One aspect I omitted from these tests is green screens and keying. I'm not very experienced with it and don't even own one anymore. But the consensus is that 4:2:2 will give you a notable improvement in your ability to key over 4:2:0. So if that's what you plan to do a lot of, you'll probably find a more significant advantage to the external on the Sony. Thanks for watching! 😃👍
Gerald Undone This is a very important omission. If you want to break chroma sub sampling the examples in this video are not the way to do it. You need to shoot strong primary colors with sharp contrast between them. You will quickly see the edges become jagged and blocky. More color resolution will direct equal less of those artifacts. When keying out blue or green screen it’s very obvious.
I do green screening and from my tests/what I think I know, your going to want as much info as you can to work with. But given the choice between 4:2:2 or higher resolution, the resolution has more of an effect. More pixels density, any problems you have become smaller. But all in all, the more you have the better. Ideally 4:4:4 8k would be best!
I wish Gerald made videos focused on the DP and filmmaking aspect. For example; How to light a scene, how colors and lighting decisions complement the subject and the environment, Tips and tricks to get the audience's attention on a specific object or aspect of the scene. He is incredibly technical in all of his videos, and I love that...but I wish he could work a little bit more towards the artistic use of cameras and other gear. I'm sure he'd be great at those as well. His genius can't go unnoticed, and it makes him a professor on everything he reads and searches about.
Gerald - thank you for your extreme clarity throughout your channel. I would love to know what you think about Canon’s C200, and their Cinema Raw Light vs everyone else’s 10 bit 4:2:2. If you have a moment. Thanks again. - David
To give a slightly more straightforward explanation of chroma subsampling: 4:4:4 - Full colour resolution 4:2:2 - Full vertical colour resolution, half horizontal colour resolution 4:2:0 - Half colour resolution So if you're shooting at 2160p with 4:2:0, the luma channel will be at full res, but the chroma channels will be in 1080p. In 4K and higher, the difference is usually pretty subtle unless you're zooming in to 300% or doing some sort of keying. The major difference comes more from codecs with less chroma subsampling also usually using higher bit rate All-I compression and having 10bit or higher bit depth. In the case in this video, the lower bit rate and interframe compression on the internal codec is the major visual difference.
Actually 4.2.0 is a quarter of full colour resolution When ntsc was designed it was decided to have each colour at one quarter of the luminance resolution to reduce bandwidth This was referred to as 4.1.1 4 being luminance 1 being red 1 being blue The pal system had the same specification but achieved it differently and to prevent any confusion called it 4.2.0 I.e. added the 2 ones together If you think about it 4.2.0 is technically nonsense Red would be 2 and blue 0 A tv system would not work with blue equal to 0 So 4.2.0 was purely done to differentiate it from 4.1.1 A Tv design engineer
@@johnwilcox7826 As far as i have understood, the difference between NTSC DV 4:1:1 and PAL DV 4:2:0 is that 4:1:1 has 1/4 resolution horizontally, but full vertical resolution. On 4:2:0, resolution is halved both horizontally and vertically. In other words, both have 1/4 color resolution but you could think 4:1:1 uses a horizontal squeeze to 25% to accomplish this, where 4:2:0 uses 50% scaling in both axes. 4:4:4 xxxx xxxx xxxx xxxx 4:2:2 x.x. x.x. x.x. x.x. 4:1:1 x... x... x... x... 4:2:0 x.x. .... x.x. .... The 4:1:1 and 4:2:0 naming comes from using the eight-pixel block for the naming as seen in the video. In 4:1:1, both the top and bottom rows only have one color sample, and as Gerald explained, in 4:2:2 both rows have two samples and in 4:2:0 top row has two samples that are copied to the bottom row, which has zero unique color samples. Visually, the two look slightly different. 4:2:0 looks better for progressive and nonmoving images, but AFAIK may cause motion problems with interlaced footage, as the 2nd field actually used copied color information from the 1st field. 4:1:1 looks mushier, but as both fields have unique color information, there are no issues with interlacing.
@@johnwilcox7826 If 4:1:1 meant luma:red:blue as you suggest that would make a lot of sense, but for historical reasons it isn't how it actually works. (Gerald has it correct.) It dates back to the way standard definition analog video was converted to digital in ancient times. Analog scan lines within a video field were digitized in pairs (which makes sense if you dig into the details of analog television), and the three numbers refer to a multiple of a base sampling frequency, in order: luma sampling on every line : chroma sampling on the upper line : chroma sampling on the lower line. In 4:1:1, for example, luma is sampled at 4 times the base frequency on every line and chroma at the base frequency on every line. In 4:2:0 chroma was sampled at twice the base frequency on the upper line of each pair (within a field) and not at all in the second line. NTSC actually used both conventions. Sony DVCAM, for example, used 4:1:1 for NTSC, but Panasonic's system used 4:2:0. NTSC DVDs are all 4:2:0. (PAL/SECAM is indeed pretty much 4:2:2 across the board, except for Brazil's PAL-M which is basically NTSC with PAL color in the broadcast signal). While I'm quibbling, resolution differences aren't as simple as a ratio of total pixels (no matter what marketing departments think) and in traditional television work vertical and horizontal resolution are defined and measure differently. In analog SD, for example, (limiting) vertical resolution is a function of the number of scan lines and horizontal is a function of bandwidth (and usually specified as the number of alternating black and white lines that can be distinguished across a width of the picture equal to the height).
Would be great an update on this matter. Since this video was recorded, 10bit 4:2:2 has almost become the standard on latest cameras. Comparing slog3 10bit 4.2.2 with even BRAW would be amazing.
Gerald - great to do a comparison, but you missed a major contextual fact. Not sure if someone else had said this in the comments already, but anyway: If you are mastering for TV / a lot of online broadcast, the amount of 4:2:2 material in your film / tv show is vital. As stated elsewhere 4:2:0 is a delivery format, not a professional acquisition format. Something may look great on your edit monitor, but that's not how it's going to play out once it's been transcoded - maybe more than once - before it shows up on HBO, Sky, BBC etc. compression compromises are multiplied down stream and become severe when delivery bitrates are again reduced to save infrastructure costs. And so when selecting your weapons of choice, you must know what the technical regulations are for the commercial client are. If you are just delivering to TH-cam / your own website, it's not going to be an issue. But if you're selling your work to others, 4:2:2 may be essential (not an option)
Not much difference. That's why 4:2:0 sampling was used in delivery format. It was more than good enough. Anyway, another good video that actually shows things! Many thanks, Gerald.
Gerald, another well done video! 😉 Your tests comport with our decision over the last two years using the GH5 NOT to bother with 8 bit or even 10 bit 4:2:2. The differences are just not meaningful for the work we do - and why I think the recorder-only, 40% less expensive Atomos Shinobi makes far more sense for most hybrid shooters than the Ninja V. With this written, I look forward to the day when you compare the Nikon Z6 12-bit ProRes RAW on the Ninja V to the 8-bit internal. At least in theory, THAT will be VERY interesting.
I don't know how many people have caught it, but the "feel free to hit the dislike button twice" thing is very clever. Love your channel, and your pacing/information density. You are one of the only channels that I don't watch at a higher speed. And thanks for making these things so simple while still being thorough.
Getting further into this pipeline (after several years). Finally digging into video now that my Nikon can shoot it internally. And ya know what? I finally understood a GU video, from start to finish. Nicely UNDONE, Gerald! Thanks for letting me hang with you over all this time.
Nothing seems crazy here, but I guess "Gerald Undone...he's methodical and precise" wouldn't be as catchy a theme song. I really appreciate the grounded-in-reality approach you have with your video topics.
No extra googling/searching for info needed with my questions, just subscription and notification on from your channel is enough, Everytime so informative with your videos. Keep creating sir, respect.
Switched from slog 3 to HLG3 thanks to you and wow have my videos that almost no one watches gotten better lol. Havent started giving HLG3 video to clients yet as I am learning it.
Cody Mchardy i completely agreed! Gerald did a great job with the explanation that I ventured out and filmed an entire music video with HLG3 and i gotta admit...grading is easier...less artifacts when grade heavily. I’m glad you like it too Cody!
NathanHoang yeah. It just requires more effort in getting it right in camera. But that’s what we need to do is get better. So it’s been a new challenge for me
you are the most unbiased reviewer on youtube. greatly appreciated. plus you talk at the speed of light so i have more time in my life while still gaining your information.
Thanks for doing these tests. I have run these tests several times over the years and I have come to the same confusion with the Sony cameras, the internal XAVCS codec when using 4k (NOT 1080, you can see the difference at 1080). The internal grades just as well as the 4k 4:2:2 from the Atomos. I am sure you will get some hate for doing these tests as I have in the past.
The sony 4k 60mbs and 100mbs still wonder how much difference there is as many videos arent really detailed test so hard to really tell the difference.
5:42-6:23 I see a green shift in the greys when going to external. Internal might be tinted 1-2 into magenta, but external recorder is 2-3 in the green by comparison
True. Some people suggest 4:2:0 will produce serious banding issues when really the thickness of these bands still could not be more than two pixels on each axis, given that colour information of maximum 2x2 pixels is merged, but the number of available colours is not reduced, thus not affecting subtle gradients like blue skies in any notable way, but rather mostly areas in which notably different colours are packed right next to each other, making every pixel's colour count. It's actually immensely clever. This method essentially means if you record in 4K your colour information's resolution will be in Full HD. If you downsample your footage to Full HD, your colour information in the downsampled footage will be 4:4:4.
- I tried to read all the previous comments before posting this and I don't think this has been mentioned, but no guarantees. . . . :-) - I think the highest need for Y'CBCR "better than 4:2:0" is when the target output is not "pixel for pixel" compared to the camera data. The cases are: 1. Stabilization, 2. lens corrections and 3. compositional cropping. - The next most important is probably for green-screen type overlay effects, which you have mentioned. - Personally my biggest need is for stabilization, which I use fairly often. In those cases, data from a pixel is stretched to another pixel. Anything you can do to make the original pixels "higher quality" will be amplified during such a process.
Hello Gerald, I am Guido and would first like to thank you for all the effort you have made. On the other hand, you also expect an effect, namely "banding", which is only available at bit depth, that is, an unsure progression, for example in the sky, etc. This is not what color subsampling is, and that's why you have it can't find any differences. color or chroma subsampling shows up in the compression of the colors and often only comes to light when you are heavily graded. A bit unattractive with all these comparisons is that when I have 422, I automatically always have 10 bits available. This value in turn influences a test result visually from more than just 420 vs 422 and that makes everything so confusing. If it was really just a question of color subsampling, you have to look for color artefacts and not for banding or brightness gradients in the colors. In contrast, a high color subsampling value ensures clean colors, especially on the hems and edges, where the compression often attacks particularly strongly. That's why the results of an Alexa look so much more homogeneous instead of an Sony a6000. I attached a very good link to you here, which shows the differences well. I will also try to make a difference with my A7III & Shogun in the near future, because I am convinced that this is perhaps an even more important parameter than 8Bit vs 10Bit. Please stay healthy and many thanks for your great work, please keep it up. :-) Greetings Guido www.niwa.nu/2013/05/why-chroma-sub-sampling-is-important/
Hi Gerald, I normally watch TY on my tv which prevents me from leaving comments. I wanted to pull up one of your videos from my computer specifically so that I could leave a comment and tell you how awesome your videos are. So helpful and down to earth, someone who doesn't appear to be full of himself or trying to be something he isn't. I appreciate the effort you put into videos. Every time I see one in my recommended, I read the title and think "hmm, sounds like that is right up my alley." Great combination of helpful and interesting. Keep up the great work! Jake
This was a discussion way over my head as I try to evaluate specs on HDMI audio extractors with DIP switches for EDID modes…however I HAD to leave a thumbs up for that parting comment “If not hit the dislike twice”. First I ever heard that said and it was well used…very smooth! Thanks for the smile
Years ago I came to similar conclusions on a purely theoretical basis (including, as other comments point out, the advantage of 4:2:2 in chromakeying). One important point that's easy to overlook: The difference is 4:2:2 versus 4:2:0 is only in vertical resolution, which is of huge practical significance in interlaced standard definition (in which each picture is a field, not a frame) but less so in progressive video and especially in higher resolutions. Consider this: In 4K cinema (DCI), luma is 4096x2160 pixels, 4:2:2 chroma is 2048x2160, and 4:2:0 chroma is 2048x1080. That is, chroma resolution in 4K 4:2:0 is identical to chroma resolution in 2K 4:4:4. As a practical matter, even today a lot of visual effects work (including green- and blue-screen and manual roto) is still done in 2K and then uprezzed to "4K." If you can pull a good matte/key in 2K 4:4:4 you can pull an equally good one (in fact, a slightly better one) in 4K 4:2:0. [Edited to correct a dumb mistake in the numbers for 4K 4:2:2.] I'm not suggesting that 4:2:2 isn't better than 4:2:0. It is! But at 4K and above the *practical* difference isn't all that huge.
Just came across you channel, Gerald. Such relevant content for what a lot of us hybrid shooters question and struggle with. Thank you for keeping it clear and clean. There are a lot of opinion based channels on youtube...yours is a breath of fresh air.
Moving from a GH5 for video to an a7r3, I have noticed no difference losing 10bit and switching between internal and external. I also watched a lot of videos and they all need to be pushed far to start seeing anything. So depending on the grade or how close you get in camera I think it’s overrated. Honestly if I cared that much, I would go to a cinema camera with raw.
Nice video. Would have loved to see higher ISO/gain tests. We know that Sony applies two stages of noise reduction on most of their cameras. The first stage is the sensor readout cleanup. The second stage happens in the XAVC H.264 encoder chip itself. In theory, the higher gain/ISO you pick, the more and noise reduction is applied. Also, in theory, the HDMI will only have the stage 1 noise reduction applied and not the second XAVC hit. So,....at higher gain, the HDMI "should" have more and more noise but also LESS noise reduction artifacts and possibly more detail. Would be interesting to find out.
The noise pattern you see it's from h.264 compression method, not form the chroma subsampling, Sony implementation of h264 is horrendous. The only advantage of going for the Atomos 8bit 422 is getting rid of the compression of h264.
@@thatchinaboi Yes, of course bitrate has a mayor play. Prores has much more bitrate, but it also is a different codec that works well with motion. But the 100mb/s h264 on Sony is not equal to the 100mb/s in Panasonic. They implement the codec in diferent ways, Panasonic gets better quality in the same bitrate.
@@MatiasRispau So the point is Gerald is wrong in saying there is only a very slight advantage in some extreme situations. Interesting how no one else pointed this out. 😂
@@thatchinaboi No, Gerald is not wrong. He is very technical about it and pointed very good aspects. I'm just adding that the noise issue was from other thing. But also Prores has other benefits apart from 4:2:2. Codec compression quality and motion is one. But, and this is a BIG but... Compared to Panasonic internal 10 bit 4:2:2 h264 codec. The difference is almost none, in that situation i see no point of using an external recorder if you just want better IQ.
@@MatiasRispau Not seeing any differences in the motion rendering is down to what type of scene you are shooting. Try shooting a very busy scene with lots of little motion. (A river or trees swaying in the wind) And yes, Gerald is wrong for not mentioning the benefit of recording in pores, as it gives the wrong impression to people that there is little to no benefit. :)
You've made 7 videos in February alone - on fire! It would be interesting to see a recording comparison between the A7M3 and the Fuji XT3. 10 Bit 4:2:2 vs. 8 Bit 4:2:2, 10 Bit 4:2:0 vs. 8 Bit 4:2:0.
I really appreciate you answering all these questions in a somewhat scientific way. The amount of effort you put in is obvious! I'll be using your affiliate links from now on as a way to show my appreciation!
The difference occurs more when you try to film in Slog2 S-Gamut with the A7s with 8bit internal recording, then try to grade it with LUTs (converting Slog2 to Rec709 then grading with a LUT) Even without zooming in or extreme grading, the colors are really blocky and color banding is really awful. Thus the reason I'm considering an external recorder (as well as the 4K capabilities)
Awesome video... I've been considering the ninja V(+a6500) as an alternative to an a6400 for my home setup. Basically, I want to have a monitor and not lose my face tracking in 4k and those seem like the two ways I can accomplish that right now. Did you get a chance to run any grading tests between external and internal with picture profiles applied?
Chroma sub-sampling MIGHT cause more banding but primarily you’ll see issue where two high saturated colors meet. Say a green-screen meets a red and blue shirt. Or green leaves meet a red or blue car. Banding will primarily increase with lower color bit depths (8-bpp vs 10-bpp) and lower fidelity codecs such as h264 vs ProRes.
This video was a godsend... SERIOUSLY! for the past week i have been stressing over whether or not i should get one. All i have gathered from most videos is that i should get one if i deliver professional work and need better codecs for better quality. Now i finally have a clear idea of what i would be getting myself into.
Thanks for your reply! i think i will be getting one as well... makes no sense shooting something that may end up being unusable because of the bit depth... appreciate your info.
i was going to go with the Ninja V for my Sony A6500 / A7iii but i think i might save some dough and get the shinobi instead....Thanks for this well organized and informative video. Much appreciated, Gerald.
Hey Gerald, using the Atomos made a huge difference in Chroma Key recordings I had to do in some of my works, so, that was an advantage in that sense. . You should try to do some internal recordings with a green screen vs with the atomos, and you will see the difference in the way it keys all the video information. Thank you for all the videos, I've learned a lot from you. The best to you!
Right now you are the most useful channel there is when it comes to providing actual useful knowledge about cameras/video, especially sony. I hope you keep it up and grow big cause you really deserve it!
I've done similar tests myself. When it does make a difference to go with the atomos is when you have skies in your shots and you want to add some dramatic flair to the by bumping up the contrast with a power window. Overcast skies can be very low contrast and it takes a very strong curve to make them look interesting.
I don´t know how long you´ve been around on TH-cam but your videos started to pop out in my recommended just recently and you are amazing! You provide exactly that kind of information I´ve been looking for and the explanations are super clear sooo... Subbed!
I liked this test, but you're missing out on something really important here: Data rates and Gop's. If you're recording externally with an Atomos in DNxHR, you will notice FAR much lower artefacts in pans and fast moving images. For me this is the most important reason to record externally. (I'm using A7sII + Ninja Inferno). Grading an internally recorded .mp4 with codec limitations leads to horrible results in this case.
Even though there is not much of a difference between the internal and external video samples in terms of image quality of the static images, there is still the advantage of bit rate. Having higher bit rate will render motion better, especially in scenes with lots of movement (as opposed to fast movement). @Gerald Undone
100M @30hz vs 100M @60hz: 30hz would have more headroom for 4.2.2, right? The extra 50M per second (roughly, considering resolution settings) would allow the headroom for more color information from 4.2.2? Maybe, saw a balance with 100M to save on battery life and storage space.
great test - you have also be aware of the codec compression, XAVC S is Long GOP(100 Mbit) and ProRes is Intra(600-800 Mbit), that means, regarding your test the XAVC S is really good
Chroma subsampling is by nature a spatial resolution issue, not a bit depth resolution issue. As such, it will only affect high resolution in space, i.e. edges and details, and its telltale artifact is aliasing (seen as a stairstepping effect) or image softness, depending on the the antialiasing filter strength chosen by the camera maker. Banding is an artifact of insufficient bit depth. Different dimension entirely. What variations in banding you saw here were likely a function of what 4:2:0 was doing to sensor readout noise - ironically, 4:2:0 acts as a minor noise filter. So, 4:2:2 benefits you wherever you need that extra edge definition: shots requiring FX/post work (keying and compositing), or upscaling headroom. But for heavy grading applications where banding looms large, you want bit depth and dynamic range, not 4:2:2.
I was considering buying a ninja V to record 4:2:2 footage with my A7III but seeing only such a subtle difference, especially after YT compression, I decided to let it be. You just saved me a couple hundered bucks
I really appreciate your efforts to help others. This video earned you my subscription. Love the ease with which you clarified such a big deal ( for which no one extend the hands to help.)
One thing your video did was make clear to me that if your end state is posting on TH-cam, don't worry about 4:2:0 versus 4:2:2! I could tell absolutely no difference between the two examples on the TH-cam display even viewing max resolution.
TH-cam converting content to SDR 8-bit BT.709 searh in your browser 1. TH-cam Recommended upload encoding settings 2. TH-cam Upload High Dynamic Range (HDR) videos
The modifications applied in these tests deal with pixel luminance. It is stored per pixel in both 422 and 420. To see the difference between 422 and 420 CHROMA modification should be applied. Extreme modifications to pixel CHROMA reveal the difference between 422 and 420.
Thanks for the video! One mistake I guess though in minuter 1:36 with 4:2:2 diagram. The color of the second row blocks should be light blue and dark blue I guess. Because the sampled pixels on the second row will be shifted 1 pixel so that a more homogeneous sampling is achieved. Also, another problem with this diagram is that the final values of the pixels which are not sampled are not copied from sampled pixels, they are interpolated! So, it's in fact better to show them in the diagram with white color. Later on, those values will be assigned by interpolating the sampled values.
If Sony could make the a7 III at least output 10-bit 4:2:2 (or RAW) to an external recorder to at least be comparable to the Nikon Z6 II, that would be huge. Sony could do it; they choose not to. Hopefully, the a7 IV will do 10-bit 4:2:2 internally because the Canon R6 and Lumix S5 can.
@@MrVh78 yes. The quality of the grain you get betwee 420 and 422 quite visible. Codec settings for delivery is also important. I do deliveries in h265 and they have a special `-tune grain` argument to retain as much grain as possible. Well, if footage endup on the TH-cam, 420 or 422 does not play much difference ofc :)
One more thing I really like in this presentation how you showed more of your personality and I loved your outro too. Basically you've converted me and I'm a die hard fan now, no jokes. Cheers! Looking forward to more vids
+ bit rate difference is another benefit of recording externally! I find a lot less artefacts, and clearing the rolling shutter on my a6500 for instance gives much better results.
I don't know how much TH-cam affected it, but it seemed to me that I couldn't see much difference between most of the examples EXCEPT the ones on the umbrellas. I definitely noticed that the 4:2:0 umbrella examples had a bit of a reddish cast to them and the 4:2:2 ones didn't. The 4:2:2 umbrella examples had more natural-looking color than the 4:2:0 ones..
Whether you are shooting internally or externally both are recorded in 8-bit which means they are both are very limiting in terms of creative colour grading.
Gerald keeps me learning, i appreciate it all as I want to be film maker. I honestly use the atamos cause i can see clearer on it, the peeking and even better the storage. Adapts to any PC, laptop, and works. Well now with the new v5 firmware with Sony and its apps, well see.
Always get the Atomos instead of the Small HD Focus. Almost the same price and the Atomos has way more features instead of just being a monitor. Good video.
in theory , XAVC HS 4K is 265 codec ,same 100Mb contains more information than 264 100M, so... I go with XAVC HS 4K currently even I can’t see the difference at all LOL cheers
Phenomenal video man! You actually talked me out of buying a Ninja V for my X-S10. I could not tell that much of a difference enough to even buy a used Ninja for $250 on CL and think I'll buy a lens instead. Thank you and yup, consider turning off that tally light next time mate.
Atomos only good for pro res to edit inside davanci. And export times. Other than that pretty much not needed. I did notice 35mbit h264 mpeg was smoother than prores. Pro res us very grainy and needs noise reduction in post.
Thank you Gerald. As I mentioned previously, I am not a Sony user, but I love tech stuff and I see how this topic can relate to other brands. 50K subs before you know it.
I found this info very worthwhile, having a Fuji X-T4. Haven't even shot video yet but am encouraged that I can manipulate the colour grade so that it will not 'break". I'll be using DR 18. Thx Gerald...get undone more - ppl love it!👍
This was the video I was looking for. Thank you for your insights. I now know buying an Atomos recorder for my A7SII is not worth it. Thnxs buddy... Great video
I have the impression that you live in a quiet purple cave in the middle of nowhere and all you do all day is tests cameras and do experiments to get the best image quality out of everything... Love your videos and how clear you are when explaining everything. Thanks for the amazing content! PS. will you be releasing any purple merch in the future?
That's pretty much exactly right! Haha. 👍😜 Thanks for the kind words and support! And yeah, I do have some merch available. Haven't updated the designs in a while though: www.redbubble.com/people/geraldundone/works/27198329-lets-get-undone?asc=t&p=t-shirt
One aspect I omitted from these tests is green screens and keying. I'm not very experienced with it and don't even own one anymore. But the consensus is that 4:2:2 will give you a notable improvement in your ability to key over 4:2:0. So if that's what you plan to do a lot of, you'll probably find a more significant advantage to the external on the Sony. Thanks for watching! 😃👍
Gerald Undone This is a very important omission. If you want to break chroma sub sampling the examples in this video are not the way to do it. You need to shoot strong primary colors with sharp contrast between them. You will quickly see the edges become jagged and blocky. More color resolution will direct equal less of those artifacts.
When keying out blue or green screen it’s very obvious.
Thank you for this important information. Whats the matter with Slog2/Slog3? The same?
I do green screening and from my tests/what I think I know, your going to want as much info as you can to work with. But given the choice between 4:2:2 or higher resolution, the resolution has more of an effect. More pixels density, any problems you have become smaller. But all in all, the more you have the better. Ideally 4:4:4 8k would be best!
That’s exactly what I was about to mention. :)
Very interesting video. Does it also mean that there is no noticeable difference in image quality between XAVC-S and ProRes HQ with a Sony ?
You are the film school I couldn't afford. Thanks Gerald.
Gerald. This. Thank you.
I wish Gerald made videos focused on the DP and filmmaking aspect. For example; How to light a scene, how colors and lighting decisions complement the subject and the environment, Tips and tricks to get the audience's attention on a specific object or aspect of the scene. He is incredibly technical in all of his videos, and I love that...but I wish he could work a little bit more towards the artistic use of cameras and other gear. I'm sure he'd be great at those as well. His genius can't go unnoticed, and it makes him a professor on everything he reads and searches about.
It's worse than that... He's better than the film school I'm about to graduate
Gerald - thank you for your extreme clarity throughout your channel. I would love to know what you think about Canon’s C200, and their Cinema Raw Light vs everyone else’s 10 bit 4:2:2. If you have a moment. Thanks again. - David
Dorian Bury - Wow dude...! Glad I saw your post.😧
To give a slightly more straightforward explanation of chroma subsampling:
4:4:4 - Full colour resolution
4:2:2 - Full vertical colour resolution, half horizontal colour resolution
4:2:0 - Half colour resolution
So if you're shooting at 2160p with 4:2:0, the luma channel will be at full res, but the chroma channels will be in 1080p.
In 4K and higher, the difference is usually pretty subtle unless you're zooming in to 300% or doing some sort of keying. The major difference comes more from codecs with less chroma subsampling also usually using higher bit rate All-I compression and having 10bit or higher bit depth. In the case in this video, the lower bit rate and interframe compression on the internal codec is the major visual difference.
Nice summary! Cheers.
Actually 4.2.0 is a quarter of full colour resolution
When ntsc was designed it was decided to have each colour at one quarter of the luminance resolution to reduce bandwidth
This was referred to as 4.1.1
4 being luminance
1 being red
1 being blue
The pal system had the same specification but achieved it differently and to prevent any confusion called it 4.2.0
I.e. added the 2 ones together
If you think about it 4.2.0 is technically nonsense
Red would be 2 and blue 0
A tv system would not work with blue equal to 0
So 4.2.0 was purely done to differentiate it from 4.1.1
A Tv design engineer
@@johnwilcox7826 That's very interesting. Thanks for the insight and clarification!
@@johnwilcox7826 As far as i have understood, the difference between NTSC DV 4:1:1 and PAL DV 4:2:0 is that 4:1:1 has 1/4 resolution horizontally, but full vertical resolution. On 4:2:0, resolution is halved both horizontally and vertically. In other words, both have 1/4 color resolution but you could think 4:1:1 uses a horizontal squeeze to 25% to accomplish this, where 4:2:0 uses 50% scaling in both axes.
4:4:4
xxxx
xxxx
xxxx
xxxx
4:2:2
x.x.
x.x.
x.x.
x.x.
4:1:1
x...
x...
x...
x...
4:2:0
x.x.
....
x.x.
....
The 4:1:1 and 4:2:0 naming comes from using the eight-pixel block for the naming as seen in the video. In 4:1:1, both the top and bottom rows only have one color sample, and as Gerald explained, in 4:2:2 both rows have two samples and in 4:2:0 top row has two samples that are copied to the bottom row, which has zero unique color samples.
Visually, the two look slightly different. 4:2:0 looks better for progressive and nonmoving images, but AFAIK may cause motion problems with interlaced footage, as the 2nd field actually used copied color information from the 1st field. 4:1:1 looks mushier, but as both fields have unique color information, there are no issues with interlacing.
@@johnwilcox7826 If 4:1:1 meant luma:red:blue as you suggest that would make a lot of sense, but for historical reasons it isn't how it actually works. (Gerald has it correct.) It dates back to the way standard definition analog video was converted to digital in ancient times.
Analog scan lines within a video field were digitized in pairs (which makes sense if you dig into the details of analog television), and the three numbers refer to a multiple of a base sampling frequency, in order: luma sampling on every line : chroma sampling on the upper line : chroma sampling on the lower line. In 4:1:1, for example, luma is sampled at 4 times the base frequency on every line and chroma at the base frequency on every line. In 4:2:0 chroma was sampled at twice the base frequency on the upper line of each pair (within a field) and not at all in the second line. NTSC actually used both conventions. Sony DVCAM, for example, used 4:1:1 for NTSC, but Panasonic's system used 4:2:0. NTSC DVDs are all 4:2:0. (PAL/SECAM is indeed pretty much 4:2:2 across the board, except for Brazil's PAL-M which is basically NTSC with PAL color in the broadcast signal).
While I'm quibbling, resolution differences aren't as simple as a ratio of total pixels (no matter what marketing departments think) and in traditional television work vertical and horizontal resolution are defined and measure differently. In analog SD, for example, (limiting) vertical resolution is a function of the number of scan lines and horizontal is a function of bandwidth (and usually specified as the number of alternating black and white lines that can be distinguished across a width of the picture equal to the height).
Would be great an update on this matter. Since this video was recorded, 10bit 4:2:2 has almost become the standard on latest cameras. Comparing slog3 10bit 4.2.2 with even BRAW would be amazing.
This video made my subscribe man. Great info and presentation Gerald 🙌
Happy you did! Appreciate the support! 😃🙏
Same Here.
I shall do your portraits on instagram.com/ghrabbath
You could possibly start your next video with, "... I'm a geek so you don't have to be."
That is oddly enough, my tagline :)
If you're here though, you're probably a bit of a geek
Gerald - great to do a comparison, but you missed a major contextual fact. Not sure if someone else had said this in the comments already, but anyway:
If you are mastering for TV / a lot of online broadcast, the amount of 4:2:2 material in your film / tv show is vital. As stated elsewhere 4:2:0 is a delivery format, not a professional acquisition format. Something may look great on your edit monitor, but that's not how it's going to play out once it's been transcoded - maybe more than once - before it shows up on HBO, Sky, BBC etc. compression compromises are multiplied down stream and become severe when delivery bitrates are again reduced to save infrastructure costs. And so when selecting your weapons of choice, you must know what the technical regulations are for the commercial client are.
If you are just delivering to TH-cam / your own website, it's not going to be an issue. But if you're selling your work to others, 4:2:2 may be essential (not an option)
You rock! You have definitely risen to the top of my TH-cam watch list.
I'm honoured! Thank you. 😃🙏
Not much difference. That's why 4:2:0 sampling was used in delivery format. It was more than good enough. Anyway, another good video that actually shows things! Many thanks, Gerald.
Gerald, another well done video! 😉
Your tests comport with our decision over the last two years using the GH5 NOT to bother with 8 bit or even 10 bit 4:2:2. The differences are just not meaningful for the work we do - and why I think the recorder-only, 40% less expensive Atomos Shinobi makes far more sense for most hybrid shooters than the Ninja V. With this written, I look forward to the day when you compare the Nikon Z6 12-bit ProRes RAW on the Ninja V to the 8-bit internal. At least in theory, THAT will be VERY interesting.
Thanks for the kind words and comment!
I too am very interested in that comparison. Can't wait for that firmware to drop! 😃👍
I don't know how many people have caught it, but the "feel free to hit the dislike button twice" thing is very clever. Love your channel, and your pacing/information density. You are one of the only channels that I don't watch at a higher speed. And thanks for making these things so simple while still being thorough.
Thanks so much! Glad to hear you're enjoying the videos. Thanks for taking the time to say so. Cheers!
Getting further into this pipeline (after several years). Finally digging into video now that my Nikon can shoot it internally. And ya know what? I finally understood a GU video, from start to finish. Nicely UNDONE, Gerald! Thanks for letting me hang with you over all this time.
Nothing seems crazy here, but I guess "Gerald Undone...he's methodical and precise" wouldn't be as catchy a theme song. I really appreciate the grounded-in-reality approach you have with your video topics.
Haha. Exactly! It's all about getting it stuck in your head. 😜👍
And thanks! I really appreciate that.
Useful tests there Gerald, i enjoyed that...
Thanks, Gordon! Cheers.
No extra googling/searching for info needed with my questions, just subscription and notification on from your channel is enough, Everytime so informative with your videos. Keep creating sir, respect.
Switched from slog 3 to HLG3 thanks to you and wow have my videos that almost no one watches gotten better lol. Havent started giving HLG3 video to clients yet as I am learning it.
Cody Mchardy i completely agreed! Gerald did a great job with the explanation that I ventured out and filmed an entire music video with HLG3 and i gotta admit...grading is easier...less artifacts when grade heavily. I’m glad you like it too Cody!
NathanHoang yeah. It just requires more effort in getting it right in camera. But that’s what we need to do is get better. So it’s been a new challenge for me
same here it’s because oh him i changed from slog2 to hlg3
If you also use an Fs7 camera with slog 3 and will Edit this two formats together would you still use HLG3 on A7iii?
yes hlg3, then apply in post two luts: first lut to convert hlg3 to slog3 and the last lut, lut for slog3. That match perfectly fs7 and a7rIII.
you are the most unbiased reviewer on youtube. greatly appreciated. plus you talk at the speed of light so i have more time in my life while still gaining your information.
Haha. Thank you very much! Really appreciate it. Cheers!
Cool, and yeah once you take TH-cam compression into account I feel like there's no real difference between the two specifically for TH-cam videos.
The smartest guy on You Tube!
Thanks for doing these tests. I have run these tests several times over the years and I have come to the same confusion with the Sony cameras, the internal XAVCS codec when using 4k (NOT 1080, you can see the difference at 1080). The internal grades just as well as the 4k 4:2:2 from the Atomos. I am sure you will get some hate for doing these tests as I have in the past.
Hey, Dave! Nice to see ya. Thanks for sharing your findings. Cheers!
The sony 4k 60mbs and 100mbs still wonder how much difference there is as many videos arent really detailed test so hard to really tell the difference.
This channel is refreshing. Coming from a film geek to another
5:42-6:23 I see a green shift in the greys when going to external. Internal might be tinted 1-2 into magenta, but external recorder is 2-3 in the green by comparison
True. Some people suggest 4:2:0 will produce serious banding issues when really the thickness of these bands still could not be more than two pixels on each axis, given that colour information of maximum 2x2 pixels is merged, but the number of available colours is not reduced, thus not affecting subtle gradients like blue skies in any notable way, but rather mostly areas in which notably different colours are packed right next to each other, making every pixel's colour count.
It's actually immensely clever. This method essentially means if you record in 4K your colour information's resolution will be in Full HD. If you downsample your footage to Full HD, your colour information in the downsampled footage will be 4:4:4.
- I tried to read all the previous comments before posting this
and I don't think this has been mentioned, but no guarantees. . . . :-)
- I think the highest need for Y'CBCR "better than 4:2:0" is when
the target output is not "pixel for pixel" compared to the camera
data. The cases are: 1. Stabilization, 2. lens corrections and
3. compositional cropping.
- The next most important is probably for green-screen type overlay
effects, which you have mentioned.
- Personally my biggest need is for stabilization, which I use
fairly often. In those cases, data from a pixel is stretched
to another pixel. Anything you can do to make the original
pixels "higher quality" will be amplified during such a process.
Once I saw the explanation of 4:4:4, 4:2:2 and 4:2:0 I immediately subbed and liked the video. Your simple way of explaining is terrific! Thx man.
Thanks so much! Happy to have a new subscriber and I'm glad you liked the video. Cheers!
Hello Gerald,
I am Guido and would first like to thank you for all the effort you have made.
On the other hand, you also expect an effect, namely "banding", which is only available at bit depth, that is, an unsure progression, for example in the sky, etc. This is not what color subsampling is, and that's why you have it can't find any differences.
color or chroma subsampling shows up in the compression of the colors and often only comes to light when you are heavily graded. A bit unattractive with all these comparisons is that when I have 422, I automatically always have 10 bits available.
This value in turn influences a test result visually from more than just 420 vs 422 and that makes everything so confusing. If it was really just a question of color subsampling, you have to look for color artefacts and not for banding or brightness gradients in the colors.
In contrast, a high color subsampling value ensures clean colors, especially on the hems and edges, where the compression often attacks particularly strongly.
That's why the results of an Alexa look so much more homogeneous instead of an Sony a6000.
I attached a very good link to you here, which shows the differences well. I will also try to make a difference with my A7III & Shogun in the near future, because I am convinced that this is perhaps an even more important parameter than 8Bit vs 10Bit.
Please stay healthy and many thanks for your great work, please keep it up. :-)
Greetings Guido
www.niwa.nu/2013/05/why-chroma-sub-sampling-is-important/
Hi Gerald, I normally watch TY on my tv which prevents me from leaving comments. I wanted to pull up one of your videos from my computer specifically so that I could leave a comment and tell you how awesome your videos are. So helpful and down to earth, someone who doesn't appear to be full of himself or trying to be something he isn't. I appreciate the effort you put into videos. Every time I see one in my recommended, I read the title and think "hmm, sounds like that is right up my alley." Great combination of helpful and interesting. Keep up the great work!
Jake
This was a discussion way over my head as I try to evaluate specs on HDMI audio extractors with DIP switches for EDID modes…however I HAD to leave a thumbs up for that parting comment “If not hit the dislike twice”. First I ever heard that said and it was well used…very smooth! Thanks for the smile
I don't get why is the 4:2:0 "compression" copying the first pixel's color data instead of averaging the 4 tones.
Years ago I came to similar conclusions on a purely theoretical basis (including, as other comments point out, the advantage of 4:2:2 in chromakeying). One important point that's easy to overlook: The difference is 4:2:2 versus 4:2:0 is only in vertical resolution, which is of huge practical significance in interlaced standard definition (in which each picture is a field, not a frame) but less so in progressive video and especially in higher resolutions. Consider this:
In 4K cinema (DCI), luma is 4096x2160 pixels, 4:2:2 chroma is 2048x2160, and 4:2:0 chroma is 2048x1080. That is, chroma resolution in 4K 4:2:0 is identical to chroma resolution in 2K 4:4:4. As a practical matter, even today a lot of visual effects work (including green- and blue-screen and manual roto) is still done in 2K and then uprezzed to "4K." If you can pull a good matte/key in 2K 4:4:4 you can pull an equally good one (in fact, a slightly better one) in 4K 4:2:0. [Edited to correct a dumb mistake in the numbers for 4K 4:2:2.]
I'm not suggesting that 4:2:2 isn't better than 4:2:0. It is! But at 4K and above the *practical* difference isn't all that huge.
Just came across you channel, Gerald. Such relevant content for what a lot of us hybrid shooters question and struggle with. Thank you for keeping it clear and clean. There are a lot of opinion based channels on youtube...yours is a breath of fresh air.
Thanks for this. Really appreciate it. 🙏😃
Moving from a GH5 for video to an a7r3, I have noticed no difference losing 10bit and switching between internal and external. I also watched a lot of videos and they all need to be pushed far to start seeing anything. So depending on the grade or how close you get in camera I think it’s overrated. Honestly if I cared that much, I would go to a cinema camera with raw.
What type of color grading or corrections do you apply? Did you notice any difference when using green screen work?
Andru Casel I didn’t do any green screen, I got as close to in camera, then basic adjustments.
Nice video. Would have loved to see higher ISO/gain tests. We know that Sony applies two stages of noise reduction on most of their cameras. The first stage is the sensor readout cleanup. The second stage happens in the XAVC H.264 encoder chip itself. In theory, the higher gain/ISO you pick, the more and noise reduction is applied. Also, in theory, the HDMI will only have the stage 1 noise reduction applied and not the second XAVC hit. So,....at higher gain, the HDMI "should" have more and more noise but also LESS noise reduction artifacts and possibly more detail. Would be interesting to find out.
The noise pattern you see it's from h.264 compression method, not form the chroma subsampling, Sony implementation of h264 is horrendous. The only advantage of going for the Atomos 8bit 422 is getting rid of the compression of h264.
What about bit rate, which effects the rendering in motion? (especially scenes with lots of motion)
@@thatchinaboi Yes, of course bitrate has a mayor play. Prores has much more bitrate, but it also is a different codec that works well with motion. But the 100mb/s h264 on Sony is not equal to the 100mb/s in Panasonic. They implement the codec in diferent ways, Panasonic gets better quality in the same bitrate.
@@MatiasRispau So the point is Gerald is wrong in saying there is only a very slight advantage in some extreme situations. Interesting how no one else pointed this out. 😂
@@thatchinaboi No, Gerald is not wrong. He is very technical about it and pointed very good aspects. I'm just adding that the noise issue was from other thing. But also Prores has other benefits apart from 4:2:2. Codec compression quality and motion is one.
But, and this is a BIG but... Compared to Panasonic internal 10 bit 4:2:2 h264 codec. The difference is almost none, in that situation i see no point of using an external recorder if you just want better IQ.
@@MatiasRispau Not seeing any differences in the motion rendering is down to what type of scene you are shooting. Try shooting a very busy scene with lots of little motion. (A river or trees swaying in the wind) And yes, Gerald is wrong for not mentioning the benefit of recording in pores, as it gives the wrong impression to people that there is little to no benefit. :)
You've made 7 videos in February alone - on fire! It would be interesting to see a recording comparison between the A7M3 and the Fuji XT3. 10 Bit 4:2:2 vs. 8 Bit 4:2:2, 10 Bit 4:2:0 vs. 8 Bit 4:2:0.
I really appreciate you answering all these questions in a somewhat scientific way. The amount of effort you put in is obvious! I'll be using your affiliate links from now on as a way to show my appreciation!
Thanks so much! That's very thoughtful and appreciated. 😃🙏
You're always doing the videos that we really care for
I love the effort you put into all your videos. All the small things. You never seem to rush anything.
I guess it's better to invest in a camera that can shoot raw video rather than investing in this Atomos recorder. Thank you
The difference occurs more when you try to film in Slog2 S-Gamut with the A7s with 8bit internal recording, then try to grade it with LUTs (converting Slog2 to Rec709 then grading with a LUT)
Even without zooming in or extreme grading, the colors are really blocky and color banding is really awful.
Thus the reason I'm considering an external recorder (as well as the 4K capabilities)
When there is a lot of movement in the shot and the camera is for example panning, the XAVC-S compression and macro blocking is much more obvious.
Awesome video... I've been considering the ninja V(+a6500) as an alternative to an a6400 for my home setup. Basically, I want to have a monitor and not lose my face tracking in 4k and those seem like the two ways I can accomplish that right now. Did you get a chance to run any grading tests between external and internal with picture profiles applied?
I have! And with the exception of S-Log, which has that bug I pointed out in the other video, it's a pretty similar process and result.
Have you looked into the new atomos shinobi? wonder if you can still have the ability for face tracking in 4k with it.
LouJo - unfortunately if you’re doing any internal recording with a monitor plugged in, you lose face tracking in 4k 😔
Chroma sub-sampling MIGHT cause more banding but primarily you’ll see issue where two high saturated colors meet. Say a green-screen meets a red and blue shirt. Or green leaves meet a red or blue car. Banding will primarily increase with lower color bit depths (8-bpp vs 10-bpp) and lower fidelity codecs such as h264 vs ProRes.
This video was a godsend... SERIOUSLY! for the past week i have been stressing over whether or not i should get one. All i have gathered from most videos is that i should get one if i deliver professional work and need better codecs for better quality. Now i finally have a clear idea of what i would be getting myself into.
Thanks for your reply! i think i will be getting one as well... makes no sense shooting something that may end up being unusable because of the bit depth... appreciate your info.
i was going to go with the Ninja V for my Sony A6500 / A7iii but i think i might save some dough and get the shinobi instead....Thanks for this well organized and informative video. Much appreciated, Gerald.
On the umbrella test the colour shifted dramatically on my Samsung QN90B TV, the internal had a red tinge the external a green tinge.
I neither own a Sony nor an Atomos but I still found this fascinating!
Thanks for saying so! Happy to hear that. 😃🙏
Same here!
Hey Gerald, using the Atomos made a huge difference in Chroma Key recordings I had to do in some of my works, so, that was an advantage in that sense. . You should try to do some internal recordings with a green screen vs with the atomos, and you will see the difference in the way it keys all the video information. Thank you for all the videos, I've learned a lot from you. The best to you!
Right now you are the most useful channel there is when it comes to providing actual useful knowledge about cameras/video, especially sony. I hope you keep it up and grow big cause you really deserve it!
Thanks so much! That's a huge compliment. Cheers! 😃👍
I've done similar tests myself. When it does make a difference to go with the atomos is when you have skies in your shots and you want to add some dramatic flair to the by bumping up the contrast with a power window. Overcast skies can be very low contrast and it takes a very strong curve to make them look interesting.
Can someone please tell me how to even get 4k60p on this exact setup!?? Only getting 24...
I don´t know how long you´ve been around on TH-cam but your videos started to pop out in my recommended just recently and you are amazing!
You provide exactly that kind of information I´ve been looking for and the explanations are super clear sooo... Subbed!
I'm glad I was recommended! Thanks for subscribing. Looking forward to future comments. Cheers! 😃
I liked this test, but you're missing out on something really important here: Data rates and Gop's. If you're recording externally with an Atomos in DNxHR, you will notice FAR much lower artefacts in pans and fast moving images. For me this is the most important reason to record externally. (I'm using A7sII + Ninja Inferno). Grading an internally recorded .mp4 with codec limitations leads to horrible results in this case.
Even though there is not much of a difference between the internal and external video samples in terms of image quality of the static images, there is still the advantage of bit rate. Having higher bit rate will render motion better, especially in scenes with lots of movement (as opposed to fast movement). @Gerald Undone
That's a fair point. Thanks for sharing!
100M @30hz vs 100M @60hz: 30hz would have more headroom for 4.2.2, right? The extra 50M per second (roughly, considering resolution settings) would allow the headroom for more color information from 4.2.2? Maybe, saw a balance with 100M to save on battery life and storage space.
Are the A7m3 and A7r3 capable of outputting 10 bit and Sony just has not “activated “ that hdmi feature or they built 8 bit circuitry??
I realize this is an "old" video but this is great info for us that haven't upgraded our sony gear (yet).
great test - you have also be aware of the codec compression, XAVC S is Long GOP(100 Mbit) and ProRes is Intra(600-800 Mbit), that means, regarding your test the XAVC S is really good
Great video. Am I the only one who thought the internal shots actually withstood TH-cam’s compression better?
At 5:07 if you look at the rubik's cube you can see some colour changes on the green patches, especially at the upper and lower patches.
Nice video. I have a xt3 and i was wondering the same think. Finaly not any difference in the real life. Keep shooting
Chroma subsampling is by nature a spatial resolution issue, not a bit depth resolution issue. As such, it will only affect high resolution in space, i.e. edges and details, and its telltale artifact is aliasing (seen as a stairstepping effect) or image softness, depending on the the antialiasing filter strength chosen by the camera maker.
Banding is an artifact of insufficient bit depth. Different dimension entirely. What variations in banding you saw here were likely a function of what 4:2:0 was doing to sensor readout noise - ironically, 4:2:0 acts as a minor noise filter.
So, 4:2:2 benefits you wherever you need that extra edge definition: shots requiring FX/post work (keying and compositing), or upscaling headroom. But for heavy grading applications where banding looms large, you want bit depth and dynamic range, not 4:2:2.
I was considering buying a ninja V to record 4:2:2 footage with my A7III but seeing only such a subtle difference, especially after YT compression, I decided to let it be. You just saved me a couple hundered bucks
Yeah, I wouldn't do it for image quality, but I do still really enjoy/recommend the impact it has on the workflow and shooting ease.
I really appreciate your efforts to help others. This video earned you my subscription. Love the ease with which you clarified such a big deal ( for which no one extend the hands to help.)
Thank you! Always glad to have a new subscriber! 😃🙏
Gerald Undone - Myth buster of camera channels 😄👍 great job, thx man
One thing your video did was make clear to me that if your end state is posting on TH-cam, don't worry about 4:2:0 versus 4:2:2! I could tell absolutely no difference between the two examples on the TH-cam display even viewing max resolution.
TH-cam converting content to SDR 8-bit BT.709
searh in your browser
1. TH-cam Recommended upload encoding settings
2. TH-cam Upload High Dynamic Range (HDR) videos
Late but this is the first time I've understood colour subsampling. Thanks!
Bro i love all the technical content you keep putting out you instantly be cause one pf my favorite youtubers bro keep it up
Thanks, mate! I'm honoured. 😃🙏
Banding occurs when contrast is too low. Your umbrella is a high contrast shot. You can try light on the wall without directly shoot the light source.
Gerald, you deserve so much more Subscribers, for real!!
Thank so much! 😃👍
The modifications applied in these tests deal with pixel luminance. It is stored per pixel in both 422 and 420. To see the difference between 422 and 420 CHROMA modification should be applied. Extreme modifications to pixel CHROMA reveal the difference between 422 and 420.
Thanks for the video! One mistake I guess though in minuter 1:36 with 4:2:2 diagram. The color of the second row blocks should be light blue and dark blue I guess. Because the sampled pixels on the second row will be shifted 1 pixel so that a more homogeneous sampling is achieved.
Also, another problem with this diagram is that the final values of the pixels which are not sampled are not copied from sampled pixels, they are interpolated! So, it's in fact better to show them in the diagram with white color. Later on, those values will be assigned by interpolating the sampled values.
If Sony could make the a7 III at least output 10-bit 4:2:2 (or RAW) to an external recorder to at least be comparable to the Nikon Z6 II, that would be huge. Sony could do it; they choose not to. Hopefully, the a7 IV will do 10-bit 4:2:2 internally because the Canon R6 and Lumix S5 can.
One thing is not mentioned here is film grain rendering. 4:2:2 makes huge difference for me detail-wise.
@@MrVh78 yes. The quality of the grain you get betwee 420 and 422 quite visible. Codec settings for delivery is also important. I do deliveries in h265 and they have a special `-tune grain` argument to retain as much grain as possible.
Well, if footage endup on the TH-cam, 420 or 422 does not play much difference ofc :)
One more thing I really like in this presentation how you showed more of your personality and I loved your outro too. Basically you've converted me and I'm a die hard fan now, no jokes. Cheers! Looking forward to more vids
Thanks! Looking forward to future comments. 😃🙏
+ bit rate difference is another benefit of recording externally! I find a lot less artefacts, and clearing the rolling shutter on my a6500 for instance gives much better results.
"Feel free to hit the dislike button twice."
This is genius haha! Laughed so hard.
Thanks, Burke! Cheers. 😃👍
I don't know how much TH-cam affected it, but it seemed to me that I couldn't see much difference between most of the examples EXCEPT the ones on the umbrellas.
I definitely noticed that the 4:2:0 umbrella examples had a bit of a reddish cast to them and the 4:2:2 ones didn't. The 4:2:2 umbrella examples had more natural-looking color than the 4:2:0 ones..
Whether you are shooting internally or externally both are recorded in 8-bit which means they are both are very limiting in terms of creative colour grading.
Gerald keeps me learning, i appreciate it all as I want to be film maker. I honestly use the atamos cause i can see clearer on it, the peeking and even better the storage. Adapts to any PC, laptop, and works. Well now with the new v5 firmware with Sony and its apps, well see.
Always get the Atomos instead of the Small HD Focus. Almost the same price and the Atomos has way more features instead of just being a monitor. Good video.
Wow I thought until now that the image quality would be so much better with 4:2:2, but that is awesome to know. Thank you.
Totally love your channel! Found it very recently and I really like your reasoning/scientific approach. Cheers from Sweden
Thanks, Magnus! Cheers from Canada! 😃🙏
Hey Gerald, can you do same with a7s3 (no need for atomos) just curious if 420 is enough with 10 bit. as 422 is not supported by NVENC.
Same user of a7s3, I tried 420 10bit HS 4K and 422 10bit S 4K , I can’t see the difference ....
in theory , XAVC HS 4K is 265 codec ,same 100Mb contains more information than 264 100M, so... I go with XAVC HS 4K currently even I can’t see the difference at all LOL cheers
Thank You. Not what I expected would happen.
Fantastic video. I have had that exact question. I have an A6500, but now I know that I don't have to upgrade right away. Thank you!
Good comparison video! I think the blockiness comes from the different compression not subsampling..
It's quite possible, yeah. Good point.
Phenomenal video man! You actually talked me out of buying a Ninja V for my X-S10. I could not tell that much of a difference enough to even buy a used Ninja for $250 on CL and think I'll buy a lens instead. Thank you and yup, consider turning off that tally light next time mate.
Amazing as always, honest, thorough, realistic ! Thank you, you are a great reference
Atomos only good for pro res to edit inside davanci. And export times. Other than that pretty much not needed. I did notice 35mbit h264 mpeg was smoother than prores. Pro res us very grainy and needs noise reduction in post.
Thank you Gerald. As I mentioned previously, I am not a Sony user, but I love tech stuff and I see how this topic can relate to other brands. 50K subs before you know it.
I appreciate the weekly encouragement, Phil. Feel free to never stop! 😃🙏😜
Excellent explanation and test. Keep up the great work!
Ur content is awesome and worth visiting one year later😊
Another advantage is to get rid of the 30min limit recording time here n Europe, which is useful for a show or an interview for example
I found this info very worthwhile, having a Fuji X-T4. Haven't even shot video yet but am encouraged that I can manipulate the colour grade so that it will not 'break". I'll be using DR 18. Thx Gerald...get undone more - ppl love it!👍
Yup. And that's why the Atomos Shinobi is great! Beats the SmallHD focus in price and performance.
bit depth data makes more difference than chroma subsampling
This was the video I was looking for. Thank you for your insights. I now know buying an Atomos recorder for my A7SII is not worth it. Thnxs buddy... Great video
I have the impression that you live in a quiet purple cave in the middle of nowhere and all you do all day is tests cameras and do experiments to get the best image quality out of everything...
Love your videos and how clear you are when explaining everything. Thanks for the amazing content!
PS. will you be releasing any purple merch in the future?
That's pretty much exactly right! Haha. 👍😜
Thanks for the kind words and support!
And yeah, I do have some merch available. Haven't updated the designs in a while though: www.redbubble.com/people/geraldundone/works/27198329-lets-get-undone?asc=t&p=t-shirt
What about chroma keying and green screen compositing?