chauffeurdevan

    Not exactly.   First, each Cb and Cr (could also be UV) channel represent the deviation from gray on an axis ( from wikipedia : CB/PB and CR/PR are deviations from grey on blue–yellow and red–cyan axes, whereas U and V are blue–luminance and red–luminance differences.) (The LAB colorspace is using the same principle)   So for example, the CB/PB channel representation in 8bit is at 0 - blue full saturated, 64 - blue half saturated, 128 - grey, 256 - yellow full saturated.   Second, in the 4-2-2 relationship, the 2-2 doesn't not represent the channels, but samples per row. The sampling formula always seems to be represented on by 4 pixel columns by 2 row. So in a 4-2-2 sampling, for a region of 4 pixel wide (4-x-x) by 2 (the second row - not related to the 2s in 4-2-2), you have on the first row, 2 samples of chroma (for both U and V, or Cb and Cr), and on the second row, 2 samples.   So, on a 4-2-0, on first row, 2 chroma samples of each channels, but no sampling on the second row (it is like have half resolution on both x and y axis of chroma versus luma, eg. 1920x1080 luma resolution, 960, 540 chroma resolution).   In 4-1-1, you have one sampling of chroma on each row (so for a 1920x1080 luma resolution, you get a chroma resolution of 480x1080).   in 3-1-1 sampling, like XDCAM EX and HDCam HD, you have for each 4 pixels of luma in source, only 3 samples of luma at the output, that is why those codec are 1440x1080 with a 1.33 pixel aspect ratio.

Smoothness in the bottom image ? What is see is an just an awful lot of noise reduction.   On an image like that, you don't/can't really see the advantage of 4:2:2 over 4:2:0 or 4:1:1.   4:2:2 is not about detail in luminosity, but in chroma. As there is not much chroma information in this image, there is almost no chroma contrast/details/sharpness. We can't say for sure that it is 4:2:2.   The compression artefacts are 4:2:2 but just hoping that the camera is not wrapping 4:2:0 information in a 4:2:2 format.

Hum, not sure about being 4-2-2. Looked at the UV channels but there is not much color details in there besides chroma noise.   If you had the possibility to shot something like a red diagonal line it would be nice.

By the way, anybody tried to convert CinemaDNG to the lossy compression of the DNG available since DNG 1.4 - as a proxy copy ?   http://blogs.adobe.com/lightroomjournal/2012/10/dng-1-4-specification-notes.html   I only tried it on stills and I wonder if it could have an improve the workflow....

The 3rd and 4th XLR are not for audio. They are 12V DC Power XLR in and out plugs. So there is "just" only 2 audio XLR input. Personnally I'm perfectly fine with that.

  Just to note that to be able to use the F3 as 10bit 4:4:4 or 10bit 4:2:2 in 60p, you need an recorder that have dual link hd-sdi or 3g-sdi....

  Besides really heavy compositing, I don't care much about 32bit. 16bit should be enough for most grading. But that is another thread.   Try applying some noise (not grain) at a really low level (+/- 0,5% - but should be 100/256 = 0,39%), this will dither the image and reduce banding.   Try it before, and try it after grading effects.   Dithering will be even more important when you'ill reduce from 4k to 1080p

  The problem with a those Log color space on most cam, is that it is 8-bit. There is a lot of bits reserved for highlights but not much for mid-tone and almost nothing for shadow. It is almost useless in 8bit.   Even more, if your shadows and highlight are not using the full dynamic range and you stretch them a lot in post, you'll get a lot of banding.   I would use 8-bit Log only in a controlled environnement (eg. studio) where the highlights and shadows use the entire dynamic range.   I really think that Log should be at least 10bits.

Wow, even Black Magic have some license agreement against modifying, reverse engineering, decompiling,etc... for the Cinema Camera :   http://www.blackmagicdesign.com/support/detail/register/download/?sid=27541&pid=27542&fid=33840&did=33847&os=win&alt=true&sdk=false&beta=false&archived=false   License.rtf inside the installer   6. The Software is owned by Blackmagic Design or its suppliers. Title, ownership rights, and intellectual property rights in and to the Software shall remain in Blackmagic Design and/or its suppliers. You agree to abide by the copyright law and all other applicable laws of the Australia, United States and other applicable jurisdictions. You acknowledge that the Software in source code form remains a confidential trade secret of Blackmagic Design and/or its suppliers. You may not, permit unlicensed persons to use your copy of the Software, modify, translate, reverse engineer, decompile, disassemble, or create derivative works based on the Software, copy the Software, rent or lease, rights to the Software, or remove any proprietary notices or labels on the Software.  

http://web.canon.jp/imaging/eosd/firm-e/eosdigital7/firmware.html and others...   You shall not assign, sublicense, sell, rent, lease, loan, convey or otherwise transfer to any third party, or copy, duplicate, translate or convert to another programming language the Software, except as expressly provided herein. You shall not alter, modify, disassemble, decompile or otherwise reverse engineer the Software and you also shall not have any third party to do so. You shall not modify, remove or delete a copyright notice of Canon contained in the Software. Except as expressly provided herein, no license or right, express or implied, is hereby conveyed or granted by Canon to you for any intellectual property of Canon. 2. OWNERSHIP AND COPYRIGHT: The Software is copyrighted and owned by Canon. You agree and acknowledge that Canon transfers neither ownership interest nor intellectual property in the Software to you under this Agreement or otherwise, and that Canon retains all right, title and interest to the Software.    

  Absolutely, but some condition needs to be respected. Downscaling needs to be a proper algorithm using some kind ok interpolation (bicubic, lanczos, etc..), not a nearest neighbor one. And the math should be done in an higher bit-depth space (idealy 32-bit or more).   Here is an example of a 1-bit b&w image downscaled by 25% (for the example, it have been upscaled back without interpolation): http://www.digitalartform.com/archives/2012/10/lower_bit-depth.html

I simply disagree, on about every point.   First, the price should not be set on production (and must never be the only parameter on price settings. Even more, this is not a product from mass production. When marketing a product like that, the thinking is not of creating a bestseller - you should not set it a price low enough that is will sell five times more. It will cause more problem and more loses on medium/long term than the there is benefits to. Why ? Manufacturing.   This camera is produced in small quantity. There is a small part of a factory dedicated to it. If your price is too low to be true and you sell much more than you can currently produce, your first problem will be the same problem as with the Black Magic Cinema Camera, a lot of unsatisfied people waiting way too long for it (however on the bmcc the price/quality is way too good for customers to be too much of a problem). So what's now ? You wan't to increase production. How do you do it ? You reduce the production of another camera to replace it with this one ? This is a high end camera so you are not reducing whatever Powershot manufacturing line to do it, so you'ill probably use the 5D line, but is it a good solution ? I don't think so. If not, maybe you can create a new factory, train new people just for a few thousands cameras (that is less profitable cause you reduced the price, remember ?)   Also, I don't consider not exploiting the full ability of a component, crippling. It is in fact maximizing the manufacturing process by reducing the pieces count to a minimum as you reuse them in many products. That maximization help cutting the cost, that reduce the price at the end for the consumer. (12k for a 4k camera, how much was it 5 years ago ?) Does the iPhone 4S really that different from an iPod Touch to cost 2.5x more ?   As for firmware. I believe that software can be sold to whatever price. Be it a single line of code, or thousands of classes and functions. There is a lot of software that is using the same exact code for the free version than the Pro version that sells for many thousands. It executes in a different way cause in the serial you set a 3 instead of a 4. I really do think that the 1DC is different enough, and the added value (yes, I'm talking about you, single line of code) gives a much more interesting product than the 1DX - Which camera for less than 12k gives you 4k 4:2:2 8bit (that you can convert to 1080p 12bit 4:4:4 !) Will I buy no. I'm not DP, just an Art Director. Would I buy it for 6k ? Hell no, if they sell it at 6k, I want Sony to sell its new F5 at 6k also !

I saw the Hobbit yesterday in 3D/48 and actually really enjoyed the 48fps. It provides nice fast camera movement without the usual 24p stuttering.   What I really dislike however was the really poor grading in the highlights - too much dynamic there, should be more compressed. I really thinks that it was the element that gave the TV look more than the framerate. Even more, the highlight were really badly balanced between live action and CGI.   I fuond that it really missed a beauty pass on most scene - mainly on the atmospheric side. For example, in the cave with all hundreds of orcs, there was no fading, no dust in air covering the orcs far behind. The result is really more theatrical than cinematic.

I don't know. Are they really doing it ? Without any problem ? Here is the footnote from the official Canon 1D-C page : [url="http://cinemaeos.usa.canon.com/products.php?type=Camera-1DC"]http://cinemaeos.usa...type=Camera-1DC[/url] [b]* This device has not been authorized as required by the rules of the Federal Communications Commission. This device is not and may not be offered for sale or lease, or sold or leased, until authorization is obtained.[/b] If Canon cannot deliver this high level DSLR camera I don't see a lower end from anybody going in that territory anytime soon. [quote name='Germy1979' timestamp='1349785661' post='19522'] The 1-DC can do it. I wouldn't even say this if Andrew hadn't interviewed the Canon guy that confirmed it was just a badass 1-DX. The sensor has to be downplayed to a lower megapixel count to be optimal for video, but it's supposedly the same 18mp sensor as the 1-DX with a different firmware telling it to work a different way... [/quote]

[quote name='sanveer' timestamp='1349719677' post='19490'] 4 years later, and the video quality, in DSLRs, is exactly the same, albeit some minor, cosmetic changes. [/quote] I do not expect to have better video quality in a DSLR anytime soon. The concept in itself is far from perfect. To process those high resolution sensors, you would need better DSP chips. At the same time those chips generate so much heat to process all that data that this heat cannot dissipate out of those small body. They are just not created for that. The easiest solution would be to put in them a low megapixels count in them, kinda 4k, but as they are marketed as DSLR, it would not be a good success as it will not generate enough sales on the photo side (where most the sales are in DSLR). So I will forget a breakthrough in the video quality of DSLRs for a few years. By now, I thought that we would already have some camera likes the C300 from the competition. I really think this is where the future belong as it is the best form factor of any camera I ever saw.

From this grab, it seems it is just your monitor that is clipping. Non-calibrated Apple screens are really bad, they are too contrasty, a crush the black by a big margin. Not speaking of being glossy glass. On my calibrated Nec LCD2490WUXi, the left image is correct. No black crushed and - as there is any - no highlight clipped. You should really look to calibrate your screen, or get a better one. [quote name='EOSHD' timestamp='1348309262' post='18887'] For reference this is the levels problem I am still having with the GH2 after all these years, and now GH3. Why does my Mac not recognise that GH2 footage should be 16-235 not 0-255? Quicktime treats it as 0-255 as well, crushed blacks and blown highlights. Full resolution screen grab: [url="http://www.eoshd.com/uploads/gh2-levels.jpg"]http://www.eoshd.com.../gh2-levels.jpg[/url] [/quote]

sorry double post...

Actually it depends. When you have hard color edge - for example a red circle over a black (or green) background or a fin color line - the better chroma sampling you have the best it is. So it really depends on the shots. A lots of the color moire that people on this website are seeing is actually bad chroma sampling like 4:2:0. For computer graphics, anything less than 4:4:4 is really annoying. [quote name='Axel' timestamp='1347658179' post='18064'] I am not so sure what you expect of 4:2:2. You won't [u]see[/u] any difference compared to 4:2:0. It is color resolution actually. 4:2:0 is a compression factor that uses the [u]inability[/u] of the human eye to distinguish colors with the same accuracy as subtle brightness changes. It can't be downsampled well for different TV resolutions (SD, HDready, FullHD, interlace, progressive), that's why the broadcasters officially don't support it. There was a myth that you couldn't color-key with it. Now look at all the greenscreen videos shot with DSLRs. [/quote] In fact, what you describe is a linear color space. (Gamma 1.0) At 8bit, it is not really a good idea, as you'll get a lot of posterization in the shadows. Most image/video codec work in a logarithmic color space, mainly a 2.2 gamma curve. According to wikipedia, it gives you about 11-stops of dynamic range - similar to 11bit linear - that gives about 24-steps for each stop. As the GH2 seems to have 12 bits raw, you're not losing much dynamic range in the conversion (the first stop have only one step). A 10-bit log color space with gamma 2.2 should be similar to 14bit linear, and 12bit log to 16bit linear. [quote name='Axel' timestamp='1347658179' post='18064'] 10-bit is different. It is more related to the dynamic range. 1-2-4-8-16-32-64-128-256 describe the stops an 8-bit video can store. Every time the luminance doubles is one stop, makes 9 stops ideally. Due to the limitations of electronic signals, the range is mostly between 16 and 235, between 4 and 5 stops. HDR 8-bit means scaling the values through a curve, remapping them. But then, of course, the whole image is just flatted down, gradients that should have at least 30 steps to appear smooth get only 10, and banding occurs. Let's assume the GH2 could hold 8 stops by various tricks. If the GH3 has considerably better dynamic range, say 10 stops, how can they be stored in 8-bit, when 10-bit can only store 11? [/quote] Unless your doing a really really subtle gradient (or generating computer graphics), you should not see the difference. So you probably have amazing eyes, or a bad 8bit monitor. Where the 10bit shine is on grading - you'll have less posterization if you manipulate too much the luminosity. [quote name='Axel' timestamp='1347658179' post='18064'] It's easy to see a difference between 8-bit and 10-bit - on a 10-bit monitor. Alas, Panasonic may rightly suppose that most GH3 buyers don't own a 10-bit monitor. So perhaps the banding issue may not be solved. [/quote]

Since march, S-Log is a free firmware update.

[quote name='EOSHD' timestamp='1346287275' post='16905'] Interesting. So 10bit raw for example, is only 10 stops? That doesn't seem right to me, but there seems a lot of logic in what you're saying, do continue... [/quote] Please understand that this apply only to linear raw capture, not to processed data like 8bit with gamma 2.2 (from wikipedia : the dynamic range of an 8-bit JPEG file is about 11 stops) or else. So let say we have a 1 bit raw file. So for each pixel you have either light or darkness. Let's get a 2 bit raw : At 0, you have black at 1, you have 50% brightness (first stop) at 2, you have 50% brightness at 3, you have 100% brightess (second stop) 3bit RAW : At 0, you have black at 1, you have 25% brightness (first stop) at 2, 37,5% at 3, you have 50% brightness (second stop) at 4, 62.5% at 6, 75% at 7, 87.5% at 8, you have 100% brightess (second stop) In fact in linear, your first stop have only one step. As real sensor have noise, you generaly lose a stop or two. That's why sometime the sensor output 14-stops of data, but the manufacturer truncate the two first LSB (least significante bit) of the raw to lighten the weight of the data. eg. Olympus, probably BMCC.

[quote name='markm' timestamp='1346269429' post='16880'] So Canon derives 12 stops by using a gamma curve Surely that means compressing other stuff like maybe skin tones on higher exposures? Think somehow I'd be happier with 14 stops of raw. [/quote] Hmm, raw is linear, for each bit you have a stop. An 12-bit raw from an ideal sensor will have 12 stops. So the BMCC have max 12 stops of dynamic range. Is it bad ? Absolutly not. People want 14-stops but they forget that for each stop you double the quality and the range. But the BMCC is still far from a Phase One back, or let alone a Nikon D4.