A couple minutes in and I'm like: What on earth am I watching with all this math?!?! End of the video: I had to hit the subscribe button! This was very well explained!
I've been waiting FOREVER for Canon to answer this. If it's line skipping, my pre-order might be cancelled for the a7S III. Great explanation of how the downsampling modes work.
Yes I've also tried hard to get this answered, unsuccessfully. Maybe the local offices outside Japan don't know, or it's still in progress before the final release firmware, or some other reason. If it does line skipping, I'll also have to cancel my preorder and switching to Sony. I simply need to be able to record good quality 4k long-form video. IPB-only on the R6 just doesn't cut it, plus it line skips on the 120p 1080 as we saw on that boat clip, so that's a no-go too. Will be interesting few weeks ahead!!!
Just as I commented in another video regarding this. The R5 is a great stills camera with some cine capabilities, not a compact cine mirrorless with some stills function. Canon isn't insane enough to kill their Cine line
Canon should at some point, hopefully, see the advantage of selling any Canon camera to the end-user, rather than selling them a Sony, Panasonic, Nikon, or Fujifilm. The R5 is the first model in a while that suggests they might have finally caught on, body-wise. No complaints about the RF glass, other than price. Sony, too, seems to have kept the A-series crippled in video. But there's a difference. Sony has long had the technology to offer DCI formats, 4:2:2, 10-bit, or even raw recording. They chose not to. To avoid competition with CineAlta or to sell more Panasonics, it's hard to say for sure :-) I was pretty sure Canon was doing the same thing, intentionally wielding the Cripple Hammer so as not to compete too closely with Cinema EOS. But if you look at all the issues in Canons before these new models: no IBIS, cropped 4K, no DPAF in 4K, etc. it pretty much adds up to something else: Canon was just behind on technology. Old tech means more heat. IBIS makes it harder to cool a chip. Cropping lowers both the readout demands (sensor heat) and the image signal processing load (processor speed). All the specs suggested that Canon had finally licked these problems. The reality says that they've made progress, but haven't eliminated these issues entirely. So perhaps the older models were as good as they could have been given Canon's tech. Crippled relative to other competition, but not intentionally so.
That’s because stupid Canon hyped the R5 and R6 for video capabilities and being video centric. Now they have to eat the shit. And by the way, also the sensor of the R5 is just of old technology, it’s not even a BSI or stacked sensor. Especially the last point is resulting in a massive and ugly Rolling Shutter effect, it’s even worst than on 1DX Mark III in LV Mode.
Stefan1968ful From a marketing point of view, you would want to hype the video capability since it is a stills camera. A lot of consumers understand this, because if they were going all in on video they would have included a fan. These video features are a bonus
Stefan1968ful Sony fanboys have the least IQ of all. The Sony stacked sensor was developed to have more speed at 10fps mechanical and 20 fps electronic (12 bit compressed) in the A9. The new sensor in the canon 1dx mkiii destroys it in speed by doing 20fps mechanical (14 bit uncompressed). The A9ii buffer can do 239 frames in RAW while the 1dxmkiii can do unlimited. At 10fps mechanical, The Sony a7R IV buffer supports a maximum of 68 JPEG, 68 compressed RAW, or 30 uncompressed RAW images while the R5 again stomps it. At 20 fps, it'll write at least 84 Raws + Large/Fine JPEG files to a CFexpress card, a number that increases to 170 shots if you just shoot JPEG. In 12 fps mode you'll get 90 Raw + Large/Fine HEIFs, 160 Raws with JPEGS and 180 Raw files. Talk about processing speed. While Canon has the edge in speed of shooting and writing it to the cfexpress, sony was a tiny bit faster in AF locking. Neither are a whole lot better than each other. Just different sensors. Second thing is you probably didn’t pay attention to the photo specs of the R5. Even the R5 does 12fps mechanical and 20 electronic matching the A9mkii. That is a BIG DEAL. While Sony makes either a photo or video centric camera, canon does a true hybrid. Again both Sony and Canon are awesome cameras for different people. Fanboys won’t be able to do jack sh** with either cameras and can only complain 🤷🏽♂️
Stefan1968ful Ahahaha you didn’t insult any camera or hurt anyone bro... but you just made a fool out of yourself. I just spoke logic and facts, which is too much for fanboys to comprehend.. it’s ok 😊
Comparing apples and oranges. If I am a still photographer who occasionally shoots video, I want the R5. If I am a videographer who occasionally takes a still, I want the Sony. But they really don't compete.
Canon R5, R6 are mirrorless continuation of 5D, 6D famous series of photo cameras with video functions. Yes man. Didn't you really got that these are fantastic PHOTO cameras with great video functions?! While Sony A7s III has 12 Mp sensor. No photographer in 2020 wants a 12 Mp camera. So Sony's is clearly a VIDEO oriented camera, although, I suppose, it can take some photos too. So it's just not quite correct to compare a PHOTO camera with a VIDEO camera in domain of shooting video. And you had to point that out, at least. I do hope that all the insane vloggers who need so much to shoot their talking heads for hours uninterruptedly in SQ 4k to enjoy their own outstanding beauty in super crisp details, that they all will forget Canon R5 and go for Sony's camera. It's already too many orders placed on R5. So let the insane vloggers forget this fantastic photo camera and use Sony's for shooting for hours and hours on their clamorous mediocre stuff. The less insane people in Canon's world, the better. In spite of all the inadequate shouts about R5's "problems", it's also very clear to me that people with hands and brain will be able to create just fantastic video content with R5 & R6. Some people find tremendous faults with every camera; some people shoot tremendous art with every camera...
You are absolutely right that the R5 is an amazing photo camera, and paired with the new crazy sharp RF 50 mm f/1.2 or RF 85 mm f/1.2 plus the new superb eye tracking dual pixel AF, it will produce some mind bending portraits, while as you say, the A7s III is built to be a video camera only and there are even rumors that Sony disable an interpolation based, higher MP photo mode, I suppose to not get reviews of "bad photos". Thanks for the comment!
First of all,plenty of people are still shooting with 12MP cameras,because they are still being purchased,and that's many brands,not just Sony.Next,Sony happens to be the single scapegoat for bitter,and petty users of users of the so-called 'icons',and that's because Sony mopped the floor with them,in innovation,for the past 10 years or so.Adding to that,aside from the SD/DSLR series and 1DX2/3,Canon fell behind in innovation,except for the M%0,which was the leading mirrorless camera worldwide,which helped hide the failures of that EOS R fiasco.The problem of a camera overheating and possibly being damaged is not some 'inadequate' concern,unless the shooter has enough funds to just buy multiple cameras,no matter what brand.It's quite hilarious to get smoked for years in innovation from a company like Sony,who makes Walkmans,TVs,DATs,MiniDisc,clock radios,and game consoles,and was NOT a interchangeable lens camera company.Plus,cameras that do great video are not just for vloggers,and there's a ton of ways to use those features. Insane,is making general and inaccurate comments,with only opinion to back it up.Let's say I get a Sony to do video.I have a my own live project studio where i have made many mixtape projects,and I might want to stream live,just spinning or a live session with other DJs,recording artists,etc.That's not mediocre just because ONE person does not or cannot do the same.I'm going to look at the Canon R5,yet never because you nor anyone else has an opinion about it.It's my choice,my money,and the right tool for the job that matters..I'm speaking out to anyone who THINKS they know it all,who THINK their opinions matter when they are not buying it,and who have some kind of twisted,petty,and childish attachment to inanimate objects,so they think they can judge someone else's choices.No,I am NOT a Sony shooter at present,and yes I do own Canon cameras,along with other brands.I am not stuck on brands,and yes,a Sony/Canon/Nikon 12MP camera can take GREAT stills,in the hands of a competent,talented,and intelligent shooter.
@@michaels8597 Your stills argument falls flat for wildlife photography where more mp's are always welcome. The R5 will convert many 5 series DSLR users to mirrorless.
Thanks mate, love your handle. In fact my patents often came when I stumbled upon a tech problem before anyone else. The solution was obvious. Cheers mate
Really great Technical deep dive! on the different ways downsampling is achieved in these cameras. It's a great time for people in market for a new camera, so so many choices and there will surely be a great fit for mostly everyone. There is no one perfect camera. But we are getting one step closer! Thanks for the shout out appreciate that! Looking forward to the next video!
Thanks Rafael for stopping by and thanks for all your amazing content! Learning a lot from you and I have your videos on notification. I think all the camera manufacturers planned to top the line mid 2000 because the summer olympics were scheduled for this time. Everyone who can afford traveling to Tokyo to watch them, can afford to buy a new camera :-) Development cycles are long, so whole Olympics are delayed, it's still Christmas for us camera lovers! :-)
Great video. But just FYI, The R6 does 4k by oversampling 5k and not line skipping or pixel binning :) The R5 however does Pixel binning in some modes which makes it the perfect hybrid for most people. 8K RAW and ALL-I / IPB up to 30p 4K HQ from 8K sensor readout up to 30p 4K pixel binned mode up to 30p 4K / 60p - also pixel binned 4K / 120p - also pixel binned www.eoshd.com/8k/a-closer-look-at-the-canon-eos-r5-lower-quality-4k-mode-to-avoid-thermal-cut-off/
Hi Ali, thanks mate. Yes I saw that article and I tried to dig into the source. My conclusion was that the R5 pixel binning mentioned there was still rumor based - but I also hope it's true :-) I also asked one Canon representative and they said they don't have that information yet. The R6 line skipping I showed in this video was indeed line skipping when the R6 downsampled the 5k sensor to FHD (2k) HFR. You are right that the R6 does not line skip to get from 5k to 4k. Cheers and thanks for the comment
I honestly think the r5 is the ultimate hybrid camera! It's not perfect but the closest thing to the dream hybrid camera. Wish it had 6k 50fps and 1080p 240fps though as it should be able to handle it. Also with you could shoot raw in slow motion however I think this camera will be seeing firmware updates that will make it even better and it's a godly hybrid camera!
Haha then we share that experience too. Before this video I got like 3 views - with 2 out of those probably being my wife and my old mother - and as you can see, I put quite a lot of effort into some of them. So the YT algo was not exactly my best friend. Let's see if we can make up now
Hey I just stumbled on this video and wow, it's so great to see a really nice technical site that cuts through subjective BS and sticks to the science and explains it simply and succinctly. For 99% of users R5 will be a perfectly fine camera for video and not many people buying R5 are cross-shopping it with A7sIII. I have less than zero interest in a 12MP FF camera in 2020. R5 is a massively better all-round camera, A7sIII is a one trick pony for serious video and lowlight users.
Thanks mate. Now that we have both cameras out and tested, turns out this simplified calculation was almost exactly spot on. I made a part 2 (the a7S III vid) where I elaborated a bit on this. Now we know that the R5 doesn't pixel bin, so the best we can get is the 8k downsampled, 4 S/N in the example, vs 20 for the bigger pixels, i.e. a factor 5 difference. That's almost precisely the measured value: 2 stops of noise limited difference in dynamic range. And that means Sony pretty much won, for video.
Hello, About the last Fujifilm camera, does it mean that XT-4 is also using line skipping or pixel binning? So, does XT-4 have the same hoverheating and noise problems?
Can pixel binning be added in firmware or is it hardware dependent ? Also it might be limited to line skipping so that mode remained unaffected by heat
It's hardware dependent in the sense that it needs to be supported by the sensor. Then of course the firmware can opt to use it or not, so the only remaining hope is that they haven't switched it on. Slim chance though. If it was supported in the sensor, I think they would have used it from day 1. So no.
Nop. The only new fact since I made the video is some actual test footage here: vimeo.com/439692505 Unfortunately unfortunately unfortunately the guy tested every mode EXCEPT the Log mode in 4k non-HQ, where it would have been immediately visible if it does line skipping or pixel binning (because the noise in the dark areas would either have been a lot better or a lot worse). But now since it's such a bright sunny mid-day, you can't see anything in the shadows really. It does confirm that the full sensor 8k readout downsampled to 4k in software preserves the thin spatial details though because the "HQ" mode is visibly sharper, though I haven't tested yet how much difference it is vs adding a little bit of sharpening on the non-HQ mode.
Pixel binning would lead to debayering artifacts similar to line-skipping (they don't use a quad bayer layout), so a full sensor read-out definitely would be preferable. But maybe they bin at least the dual-pixel pairs.
On the mobile phone tech side, we've used pixel binning for almost 10 years. It is possible. But you are still right because turns out the R5 doesn't pixel bin. Hence the 4k non-HQ mode is rather useless, and hence the whole camera drops in usefulness for video to almost EOS-R level.
Thanks mate. They didn't get back, but I know the answer. The R5 doesn't pixel bin in the non-HQ mode. It's a type of line skipping or at least discarding part of the data. So the lowest noise 4k mode on the R5 is the "HQ" mode, meaning it's a factor 5 difference in signal-to-noise between the R5 and the bigger pixel design like the Sony a7S III. A factor 5 corresponds to a roughly 2 stops, which is exactly the usable dynamic range difference measured. I actually made a follow-up video with that, called "divorce"....
Great video! def goes deeper than anything else I saw on the subject. I just posted it as an announcement on the FB A7iii for video shooters group - sending 28,000 members your way. Bravo
Thanks Elliot for those kind words and for sharing on the video shooter's group. Appreciate it immensely! Thank you! Also noting that you were an early adopter of RUclips, well done staying at the forefront of tech.
I love your technical and detailed review!!! I am an electrical engineer and absolutely appreciate your incite and analysis. I hope to see much more of these from you. Can you tackle the subject of anti-alias filtering? And in particular why does the high end Nikon and Sony not have one but Canon continues to use one? I would love to hear your technical view on this one. Other youtubers (and I must say much less technical than you) seem to either avoid this topic or simply don't know enough to talk about it in much detail.
WELCOME, a fellow engineer! YES! Yes anti-alias filter is a super interesting topic. Another one I started on is the lens IS vs IBIS. Maybe you saw the jarring corner wobble on Peter McKinnon's latest vlog, but seems you can't disable the R5 IBIS separately from the lens. Switching the lens IS button to off will apparently disable the IBIS too (!) If that's true that's... odd and makes that topic less useful in practical terms. Will definitely look more into the antialias filters, researched it a bit long time ago (loooong time ago lol) great idea mate!
@@ctoLarssonTECH Wow great idea to do IS vs IBIS. Another topic that has mostly fluffy marketing hype around the words with little real technical analysis. I will look forward to that one. On the topic of anti-aliasing it's another one with words being used that mean different things to engineers (like low pass filtering, oversampling,...etc). What does it really do and why is Canon the only professional line of cameras that really say they use them? I tried once renting a Canon 5DSR to compare to a Canon EOS 5DMIV but then the sensors being 45MP vs 30Mp was not really a fair comparison. Anyway, I ramble now. Looking forward to more cool stuff from you. Keep it up.
Canon should just do like Sony and make one camera biased to video and the other to stills. That was smart on Sony’s part. There are just too many physical limitations to trying to use the same sensor for optimum stills Image quality and optimum video image quality. Its like using a computer with a integrated graphics as opposed to discrete graphics. Expecting the CPU and GPU to share the same system memory and perform well. Create a discrete senso put it in a discrete camera.
Just found your channel and subscribed - more videos like this would be greatly appreciated - especially a follow-up video when the production models come out.- I love my EOS R and 3 RF lenses - so I pre-ordered the R5 - I hope this is not a very expensive mistake - Great Video!
Thanks mate. Haha yes I hope so too because I've also pre-ordered one. Tried to get hold of Canon Sweden but they're all on vacation haha. Anyway, they Canon engineering might still be working on the final firmware, so I want to test the production version after sale start not the pre-release firmware around today.
One thing I don't like about Canon is that they don't clarify how the different modes are read from the sensor. Panasonic does this, and I appreciate it so much! Let's hope the Canon R5 is truly tuned and ready for videographers, though by seeing the IBIS, it is not. Let's also hope the Sony A7SIII actually does pixel-to-pixel 4K readout, with a stacked, backside illuminated sensor.
Agree, it wouldn't hurt to release some technical details. Yes that corner wobble in Peter McK's latest vlog that came after I made this video was actually even worse than the first one. And then came another issue, I heard yesterday from one Canon rep that disabling the Lens IS with the button on the lens is how you disable the IBIS. That would be absolutely ridiculous, because using Lens IS only (without IBIS) would be the solution to any IBIS wobble problem. In that case you can't do that. But that sounds so silly that I wonder if it's really true.
@@ctoLarssonTECH Sony has the same issue with disabling IBIS and Lens Stabilization; disabling it on either disables both. :( @geraldundone is doing some chart tests and seems to have found the R5 is using line skipping in 4K/60 and 4K/120 modes... not pixel binning. (another sad face)
The R5 is 8.2K (or 8192) pixels wide. My guess is that the 4K30p Full Sensor unlimited mode is using simple 2x2 pixel binning of the 8K30p image. That math is possible for ALU (i.e. add together 4 integers and then shift 2 bits right to get the divide by 4). But 2x2 only requires 7680 pixels so they may be skipping lines. For the HQ mode, they are sampling the full 8192 pixes down to 3840. This requires divison operations, which uses more transistors and CPU cycles. (This logic seems to explain overheating issues with R6... 5.1K is about 4/3rd 3.84K so a lot more floating point is required to produce 4K with the R6.). As to why line skipping 8192 down to 3840 is possible but not 5.1K down to 3840, I think it has something to do with the Nyquist rate where you need at least twice the frequency to recover the sample.
The difference of pixel binning vs scaling it down in software/hardware by shifting bits or dividing, is that the pixel binning aggregation happens *before* the amplifier, and since the noise is mainly introduced at the amplifier stage, that's the big difference. Now that we have the data, the R5 doesn't pixel bin, which means the best read-out remains the "HQ mode" which is a full 8k readout. The calculation in this video turned out pretty spot on, because the estimated 5x better S/N with the bigger pixels vs the HQ mode is almost exactly the result (2+ stops in noise limited dynamic range) between the a7S III and the R5 HQ. I made a 2nd part where I touched on this.
Hmmm, that might work with a B&W sensor, but we're dealing with a Bayer filtered colour sensor here... All binning needs to be done in digital form so as to preserve the colour information!
You bin them per color. The illustrations would have been too messy and hard to follow if I drew 3 of each one. In mobile phones we've done pixel binning (with color) for almost 10 years. There are even phones today like the Galaxy S20 Ultra that dynamically shifts between a full sensor readout giving 108 MP in bright lighting conditions (where S/N ratio is not so important) and a 3x3 pixel binned 12 MP mode (108/9=12) in dim lighting where you need to boost the signal without introducing too much noise. But like I said, I have no idea if this sensor can do it or not. The sensor has to be built for it.
You calculate that pixel binning yields a signal / noise ratio of 8, which is better than the than the 4 s/n ratio you estimate for 4k down-sampled from a full readout of a high megapixel sensor. This conflicts with many reviewers' notions that binning and line skipping are inferior and yield comparable problems like moiré. Such reviewers also often claim that 1080p video down-sampled from 4k files can be an improvement on 1080p recorded using a sensor with fewer than 8 megapixels. According to your perspective that down-sampling would also yield a worse s/n ratio. Comparison of actual samples might offer proof, or not. After all, how can anyone peep pixels at usual video frame rates with the usual preferred motion blur? Most of the R5 / R6 sensor heat problem could have been solved with a small ventilation duct and fan, which the user might have the option to turn off and seal shut in wet or dusty settings. Even with the camera body sealed to use copper instead of magnesium for the surfaces below the flip screen or its back side would improve heat dispersion. But heat may be a rare problem in Sweden, right?
Turns out this simplified calculation was almost exactly spot on. Now we know that the R5 doesn't pixel bin, so the best we can get is the 8k downsampled, 4 S/N in the example, vs 20 for the bigger pixels, i.e. a factor 5 difference. That's almost precisely the measured value: 2 stops of noise limited difference in dynamic range. And that almost means Sony won. I made a part 2 (the a7S III vid) where I elaborated a bit on this.
You bin the red color from all 4 pixels separately, the green color separately and the blue color separately. Your mobile phone probably uses it. For example the Samsung S20 Ultra even shifts dynamically between a 108 MP mode and a 12 MP 3x3 pixel binned mode. In daylight, it can do 108MP photos for more effective cropping without much noise. In low light, it will dynamically shift to a 12MP mode, using 3x3 pixel binning.
@@ctoLarssonTECH But each pixel in a Bayer array sensor only captures one color. In a 2x2 grid, 1 gets red, 1 gets blue, and two get green. If you're just summing signal from these four pixels, then you would lose which pixel (red, green blue) collected what light. Maybe there are some intermediate steps?
@@RossMcLendon Uff, right. However, there are new technologies introduced with 1DX Mk III with the new low-pass filter. I'm not sure how that comes into play, does it help things or does it complicate things further. Points 4 and 5 in this document say some things about the new technologies in 1DX Mk III sensor, but it doesn't tell me what was changed. www.canon.fi/pro/stories/eos-1d-x-mark-iii-vs-mark-ii/ "But with this new system we are actually sub-sampling it 16 times and using a complicated algorithm to average it out. This then gives us that happy medium between having a low-pass filter and having no low-pass filter. The new design actually gives a slightly higher resolution than the Mark II because it's resolving more detail, while still reducing any chance of moiré."
@@ctoLarssonTECH It's a Bayer sensor. A 2x2 pixel cell has one red pixel one blue, and two green. Samsung's 108 megapixel sensor is a Nonacell chip -- it's got 9x9 blocks of red, green, and blue. So yeah, that's binned on-chip. Their other sensors are Tetracell (same as Sony's Quad Bayer), so you have 2x2 blocks of the same color, and those are easily binned on-chip. But for binning on a full Bayer sensor, which is as far as anyone knows or expects what Canon's using here, you have RGBG pixels in that 2x2 bin. It needs to be binned digitally, after ADC, before video compression. Even that's an advantage, as it saves the need to do Bayer interpolation. But you're saving image processing work, not sensor readout work. There is some supposition that Sony's using a Quad Bayer sensor on the A7SIII. Mostly because there was a leak of a Sony full frame (aka, 2.3") sensor with a quad bayer color filter array (otherwise it's basically the A7RIV sensor).
Din video dök upp efter att jag sökt på Canon R5 och hoppsan, en svensk video! Visste inte ens att det fanns för detta :D Kul! Prenumeration givet! Har du fått svar på om det är 2by2 pixelbinning?
Oj vad roligt med en svensk!!! Ja vi verkar inte ha haft framfötterna med oss riktigt på detta område, vi får fixa det. Försökte faktiskt kontakta Canon Sverige också och se om jag kunde låna en kamera och göra en uppföljningsvideo. Som man kunde gissa så är det semester haha. Fick tag på någon ute på en båt till slut, får se om de återkommer 😆 Inte fått svar på pixel binning, tror inte de vet. Var sitter du? Jag sitter i Stockholm.
@@ctoLarssonTECH Ja, kunnigt folk i Sverige verkar det vara få utav. Intressant video. Kikade även på din långa video där du jämför stabiliseringen, men hittade inte uppföljningsvideon där du summerade dina tankar? :) Jag sitter i Umeå! :) Har beställt 2st R5 av Cyberphoto så vi får se när de blir tillgängliga. Själv nyttjar jag väldigt lite video, utan är mestadels en stillbildsfotograf, men kommer nog börja filma mer. Speciellt norrsken så har en a7s2 som ska säljas till förmån för R6an.
Great presentation of the concept of SNR. Have the EOS R, love the EOS R admittedly more due to RF lens, but thrilled Sony is pushing the envelope!!! The new Alpha S will be very good in low light. Physics is physics, big pixels mean better light gathering and more efficient/ easier readout...
Yes it's primarily an optical issue that can only be reduced not solved by moving the sensor. Interestingly, with the lens IS in the RF 15-35 at 15mm with the EIS in the EOS R, it does not wobble, see my video on "96 stabilization..."
The issue is that when you move the sensor you create paralax artifacts that cannot be corrected. Essentially you sometimes can see behind foreground objects, other times not, and that results in the foreground appearing to wobble. All mechanical stabilization methods will do this, both in lens methods and IBIS. To see the effect, hold a finger in front of your face with one eye closed, then move your head (the sensor) from side to side, sort of like IBIS is doing. Even though your finger and the background are not moving, your perspective is, and consequently it looks like your finger is wobbling against the backdrop. This will only happen with wide angles because your subject matter is normally close enough to the lens for paralax to become noticeable. At long focal lengths the subject matter is far enough away for the differences in perspective to be too small to be observable.
I have question, just bought R5($4300) with RF 24-70 f2.8, unfortunately not able to get any sharp images at all,and find out My R5 r5 sensor misalignment, I’m not sure is really misalignment or because not latest firmware , I’m thinning return my R5 and buy new R6 or just replace with another R5, any thoughts? Thanks
I really appreciate the explanation but the cameras are already out there and we have seen what the video quality looks like. That is what matters at the end.
Great video but something doesn't pass the smell test. Why would they call it HQ high quality mode if it were lower quality than the normal pixel binned mode? It must be preserving more color information or something in HQ mode.
Valid point. Pixel binning is widely used though, in mobile phone tech we've used it for soon 10 years. The con of using full sensor read-out is worse noise than pixel binning (and processing power.= heat). The pro is that the spatial information is better, so with smart algo you can make a sharp hair look like a sharp hair also when downsampled, while it will look more like a blur in pixel binning. That could warrant a "HQ" stamp. Still, we don't know if the R5 does pixel binning in normal mode. Maybe it doesn't. Then that would explain the "HQ" stamp on the full sensor read-out.
Very eloquently explained. Thanks for putting this together and I hope this reaches wider audiences. I was thinking along the same lines : the R5's hardware in a larger XA style hybrid body to accommodate the power and cooling needs of prolonged 8K shooting would be the ticket. As for now, I just hope the R5 is a stellar stills camera, I hope it is a stills camera at all!n
Thanks Shaun for the kind words. Yes I think it's a very small step to take the R5, put it in a bigger body, add some 20-50% computing power and Canon got themselves a killer video beast with the RF system. I think the R5 will be a stunning still camera, especially paired with the new sharp RF f/1.2 lenses. The eye AF plus the RF system AF speed will make that shallow depth of field a lot more usable.
@@ctoLarssonTECH If Canon does not do this, a very clever DIYer might and find themselves with some serious popularity. Who will do it first? Maybe the Magic Lantern crew? Anyways, I hope to switch to the R5 soon (from 5D Mark IV) and get used to it in time for the world to thaw so I can use it! Strictly a stills shooter. I'll never forget my old 5D Mark 3 and 50mm F1.2 mis-focusing time after time late at night in Cologne Germany, the 5D Mark IV really tamed that lens and put it to work, I'm expecting an even greater increment of improvement with the R5.
It's an optical phenomena when you move the sensor, you can't keep both the face and the background stable (without using software). That's why the lens IS + EIS doesn't produce the same result as sensor OIS. Check my vid with the same RF 15-35, "96 ways to stabilize". Cheers mate
Thank you, really appreciate those words. I've now made a part 2, the "Divorce" video, hope it can add some further insights regarding the video capabilities and also the company outlook of Canon Inc vs Sony
some idiots have been complaining about "pixel binning" for the last 6 years (e.g. in A7r2), while actually they were complaining about the picture quality of line-skipped video. I hope this video will set them straight and expose them for frauds that they are.
Yep that's true. Also they probably don't know how many mobile phone cameras rely on pixel binning to produce such great pictures from so small sensors.
Wow, I wasn't expecting a maths lesson... my head is fried 🥴 Do the the cameras shoot crisp clear images... end of lol we've had ground breaking visuals for around 20 years give or take. Everybody seems to be splitting hairs all of a sudden... P.s please remember that the racing car scene in iron man 2 was shot on a CANON 5DMKII 👀 Just give me 4k full frame 24 to 120p and I'm happy... come on sony, I'm waiting 😄
Thank you for the encouragement, I continue to grind on. Made one on the Sony a7S III two days back. I also have two more channels (investment related).
20 fps cannot be sustained long enough to be usable for video, so even though it is there it is still a long way from being usuable. Also, most of the heat is coming from the processor, not other parts of the camera. If you don't believe me, take a RX100V and record in 4K until the camera overheats, then feel where it is getting hotest. By far the hotest part of the camera is at the rear directly above where the LCD hinge is, and that is exactly where the processor is located. The processor is the part that is getting too hot and shutting down (more accurately, it is throtling and can no longer sustain reliable encoding of the video stream)
Right, correct today 20 fps cannot be sustained long enough in the R5. What I mean is that a mere 20% improvement is nothing, if moving from that form factor into the C500mkII, C300mkIII form factor. Hence they already have the technology to do that and it will come very soon. Interesting with the location of the processor. I didn't know that. Right, that would be the place to put some contact gel and try to lead off the heat.
@@ctoLarssonTECH It still would not last longer than part of a minute though. The main reason those burst modes work is because the raw data is just being stored in the buffer, the processor can't come even lose to handling it in real time. So after a certain number of frames the buffer fills and frame rates drop drastically. That 20 fps does not last forever, it lasts until the buffer fills. The real sustained frame rate is more like 6 or 7 fps.
Large sensors are far more likely to have heat issues than tiny 1" chips. Most full frame cameras doing video have had some limits set by the sensor itself, independent of the signal processing. Now of course, the amount of signal processing needed for a 4K raw image is the same on an RX as it is on a full frame pro body, but the pro body is also likely to have far better cooling of the image processor. The other thing -- IBIS also makes it more difficult to cool the sensor, which seems to be the main reason Canon didn't have it until now. This is an upgrade in Canon's sensor technology, but it doesn't cure all ills.
Hej Anders, this is Travis from Allt för Sverige season 2. I enjoyed the video and your attention to detail. I think this channel is going somewhere. Subscribed. I pre-ordered the R5 for my photography and video work. I hope we are pleased with the camera and it's handling of 8k to 4k. Canon needs to come through with some competitive full frame 4k here. I also hope the IBIS wobble I am seeing in Peter McKinnon videos can be worked around. I'm not too worried about the overheating. I don't expect to be filming long clips of 8K or 4K 120. Did you order the camera? Who will get their's first!? I have ordered through my local camera shop instead of online in hopes I can get my hands on one in early August. good luck! tack, hej.
Hello Travis, thanks a lot for those kind words. Really appreciate it. And thanks for the subscribe. Yes I also pre-ordered the R5. There is zero doubt that it will be an absolute killer camera for photography with the f/1.2 RF lenses. The eye AF is super fast with the RF lenses, which will make that kind of razor thin depth of field work in practice. HOWEVER, that IBIS wobble in PM's vlogs however has me really scared, for video. That was horrible, I've seen his amazing vlogs for years and never seen something looking that bad leave his hands. Even worse, yesterday I heard a Canon rep say that that you can only disable the IBIS by disabling the lens IS, with the button on the lens. But that sounds so stupid that I wonder if it's true. (Otherwise we could just go back to Lens only IS for video, which doesn't wobble.) So right now I'm in stand-by mode. I do mainly video. If the R5 doesn't pixel bin and has an IBIS that doesn't work for video, and Sony incorporates their Venice color science in the A7s III, I'm going to bite the bullet and switch to Sony, which would such a bit since I have a lot of lens. In Sweden the pre-orders aren't binding. Let's monitor this next weeks!
@@ctoLarssonTECH yes many of us Canon folk have held back on the emigrating to Sony in hopes for a camera like this. I still have fingers crossed Canon has come through! Thanks
@@ctoLarssonTECH hej they called and told me my R5 will be arriving in store this Wednesday but I cannot pick up until Thursday per Canon rules on release. I would expect the same for other official Canon retailers. I have my fire extinguisher 🔥and dramamine ready!
@@ctoLarssonTECH have you heard about the scotch tape toothpick hack on R5??? the overheating saga has me on the edge of my seat. I hope the happy ending results in the R5 I had hoped for; for video. For photography, it has been a dream! Shooting a wedding with it this Sunday.
Most web ressources say hardware bining is not possible with CMOS sensors, because each pixel has its own readout. That's one of the big differences from CCD to CMOS. What you described makes perfectably sense for CCD sensors, but CMOS sensors do not shift lines or pixel before readout.
Let's see. On the mobile phone camera side we've done pixel binning for almost 10 years. The Samsung S20 Ultra even dynamically shifts between 108 MP full sensor read-out in good lighting conditions and a 12 MP 3x3 pixel in lower light. There are many different sensor types.
Once I get the production version of the camera, we can test this out. If the R5 does pixel binning on the 4k "Normal mode", that mode will probably have *better* low light performance than the "High Quality" mode and should be a good fit for your wedding work, which I imagine contains a lot of low light evening work? If it doesn't, the R6 will be a better choice. Having said that, I know other wedding photographers who will go all out for low light performance, and for that the A7s iii will beat both of them, while having some other compromises like no real good still photos, and (IMO) not as good lenses as the new RF lineup from Canon.
Thanks mate. Turns out the calculations were pretty spot on, now that we have the camera tests available. The R5 doesn't pixel bin, and the 5 times better signal/to noise with the 4x bigger pixels on the Sony vs the R5 is pretty much what we got, about 2+ stops better. I made a part 2 with the video comparison (the "divorce" video)
If the sensor supports it on low-level hardware, it can be possible to add in firmware later (even for the R6 example I showed in the video, which was 5k sensor downsampling to 2k 120fps in R6)
In order to take advantage of the improvements as shown, the sensor would need to support on-chip pixel binning, since they'd need to combine the output of four pixels into the analog-to-digital converter to see the SNR reduction as shown. What's termed "readout noise" in modern sensors is actually the noise leading up to the ADC stage. Once you're digital, there's no more noise. In past cameras (Canon's 5DIII, for example), the sensor actually read out in analog, routed that analog pixel bus across a PCB to the image signal processor, which did the ADC part. So readout and analog noise were once pretty identical. But that's fortunately ancient history. On a CCD, you technically move charge without adding noise until the charge to voltage conversion and the ADC stage. In a CMOS sensor, you generally don't move around charge, but voltage. However, I image if you had 2x2 pixel cell that you'd like to have read as a bin, it would be simple enough to merge the four charge wells for the pixels and read them out as one superpixel. I'm pretty sure my phone does this :-) The problem with that is that you're binning a 2x2 pixel cell, which is RGBG. So you can't actually process that in analog with a single ADC. So I don't really see this being an on-chip option, as it is on the recent Quad Bayer sensors used in phones (and probably cameras before too long). It's of course possible to do pixel binning after the ADC step and reduce overall luminance noise by a factor of 2, and that certainly could be done in firmware.
I worked in mobile tech for many years, these problems are the same. Plus pixel binning was a key tech to get low light photos out of high res mobile phone sensors already many years back.
Such an expert! Could you please give me your opinion?🙏🏽🙏🏽🙏🏽 I‘m wondering whether an 8k sensor would make a line skipped 4K image look better than a f. e. 5.1k sensor (R6)? Or is it the opposite? No one can tell - I hope you can explain!🙃🙏🏽
The R6, with its larger pixels, is going to better signal-to-noise ratio than any of the read-out modes of the R5, which means that it will have better low-light performance. This means you can run higher ISO on the R6 than on the R5 without ruining the video with noise. (This is assuming the sensors are of similar tech generation.) The R5 on the other hand, will have better spatial resolution if using the 4k HQ mode, which will downsample the 8k read-out and in doing so it will try to maintain thin details. That will make the R5 4k video "sharper" than the R6 4k video. So "better" here depends on what you prioritize. What I really dislike about the R6 is that it does IPB only, no All-I, which slows down the edit and can destroy some footage of grass and water etc. Also as you saw in that boat sequence, the R6 seems to lineskip to do the 120p 1080 and that is bad. Hope helped you somewhat!
CTO LARSSON - TECH hey it’s me again ✌🏽 as we do know now, the R5 does use line skipping in normal 4K... do you know if it’s technically possible for canon to switch to 2x2 pixel binning via firmware update sometime in the future? And furthermore, do you think that the camera can be cooled enough by external gear (tilta fan, RF Adapter with fan, etc.) so that it can compete with the A7SIII? Wish you all the best!
do you think the info from EOSHD answers the pixel binning questions? www.eoshd.com/8k/a-closer-look-at-the-canon-eos-r5-lower-quality-4k-mode-to-avoid-thermal-cut-off/#:~:text=In%20order%20to%20achieve%20the,work%20for%20the%20image%20processor.
I know there are rumors like the link below on the EOSHD forum, but I couldn't confirm them. I checked with one Canon representative and they hadn't gotten this information yet. The final firmware for release isn't done yet, maybe in progress still. The good news is that no one has observed the characteristic aliasing, so I'm quite hopeful that it will turn out to be pixel binning. Fingers crossed
Cast magnesium alloys, while that's a wide range of materials, have a proof stress (where it dents) of typically 75-200 MPa and a tensile strength (breaks) of 135-285 MPa. That's not much. It will blend.
From eoshd.com: “In order to achieve the higher frame rates, the sensor uses 2×2 binning, similar to the 4K video mode on the 60 megapixel Sony a7R IV. The 4K 30p non-HQ mode is read from the sensor in the same way. Since this has a much lower frame rate, there’s less heat and less work for the image processor. Thus the 4K 30p (and 24/25p) pixel binned non-HQ mode does not have a thermal cut off.”
That implies pixel binning off-sensor. It's certainly the case that a 4x4 pixel binning uses less (virtually no) image processing than debayering and then a whole sensor downrezzing. That may be what's at work here. People complain about Sony's A7R series doing pixel binning rather than whole sensor scaling, so this seems a likely explanation. There seems to be some belief that a 6-8K image rendered to 4K and compressed to 100Mb/s AVC is better than a 4K image compressed to 100Mb/s AVC. I suppose it depends on the full resolution downscaling algorithm. Though the pixel binning ought to have less color error since there is no Bayer interpolation needed.
I can’t see the point @ 7:15 about RAW video. Olympus has 60 fps RAW stills on OMD EM1 2 but they never even did 4K 60 fps video in this cam, not mention the RAW.
The point I was alluding to there was how close they are to delivering the same quality as a still photo in every frame in a 24 fps video. It's basically just 20% off, which is peanuts if moving to a bigger form factor. So I'm certain that they are already in progress of readying a cinema camera with that as we speak.
@@ctoLarssonTECH Agreed. If you jump over to CanonRumors.com, you'll see both rumours of a new cinema camera with RF mount, as well as new RF mount adapters. Both are rumoured to be coming in 2020, so be careful if you're pulling the trigger on the R5. If you only need video, you might be wise to wait.
interesting. kinda technical, but I get it. the gh5S also used 12mp stacked i think, same thing. well thanks for the info. Marty of Marty Sender RUclips peace out
good video but man towards the end of it i have to question your wisdom and mostly ask if you have ever used Sony system at all when you mentioned Sony lenses are not as good as Canon RF LOL. Sony has GM lenses on pair with Canon RF and on top of that Sony has ton of E mount fully compatible specially designed fantastic lenses for stills and video from Zeiss Loxia, Batis, Voigtlander, Tamron, Sigma and other.
You are right, that was a very sloppy statement. If there was an Edit button on youtube videos I would have rephrased it. What I meant was that there exists no lens equivalents to the RF 15-35 f/2.8 (Lens) IS or to the RF 70-200 f/2.8 IS (that compact) on the Sony system.
@@ctoLarssonTECH No worries, i thought you meant something like that. This Canon compact 70-200 2.8 seems really great though.Im sure other lens manufacturers will follow up with similar design.
Great technical stuff. I am not smart enough and you lost me. I only care about the final image. Let’s try it out and give it a final verdict later. I told my friends many time that Canon, Nikon, Sony, etc. all make good cameras. Buy one within your budget and enjoy. I am still amazed by Ansel Adams works. Was his camera better than today’s cameras?
Absolutely. I just love to geek out sometimes. At the end of the day you can take any camera from any manufacturer today and produce award winning content. The Nobel Prize winners in literature didn't win because they had the best typewriter.
So the Canon can be a good still camera and a good video camera, while the Sony can be a poor (12MP) still camera and an excellent video camera, as far as sensors go.
Yes I think you're spot on there mate. Canon will be killing it for photography with those RF f/1.2 together with this eye AF and sensor and super speed AF interface. The Sony will be killing it in low-light video for sure and based on this analysis, and probably in 4k video overall. But 12 MP photography in 2020? Not so great...
Your explanation is great! I found a lot of downloadable R5 material here: vimeo.com/user803551 The oversampled footage looks great, the normal (pixel binned or line skipped) video has noticeably less detail. It is interesting to note, that the normal 24p and 120p footage has the same quality, which means that 50p/60p video will also have the same quality. This also means that the 50p/60p 4K mode has better quality on the R6, which is oversampled! Canon is planning to add additional raw options to the R5, I am curios about that.
You’re the most technical reviewer out there man, you deserved million views and subscribers. Keep it up!
Thanks a lot for the kind words. I do love tech :-)
agree, this was super technical & useful, love it 👍
A couple minutes in and I'm like: What on earth am I watching with all this math?!?!
End of the video: I had to hit the subscribe button! This was very well explained!
Thanks for those kind words. School has destroyed math for generations. Math is more fun out in the real world.
I've been waiting FOREVER for Canon to answer this. If it's line skipping, my pre-order might be cancelled for the a7S III. Great explanation of how the downsampling modes work.
Yes I've also tried hard to get this answered, unsuccessfully. Maybe the local offices outside Japan don't know, or it's still in progress before the final release firmware, or some other reason. If it does line skipping, I'll also have to cancel my preorder and switching to Sony. I simply need to be able to record good quality 4k long-form video. IPB-only on the R6 just doesn't cut it, plus it line skips on the 120p 1080 as we saw on that boat clip, so that's a no-go too. Will be interesting few weeks ahead!!!
@@ctoLarssonTECH I GOTSTA KNOW! LOL
Just as I commented in another video regarding this.
The R5 is a great stills camera with some cine capabilities, not a compact cine mirrorless with some stills function.
Canon isn't insane enough to kill their Cine line
I agree, my big fear now is that I need to get both, the R5 for photos and the A7siii for video 😆😂
Absolutely correct. 5D4 replacement for me.
Canon should at some point, hopefully, see the advantage of selling any Canon camera to the end-user, rather than selling them a Sony, Panasonic, Nikon, or Fujifilm. The R5 is the first model in a while that suggests they might have finally caught on, body-wise. No complaints about the RF glass, other than price.
Sony, too, seems to have kept the A-series crippled in video. But there's a difference. Sony has long had the technology to offer DCI formats, 4:2:2, 10-bit, or even raw recording. They chose not to. To avoid competition with CineAlta or to sell more Panasonics, it's hard to say for sure :-)
I was pretty sure Canon was doing the same thing, intentionally wielding the Cripple Hammer so as not to compete too closely with Cinema EOS. But if you look at all the issues in Canons before these new models: no IBIS, cropped 4K, no DPAF in 4K, etc. it pretty much adds up to something else: Canon was just behind on technology. Old tech means more heat. IBIS makes it harder to cool a chip. Cropping lowers both the readout demands (sensor heat) and the image signal processing load (processor speed). All the specs suggested that Canon had finally licked these problems. The reality says that they've made progress, but haven't eliminated these issues entirely. So perhaps the older models were as good as they could have been given Canon's tech. Crippled relative to other competition, but not intentionally so.
But the most important point is, this is a Photo camera, but everyone seems to discussing about video capability....
That’s because stupid Canon hyped the R5 and R6 for video capabilities and being video centric. Now they have to eat the shit. And by the way, also the sensor of the R5 is just of old technology, it’s not even a BSI or stacked sensor. Especially the last point is resulting in a massive and ugly Rolling Shutter effect, it’s even worst than on 1DX Mark III in LV Mode.
Stefan1968ful From a marketing point of view, you would want to hype the video capability since it is a stills camera. A lot of consumers understand this, because if they were going all in on video they would have included a fan. These video features are a bonus
Stefan1968ful Sony fanboys have the least IQ of all. The Sony stacked sensor was developed to have more speed at 10fps mechanical and 20 fps electronic (12 bit compressed) in the A9. The new sensor in the canon 1dx mkiii destroys it in speed by doing 20fps mechanical (14 bit uncompressed). The A9ii buffer can do 239 frames in RAW while the 1dxmkiii can do unlimited. At 10fps mechanical, The Sony a7R IV buffer supports a maximum of 68 JPEG, 68 compressed RAW, or 30 uncompressed RAW images while the R5 again stomps it. At 20 fps, it'll write at least 84 Raws + Large/Fine JPEG files to a CFexpress card, a number that increases to 170 shots if you just shoot JPEG. In 12 fps mode you'll get 90 Raw + Large/Fine HEIFs, 160 Raws with JPEGS and 180 Raw files. Talk about processing speed. While Canon has the edge in speed of shooting and writing it to the cfexpress, sony was a tiny bit faster in AF locking. Neither are a whole lot better than each other. Just different sensors.
Second thing is you probably didn’t pay attention to the photo specs of the R5. Even the R5 does 12fps mechanical and 20 electronic matching the A9mkii. That is a BIG DEAL. While Sony makes either a photo or video centric camera, canon does a true hybrid. Again both Sony and Canon are awesome cameras for different people. Fanboys won’t be able to do jack sh** with either cameras and can only complain 🤷🏽♂️
Ali Mustafa I insulted you toy camera, right? You feel now butt hurt, right? :) Good luck, stupid nonce LOL
Stefan1968ful Ahahaha you didn’t insult any camera or hurt anyone bro... but you just made a fool out of yourself. I just spoke logic and facts, which is too much for fanboys to comprehend.. it’s ok 😊
Finally someone that knows the real tech specs . Instantly subscribed
Thanks a lot for the sub Matt!
Best video I have seen in a long time. I rung that bell so I can find out as soon as you know. Can’t wi to see the answer
Thanks for the kind words and for the bell! Me too, eagerly awaiting!
Comparing apples and oranges. If I am a still photographer who occasionally shoots video, I want the R5. If I am a videographer who occasionally takes a still, I want the Sony. But they really don't compete.
To some maybe, but even for occasional video shoots, the R5 will be very difficult. See 2nd part (the "divorce" video)
oh yes they do , and if you're a still photographer you buy Nikon :)
Canon R5, R6 are mirrorless continuation of 5D, 6D famous series of photo cameras with video functions. Yes man. Didn't you really got that these are fantastic PHOTO cameras with great video functions?! While Sony A7s III has 12 Mp sensor. No photographer in 2020 wants a 12 Mp camera. So Sony's is clearly a VIDEO oriented camera, although, I suppose, it can take some photos too. So it's just not quite correct to compare a PHOTO camera with a VIDEO camera in domain of shooting video. And you had to point that out, at least. I do hope that all the insane vloggers who need so much to shoot their talking heads for hours uninterruptedly in SQ 4k to enjoy their own outstanding beauty in super crisp details, that they all will forget Canon R5 and go for Sony's camera. It's already too many orders placed on R5. So let the insane vloggers forget this fantastic photo camera and use Sony's for shooting for hours and hours on their clamorous mediocre stuff. The less insane people in Canon's world, the better.
In spite of all the inadequate shouts about R5's "problems", it's also very clear to me that people with hands and brain will be able to create just fantastic video content with R5 & R6. Some people find tremendous faults with every camera; some people shoot tremendous art with every camera...
You are absolutely right that the R5 is an amazing photo camera, and paired with the new crazy sharp RF 50 mm f/1.2 or RF 85 mm f/1.2 plus the new superb eye tracking dual pixel AF, it will produce some mind bending portraits, while as you say, the A7s III is built to be a video camera only and there are even rumors that Sony disable an interpolation based, higher MP photo mode, I suppose to not get reviews of "bad photos". Thanks for the comment!
First of all,plenty of people are still shooting with 12MP cameras,because they are still being purchased,and that's many brands,not just Sony.Next,Sony happens to be the single scapegoat for bitter,and petty users of users of the so-called 'icons',and that's because Sony mopped the floor with them,in innovation,for the past 10 years or so.Adding to that,aside from the SD/DSLR series and 1DX2/3,Canon fell behind in innovation,except for the M%0,which was the leading mirrorless camera worldwide,which helped hide the failures of that EOS R fiasco.The problem of a camera overheating and possibly being damaged is not some 'inadequate' concern,unless the shooter has enough funds to just buy multiple cameras,no matter what brand.It's quite hilarious to get smoked for years in innovation from a company like Sony,who makes Walkmans,TVs,DATs,MiniDisc,clock radios,and game consoles,and was NOT a interchangeable lens camera company.Plus,cameras that do great video are not just for vloggers,and there's a ton of ways to use those features.
Insane,is making general and inaccurate comments,with only opinion to back it up.Let's say I get a Sony to do video.I have a my own live project studio where i have made many mixtape projects,and I might want to stream live,just spinning or a live session with other DJs,recording artists,etc.That's not mediocre just because ONE person does not or cannot do the same.I'm going to look at the Canon R5,yet never because you nor anyone else has an opinion about it.It's my choice,my money,and the right tool for the job that matters..I'm speaking out to anyone who THINKS they know it all,who THINK their opinions matter when they are not buying it,and who have some kind of twisted,petty,and childish attachment to inanimate objects,so they think they can judge someone else's choices.No,I am NOT a Sony shooter at present,and yes I do own Canon cameras,along with other brands.I am not stuck on brands,and yes,a Sony/Canon/Nikon 12MP camera can take GREAT stills,in the hands of a competent,talented,and intelligent shooter.
4500 euros for a photo camera? Nothing new that is not available on the market. What a way to waste money !!
@@michaels8597 Your stills argument falls flat for wildlife photography where more mp's are always welcome. The R5 will convert many 5 series DSLR users to mirrorless.
It don't know who you are... I don't know we're you from... But thanks for finding us lol
Amazing techrew. Спасибо. Hope you become famous fast!
Thanks mate, love your handle. In fact my patents often came when I stumbled upon a tech problem before anyone else. The solution was obvious. Cheers mate
Really great Technical deep dive! on the different ways downsampling is achieved in these cameras. It's a great time for people in market for a new camera, so so many choices and there will surely be a great fit for mostly everyone. There is no one perfect camera. But we are getting one step closer! Thanks for the shout out appreciate that! Looking forward to the next video!
Thanks Rafael for stopping by and thanks for all your amazing content! Learning a lot from you and I have your videos on notification. I think all the camera manufacturers planned to top the line mid 2000 because the summer olympics were scheduled for this time. Everyone who can afford traveling to Tokyo to watch them, can afford to buy a new camera :-)
Development cycles are long, so whole Olympics are delayed, it's still Christmas for us camera lovers! :-)
@@ctoLarssonTECH You are exactly right! it's a 2 year product cycle! especially for Canon. Lining up with the olympics.
Great video. But just FYI, The R6 does 4k by oversampling 5k and not line skipping or pixel binning :) The R5 however does Pixel binning in some modes which makes it the perfect hybrid for most people.
8K RAW and ALL-I / IPB up to 30p
4K HQ from 8K sensor readout up to 30p
4K pixel binned mode up to 30p
4K / 60p - also pixel binned
4K / 120p - also pixel binned
www.eoshd.com/8k/a-closer-look-at-the-canon-eos-r5-lower-quality-4k-mode-to-avoid-thermal-cut-off/
Hi Ali, thanks mate. Yes I saw that article and I tried to dig into the source. My conclusion was that the R5 pixel binning mentioned there was still rumor based - but I also hope it's true :-) I also asked one Canon representative and they said they don't have that information yet. The R6 line skipping I showed in this video was indeed line skipping when the R6 downsampled the 5k sensor to FHD (2k) HFR. You are right that the R6 does not line skip to get from 5k to 4k. Cheers and thanks for the comment
CTO LARSSON - TECH ah I see. On another note, this video of yours is going Viral. Hope you reach to 10k subs soon!
I honestly think the r5 is the ultimate hybrid camera! It's not perfect but the closest thing to the dream hybrid camera. Wish it had 6k 50fps and 1080p 240fps though as it should be able to handle it. Also with you could shoot raw in slow motion however I think this camera will be seeing firmware updates that will make it even better and it's a godly hybrid camera!
I don't see any sources linked in that article. He just refers to 'the menus'.
@@dukebozikowski3801 "r5 is the ultimate hybrid camera" at 45mp? thought ~24mp sensors were the sweet spot
This is nice to know. Please make a follow-up video when you can confirm the doubts raised were. Subscribed!
Will definitely do that. Thanks for the subscribe!
Excellent Review!! Thank you for sharing! We’re looking forward to hearing your insight when the R5 arrives!! All the best!!
Thanks mate, will keep you posted!
This is the FIRST time I was glad with the RUclips algorithm! Subbed! Thanks for this amazing video! :)
Haha then we share that experience too. Before this video I got like 3 views - with 2 out of those probably being my wife and my old mother - and as you can see, I put quite a lot of effort into some of them. So the YT algo was not exactly my best friend. Let's see if we can make up now
Hey I just stumbled on this video and wow, it's so great to see a really nice technical site that cuts through subjective BS and sticks to the science and explains it simply and succinctly. For 99% of users R5 will be a perfectly fine camera for video and not many people buying R5 are cross-shopping it with A7sIII. I have less than zero interest in a 12MP FF camera in 2020. R5 is a massively better all-round camera, A7sIII is a one trick pony for serious video and lowlight users.
Thanks for the kind words mate! Hope to add more technical content going forward
Great explanation, What I am hoping to see on R5 is the ability to recover shadows and highlight, readout speed using silent shutter.
Turns out the R5 didn't pixel bin, which means Sony won on video, by using the bigger pixels. I made a part 2 of the this, the "Divorce" video)
Great explanations of S/N ratio.
Thanks mate. Now that we have both cameras out and tested, turns out this simplified calculation was almost exactly spot on. I made a part 2 (the a7S III vid) where I elaborated a bit on this. Now we know that the R5 doesn't pixel bin, so the best we can get is the 8k downsampled, 4 S/N in the example, vs 20 for the bigger pixels, i.e. a factor 5 difference. That's almost precisely the measured value: 2 stops of noise limited difference in dynamic range. And that means Sony pretty much won, for video.
Just subscribed. Your work is amazing. You explain everything so well.
Thanks for the kind words and for the subscribe! Appreciate a lot
Hello,
About the last Fujifilm camera, does it mean that XT-4 is also using line skipping or pixel binning? So, does XT-4 have the same hoverheating and noise problems?
I haven't looked into the XT-4
Can pixel binning be added in firmware or is it hardware dependent ? Also it might be limited to line skipping so that mode remained unaffected by heat
It's hardware dependent in the sense that it needs to be supported by the sensor. Then of course the firmware can opt to use it or not, so the only remaining hope is that they haven't switched it on. Slim chance though. If it was supported in the sensor, I think they would have used it from day 1. So no.
Have you an update about this topic? Thank you for the wonderful input
Nop. The only new fact since I made the video is some actual test footage here: vimeo.com/439692505 Unfortunately unfortunately unfortunately the guy tested every mode EXCEPT the Log mode in 4k non-HQ, where it would have been immediately visible if it does line skipping or pixel binning (because the noise in the dark areas would either have been a lot better or a lot worse). But now since it's such a bright sunny mid-day, you can't see anything in the shadows really.
It does confirm that the full sensor 8k readout downsampled to 4k in software preserves the thin spatial details though because the "HQ" mode is visibly sharper, though I haven't tested yet how much difference it is vs adding a little bit of sharpening on the non-HQ mode.
Pixel binning would lead to debayering artifacts similar to line-skipping (they don't use a quad bayer layout), so a full sensor read-out definitely would be preferable. But maybe they bin at least the dual-pixel pairs.
On the mobile phone tech side, we've used pixel binning for almost 10 years. It is possible. But you are still right because turns out the R5 doesn't pixel bin. Hence the 4k non-HQ mode is rather useless, and hence the whole camera drops in usefulness for video to almost EOS-R level.
I’m looking forward the R5 full review from you. Now it’s 2021 already not sure if you would still
make it ?
I actually watched the whole things. I am giving myself a reward! awesome depth! love it. Did canon get back to you on that data?
Thanks mate. They didn't get back, but I know the answer. The R5 doesn't pixel bin in the non-HQ mode. It's a type of line skipping or at least discarding part of the data. So the lowest noise 4k mode on the R5 is the "HQ" mode, meaning it's a factor 5 difference in signal-to-noise between the R5 and the bigger pixel design like the Sony a7S III.
A factor 5 corresponds to a roughly 2 stops, which is exactly the usable dynamic range difference measured. I actually made a follow-up video with that, called "divorce"....
Great video! def goes deeper than anything else I saw on the subject. I just posted it as an announcement on the FB A7iii for video shooters group - sending 28,000 members your way. Bravo
here i am haha
Thanks Elliot for those kind words and for sharing on the video shooter's group. Appreciate it immensely! Thank you! Also noting that you were an early adopter of RUclips, well done staying at the forefront of tech.
I love your technical and detailed review!!! I am an electrical engineer and absolutely appreciate your incite and analysis. I hope to see much more of these from you. Can you tackle the subject of anti-alias filtering? And in particular why does the high end Nikon and Sony not have one but Canon continues to use one? I would love to hear your technical view on this one. Other youtubers (and I must say much less technical than you) seem to either avoid this topic or simply don't know enough to talk about it in much detail.
WELCOME, a fellow engineer! YES! Yes anti-alias filter is a super interesting topic. Another one I started on is the lens IS vs IBIS. Maybe you saw the jarring corner wobble on Peter McKinnon's latest vlog, but seems you can't disable the R5 IBIS separately from the lens. Switching the lens IS button to off will apparently disable the IBIS too (!) If that's true that's... odd and makes that topic less useful in practical terms. Will definitely look more into the antialias filters, researched it a bit long time ago (loooong time ago lol) great idea mate!
@@ctoLarssonTECH Wow great idea to do IS vs IBIS. Another topic that has mostly fluffy marketing hype around the words with little real technical analysis. I will look forward to that one. On the topic of anti-aliasing it's another one with words being used that mean different things to engineers (like low pass filtering, oversampling,...etc). What does it really do and why is Canon the only professional line of cameras that really say they use them? I tried once renting a Canon 5DSR to compare to a Canon EOS 5DMIV but then the sensors being 45MP vs 30Mp was not really a fair comparison. Anyway, I ramble now. Looking forward to more cool stuff from you. Keep it up.
Canon should just do like Sony and make one camera biased to video and the other to stills. That was smart on Sony’s part. There are just too many physical limitations to trying to use the same sensor for optimum stills Image quality and optimum video image quality. Its like using a computer with a integrated graphics as opposed to discrete graphics. Expecting the CPU and GPU to share the same system memory and perform well. Create a discrete senso put it in a discrete camera.
Yes you are so correct. I actually made a part 2 of this (the "Divorce" video) where I came to that precise conclusion! :-)
Just found your channel and subscribed - more videos like this would be greatly appreciated - especially a follow-up video when the production models come out.- I love my EOS R and 3 RF lenses - so I pre-ordered the R5 - I hope this is not a very expensive mistake - Great Video!
Thanks mate. Haha yes I hope so too because I've also pre-ordered one. Tried to get hold of Canon Sweden but they're all on vacation haha. Anyway, they Canon engineering might still be working on the final firmware, so I want to test the production version after sale start not the pre-release firmware around today.
good explanation of how pixels can be sampled. Very clear!
Happy you liked it!
You explain things very clearly. Love it.
Thanks for the kind words Henrik!
Canon R5 vs sony a7s iii
Whitch one is best im really confused 🤔🤔
One thing I don't like about Canon is that they don't clarify how the different modes are read from the sensor. Panasonic does this, and I appreciate it so much! Let's hope the Canon R5 is truly tuned and ready for videographers, though by seeing the IBIS, it is not. Let's also hope the Sony A7SIII actually does pixel-to-pixel 4K readout, with a stacked, backside illuminated sensor.
Agree, it wouldn't hurt to release some technical details. Yes that corner wobble in Peter McK's latest vlog that came after I made this video was actually even worse than the first one. And then came another issue, I heard yesterday from one Canon rep that disabling the Lens IS with the button on the lens is how you disable the IBIS. That would be absolutely ridiculous, because using Lens IS only (without IBIS) would be the solution to any IBIS wobble problem. In that case you can't do that. But that sounds so silly that I wonder if it's really true.
@@ctoLarssonTECH Sony has the same issue with disabling IBIS and Lens Stabilization; disabling it on either disables both. :( @geraldundone is doing some chart tests and seems to have found the R5 is using line skipping in 4K/60 and 4K/120 modes... not pixel binning. (another sad face)
The R5 is 8.2K (or 8192) pixels wide. My guess is that the 4K30p Full Sensor unlimited mode is using simple 2x2 pixel binning of the 8K30p image. That math is possible for ALU (i.e. add together 4 integers and then shift 2 bits right to get the divide by 4). But 2x2 only requires 7680 pixels so they may be skipping lines. For the HQ mode, they are sampling the full 8192 pixes down to 3840. This requires divison operations, which uses more transistors and CPU cycles. (This logic seems to explain overheating issues with R6... 5.1K is about 4/3rd 3.84K so a lot more floating point is required to produce 4K with the R6.). As to why line skipping 8192 down to 3840 is possible but not 5.1K down to 3840, I think it has something to do with the Nyquist rate where you need at least twice the frequency to recover the sample.
The difference of pixel binning vs scaling it down in software/hardware by shifting bits or dividing, is that the pixel binning aggregation happens *before* the amplifier, and since the noise is mainly introduced at the amplifier stage, that's the big difference.
Now that we have the data, the R5 doesn't pixel bin, which means the best read-out remains the "HQ mode" which is a full 8k readout. The calculation in this video turned out pretty spot on, because the estimated 5x better S/N with the bigger pixels vs the HQ mode is almost exactly the result (2+ stops in noise limited dynamic range) between the a7S III and the R5 HQ. I made a 2nd part where I touched on this.
Wow🥳 I just learned more than the camera manufacturers or anyone else has ever explained to me!!
Thanks mate for the kind words. I guess the true experts among the camera manufacturers sit in Japan. We'll have to brush up on our Japanese 😂
oh god, what a wonderful channel. Super well done :-) keep going
Thank Sir, what a fantastic comment to get. Thank you. You have some beautiful videos yourself. I'll have to improve on my Italian :-)
Hmmm, that might work with a B&W sensor, but we're dealing with a Bayer filtered colour sensor here... All binning needs to be done in digital form so as to preserve the colour information!
You bin them per color. The illustrations would have been too messy and hard to follow if I drew 3 of each one. In mobile phones we've done pixel binning (with color) for almost 10 years.
There are even phones today like the Galaxy S20 Ultra that dynamically shifts between a full sensor readout giving 108 MP in bright lighting conditions (where S/N ratio is not so important) and a 3x3 pixel binned 12 MP mode (108/9=12) in dim lighting where you need to boost the signal without introducing too much noise.
But like I said, I have no idea if this sensor can do it or not. The sensor has to be built for it.
Great video! Thank you for this information! 😁😁🙌
Thanks, I almost made a part 2 now (the sony a7S III video). Hope can add some more insights
You calculate that pixel binning yields a signal / noise ratio of 8, which is better than the than the 4 s/n ratio you estimate for 4k down-sampled from a full readout of a high megapixel sensor. This conflicts with many reviewers' notions that binning and line skipping are inferior and yield comparable problems like moiré. Such reviewers also often claim that 1080p video down-sampled from 4k files can be an improvement on 1080p recorded using a sensor with fewer than 8 megapixels. According to your perspective that down-sampling would also yield a worse s/n ratio. Comparison of actual samples might offer proof, or not. After all, how can anyone peep pixels at usual video frame rates with the usual preferred motion blur?
Most of the R5 / R6 sensor heat problem could have been solved with a small ventilation duct and fan, which the user might have the option to turn off and seal shut in wet or dusty settings. Even with the camera body sealed to use copper instead of magnesium for the surfaces below the flip screen or its back side would improve heat dispersion. But heat may be a rare problem in Sweden, right?
Turns out this simplified calculation was almost exactly spot on. Now we know that the R5 doesn't pixel bin, so the best we can get is the 8k downsampled, 4 S/N in the example, vs 20 for the bigger pixels, i.e. a factor 5 difference. That's almost precisely the measured value: 2 stops of noise limited difference in dynamic range. And that almost means Sony won. I made a part 2 (the a7S III vid) where I elaborated a bit on this.
How would you retain color info if you pixel binned 2x2 pixels?
You bin the red color from all 4 pixels separately, the green color separately and the blue color separately.
Your mobile phone probably uses it. For example the Samsung S20 Ultra even shifts dynamically between a 108 MP mode and a 12 MP 3x3 pixel binned mode. In daylight, it can do 108MP photos for more effective cropping without much noise. In low light, it will dynamically shift to a 12MP mode, using 3x3 pixel binning.
@@ctoLarssonTECH But each pixel in a Bayer array sensor only captures one color. In a 2x2 grid, 1 gets red, 1 gets blue, and two get green. If you're just summing signal from these four pixels, then you would lose which pixel (red, green blue) collected what light. Maybe there are some intermediate steps?
@@RossMcLendon Uff, right. However, there are new technologies introduced with 1DX Mk III with the new low-pass filter. I'm not sure how that comes into play, does it help things or does it complicate things further.
Points 4 and 5 in this document say some things about the new technologies in 1DX Mk III sensor, but it doesn't tell me what was changed.
www.canon.fi/pro/stories/eos-1d-x-mark-iii-vs-mark-ii/
"But with this new system we are actually sub-sampling it 16 times and using a complicated algorithm to average it out. This then gives us that happy medium between having a low-pass filter and having no low-pass filter. The new design actually gives a slightly higher resolution than the Mark II because it's resolving more detail, while still reducing any chance of moiré."
@@ctoLarssonTECH It's a Bayer sensor. A 2x2 pixel cell has one red pixel one blue, and two green. Samsung's 108 megapixel sensor is a Nonacell chip -- it's got 9x9 blocks of red, green, and blue. So yeah, that's binned on-chip. Their other sensors are Tetracell (same as Sony's Quad Bayer), so you have 2x2 blocks of the same color, and those are easily binned on-chip.
But for binning on a full Bayer sensor, which is as far as anyone knows or expects what Canon's using here, you have RGBG pixels in that 2x2 bin. It needs to be binned digitally, after ADC, before video compression. Even that's an advantage, as it saves the need to do Bayer interpolation. But you're saving image processing work, not sensor readout work.
There is some supposition that Sony's using a Quad Bayer sensor on the A7SIII. Mostly because there was a leak of a Sony full frame (aka, 2.3") sensor with a quad bayer color filter array (otherwise it's basically the A7RIV sensor).
Din video dök upp efter att jag sökt på Canon R5 och hoppsan, en svensk video! Visste inte ens att det fanns för detta :D Kul!
Prenumeration givet! Har du fått svar på om det är 2by2 pixelbinning?
Oj vad roligt med en svensk!!! Ja vi verkar inte ha haft framfötterna med oss riktigt på detta område, vi får fixa det. Försökte faktiskt kontakta Canon Sverige också och se om jag kunde låna en kamera och göra en uppföljningsvideo. Som man kunde gissa så är det semester haha. Fick tag på någon ute på en båt till slut, får se om de återkommer 😆 Inte fått svar på pixel binning, tror inte de vet. Var sitter du? Jag sitter i Stockholm.
@@ctoLarssonTECH Ja, kunnigt folk i Sverige verkar det vara få utav. Intressant video. Kikade även på din långa video där du jämför stabiliseringen, men hittade inte uppföljningsvideon där du summerade dina tankar? :)
Jag sitter i Umeå! :)
Har beställt 2st R5 av Cyberphoto så vi får se när de blir tillgängliga. Själv nyttjar jag väldigt lite video, utan är mestadels en stillbildsfotograf, men kommer nog börja filma mer. Speciellt norrsken så har en a7s2 som ska säljas till förmån för R6an.
The best r5/a7s3 video I've seen so far. Subed immediately
Thanks a lot for the sub. Will keep you posted on how this develops!
Great presentation of the concept of SNR. Have the EOS R, love the EOS R admittedly more due to RF lens, but thrilled Sony is pushing the envelope!!! The new Alpha S will be very good in low light. Physics is physics, big pixels mean better light gathering and more efficient/ easier readout...
Yes I'm also very excited about that. It will be low light performance like we've never seen in that size body. Great times!!
It seems IBIS and super wide has problems across the board at the current tech level. The Fujifilm X-T4 has the same wobble with the 8-16mm lens @ 8mm
Yes it's primarily an optical issue that can only be reduced not solved by moving the sensor. Interestingly, with the lens IS in the RF 15-35 at 15mm with the EIS in the EOS R, it does not wobble, see my video on "96 stabilization..."
The issue is that when you move the sensor you create paralax artifacts that cannot be corrected. Essentially you sometimes can see behind foreground objects, other times not, and that results in the foreground appearing to wobble. All mechanical stabilization methods will do this, both in lens methods and IBIS.
To see the effect, hold a finger in front of your face with one eye closed, then move your head (the sensor) from side to side, sort of like IBIS is doing. Even though your finger and the background are not moving, your perspective is, and consequently it looks like your finger is wobbling against the backdrop. This will only happen with wide angles because your subject matter is normally close enough to the lens for paralax to become noticeable. At long focal lengths the subject matter is far enough away for the differences in perspective to be too small to be observable.
I have question, just bought R5($4300) with RF 24-70 f2.8, unfortunately not able to get any sharp images at all,and find out My R5 r5 sensor misalignment,
I’m not sure is really misalignment or because not latest firmware , I’m thinning return my R5 and buy new R6 or just replace with another R5, any thoughts?
Thanks
I really appreciate the explanation but the cameras are already out there and we have seen what the video quality looks like. That is what matters at the end.
Thanks mate. I have only seen footage from pre-production units, so might not be same firmware on the launch day. We'll find out next week!
I want 4k raw video from a keyring with no overheating.
You already got it - in the phone, plus dual exposure
I love that canon releases a camera and everyone starts talking about sony.
Haha yes that's true. I think the A7s III will be super exciting for reasons explained in this video!
Great stuff. Very nice explanation of this tricky subject matter.
Thank you for those kind words!
Great video but something doesn't pass the smell test. Why would they call it HQ high quality mode if it were lower quality than the normal pixel binned mode? It must be preserving more color information or something in HQ mode.
Valid point. Pixel binning is widely used though, in mobile phone tech we've used it for soon 10 years. The con of using full sensor read-out is worse noise than pixel binning (and processing power.= heat). The pro is that the spatial information is better, so with smart algo you can make a sharp hair look like a sharp hair also when downsampled, while it will look more like a blur in pixel binning. That could warrant a "HQ" stamp. Still, we don't know if the R5 does pixel binning in normal mode. Maybe it doesn't. Then that would explain the "HQ" stamp on the full sensor read-out.
Very eloquently explained. Thanks for putting this together and I hope this reaches wider audiences. I was thinking along the same lines : the R5's hardware in a larger XA style hybrid body to accommodate the power and cooling needs of prolonged 8K shooting would be the ticket. As for now, I just hope the R5 is a stellar stills camera, I hope it is a stills camera at all!n
Thanks Shaun for the kind words. Yes I think it's a very small step to take the R5, put it in a bigger body, add some 20-50% computing power and Canon got themselves a killer video beast with the RF system. I think the R5 will be a stunning still camera, especially paired with the new sharp RF f/1.2 lenses. The eye AF plus the RF system AF speed will make that shallow depth of field a lot more usable.
@@ctoLarssonTECH If Canon does not do this, a very clever DIYer might and find themselves with some serious popularity. Who will do it first? Maybe the Magic Lantern crew? Anyways, I hope to switch to the R5 soon (from 5D Mark IV) and get used to it in time for the world to thaw so I can use it! Strictly a stills shooter. I'll never forget my old 5D Mark 3 and 50mm F1.2 mis-focusing time after time late at night in Cologne Germany, the 5D Mark IV really tamed that lens and put it to work, I'm expecting an even greater increment of improvement with the R5.
@@ctoLarssonTECH www.canonrumors.com/there-is-another-rf-mount-camera-coming-in-2020-cr2/ this might be it!
Great video, interested to see if Canon reply.
They didn't and I'd say they have now lost
The Peter McKinnon video corner frame wobble was lens distortion. Not the sensor stabilation.
It's an optical phenomena when you move the sensor, you can't keep both the face and the background stable (without using software). That's why the lens IS + EIS doesn't produce the same result as sensor OIS. Check my vid with the same RF 15-35, "96 ways to stabilize". Cheers mate
@@ctoLarssonTECH Is that the same thing that's happening with Tilt-Shift lenses, except with the lenses it's by design, but here it's a side-effect?
You explain very well
Thanks for the kind words
Gerald Undone 2.0 :)
😆 Second time I've heard that. I love his channel, so a great compliment. Thanks mate
Great explanation, thanks for the video!
Thanks for the kind words
Amazing in depth video ! Appreciate it!
Thanks mate, it's fascinating stuff!
Great video and really interesting info... thank you! 👍
Thank you, really appreciate those words. I've now made a part 2, the "Divorce" video, hope it can add some further insights regarding the video capabilities and also the company outlook of Canon Inc vs Sony
Incredible Work ✌️
Thanks mate
some idiots have been complaining about "pixel binning" for the last 6 years (e.g. in A7r2), while actually they were complaining about the picture quality of line-skipped video. I hope this video will set them straight and expose them for frauds that they are.
Yep that's true. Also they probably don't know how many mobile phone cameras rely on pixel binning to produce such great pictures from so small sensors.
Is the a7r4 FF4k and FF1080p line skipping or pixel binning
Good question. I don't know. Actually I thought of that when making the video but never got around to researching it.
@@ctoLarssonTECH I failed to get that anywhere online. And i dont know how I could testify it.
So in short Canon overhyped video capabilities to wrap a good still camera in very expensive price tag
Exactly exactly! That was my precise conclusion in the part 2 of this vid (the "Divorce" video)
Amazing video, you deserve more subscribers. Im already subscribe
Thanks for the sub!
Wow, I wasn't expecting a maths lesson... my head is fried 🥴
Do the the cameras shoot crisp clear images... end of lol we've had ground breaking visuals for around 20 years give or take. Everybody seems to be splitting hairs all of a sudden...
P.s please remember that the racing car scene in iron man 2 was shot on a CANON 5DMKII 👀
Just give me 4k full frame 24 to 120p and I'm happy... come on sony, I'm waiting 😄
Math is more fun when it's in the real world :-) True, just give me great image and I'm happy. The Sony will be a beast for video, no doubt.
Nice video. If SAR rumors are true, maybe add a follow up video about how quad-bayer sensor might work in terms of processing video vs still image?
Yep looking forward to digging deeper into that subject. Thanks!
Thank for your insight in to the R5 R6 Arewsome! Cheers Derek
Thanks Derek, appreciate it! Exciting weeks ahead with both this one and the A7s III around the corner!
This was very informative video!subbing :) please keep makin videos
Thank you for the encouragement, I continue to grind on. Made one on the Sony a7S III two days back. I also have two more channels (investment related).
20 fps cannot be sustained long enough to be usable for video, so even though it is there it is still a long way from being usuable.
Also, most of the heat is coming from the processor, not other parts of the camera. If you don't believe me, take a RX100V and record in 4K until the camera overheats, then feel where it is getting hotest. By far the hotest part of the camera is at the rear directly above where the LCD hinge is, and that is exactly where the processor is located. The processor is the part that is getting too hot and shutting down (more accurately, it is throtling and can no longer sustain reliable encoding of the video stream)
Right, correct today 20 fps cannot be sustained long enough in the R5. What I mean is that a mere 20% improvement is nothing, if moving from that form factor into the C500mkII, C300mkIII form factor. Hence they already have the technology to do that and it will come very soon.
Interesting with the location of the processor. I didn't know that. Right, that would be the place to put some contact gel and try to lead off the heat.
@@ctoLarssonTECH It still would not last longer than part of a minute though. The main reason those burst modes work is because the raw data is just being stored in the buffer, the processor can't come even lose to handling it in real time. So after a certain number of frames the buffer fills and frame rates drop drastically. That 20 fps does not last forever, it lasts until the buffer fills. The real sustained frame rate is more like 6 or 7 fps.
Large sensors are far more likely to have heat issues than tiny 1" chips. Most full frame cameras doing video have had some limits set by the sensor itself, independent of the signal processing. Now of course, the amount of signal processing needed for a 4K raw image is the same on an RX as it is on a full frame pro body, but the pro body is also likely to have far better cooling of the image processor. The other thing -- IBIS also makes it more difficult to cool the sensor, which seems to be the main reason Canon didn't have it until now. This is an upgrade in Canon's sensor technology, but it doesn't cure all ills.
Ohhh man.... You are awesome... Subscribed you... Thanks for this video
Wow, thanks for those words mate. Really means a lot!!
Sir one more thing.... I m a photographer and i do video and photograph both... Suggest me a camera should i wait for A7S3 OR go for R6
Excellent! I was looking for a more technical channel centered around photography.
Super happy if I could bring some value mate. Thanks!
Ah thanks! learned something new today! You gained a subscriber :)
Thanks a lot for the sub!
I don't think pixel binning can be done on standard bayer sensor unless Canon use a quad bayer.
Yes I'm also not sure. I know the rumors site wrote they do, but I'm not so sure. The sensor needs to be designed for it
dude u dive too deep in detail and explaination, great video !
Love tech lol :-)
Great video! You got a new subscriber, also from Sweden =)
Happy to hear that!! Thanks a lot. Will check out your channel too! Where are you based?
@@ctoLarssonTECH På västkusten. Halmstad ungefär =)
Och du?
Hej Anders, this is Travis from Allt för Sverige season 2. I enjoyed the video and your attention to detail. I think this channel is going somewhere. Subscribed. I pre-ordered the R5 for my photography and video work. I hope we are pleased with the camera and it's handling of 8k to 4k. Canon needs to come through with some competitive full frame 4k here. I also hope the IBIS wobble I am seeing in Peter McKinnon videos can be worked around. I'm not too worried about the overheating. I don't expect to be filming long clips of 8K or 4K 120. Did you order the camera? Who will get their's first!? I have ordered through my local camera shop instead of online in hopes I can get my hands on one in early August. good luck! tack, hej.
Hello Travis, thanks a lot for those kind words. Really appreciate it. And thanks for the subscribe. Yes I also pre-ordered the R5. There is zero doubt that it will be an absolute killer camera for photography with the f/1.2 RF lenses. The eye AF is super fast with the RF lenses, which will make that kind of razor thin depth of field work in practice.
HOWEVER, that IBIS wobble in PM's vlogs however has me really scared, for video. That was horrible, I've seen his amazing vlogs for years and never seen something looking that bad leave his hands. Even worse, yesterday I heard a Canon rep say that that you can only disable the IBIS by disabling the lens IS, with the button on the lens. But that sounds so stupid that I wonder if it's true. (Otherwise we could just go back to Lens only IS for video, which doesn't wobble.)
So right now I'm in stand-by mode. I do mainly video. If the R5 doesn't pixel bin and has an IBIS that doesn't work for video, and Sony incorporates their Venice color science in the A7s III, I'm going to bite the bullet and switch to Sony, which would such a bit since I have a lot of lens.
In Sweden the pre-orders aren't binding.
Let's monitor this next weeks!
@@ctoLarssonTECH yes many of us Canon folk have held back on the emigrating to Sony in hopes for a camera like this. I still have fingers crossed Canon has come through! Thanks
@@ctoLarssonTECH hej they called and told me my R5 will be arriving in store this Wednesday but I cannot pick up until Thursday per Canon rules on release. I would expect the same for other official Canon retailers. I have my fire extinguisher 🔥and dramamine ready!
@@ctoLarssonTECH have you heard about the scotch tape toothpick hack on R5??? the overheating saga has me on the edge of my seat. I hope the happy ending results in the R5 I had hoped for; for video. For photography, it has been a dream! Shooting a wedding with it this Sunday.
Excellent commentary - I’ve subscribed!
Thanks for the kind words and for the sub!
Very informative video thank you!
Thank you for those kind words
Most web ressources say hardware bining is not possible with CMOS sensors, because each pixel has its own readout. That's one of the big differences from CCD to CMOS. What you described makes perfectably sense for CCD sensors, but CMOS sensors do not shift lines or pixel before readout.
Let's see. On the mobile phone camera side we've done pixel binning for almost 10 years. The Samsung S20 Ultra even dynamically shifts between 108 MP full sensor read-out in good lighting conditions and a 12 MP 3x3 pixel in lower light. There are many different sensor types.
Sir my work wedding & pre wedding cinematic right now I am using canon mark 4 so al go for r5 aur r6 please tell me?
Once I get the production version of the camera, we can test this out. If the R5 does pixel binning on the 4k "Normal mode", that mode will probably have *better* low light performance than the "High Quality" mode and should be a good fit for your wedding work, which I imagine contains a lot of low light evening work? If it doesn't, the R6 will be a better choice. Having said that, I know other wedding photographers who will go all out for low light performance, and for that the A7s iii will beat both of them, while having some other compromises like no real good still photos, and (IMO) not as good lenses as the new RF lineup from Canon.
I think I understood about 75% of this hah. Considering I had no context coming in well done.
Thanks mate. Turns out the calculations were pretty spot on, now that we have the camera tests available. The R5 doesn't pixel bin, and the 5 times better signal/to noise with the 4x bigger pixels on the Sony vs the R5 is pretty much what we got, about 2+ stops better. I made a part 2 with the video comparison (the "divorce" video)
Can this binning 2*2 be fixed in a firmware update thereby making Canons both cameras unbeatable.. ? Max tech said may be it's possible to fixx
If the sensor supports it on low-level hardware, it can be possible to add in firmware later (even for the R6 example I showed in the video, which was 5k sensor downsampling to 2k 120fps in R6)
Unlikely. That sort of stuff is done in hardware in the image signal processor for the most part. Firmware runs the processor, not the ISP.
In order to take advantage of the improvements as shown, the sensor would need to support on-chip pixel binning, since they'd need to combine the output of four pixels into the analog-to-digital converter to see the SNR reduction as shown. What's termed "readout noise" in modern sensors is actually the noise leading up to the ADC stage. Once you're digital, there's no more noise. In past cameras (Canon's 5DIII, for example), the sensor actually read out in analog, routed that analog pixel bus across a PCB to the image signal processor, which did the ADC part. So readout and analog noise were once pretty identical. But that's fortunately ancient history.
On a CCD, you technically move charge without adding noise until the charge to voltage conversion and the ADC stage. In a CMOS sensor, you generally don't move around charge, but voltage. However, I image if you had 2x2 pixel cell that you'd like to have read as a bin, it would be simple enough to merge the four charge wells for the pixels and read them out as one superpixel. I'm pretty sure my phone does this :-)
The problem with that is that you're binning a 2x2 pixel cell, which is RGBG. So you can't actually process that in analog with a single ADC. So I don't really see this being an on-chip option, as it is on the recent Quad Bayer sensors used in phones (and probably cameras before too long). It's of course possible to do pixel binning after the ADC step and reduce overall luminance noise by a factor of 2, and that certainly could be done in firmware.
Are you working in R&D of camera or just putting theory and self logic?
I worked in mobile tech for many years, these problems are the same. Plus pixel binning was a key tech to get low light photos out of high res mobile phone sensors already many years back.
Good stuff, thanks for the in depth look
Thanks a lot for the watch and the kind words mate!
Such an expert! Could you please give me your opinion?🙏🏽🙏🏽🙏🏽 I‘m wondering whether an 8k sensor would make a line skipped 4K image look better than a f. e. 5.1k sensor (R6)? Or is it the opposite?
No one can tell - I hope you can explain!🙃🙏🏽
The R6, with its larger pixels, is going to better signal-to-noise ratio than any of the read-out modes of the R5, which means that it will have better low-light performance. This means you can run higher ISO on the R6 than on the R5 without ruining the video with noise. (This is assuming the sensors are of similar tech generation.)
The R5 on the other hand, will have better spatial resolution if using the 4k HQ mode, which will downsample the 8k read-out and in doing so it will try to maintain thin details. That will make the R5 4k video "sharper" than the R6 4k video.
So "better" here depends on what you prioritize.
What I really dislike about the R6 is that it does IPB only, no All-I, which slows down the edit and can destroy some footage of grass and water etc. Also as you saw in that boat sequence, the R6 seems to lineskip to do the 120p 1080 and that is bad.
Hope helped you somewhat!
CTO LARSSON - TECH hey it’s me again ✌🏽 as we do know now, the R5 does use line skipping in normal 4K... do you know if it’s technically possible for canon to switch to 2x2 pixel binning via firmware update sometime in the future? And furthermore, do you think that the camera can be cooled enough by external gear (tilta fan, RF Adapter with fan, etc.) so that it can compete with the A7SIII?
Wish you all the best!
Really like your video, when would you think you will get the information whether 4k30 is pixel binning.
do you think the info from EOSHD answers the pixel binning questions? www.eoshd.com/8k/a-closer-look-at-the-canon-eos-r5-lower-quality-4k-mode-to-avoid-thermal-cut-off/#:~:text=In%20order%20to%20achieve%20the,work%20for%20the%20image%20processor.
Hello卜水 yes.
I know there are rumors like the link below on the EOSHD forum, but I couldn't confirm them. I checked with one Canon representative and they hadn't gotten this information yet. The final firmware for release isn't done yet, maybe in progress still. The good news is that no one has observed the characteristic aliasing, so I'm quite hopeful that it will turn out to be pixel binning. Fingers crossed
Eye opening perspective! Thank you
Thanks a lot, appreciated
People think physics doesn’t apply, how many times can you tell them and still be ignorant.
True, love physics!!!! Explains the world.
But will it blend?
Cast magnesium alloys, while that's a wide range of materials, have a proof stress (where it dents) of typically 75-200 MPa and a tensile strength (breaks) of 135-285 MPa. That's not much. It will blend.
@@ctoLarssonTECH well at least the magnesium alloy body should manage the already famous overheating just fine.
From eoshd.com: “In order to achieve the higher frame rates, the sensor uses 2×2 binning, similar to the 4K video mode on the 60 megapixel Sony a7R IV.
The 4K 30p non-HQ mode is read from the sensor in the same way.
Since this has a much lower frame rate, there’s less heat and less work for the image processor. Thus the 4K 30p (and 24/25p) pixel binned non-HQ mode does not have a thermal cut off.”
Thanks mate. I've seen that, but have not been able to confirm that rumor yet. Fingers crossed
CTO LARSSON - TECH Great! Will be looking forward to your follow up video 👍
That implies pixel binning off-sensor. It's certainly the case that a 4x4 pixel binning uses less (virtually no) image processing than debayering and then a whole sensor downrezzing. That may be what's at work here. People complain about Sony's A7R series doing pixel binning rather than whole sensor scaling, so this seems a likely explanation. There seems to be some belief that a 6-8K image rendered to 4K and compressed to 100Mb/s AVC is better than a 4K image compressed to 100Mb/s AVC. I suppose it depends on the full resolution downscaling algorithm. Though the pixel binning ought to have less color error since there is no Bayer interpolation needed.
You got me lost in like 1 minute, sorry LOL
The second part ("Divorce" video) is a lot clearer and easier to follow. Hope will add you some value.
Bra jobbat Larsson!
Tack så mycket Ruzdi!
I can’t see the point @ 7:15 about RAW video. Olympus has 60 fps RAW stills on OMD EM1 2 but they never even did 4K 60 fps video in this cam, not mention the RAW.
The point I was alluding to there was how close they are to delivering the same quality as a still photo in every frame in a 24 fps video. It's basically just 20% off, which is peanuts if moving to a bigger form factor. So I'm certain that they are already in progress of readying a cinema camera with that as we speak.
@@ctoLarssonTECH Agreed. If you jump over to CanonRumors.com, you'll see both rumours of a new cinema camera with RF mount, as well as new RF mount adapters. Both are rumoured to be coming in 2020, so be careful if you're pulling the trigger on the R5. If you only need video, you might be wise to wait.
interesting. kinda technical, but I get it. the gh5S also used 12mp stacked i think, same thing. well thanks for the info. Marty of Marty Sender RUclips peace out
Thanks Marty, happy if it could bring some new learnings for you
good video but man towards the end of it i have to question your wisdom and mostly ask if you have ever used Sony system at all when you mentioned Sony lenses are not as good as Canon RF LOL. Sony has GM lenses on pair with Canon RF and on top of that Sony has ton of E mount fully compatible specially designed fantastic lenses for stills and video from Zeiss Loxia, Batis, Voigtlander, Tamron, Sigma and other.
You are right, that was a very sloppy statement. If there was an Edit button on youtube videos I would have rephrased it. What I meant was that there exists no lens equivalents to the RF 15-35 f/2.8 (Lens) IS or to the RF 70-200 f/2.8 IS (that compact) on the Sony system.
@@ctoLarssonTECH No worries, i thought you meant something like that. This Canon compact 70-200 2.8 seems really great though.Im sure other lens manufacturers will follow up with similar design.
Excellent factual commentary.
Thanks Kurt!
Sincere thanks to you! I've learnt sth today!
Happy if I could be of service Tomasz
Great technical stuff. I am not smart enough and you lost me. I only care about the final image. Let’s try it out and give it a final verdict later. I told my friends many time that Canon, Nikon, Sony, etc. all make good cameras. Buy one within your budget and enjoy. I am still amazed by Ansel Adams works. Was his camera better than today’s cameras?
Absolutely. I just love to geek out sometimes. At the end of the day you can take any camera from any manufacturer today and produce award winning content. The Nobel Prize winners in literature didn't win because they had the best typewriter.
@@ctoLarssonTECH Don't get me wrong. The technical part you presented is good.
So the Canon can be a good still camera and a good video camera, while the Sony can be a poor (12MP) still camera and an excellent video camera, as far as sensors go.
Yes I think you're spot on there mate. Canon will be killing it for photography with those RF f/1.2 together with this eye AF and sensor and super speed AF interface. The Sony will be killing it in low-light video for sure and based on this analysis, and probably in 4k video overall. But 12 MP photography in 2020? Not so great...
Your explanation is great! I found a lot of downloadable R5 material here: vimeo.com/user803551 The oversampled footage looks great, the normal (pixel binned or line skipped) video has noticeably less detail. It is interesting to note, that the normal 24p and 120p footage has the same quality, which means that 50p/60p video will also have the same quality. This also means that the 50p/60p 4K mode has better quality on the R6, which is oversampled! Canon is planning to add additional raw options to the R5, I am curios about that.
Wow, great resource. I'm going to go check that out right now! Thanks mate
Great information. Good job.
Thanks for those words