brainiac wrote:
It's not entirely clear to me that pixel DR and image DR are exactly the same thing when factors like noise, Bayer interpolation, colour filters, and so on, are involved.
The extra DR would have to be reflected as extra information at the pixel level, and this information would be detectable.
brainiac wrote:
Think of tape bias: a noise deliberately applied to the medium in order to extend DR.
That's interesting! Doesn't seem to be applicable, though, from what I've just read. Tape bias operates by pushing the signal into a better part of the transfer function (so says Wikipedia plus another site). Here we're talking about whether naturally occuring random noise in a single subsensor, combined with other random subsensor noise, can somehow generate picture information. In neither situation, anyway, do you have signal information being created from nothing but random noise.
Now I will retire from the field, as I am out of my depth. It still makes sense to me that the pixel-level DR is also the maximum achievable sensor DR, since you can't get something from nothing.
Personally I don't lust after high megapixel yields, 10 or 12 is fine by me. I just wish someone would bring out a sensor that doesn't band in the shadows at high ISO.
Banding holds me back more than sensor yield or lens quality, in fact now that we have Live View (for accurate manual focus), and DNG (for RAW archiving), seems to me that banding is the "last great digital problem".
jvarszegi wrote:
Now I will retire from the field, as I am out of my depth. It still makes sense to me that the pixel-level DR is also the maximum achievable sensor DR, since you can't get something from nothing.
That's how I would see it too. Anything done to the signal after capture can not increase the DR, although it may map it to more useful areas via use of tone curves such as is done with HTP and DRO.
Here's an excerpt that's relevant and then check his figure 7.
The pixel size in digital cameras also affects dynamic range. Dynamic range is defined here to be the maximum signal divided by the noise floor at each ISO. The noise floor is a combination of the sensor read noise, analog-to-digital conversion limitations, and amplifier noise. These three parameters can not be separated when evaluating digital cameras, and is generally called the read noise. The measured read noise near unity gain is essentially equal to sensor manufacturer's published specifications for read noise, so the zero signal case is read noise limited. As you might have surmised by now, with the larger pixels collecting more photons, those larger pixels also have a higher dynamic range. ...Show more →
That is basically HDR superposition. The idea must have occurred to just about everyone in the past, even me. It's crazy that patents are granted for such things.
David Baldwin wrote:
Banding holds me back more than sensor yield or lens quality, in fact now that we have Live View (for accurate manual focus), and DNG (for RAW archiving), seems to me that banding is the "last great digital problem".
Nikon certainly beats Canon for minimal banding at high iso, but not by a huge margin.
David Baldwin wrote:
Personally I don't lust after high megapixel yields, 10 or 12 is fine by me. I just wish someone would bring out a sensor that doesn't band in the shadows at high ISO.
Banding holds me back more than sensor yield or lens quality, in fact now that we have Live View (for accurate manual focus), and DNG (for RAW archiving), seems to me that banding is the "last great digital problem".
Besides the banding there are also the ugly purple/green colored blotches that seem to be more dominant in the 50D vs 40D/450D. These blotches aren't correctable with NR software like chroma and luminance noise speckles and therefore potentially more harmful to a picture IMHO. They will be noticeble in prints as well.
Pixel Perfect wrote:
Chroma noise has always been a Canon weakness, even when they were delivering better luminance noise than Nikon.
It certainly is a problem, but I don't think the Nikon sensor does any better, it just has much more sophisticated noise reduction. To apply Nikon-style noise reduction, try the following recipe:
- duplicate base layer
- run smart blur on the dupe layer
- set smart blur settings to 8, 16, 24 or 32 pixels, high quality, overlay edge mode, and set threshold so that the right quantity of edges show in the preview (the white lines will be where you are not removing noise/colour detail)
- use wand, tolerance 0, non-contiguous, not all layers to select all white pixels
- delete
- feather by a pixel or so
- delete again
- gaussian blur by 2 pixels
- set layer blending mode to colour
- set layer blending options to blend if grey 0, 100/200 or thereabouts (option click the white point slider to split it into the upper and lower graduation points), so that only the shadows are affected
- if necessary un-apply by painting black on a layer mask
Obviously this can all be made into an action so it's one click.
Pixel Perfect wrote:
Chroma noise has always been a Canon weakness, even when they were delivering better luminance noise than Nikon.
Noise in the raw data is neither chroma nor luma. It only becomes so after demosaicing. If one camera is showing a different ratio of chroma/luma noise than another, it is due to differences in the raw conversion, not the raw data.
ejmartin wrote:
Noise in the raw data is neither chroma nor luma. It only becomes so after demosaicing. If one camera is showing a different ratio of chroma/luma noise than another, it is due to differences in the raw conversion, not the raw data.
Ok, but Canon's conversion seems to be worse than Nikon's for chroma across many converters.
Pixel Perfect wrote:
Ok, but Canon's conversion seems to be worse than Nikon's for chroma across many converters.
If there is an issue, it may be with the spectral response of the color filters. If there were poor color separation in Canon's choice of filters, that could potentially lead to more chroma noise as small color differences were amplified to generate proper color saturation.
People's apparent general apathy over the Canon 50D's extra pixels should alert Canon that their marketing department may be barking up the wrong tree. Cost and conspicuous value, not megapixellage, are the concerns now.
Yes, until the economic downturn improves Canon should concentrate on reducing price. There doesn't seem to me to be a general clamour for increased megapixels, but money all over the world is tight. Last Photokina there were rumours that Canon were going to issue more than one FF body, the 5D2 they did issue, but its less highly specified, cheaper FF brother did not appear. That missing camera represents a missed opportunity for Canon, the lack of a modern entry level FF with say 12-15 MP, is very disappointing to many of us. Perhaps Canon could have taken the original 5D, put Live View and a dust shaker in it, and sold it at a bargain price (on the basis that the R&D on such a camera would be very small, the sensor could have remained the same). I would certainly have bought such a camera, I suspect I am not alone. The video boys would still pony up for the 5D2..
Let Canon start the push for sanity by issuing an modern austerity full frame body, at a half or two thirds the price of the cost and sensor resolution of the 5D2 (which here in the UK may go for a whopping £2300 inc sales tax). £1150 for the more basic 5D upgrade seems more realistic in a country where next year unemployment is expected to climb in the millions. Even those of us with good jobs are getting nervous about paying for luxury items, and top dollar for top specifications will not result in many sales IMHO.
IMHOP, the pixels are always one part of the equasion, albeit, after 6mp a smaller one, but I feel like they are doing a great job allowing you to choose how many pixles you use with the different file sizes.... just because they are there doesn't mean that every wedding photog should have to get a new 10tb hard drive after each week of work..... but it breeds innovation, less pixel noise if done right, and who knows what else. perhaps the day will come that a ff camera isn't really that special? Who knows. I still want my 5dMkII (but will be settling on something less)
David Baldwin:
Perhaps Canon could have taken the original 5D, put Live View and a dust shaker in it, and sold it at a bargain price (on the basis that the R&D on such a camera would be very small, the sensor could have remained the same). I would certainly have bought such a camera, I suspect I am not alone. The video boys would still pony up for the 5D2..
Exactly! What really surprises me that the price of 5D is still high ($2000). I thought that it would be no higher than $1500 at the end of this year. What seems to be happening is that 5D is becoming more difficult to get (especially in Canada)
I wonder if Canon's apparent decision not to bring out an affordable Ff body is in fact designed to allow them to keep the 50D price high. If so that's their decision.
But if thats the way they are thinking they are unlikely to get any of my money in 2009. IMHO the 50D isn't much of an upgrade to my 40D, and the 5D2 isn't priced or specified for the likes of me.
I can't be the only reasonable well off keen amateur that at present doesn't have a Canon body fulfilling their needs, or the prospect of one. Surely Canon need to promote affordable FF photography so that we go out and buy lots of L lenses. I've got 2 already, but until the body situation is rectified I have no reason to buy anymore.
I don't care about megapixels frankly - the 10 megapixels of my 40D provide surprising amouts of detail, but I need Ff to produce shallower depth of field and 35mm type perspectives. Megapixels now have exactly zero capacity for making me reach for my cheque book.
Tentacle wrote:
Then again, as someone mentioned, the processor GHz wars ended. Intel gambled big on an architecture that was trading off instructions-per-cycle against higher clock speeds, and they ran head-on into a thermal barrier.
The Pentium 4 a.k.a. NetBurst architecture was meant to hit 6+ GHz. They stranded at 3.6. And they went back to an older CPU architecture and build on that to deliver us the Intel Core Duo, Core Quad and Core i7 processors. At lower clock frequencies... So who knows...
My thoughts exactly.
For a time when the "MHz War" was raging, if you wish to call it that, AMD had the computationally faster processors with lower clock speeds than the Intels, so AMD branded their CPUs as "this is the speed the CPU would be if it were an Intel" in order to compete in the marketplace to win over the technically ignorant consumers.
I can see something analogous happening with camera sensors.
David Baldwin wrote:
I wonder if Canon's apparent decision not to bring out an affordable Ff body is in fact designed to allow them to keep the 50D price high.
I'd be happy with an affordable 1.3x or even Nikon's 1.5x sensor size.
I had a 50D for a few weeks and returned it (I had a 40D,30D,20D) I dont think its of much use to landscape photographers and I only use L glass, I think using it for portrait work it could be a great camera just not stopped down. So my 5D will suffice till the Mk II is proven.
