Wow, my head is spinning after reading this thread. My brain is clearly short of a few bits.
Tentacle wrote:
Processing headroom and prevention (further reduction) of quantisation errors in shadow parts.
Does this mean that there will be less noise introduced by white balance adjustment and / or the possibility to make even greater white balance adjustment than is possible with 12 bits? This is of particular interest to me as an underwater photographer.
Does this mean that there will be less noise introduced by white balance adjustment and / or the possibility to make even greater white balance adjustment than is possible with 12 bits? This is of particular interest to me as an underwater photographer.
Thanks!
The deeper you get, the more blue the light becomes, red light is absorbed more. Now, if you correct the WB to restore normal white, it means that the red channel gets boosted by a significant amount. (And the green channel to a lesser extent.) Having 14 bits instead of 12 bits means you've got 4 times the tonal graduations, so you can restore an even weaker red channel before posterisation becomes an issue.
(Keep in mind that this is just an educated guess on my part. I know the advantage will be there, I can't estimate how big it actually is.)
hankk wrote:
IIRC, photons are binary-its either there or not. So you are measuring a number of photons with either film or sensor. The problem is, we can't count individual photons and so have to aggregate them.
A single photon is either there or not. But you're talking about many photons per unit time, producing an overall exposure intensity (often measured in candelas per square meter).
Its probably better to say that film has many more discrete levels, and the processing can non-linearly amplify these levels to the range where the human eye can differentiate them.
On a sensor, there smallest difference is 1 (assume a scale of 0 to 16K), while in this same step ( 1/16K of the total range) film has thousands of steps (think of the difference between a 14 bit number and a 32 bit number -- both can have the same bottom and top level, but the 32 bit number has many more steps).
I think this is where the comparison between film and digital completely falls apart, for a couple reasons. First of all, at the ultrastructural level, the density of film is accumulated according to an exposure-response curve that is very seldom linear. This is particularly true of older emulsions in which there is variable grain size. Films have a flatter 'toe' than the shadow end of digital, and they have a far, far flatter 'shoulder'. So above zone 9, 10, 11, your shoulder continues to trail out. Different developing techniques can alter the slope of the curve and where the shoulder and toe begin (with more flexibility in the shoulder, of course).
At the microscopic level, film grains of course are not simply on or off. They assume different size and shape depending on the emulsion, the intensity of exposure, and the developer chemistry. They can clump into particles of various sizes depending on the developer chemistry. They can be toned with selenium or gold, which will disproportionately increase highlight density. And they can be developed with staining developers that mask grain structure and give a compensating effect in highlights.
So there really are no discrete levels at all with film, either at the level of the grain or at the level of the exposure-density curve. The chemistry and physics of a single film grain just can't really be compared to a single photodiode or (ultimately) pixel.
I mean, I just developed some sheet film shot with Tmax 100, Tech Pan, and HP5+, all with vastly different emulsions, and all shot on 4x5 at the same time (I was doing a comparison). The differences in microcontrast, macrocontrast, and exposure response relationship is just beyond belief. But then again all three of these films have completely different emulsions. So I just don't think you can make any direct comparison to the data content of a bit. A bit gives you a discrete value; a film grain does not.
DrPablo wrote:
Pondria -- even if there were examples, how would you know that any visible difference was due to the bit depth instead of some other factor?
Everything else being the same, any difference should be due to the 14bit.
Canon or someone should be able to run controlled tests specifically designed for the case.
ziggg wrote:
The simple answer is with 14 bit A/D, the noise floor is effectively lower than with 12 bit, therefore raising the dynamic range, which is defined as maximum signal divided by the noise floor. The Canon 1D Mark II and 5D are limited by 12 bit A/D; the 1D Mark III, even without its sensor improvements, would benefit from 14 bit A/D conversion.
Pondria wrote:
Everything else being the same, any difference should be due to the 14bit.
Canon or someone should be able to run controlled tests specifically designed for the case.
That's a tough order, though. The fact that it's 14-bit means that you'll automatically have different data encoding, different generation of a RAW file, and different RAW conversion. You'll also certainly be talking about equipment that differs in other respects in terms of processing (like Digic III, and dual processors).
Finally, you'll need to have testing parameters that will be able to illustrate a difference. It may be that under normal or low contrast lighting, a 14-bit file will more easily posterize (when edited) because a narrower dynamic range is stretched over a wider number of values. But under high contrast lighting the 14 bit file will capture more noise-free data in highlights and shadows, especially when shadows are lightened.
So it may be a tall order to produce an unbiased comparison that illustrates what kind of conditions are necessary to reveal a difference (one way or the other -- or both).
Just like me to offer problems without solutions. Sorry
DrPablo wrote:
That's a tough order, though.
[ ... why it is tough to demonstrate the diff ...]
I do not necessarily agree with the arguments that you made. But, that's fine. Let me just say we agree.
So, if it is so tough to show the difference even with the specially designed test to demonstrate the difference, how could one see the diff on the images ? That is the main question that I asked. And I bet, we will see soon people bragging about, "Wow, look at this DR and Fine tones. 14 bit works !".
Very often the best we can do is an imperfect test, so something formal from Canon will of course be more convincing than enthusiasts on a forum.
I mean for all the discussions we have about the different 70-200 lenses, or RAW versus JPEG, or Tokina versus Canon superwides, or Nikon versus Canon, it's pretty rare that these differences are demonstrable under real life shooting conditions, especially the poorly controlled ones that we offer one another.
Just look at Dan Margulis' vociferous arguments that 16-bit editing in Photoshop offers no advantages over 8-bit. Whether he's right or wrong, the mere fact that someone of his stature and expertise can make that claim (with abundant evidence) just goes to show how subtle the differences are in practice.
One of the solutions used in industrail applications is to use Peltier modules to cool the sensors to about minus 30-60 Celcuis. This lowers the noise floor to almost zero. The drawback is that the sensors have to have moisture barriers, high battery consumpion, added weight, etc. which makes that approach impractial. Another approach is to cool with a liquid nitrogen flow which takes the shot effects to zero, Highly impractial for cameras but works well for fixed industrial sensers. As the sensor technology improves the noise floor is decreasing without the cooling, though I doubt the shot effect noise will ever reach the same level as cooling. As Director of Field Engineering for a company that produced test equipment for the semiconductor industry I worked with the early sensors that required such cooling and detected IR radiation in a totaly black incloser.
My understand is that the 14bit, linear, data is recovered in the RAW converter, and so that prints and other output medium can be acommodated, the data is converted into a nonlinear file (adjusted mostly by eye on a monitor where the data and dynamic range are compressed) and stored as a (further) uncompressed DNG, TIFF, PSD or compressed JPEG, etc.
Although there maybe printers that accept 16 bit data I don't know of any "16 bit" paper. So artful (hopefully) compression of image data is necessary when printing since the 8 to 9 stop dynamic range of a DLSR image file must by displayed by reflected light from a print with a maximum 3 stop range.
There are no 16 bit monitors-they are less than 8 bits with very limited resolution which is why we zoom in so much. LCD monitors are today's equivalent of a slide projector which was (then) the high quality enlargement/display medium of the film days. With a good projector and a beaded screen you could get huge displays to rival the biggest Plasma sets with a very wide dynamic range with a 35mm slide. For the few with 2 1/2 X 2 1/4 slides that was heaven!
So my understanding of the upshot is that DLSR encoding of original linear data at 12 and above bits provides more information that can be squeezed (EX: merge two different exposures of the same scene) into today's limited display formats. When a higher quality format evolves in the future then reprocessing of the original RAW (or a DNG encoded file) will allow more data to be displayed.
Pondria wrote:
Everything else being the same, any difference should be due to the 14bit.
Canon or someone should be able to run controlled tests specifically designed for the case.
Well, I can't help you with that because I don't have a 1D3
Alan321 - I've been following you and have to agree.
Like Pondria I would like to see examples. In another thread I stated what I would like to see: a high contrast scene taken with the same tripod mounted lens and different bodies (at the same ISO 100 setting) - the 1D III and something like the 5D.
It is my belief that the tonal range in the 1D III shot will be greater (I said "tonal", not dynamic, not stop, not absolute, not shooting, not firing...).
Now maybe my belief will be upheld or I'll find myself saying "well, there goes that theory", but I'll decide when I see the images (unless there is something about the way they are done that still gives me hope for the better images).
And if that difference is there, which is my belief from the images I have seen so far, I would like to have it. It makes a difference in current editing processes and things could change in the future as far as what could be done with those extra levels (tones) in future output devices.
And if sensors were incapable of distinguishing more than 4096 different tones or voltage levels (due to noise or whatever) then there would be no reason to change from 12-bit to 14-bit processing because of the expense involved and the extra data processing involved, not to mention all the changes needed in the software. It wouldn't be done - not cost-effective.
So I'm thinking we're actually getting something here (more tones, more different color values in the image) and I think it can be seen already, but a controlled sceen and equipment shot difference would help a lot. A nice sunset scene would probably be one of the best to see. I was very surprised by the posterization in my first 5D sunset shot (wouldn't have been if I'd thought about it I suppose).
So, anybody with a 5D and a 1D III got a nice sunset handy?
SoundHound wrote:
So my understanding of the upshot is that DLSR encoding of original linear data at 12 and above bits provides more information that can be squeezed (EX: merge two different exposures of the same scene) into today's limited display formats. When a higher quality format evolves in the future then reprocessing of the original RAW (or a DNG encoded file) will allow more data to be displayed.
That's nothing new, though. Negative film has worlds more information than can be displayed in a print. There are many ways to compress information in a way that evokes the contrast range of an original scene, or an intermediate storage medium like a digital file or a negative. Fundamentally, people will believe that a print is super high contrast if you have a range from deep black to paper white. The trick is engendering it with a curve that preseves believable tonal relationships and transitions.
Not new to you or me DrPablo-maybe new to some others. However the thrust of my comment was meant to indicate, to 12/14 bit camera owners, etc that 8 bits after the RAW conversion is enough, today, for post processing and existing display mediums.
ziggg wrote:
The simple answer is with 14 bit A/D, the noise floor is effectively lower than with 12 bit, therefore raising the dynamic range, which is defined as maximum signal divided by the noise floor. The Canon 1D Mark II and 5D are limited by 12 bit A/D; the 1D Mark III, even without its sensor improvements, would benefit from 14 bit A/D conversion.
Absolutely, which is why I posted a reference which explains why the dynamic range of current high end digital SLRs are limited by the 12 bit A/D converter and will benefit from 14 bit. Did you read it?
