maxotics

The whole idea of 4:4:4, 4:2:0 and 4:2:2 is, as has been said, the ability to create low-bandwidth video by compromising on color space, which we are less likely to notice (over contrast/sharpness).     The better color information you have, the easier to key.  Andy should write all the people who spent $5,000+ on cameras that do 4:2:2 or 4:4:4 and tell them they wasted all their money ;)   Andy, all you need to say is, yes, color keying is easier on a 4:4:4 camera, but for what you want to do, your money is better allocated on a 4:2:0 camera like the G6 and good green screens, lighting and software.  No one here has argued against that!  We've only tried to educate the OP on the issues.   You're effectively saying that some of us don't know anything about real-life shooting, only academic theories.  Also, you shoot music videos that are so busy, visually, that no one would notice green in the hair, etc.  We don't know what the OP is wanting to shoot.     It's Murphy's law that one day, Andy, you will get a job where the client will notice the difference.  Remember to report back to us ;)

Axel brings up an important technique, using the green screen for only the area directly behind your subject and hard-cutting out the rest.  That usually needs good software. Again, I am NOT saying that you can't green screen with 4:2:0, only that it is something you should understand about the process.  The software that you use is very important, and can be expensive.  No easy decisions.  You can also see there is plenty of space between Andy's subject and the green screen background.   I want to point out, other experts, that we're talking to a beginner here. I'm DEFINITELY impressed by what you've done, but you know what you're doing and have a deep understanding of the tools involved.   :)

Jiban, just in case you don't understand the issues from the beginning...   The generic term for green screen (or blue screen) is chromakey.  It means you "key" out a color (chroma), like a keyhole, and now you can see what is behind it (the chroma/color).     In order for this to work, the software must go through the image, pixel by pixel, and determine if each pixel is the chroma you have selected, green in your case.  Naturally, every pixel will not be an exact green.  Some will be dark, some light, some may be bluish from the lighting, or reddish, etc.   In the software you can set the lattitude of what is, or is not, a green.   This leads to problems.   To key out the bluish green around an edge of the screen, you may pick up a bluish-green part of someone's shirt, and then that ends up being keyed through.     Green, may reflect off the screen, onto someone's hair, and then that gets screwed up.   IN SHORT, GREEN SCREEN, IN PRACTICE, iS DIFFICULT TO DO VERY WELL.   Optimally you need   1. A large screen 2. A long distance between screen and subject to prevent spill 3. A well lit screen 4. A well list person, lit in a way that will match the visuals you'll key in.   If you think about this stuff, you'll realize that the better the camera knows a green from a not-green, the better it can apply a key.  Most video cameras throw out a lot of color information in compression, 4:2:0 (instead of 4:4:4) that one doesn't notice much in most circumstances BUT is a serious problem for green screen work.   YES, you can pull a good key with 4:2:0 video cameras, like the G6, but it is MORE difficult than a camera that doesn't compress the video image in that way.  I don't know whether it would be better to invest in a better green screen and lighting and use a G6, or invest in a camera that does 4:4:4, maybe $4,000 used and go cheap on lights, etc.  There is NO hard answer.     In short, I am giving background behind what mtheory is saying--no one in their right mind would choose a 4:2:0 camera to do professional level green screen work unless they had a perfect studio setup.   The ML setup works because the RAW video contains 14bits of true color information at every pixel location.  Yes, the workflow is monstrous, but if you end up spending all your time trying to light your screen screen to get a good image, and can't, with a 4:2:0 camera, you might not think it so bad ;)   Almost every camera your friends will own is probably 4:2:0.  Borrow on and try it first.  Don't invest money in any camera until you go through chromakey workflow at least once to understand the issues I've explained.  Then you'll know what is, or is not important to you.

I'm fairly certain Guy + Cameras + 3D printers = Divorce ;)   That doesn't stop me from fantasizing!  

I find the RAWtoGPCF has problems.  Apparently a weak debayering algo, according to 1% on the forum.  The problem I see is when light comes through leaves it debayers to blue dots.  So I wouldn't be surprised if there is a noise problem.   I've done a fair amount of experimentation and find that LMMSE and Amaze do better de-bayering for ML RAW.   Might be subjective.   I believe RAW2GPCFv113 is old, and meant to show people Cineform 422, as an introduction.  So you might also contact GoPro about your problem, since you have the premium version.

I loved my Nex7 except for this   1 Proprietary flash shoe meant adapter made the camera unwieldy (looks like they've finally smelled the coffee on that one) 2. No focus distance on lens, or display (or maybe I couldn't get to it). 3. Focus was really slow and often painful (which made distance markings desirable, so you could pull focus abstractly).   I can do 720p RAW video on the EOS-M.  I can't believe Sony couldn't do some sort of ProRes thing in the camera.  You never know with Sony, either they're throwing everything but the kitchen sink in (like the killer VX1000 years back), or they're playing proprietary add-on games.  It seems a waste that they would work so hard on translucent mirror tech (I know doesn't apply to this camera) only to skimp on the video tech.   This camera with anything close to BMPCC video, wow!

  Sorry, my tone was a tad bit professorial ;)  I've been wrong about this stuff before and assume I'll be wrong again, maybe right now!     I wouldn't be surprised if you're absolutely right and the sensor electronics save in yCbCr.  I look at the bayer pixels in RGB, so that's the way I think.  But it's true that, whichever color process, the question is how many bits of color information do you have for each pixel?  How many bits R or Cb?  Let me know if I'm wrong here.   In photography, and filmmaking, the quality of the image is determined through color (contrast in bw) and resolution.  This has been true since I was dunking photo paper in chemicals way back when.   Video compression has NOTHING to do with achieving photo quality IQ, right?  The goal is pleasing motion images within a specific bandwidth (camera/card writing speeds, iMovie, etc.)  They're not thinking fidelity, they're thinking not making compression obvious, artifacts, etc.  They do a great job at it, don't get me wrong, but there is no real solution to photo IQ except accurate pixel by pixel 24 bit color data.   Let me answer your question about what equipment to buy as a 52 year-old.   1. The client doesn't understand a thing about what you're doing, technically, and doesn't want to know (unless he can brag about it at a cocktail party). 2. Doing what you think are idiotic and simple things for your client earn your more points than being able to shoot a 30-minute interview entirely in RAW.  Most people don't see the difference.  In fact, because H.264 and the like are so prevalent, many people think of that look as the best look. I HATE THAT.  But it's the truth, right? 3. Audio, lighting, assistant, post-production, etc., are all more important than a 5D3 alone.  So you may need to keep your cameras but have them buy that stuff for you.  An improperly lit person will get you in trouble, doesn't matter what camera you have.   All that said, equipment costs are actually small for businesses.  Many business people, myself included, are happy to buy a worker something exotic tech just to make them happy.  If someone is paying you $50,000 a year and the cost would be 3 times that from a vendor, then a $3,300 5D3 is not a big price to pay.  This worked for me when I was young!  As long as I got the equipment I wanted, I didn't care that much what they paid me.     If that's true, get a 5D3.  It has a good re-sale value (good for company).  No one ever was fired for buying Canon (IBM in my day).  You can do RAW, or not.  That's the camera you want.  Moire-less video, both H.264 and RAW.  It's a monster camera.  Get a G6 too.  I think Andy is right, that's the right combination for most work.  Better to have two of those than 1 GH3 for interviews.   If your client can't afford that, then time to profit from your GH2 :)

dishe, here the "math" behind what you're feeling.  In the lowest HD resolution of 1280x720 you'd like to have 16 million colors per pixel.  That works out to 3 bytes x 921,600, times 24fps (let's say) giving you 66 megabytes per second.  Most SD cards can not be written to that fast (or cameras dump that fast); none could when the GH2 came out.   Your sensor pixels are not full color, but either r,g,b, at 14 bit on the Canons (don't know panny).  Anway, 1.7 bytes times 921,600 times 24 fps equals 38 megabytes per second (about what my EOS-M ML RAW hacked camera writes at).   The truth is, all the consumer Panasonic cameras throw out a lot of color to achieve bit rates that can be written to SD cards.  Once you start looking at higher dynamic range color it's hard to go back.  To me, there is very little difference between all the Panny consumer cameras.  So on that score, I don't think it matters a whole lot whether you use a GH2 with stock firmware or a G3 or GH3.  The latter are better, but marginally in my opinion (please don't flame me).   That said, you should get the camera that makes your overall workflow the easiest.  If you need a small size, go with the G6, can use a large camera, go GH3.  Already have a GH2, stick with that.   If what you really want is photo quality in your video, you're going to have to go with a camera that can write raw sensor data to your card.  At a consumer budget, that's Black Magic or ML hacked Canons.   The Panasonics will not do high dynamic range and the RAW cameras will not do long-take videos with reasonable file sizes.  There is no one-size fits all.  Unfortunately, nothing close.  For various technical reasons, the technological approaches are very far apart.  

And professional RAW starts at 30MBS (mega bytes per second), which would be 30x8(bits) 240mbps.  If you hang around this forum that may confuse you at first.

I agree with you JG and having worked intensively with RAW for a bit I can give more reasons why what you say is the case.   1. Line skipping.  The sensor is reading data one line at a time partly to maintain expected frame size, but also because recent chips can't read AND process a full frame 24 times a second.  The 5D3 obviously can (as Zach pointed out).  But I bet those electronics are expensive and power-hungry (and cost Canon considerable R&D).  Eventually they'll make it to other cameras, but I see no economic incentive for Canon to give people other options to the 5D3 this soon.  The camera seems on track, with non-moire video to be another classic like the 2 Canons before it.  Good for them!  The 7D, IMHO, is still a sports camera or high-end/rugged Rebel.   It was just 11 years ago that Star Wars, Attack of the Clones, was shot on 1080p, on a Sony something-or-other.  My guess is the 5D3 is either as good, or better.  Think of that, a $3,300 camera!   There's a lot of competition in digital film "cinema" cameras now.  Hard to be generous without profits!   When I first start doing this, I too, wanted something between H.264 and RAW.  What I didn't see then, and hopefully this will help others, is that Photo processing/compression is about DATA and video processing is about PERCEPTION, similar differences between WAV and MP3, in audio.   In photo processing, you'll looking to keep as much image data as possible because photographers/viewers have a low tolerance for pixelation, blotchy colors, etc.  The bar is actually pretty high now!  So I don't know how you'd reduce the image information half-way?  Cut down the number of red or blue pixels?  In any case, such a strategy would have little or no use for photography.   Video processing is about throwing out as much image data as you can and compressing the rest to create as pleasing a MOVING picture as you can, psychologically.  There is no real connection between RAW bayer data to JPG, and image data converted to a video 4:2:2 color space.  That whole #:#:# quality idea is centered on psychological compression methods for moving pictures--that is, that we're not as sensitive to low-color and dynamic range as we are to sharpness, etc.     A simple way of saying what I'm saying, is, to use audio as an example, low-bit-rate WAV is probably worse than high-bit-rate MP3 (psychologically).  They are NOT the same approaches!  You can't meet in the middle.  They are parallel approaches to image recording and display that don't meet.   There is no middle in video.  That's what my journey is leading to.  You either have full 14-bit color, or you have color-reduction and motion-base compression.     If one insisted on a middle ground, it would be 14bit 720x480 upscaled.  

ML has been steadily improving since the 50D shooter's guide.  Today, I just shot 5 minutes worth of 1280x720 RAW video on the EOS-M.  Saved to one 10GB file.  Some people are now getting continuous 1920x1080 on the 50D.    Dual ISO and ETTR have been progressing (though I'm not using them).  So has various audio features.  The MLV (Magic Lantern Video) spec will eventually bring further improvements.  The fact is, Canon doesn't need to provide better video for the 7D, hackers are doing it for free.  Will it be scary for "production"?  Of course.  But as you point out, Canon already has production ready equipment to buy/rent--if that's your need.     ML RAW is still in Alpha.  Every day it gets more powerful and more stable.  Unfortunately, there are NOT a lot of people working the frontier.  When it becomes stable enough a lot of people might jump on.  Why should Canon do anything until they see how that sorts out?     It doesn't really make any difference to me, if I'm getting a RAW file thorugh ML or through Canon's new firmware udpate :)  Naturally, I wish Canon would do it, but I wish Windows 8 had a Start button too!

  REALLY SMALL!.  I'm on the ML forum often.  I notice four main devs, Andy, 1%, gr3gg0, Alex.   Many of the devs who wrote some of the post processing software, like RAWanizer and PinkDotRemover, have left the scene.  Apparently, the current version of Davinci Resolve isn't working with cndgs.  I don't know if it's even big enough to be a "market".  Yesterday, a guy complained that he hadn't heard back from Mosaic Engineering about a VAF adapter for the 50D.  I think a LOT hinges on the success of the BMPCC.    Reasons for BM to be excited, evidence to also make them terrified.

Doesn't surprise me at all. My 50D gets warm within 30 seconds of shooting raw. I've never understood all these people complaining about the battery as if the engineers who made the camera, know nothing, and just picked up the first battery they saw at Radio Shack ;)

Andy, when I'm watching TV with my wife and some bad video comes on she says "Looks like your RAW stuff." :)

I wouldn't use Magic Lantern RAW for anything where I need to shoot continuously for more than 3 minutes. In a review of the BMPCC an engineer said the camera uses powered peltier cooling of the chip, which is why the battery can drain quickly on warm days. I believe I see hot pixels on my 50D after it's been shooting for a bit. I want a GH3 or G6 too, for all those reasons! For now I use H.264 on the EOS-M or Panasonic GF3

No one is disputing that the G6 is a wonderful camera. But taking comparison shots of buildings will not show the where RAW is better over H.264 (in fact, RAW might look worse because it isn't sharpening the image and smoothing the colors). The problem is we don't know what color those buildings are supposed to be, so any image you show, assuming the bricks aren't green, will look good. Take comparison video of a human face, however, and people will see the difference between footage shot in H.264 and RAW. Biologically, we are programmed to "read" face color/mood/health. Even if we've never seen the person before, we can tell when they look healthy and when they look less so. (though I'm not saying people can't look great in H.264) Again, I am not saying everyone with a G6 should trade it in for a 5D3. It depends on what/how you shoot. I use those cameras. But you're only hurting yourself if you won't look at the differences objectively. I love your posts. I love your videos. I also love the look of RAW.

I don't know that camera, but generally, you need to load "new" firmware(ML) into the camera and it will ask if you want it to automatically boot.  It will then boot using the autoexec.bin on the card.  

I shot this today, since we have a 16mm guy in the house.  This video shot with a $350 setup.   https://vimeo.com/76181035

  I would think, theoretically, there would be something going on.  Since the 5D3 is a full frame sensor, and it isn't line-skipping, but sampling pixels throughout the whole image, it has to be COLOR skipping.  I've been studying a lot of moire issues and it looks to me that most chromatic raw happens when color is separated by two lines, driving one to the blue pixels the other to the red.  Those other cameras should not have the issue.   Hopefully Andrew can look into at some point.  This is still one of my favorite videos.  It really shows the difference between RAW and H.264, to those who really want to see the truth ;)   https://vimeo.com/66206596

  LOL! Say it isn't so JG! :)     I've been thinking about this same subject today.  I have   a D600 + 85mm for Portraits 50D for ML RAw 2 EOS-Ms for H.264, ML RAW 720p PORTABILITY A bunch of lenses   Naturally, I think by spending a few hundred every month I'm not spending money ;)   But at this point, I'm a fool not to just sell it all and get a 5D3.  The only thing I lose is portability.   So once answer for your question   If you have the money, you want a 5D3.  Even if you are a die-hard Nikon guy, you should have one.  With ML it does so much cool stuff.  For photography that is one hell of a camera, really the second best you can buy from Canon.  The BM is NO camera.   The RAW footage I see from the 5D3 (I look at everything I can) is simply MONSTER!     Portability.  Face it, the BMPCC is the only game around, just have to wait for it to come to town.   The EOS-M RAW isn't bad, but if the workflow diff between BM and 5D3 have you concerned, it's not for you.   I'd love a G6 too, just to see what Andy keeps foaming at the mouth about ;)     In short, Markm, I doubt you can go wrong with the 5D3.  It will definitely hold its value.  Not so certain how all these RAW video cameras will fare.  They only do one thing well.   JG I haven't forgotten you.  H.264!  Man up! :)

@Cantsin, thank for you comparison of the 50D and BMPCC. Size is a big deal for me. You've confirmed what I suspected. @Zach, don't they know who you are?! They should have given you one! :) While I wait for the BMPCC I've been working on the EOS-M. I think it will end up a very decent 720p 14bit RAW camera. I plan to post new stuff this weekend.

  @ScreensPro, I'd put it this way, I'd be pretty pissed off if my nuanced story of love, betrayal and redemption looked like it was shot in an over-lit MacDonalds.  There are is a certain look and feel that the GH3 just can't shoot, period. (In any case, no one can accuse Andrew of taking sides in camera tech).  Now, with a good script, great acting, directing, would it stop that masterpiece from succeeding--I don't think so.  But wouldn't it be worth the effort to get a few extra tears in the audience if you could show the dynamic range and color nuance (the mood) you actually lit? :)   These are tools and techniques.  Who wouldn't agree that one should focus on the fundamentals?   Some think the fundamentals start with contrast and color.     What you're voicing is a frustration many people here have--that we can't spend more time shooting a real film, then experimenting with this equipment.  This is how I enjoy a little taste of film-making, pathetic as it might be.     I console myself with the thought that the stuff I figure out may help such a genius that you have in mind!

  Absolutely, and look at it from Blackmagic's POV. I'm sure they'd love to get even the littlest piece of Canon's high end market.  So while Andrew is trying to get the Japanese giants to look down at us, someone at BM is trying to get them to forget about us and look up at them ;)   I try to look at any RAW video on VIMEO, YouTube.  I read ML forums.  I believe there are, at best, maybe 20 people who shoot RAW on the 50D and maybe a few hundred, at best a few thousand, on the 5D3.   I've sent Andrew messages about focusing on EOS RAW stuff, but he always has some new more expensive camera he'd rather write about.  I don't blame him.  The real money for bloggers is in camera pornography.     People who hack cameras like me--the last kind of customer you really want ;)  Unless you want to live in a cold-water flat and have maybe 3 lenses in your collection.

Magic Lantern has 27,000 members.  Let's say there are another 23,000 videographers who read EOS-HD, DP-REVIEW or nothing.  Let's put hard-core DSLR video users, both professional, semi-professional and hobbyist at 50,000.  Canon sold 9 million DSLR like cameras this year.  I assume Nikon sold a similar amount. The U.S. is over 300 million, even 10% of that is 30 million.  Even if Canon looks at their numbers only 50,000 / 9,000,000 equals 0.5 %, that's half of one percent.    There are many ways one can try to increase market-share by 0.5%. I'm sure there are many good reasons why RAW like video is not it.  It's in the numbers.

Thanks for the review! From the perspective a photo nut who just wants to shoot family and friends and maybe some artsy footage of a garbage can 1. The 5D3 is 3x times the price of the BMPCC; they are not competitors in my budget. 2. I have not found a way to get any H.264 to approach the quality of a 4:2:2 codec, period. So the Panny G's are not an option. 3. You turn the camera on. Set ISO, Aperature, Shutter, focus and shoot. Everything after that is a lazy-boy. I use Sigma cameras for true 3-color pixel data--you're all a bunch of babies ;) 4. The only competitor near the price and size is a hacked EOS-M. They are artificially cheap now. Even if Canon decided to make the new EOS-M a competitor to the BMPCC it would have to either change the sensor or down-sample all its pixels, like the 5D3, both options requiring serious engineering. We do NOT know if an EOS-M, configured to shoot Pro-res, would NOT have all the problems the BMPCC is having in battery issues, etc. In sum, like the Sigma cameras (for full-frame IQ in your pocket), if you want to shoot film-like video at $1,000 the BMPCC is the ONLY game in town. I know you have to be balanced. That's what the forum is here for. Nothing you have said indicates the camera does not deliver on its promise. No one should let themselves get lost in the weeds on this one.