvin Illumination 

Sony A1 -- WB In Camera vs. In Post

SonyA115000CWB_1.1.2.thumb.jpg.caaef54d0878adbdf9cf9ef1aeb648d0.jpg

SonyA115000Post_1.2.1.thumb.jpg.e441c093fba22b6a84bfceeab3538490.jpg

Canon R5 C -- WB In Camera vs. In Post

CanonR5C15000_1.3.1.thumb.jpg.adaf99e8285cbfe18c874d5711f6298d.jpg

CanonR5C15000InPost_1.3.2.thumb.jpg.7f3f1e8dfb05f047dec5a6de6661a1b2.jpg

First, the Canon files are quite noisy. But this is because I accidentally underexposed them by around 2 stops and had to recover exposure in post. The custom white balance is also a bit off, because it's set based on the entire scene rather than a smaller selectable area like on the A1. So ignore that part (though, as an aside, it's worth noting that the raw files are noisy!!). I will see if I can reshoot this test later and properly expose it and white balance off a grey card, but ultimately it doesn't matter for the purpose of judging WB adjustment latitude.

Second, there is enough latitude in the Canon 12-bit raw file to adjust CWB from 5000k to 15000k and achieve identical results to setting it at 15000k in camera in the first place. The same could not be achieved with the Sony 10-bit 4:2:2 files. Here is the Davinci resolve Camera Raw tab settings and histograms so you can see the results are the same when you dial in the same Color Temperature and Tint to the second file in post.:

1831418599_ScreenShot2022-07-13at3_08_08PM.thumb.png.b83c08f2a68148f27bed4899f3d8cad6.png

1281981725_ScreenShot2022-07-13at3_07_53PM.thumb.png.a68c689885c5b5ed8b58957b25427cf1.png

30000 Kelvin Illumination

Sony A1 -- WB In Camera vs. In Post

SonyA130000CWB_1.3.1.thumb.jpg.c2441a68a3724eb4a4a0efe5fddfc37e.jpg

SonyA130000Post_1.4.1.thumb.jpg.908faceb6e23cd4f001fad9ae3099938.jpg

Canon R5 C -- WB In Camera vs. In Post

CanonR5C30000CWB_1.5.1.thumb.jpg.00b28332065e65abf7d59d9b7888714d.jpg

CanonR5C30000_1.4.1.thumb.jpg.56b6fe991b0820a82e91be003313fd14.jpg

Same result here with the Canon R5 C files -- there's enough latitude to adjust white balance from 5000k to 30000k in post and achieve identical results to setting white balance in camera. With the A1 10-bit 4:2:2 files.. not so much.

45000 Kelvin Illumination

Sony A1 -- WB In Camera vs. In Post

SonyA145000CWB_1.5.1.thumb.jpg.2a7fae187f9f9c6a45629a35ff3b41a2.jpg

SonyA145000InPost_1.6.1.thumb.jpg.09263992c2096ed88972c2f84e80c214.jpg

Canon R5 C -- WB In Camera vs. In PostCanonR5C45000_1.6.1.thumb.jpg.2a0a23c1063e5de2cad3d945276d21d0.jpg

CanonR5C45000CWB_1.6.1.thumb.jpg.3cde134f092a64f270b6c9bba5d1d0ac.jpg

Nikon Z9 -- WB In Camera vs. In Post

NikonZ945000CWB_1.1.1.thumb.jpg.7c9cbef0452e31f832741a3524442fca.jpg

NikonZ945000NoCWB_1.2.2.thumb.jpg.69cd7bb7f1df75b1e8fec5b4a9bd6ff7.jpg

Same result here with the Canon R5 C files -- there's enough latitude to adjust white balance from 5000k to 45000k in post and achieve identical results to setting white balance in camera. With the A1 10-bit 4:2:2 files.. definitely not.

Interestingly, the Nikon Z9 files also appear to preserve enough latitude to adjust the white balance from 5000k to 45000k in post, but the process is a bit more convoluted. Just dialing in 50000k and +150 tint in the Camera Raw tab doesn't do it -- I had to make white balance adjustments and to the color wheels. But the ultimate result is that I was able to get to the same end. So I'd say the Nikon Z9 raw files also work, they just need a different workflow.:

Nikon Z9 processing on file with in-camera custom white balance set:

1597038223_ScreenShot2022-07-13at5_12_47PM.thumb.png.6339f4bec2571fc42684711f7864e17c.png
Nikon Z9 processing on file with white balance dialed in in post:

234848455_ScreenShot2022-07-13at5_13_34PM.thumb.png.4877c5c9ee23ef91945e517ce39b5115.png1113664208_ScreenShot2022-07-13at5_20_13PM.thumb.png.37b183f4258479161678a8dc82e8b047.png

 

Conclusion

The Canon R5C's Cinema Raw and the Nikon Z9's N-Raw files allow you to adjust your white balance in post with similar latitude as you would get with RAW still images. This is really powerful for underwater shots. The Sony A1 meanwhile, while it does a great job with CWB when set in camera, doesn't preserve nearly the same flexibility in post if you don't get the white balance right at shooting time.

I also discovered that this flexibility is only available if you set Davinci Resolve to debayer the Raw files using Decode Quality: "Full Res -  Canon". If you use Decode Quality: "Full Res - Resolve" you can't adjust the white balance as precicesly in post. With the same file, take a look at the two different results depending on which Decode Quality setting you pick. For some reason, the green channel is amplified more when selecting the Resolve debayering algorithm. 

924968379_ScreenShot2022-07-13at3_34_26PM.thumb.png.5daa92a4368c82f7338d691384b412b9.png2047928416_ScreenShot2022-07-13at3_34_39PM.thumb.png.42c77a8942f44e884220c983eacf4bd7.png

The tradeoff is that with Full Res - Canon the files only play back at around 15 fps, while Full Res - Resolve allows real time 60 fps playback. So you probably want to keep the settings to Full Res - Resolve during editing and only render using Full Res - Canon once you're done with all your edits.

Screen Shot 2022-07-13 at 5.12.47 PM.png

NikonZ945000CWB_1.1.1.jpg

Edited by dreifish
  • Like 1

Share this post


Link to post
Share on other sites

I may have been a bit too hasty in concluding the Canon R5 C Cinema Raw files preserve the same WB latitude as setting WB in camera and setting WB in Camera is unnecessary.

On further investigation, I discovered another peculiar quirk with the way Davinci Resolve handles White Balance in the Camera Raw tab for Canon R5 C files.

Here is a file shot under 45000 kelvin light a couple of days ago. This time it was properly exposed, with custom white balance set in camera. If I develop the file using the Decode Quality:  Full Res - Resolve setting, then I get a very nice, neutral result -- as good as can be expected with such challenging light:

1078958768_WB-Canon_1.2.2.thumb.jpg.b357ee5c14facacc5347ce2399967bba.jpg

798962994_ScreenShot2022-07-13at3_55_29PM.thumb.png.82c6548de6c4b5fab26a31360a6dd991.png

But if I choose Decode Quality: Full Res - Canon, then the file turns blue!

WB-Canon_1.2.1.thumb.jpg.d391c4bd5ba187a2b17a542ebdaecab7.jpg

1601444413_ScreenShot2022-07-13at4_00_59PM.thumb.png.4f8671a4b85323f93221011a93cba433.pngIt's as if the camera now is maxed out at 15000k of WB adjustment, where as before it could accurately remember a warmer white balance when setting custom white balance in camera and debayering the file using Resolve's debayering algorithm rather than the one provided by Canon. 

 

Loading the same file in Canon's own Cinema Raw Development app gives you the right result though:

359575895_ScreenShot2022-07-13at4_25_31PM.thumb.png.a0f9d020a2c7b17c231dcb9ca85c2435.png

And more importantly, if I load its sister file (which had WB set at 5000k in camera), I can use the white balance picker in Cinema Raw Development app to achieve the same result as with the file where WB was set at 45000k in camera:

1984747938_ScreenShot2022-07-13at4_23_59PM.thumb.png.3a8c66f403b9612022dea26afe2ad284.png

So the latitude is there in the Canon R5 C files -- but the way Davinci Resolve currently debayers those files leaves some quirks.

Share this post


Link to post
Share on other sites
On 7/7/2022 at 4:57 AM, Architeuthis said:

=> Testing C-AF of a small, central, field with subsequent trekking for re-framing (in plenty of light and low light) would be a great test (I think the fancy algoritms for subject, face, animal etc. detection may not be so useful for UW)...

With establishes macro lenses (Nikon 105, Canon EF and RF 100, Sony 90, Sigma 105 (for Sony)). Comparison with GH5 and Zuiko 60mm or Pana 45mm will give a good reference for many of us...

I know, I am asking too much, but you asked for suggestions...

 

Wolfgang

 

Honestly, I would expect this to be dictacted by the lenses more so than the camera bodies. For such a simple focusing task, all three are incredibly fast (if the lens autofocus motors can keep up).

Unfortunately I didn't have access to the RF or Z-mount macros. But the Sony 90mm macro was very fast at this task on the A1. And people have complained about it being slow in the past..

  • Like 1

Share this post


Link to post
Share on other sites

Sponsors

Advertisements



×
×
  • Create New...