p.11 #3 · Digital Medium Format / 35mm equiv difference
On lighting: from what I know Annie uses Photek Softlighters - essentially umbrellas with socks that soften lighting and give more painterly quality - something I been using a long time.
Edit: They give a nice soft light, but it's a PITA to position it close to model without skewering him/her/it with the umbrella shaft or getting it in the shot...
Yep, this is exactly what I think when I hear "MF look".
It seems that megapixels, and sensor size plays a role in smoother tonality or color gradation, or what ever else you call it. D800e camera proves at least the megapixel side of tonality improvement over other FF cameras. The question I have is, how much does 16bit sensor(not processing afterwards) has to do in smoother color gradation in MFD cameras? If say, next gen FF cameras come with 16bit sensors would we see improvement in color gradation rivaling MFD evermore?
p.11 #6 · Digital Medium Format / 35mm equiv difference
To me it's about pixel pitch for sensor area at given bit depth.
Plus of course CFA data interpolation.
It remains to see Canon's new 90mpx sensor..or whatever it is anyway...
p.11 #7 · Digital Medium Format / 35mm equiv difference
16 bit AD does so close to nothing (compared to a 12-bit AD) that it's mostly an exercise in marketing. It's also a bad component choice, since most of the AD's they choose are specified to 2-2.5 LSB errors with zero signal.
Most MF backs would do very well with accurate 12-bit AD's. The newest generation justifies 14-bit convertors, but only just...
It's like measuring out portions of spongecake with a laser interferometer, to micrometer precisions. It doesn't really matter since simple stuff like a 2º change in room temperature or a few percent of air humidity will induce errors that are hundreds of times larger than the initial measurement. And the cake will move when measure it, so you can't even get the same value twice, no matter how many times you try...
The same goes for AD conversion. If the absolute lowest margin of error is 1:10,000 from the sensor (about 13 bits), then you can "measure" it with a million bits if you like - the error will still be at the 13'th bit level. All the bits after that are just random noise.
A normal pixel size today is around 5x5µm, and from that you get about 50ke- of light energy converted - enough to give a base ISO of 60-100. On that you have a base read noise of maybe 4e- (that's four single electrons per pixel!), so divide 50,000 by four, get 12,500e-. 12,500 is the same as 2^13.5, so 13.5 bits for the very best sensors we have today.
Those are about 1.5 bit higher numbers than what you have in most MF sensors, where you have about 10-15e- of read noise like in the P45+, P65+ or the H4D-series.
Truesense (Kodak) even publishes the spec sheets, here you go: http://www.truesenseimaging.com/all/download/file?fid=8.63
In that, they even specify 69.3dB of DR, that's 69.3/6 = 11.55bits of DR, which means that 12 bits is actually more than you need.
Dalsa (P65+) specify slightly better values, for the current version the say about 11.9 bits, and 50ke- FWC. 12 bits would still be more than enough in a well balanced construction.
p.11 #8 · Digital Medium Format / 35mm equiv difference
Joakim beat me to it, but I already wrote my reply, so here it is:
bgbs wrote:
The question I have is, how much does 16bit sensor(not processing afterwards) has to do in smoother color gradation in MFD cameras? If say, next gen FF cameras come with 16bit sensors would we see improvement in color gradation rivaling MFD evermore?
I am not sure of the origin of the 16-bit meme for MF Digital Back. It seems to be a long lived meme that 16 bit files from a DBMF yield superior dynamic range and tonal range over 12 bit or 14 bit raw files. Is it possible that it originally was as "16 bit TIFF files are superior over 8 bit TIFF files", and that this 16 bit superiority meme uncritically propagated into the world of 12 - 14 bits raw files?
Regarding dynamic range, the current sensor technology with the highest per pixel dynamic range is in some 24 MP FF sensors, like in the D600. With 13.5 stops DR mapped linearly into 14 bits files, one can argue that there would be a slight benefit in moving to 16 bits. The irony is that MFDB sensors that always gave 16 bits files are slightly behind that, they don't need more than 12-13 bits and would all be well served with 14.
Your question was about color gradations.
Quoting from Joaking at page 3:
"Some of the main parts that are hard to mimic is the way noise and luma detail contrast relate to the underlying color. In a Canon, that has very low green-orange separation in raw, luma detail and noise on green, orange and red colors will behave a certain way. They tend to get "flat", and hue tends to get rather flat too. Trying to increase color separation in the camera profile increases chroma noise in the file, and it also has negative effects on luma - giving lots of false detail."
The origin of the chroma and luma noise that will be amplified as described here, is still mainly photon shot noise. Even if the sensors can capture a number of photons (in the 50 000 order of magnitude) that intuitively require 16 bits files, the photon shot noise on this scale is very high in absolute numbers. Let's take one example at 18% grey of a sensor saturating at 55 555 photons, that will be 10 000 photons in average. The noise level from photon shot noise is the square root of that, or 100 photons. The signal to noise ratio is 10 000 / 100 = 100, or 40 dB. That is regarded as a clean image. But if you start to push the color separation, the noise levels will rise as Joakim describes.
If a 14 bits raw file is used to record the information from such a sensor, the quantization steps will be something like 4 photons. This means that we have 100 / 4 = 25 quantization steps to resolve the noise. At this level there is no appreciable benefit of increasing the quantization, it is all about getting enough light into a sensor beneath a suitable color filter array.
p.11 #9 · Digital Medium Format / 35mm equiv difference
Takes a little time to digest for a less technical person, but I am getting there...
Very interesting development!
ok, so far we removed the PP and lighting out of the equation.
Now we are down to essentially 12Bit delivering what we need.
What do we have left so far that makes a difference? Size?
p.11 #11 · Digital Medium Format / 35mm equiv difference
Krosavcheg wrote:
Takes a little time to digest for a less technical person, but I am getting there...
Very interesting development!
ok, so far we removed the PP and lighting out of the equation.
Now we are down to essentially 12Bit delivering what we need.
What do we have left so far that makes a difference? Size?
The glass design @ projected image/angles of incidence/transition rates.
p.11 #12 · Digital Medium Format / 35mm equiv difference
Matt: viewing on iphone atm
1. Beautiful but can be DSLR as well
2. Light is too harsh to tell - it could be film rangefinder, dslr, m43 or MF
3. I seen the kitteh shot before colour of the door and light falloff is beautiful.
p.11 #14 · Digital Medium Format / 35mm equiv difference
Viewing them on my 17' screen I can now see the crisp.
They look surreal but in a different way - as if detail was introduced later somehow.
Now kitteh looks oversharpened.
2. looks less harsh on computer screen compared to iPhone
Thought it might make an interesting comparison, as to me the Foveon images I've seen have a distinct look which presumably must be attributed to the three colour samples per photosite.
I wonder how colour sensitivity detail compares to MFD backs (I know the Foveon has issues under certain conditions).
p.11 #16 · Digital Medium Format / 35mm equiv difference
Matt
Wikipedia got a breakdown on CFA in Foveon - I would just make a muppet of myself re-writing it.
Basically same old - a pull somewhere to push somewhere else...
p.11 #17 · Digital Medium Format / 35mm equiv difference
A MF images have different look. So do motion pictures. The slightly different sensors in MF cameras or Zeiss Master Primes etc. lenses play a role in it, but I think most of the effect is in building the scene, for example manipulation of light and color coordination of props and in post processing.
p.11 #18 · Digital Medium Format / 35mm equiv difference
It seems that we FF users are screwed because FF can never match MF quality, and since we can't afford MF because we want better IQ, we are screwed even more (I'm being overly sarcastic of course).
So far I've learned that crunching more pixels on FF sensor will put stress on lenses causing yet more problems and detail issues. I wonder if Foveon is the avenue for future growth for FF cameras? With Foveon manufacturers can dial back on resolution, yet maintain detail, which puts less stress on lenses, and still maintains great color. The only issues they would need to solve is high ISO.
p.11 #19 · Digital Medium Format / 35mm equiv difference
Sigma doesnt have MF look, cause its APS-C. It doesnt have even FF look.
That said, its awesome in certain departments.. (resolution, BW work, base ISO stuff).
Quite a bit of MF look comes from lens itself, but mostly its simply cause everything is bigger. Not that obvious with digital MF as it was with analog MF.
Leibovitz shoots with Leica S system too. Not much of "real MF", but it seems it works for her... (well, her pics certainly do).
p.11 #20 · Digital Medium Format / 35mm equiv difference
bgbs wrote:
It seems that we FF users are screwed because FF can never match MF quality, and since we can't afford MF because we want better IQ, we are screwed even more (I'm being overly sarcastic of course).
So far I've learned that crunching more pixels on FF sensor will put stress on lenses causing yet more problems and detail issues. I wonder if Foveon is the avenue for future growth for FF cameras? With Foveon manufacturers can dial back on resolution, yet maintain detail, which puts less stress on lenses, and still maintains great color. The only issues they would need to solve is high ISO.
Um, not really.. Foveon in SD1M is 15 mpix true resolution, which is way harder than anything else on lens. That sensor is reason why Sigma made 35/1.4 and 18-35/1.8 .. those lens are made for that at first place (plus its used as test bench for any new Sigma lens).
Foveon isnt as far as resolution goes any different than Leica M-M. Per-pixel resolution, no interpolation of detail. Which among other things is why its so good for BW.
