Camera with the most latitude
Posted 28 January 2011 - 01:04 AM
Just wondering if anyone knows of an HD cam out there with the same latitude range as film (11-13 stops generally, I don't know a lot about HD, thus my question). I really like the results of the Canon 5D/7D but they look too 'digital' (not as 'flat' and interesting as film).
Posted 28 January 2011 - 02:19 AM
I suppose when you say ”latitude”, you really mean ”dynamic range”. The answer might be the Arri Alexa – check out the link in this thread.
The Alexa looks awesome! Aren't 'latitude' and 'dynamic range' interchangeable (synonyms)? As well, how does the RED and Canon DSLR cameras compare to the Alexa?
Posted 28 January 2011 - 03:15 AM
On www.provideocoalition.com you find a halfway professional and comparable setup where they've tested the ALEXA, the Canon 5dMk2 and the new RED:
It says 14 stops on the Alexa, 11 on the RED MX (which is the same sensor technology as the EPIC) and 9 stops on the Canon. The results can vary depending on ISO, white balance, even lens choice (less contrasty lenses brighten the dark areas and creating the illusion of additional dynamic) - but they give you an idea. I think they also correspond quite well with real-world-results. They recently did a comparison for evaluation in German public-TV and tested the Alexa in a foundry and compared it to S16 Kodak 5219 (scanned with ARRISCAN). Result: resolution and dynamic range are very close, images of the Alexa are much cleaner and with higher sensitivity.
Posted 28 January 2011 - 04:09 AM
No, but they get misused as synonyms a lot.
Aren't 'latitude' and 'dynamic range' interchangeable (synonyms)?
Dynamic range is the whole range of light intensity that your device can capture. Latitude is your "margin of error", ie. how much you can under- or overexpose before your footage is screwed. As Georg pointed out, there’s no single correct way of measuring the dynamic range of a camera or a film stock – and latitude is an even more subjective thing, as it also depends on the scene being shot.
Check out this Wikipedia article or just google it up.
Edited by Antti Näyhä, 28 January 2011 - 04:10 AM.
Posted 28 January 2011 - 07:56 AM
I just did some serious test with an ALEXA and RED MX.
My experience and test's tell me that the "printed latitude in manuals" always to high is.
Our result was
ALEXA +/- 10
RED MX +/-9,5
My experience with the Canon is something between 7,5 and 8,5 stops (depending on ISO)
An other option (budget question) is the Panavison Genesis. I haven't tested it but what I read and heard it's something around 11.
Also important to know is, that the curve from the Alexa is "moving" in relation to the ISO. (Log C, Pro Res 4by4)
For example: With a low ISO you get more detail in the shadow. So actually what one need when you're shooting Low light.
If you go higher the captured areas move up.
Art Adams did some great test on that as well: http://provideocoali...u_need_to_know/
Posted 28 January 2011 - 08:36 AM
Art Adams also did the dynamic range tests I mentioned, although he doesn't use the more sophisticated DRTC-technique, he comes to nearly to same result as ARRI itself when speaking about 13.5/14 stops - we are not even talking RAW, we're talking ProRes4444! All possible sources of error are also suppressed due to the same measurement standards (even the same lens) - and still, the diffference between ALEXA and the rest is huge and clearly visible from the samples as well as real-world tests (I haven't found any online-samples from the foundry-test, sorry).
Makes me wonder why your results are this close to each other and far apart from Art Adams/ARRI?
Mr. Adams tested the ALEXA at 800ASA, as well as the Red MX (which propably reduces DR) and 400ASA ("base" ASA 100-160?) for the Canon.
Posted 28 January 2011 - 12:07 PM
Posted 31 January 2011 - 09:03 AM
Sensor datasheet specs are generally meaningless.
Digital sensors generally need a lot of blue gain to produce warm tungsten images. This limits the dynamic range is blue. Think 10 stops in green but only 8.5 stops in blue. Even a daylight image could waste a stop in red and blue because green is very powerful. These gains ar digital in modern camcorders, so they come with highlight loss and higher middle grey quantisation.
Digital gain such as used on Red One changes dynamic range. Even though the sensor captures the full dynamic range, the dynamic is distributed in a different way around middle grey. ISO300 can be 6stops under and 3 stops over, ISO600 is 5 stops under and 4 stops over, ISO1200 is 4 stops under and 5 stops over and so on. The problem with this method is that the middle grey quality, everything that counts is progressively deteriorated by quantisation. Tonality is lost, one stop per 2x ISO boost. That means you get 1/8th the possible tones on the mid grey exposed subject for each 2x digital gain boost in post. The dynamic range over middle grey is high but it must be shaped to a very small range on the output, because we do want the mid grey to be at a certain level after all. This limits its usefulness to a degree.
Analog gain as used in most other cameras limits dynamic range. Here the problem is not quantisation, the analog gain makes sure the mid grey gets the same digitization quality (bits) regrdless of gain setting. The problem is noise level itself. The analog gain paradigm avoids quantisation but it does not get a highlight range advantage of digital gain. The over range stays fixed. ISO300 is 9 stops, 6 under and 3 over. ISO600 is 8stops, 5 under and 3 over. ISO1200 is 7 stops, 4 under and 3 over and so on.
A dynamic range test chart is only valid for fully uncompressed cameras. In this case you know that if the shade is visible it will also have texture (according to zone system). On a compressed camera it is misleading, you might see the shade but the compression only allows blocky artifacts at that exposure level, which limits the usefulness of the tone. In some dSLRs, the video is heavily quantised even 2 stops below middle grey. Ansel Adams would think they are very low dynamic range system even though they don't do that bad in dynamic range test charts.
Film is rarely delivered to the audience with more than 9 or 9.5 stops of dynamic range and most digital cameras can get that today. Even consumer models when recorded uncompressed. The problem is not dynamic range, it's latitude, the degree of error we can get in the exposure level when shooting digital.
One problem of digital is that highlights are unnaturaly clipped when the sensor saturates. Many companies have attempted to fix the issue with gamma curves and some are better than others.
A high dynamic range camera does not really solve the image quality issue and does not guarantee filmic images. You get more room for exposure error, and improved noise levels, but you still have to decice what 9 stops of range you will actually present to the audience, because there is the magic of film color, high midrange saturation, gentle highlight shaping and velvet blacks. You can't get filmic images if 15 stops are still visible in the frame. You will want skin tones that are rich and realistically saturated not zombie faces, products that looks true to their saturated logo colors, food that looks good, colorful fabrics. It doen't matter what type of material is shot with the system, the image must look good. In film, the film print or the digital equivalents take care of this problem. Very high dynamic range on the output might be useful in engineering applications, but photography is not its place. Many dSLRs achieve more than 14stops in raw output, and we can get great useless images with 14stop dynamic range if we tweak raw development. Their default jpeg setting keeps about 8-9 stops of dynamic range due to the aesthetic problem of color quality and contrast.
The problem with dynamic range is that the systems we use to view the material cannot cover much range in our viewing system. The human vision adapts but not that much, it's still a small part of our range. When we do get displays with true high dynamic range such as 1,000,000:1 that also cover an equivalently high range in the human vision range, dynamic range will be able to represent reality. I doubt realistic contrast will be welcome by the audiences though, most people prefer being blinded by a relatively blight white screen that to the actual sun. It's aesthetically more useful. In the same fashion, in a horror film we actually prefer the representation of a dungeon full of rotting bodies on screen as implied by image and sound alone. We wouldn't want to actually smell the dungeon.
Till then, high dynamic range digital is like film negative. More room to move and the choice of what to present to the viewer made in post.
I see room for truly bad images once HDR features are implemented in camcorders. Many will want to keep the range, they will see it as a solution to the photography problem, etc. Lifeless images. Instead of lighting for interior so that saturation and skin tone will look great in the interior, or light for exterior, many will simply not light. In post they will eventually discover that even though the have the range, the contrast and saturation is wrong and it takes a lot of tweaking and great knowledge on the subject of gamma and color to get what a normal exposure with a typical dynamic range camera would easily get. Since users are rarely educated in imaging and raw development tools are designed by engineers that usually ignore the photographical usefulness of their RAW tools, it might be a recipe for disaster.
Posted 30 April 2012 - 04:50 AM