p.3 #1 · Digital Medium Format / 35mm equiv difference
The color rendering of say a H4D or a P65+ is close to impossible to mimic with certain smaller cameras, Canon uses a totally different filter strategy, and so does Oly. It EXTREMELY hard to get the same effect in manual PP work, and actually impossible to do it fully automated even with the best camera color profiles.
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.
But, Canon did this for a reason - this CFA strategy gets the smallest possible penalty for shooting in fluorescent and stadium lights, something you actually do in quite a lot of sports, as a reporter, or indeed as a private user shooting your kids with the kit lens. And they get very even and flat skin tones, stuff like red rashes and so on are covered up automatically by the lowered hue resolution.
Medium format, Leica, Kodak - they all err on the "other side" of that fence. They're too well resolving in orange, and this isn't good either. The camera then gets extremely "sensitive" to the lighting conditions. Try shooting informal portraits with a P45+ in an office space with normal office lighting fixtures, without flash or natural light - the results are horrific. People look like they're in second stage corpse decomposition no matter what PP or camera profile you use.
Nikon does some stuff right, but they have an over-sensitivity to yellow, and that makes the balance very hard to nail down in people photography. But they're (IMO) the best landscape cameras, since no other camera can touch them for green hue resolution. They have extreme resolution in green-yellow, something you can see when shooting large landscapes - every bush and tree can be identified by the amount of chlorofyll-A vs Chlorofyll-B mix the plant uses. The camera can easily pick out extremely small hue differences between two very similar greens standing next to each other.
The best color "balance" for general usage that I've ever seen is actually the older Sony cameras (like the 850/900). They had their weaknesses, but they were better balanced than anything I've seen since - not counting the 15k€ specialist SML-filtered cameras we've used next to the hyperspectrals when doing art and craft history, but they're special cases.
p.3 #2 · Digital Medium Format / 35mm equiv difference
theSuede wrote:
The best color "balance" for general usage that I've ever seen is actually the older Sony cameras (like the 850/900). They had their weaknesses, but they were better balanced than anything I've seen since - not counting the 15k€ specialist SML-filtered cameras we've used next to the hyperspectrals when doing art and craft history, but they're special cases.
Where do you feel the newer 24mp FF Sony sensor falls short then? Just out of interest. I still use 850/900 full-time.
p.3 #3 · Digital Medium Format / 35mm equiv difference
Joakim, Very good post. I can always learn stuff here. And trying to understand what happen during PP my images from D700, 6D, and M9
I forget about color. But when I saw OP's link, I instantly think about M9.
Since I have used FF from three manufacture, I like the Canon least from color point of view. But I know 6D won't do justice for 'Canon color' as it might be poorest of all FF canon introduced. But I feel i am alone here as not many people complain about its color.
People always claim this is not a problem at digital age, just profile your camera, you are done. Which simply not the truth. Profile is use digital mixer to compensate RGB filer pack result, ending with significant gain at certain spectrum, with too much manipulate in PP, chrome noise will be one problem, and tonal range in color also suffer, at difficult lighting condition, this manipulate goes to extreme and IQ suffer even more.
I have very little complain about D700's color, but file is too polished/polite without bite, and I always heard good thing about Sony, but too bad there is no D900 MKII.
But I won't associate 'color' to MF to 35mm difference, as Leica M9 or DMR will give similar color output.
p.3 #5 · Digital Medium Format / 35mm equiv difference
theSuede wrote:
But, Canon did this for a reason - this CFA strategy gets the smallest possible penalty for shooting in fluorescent and stadium lights, something you actually do in quite a lot of sports, as a reporter, or indeed as a private user shooting your kids with the kit lens. And they get very even and flat skin tones, stuff like red rashes and so on are covered up automatically by the lowered hue resolution.
It's clear then: we need "films" for digital sensors, or electronically adjustable color filters.
p.3 #6 · Digital Medium Format / 35mm equiv difference
ricardovaste wrote:
Where do you feel the newer 24mp FF Sony sensor falls short then? Just out of interest. I still use 850/900 full-time.
They're a bit more flattened, less sharp - probably thinner to make the filter stack lower and increase sensor efficiency. The older FF Sony's were at the edge of being "too well resolving" in color, without stepping over the line and getting hyper-sensitive. Many cameras are in fact MORE sensitive than the eye for hue-changes, and this is of course also an error that we notice. Metameric errors are ALL deviations from human-sight, no matter if it's under-sensitivity or over-sensitivity to hue-change.
Do note that ALL color schemes are compromises, it's impossible to make a human-eye color sensitivity sensor if you want it to be usable over ISO400... The M- and L- cone receptors in the eye are so close together, similar, in "absolute color" that you need lots and lots of light to make out small chroma detail. Your eyes work together with you brain like a multi-shot noise removal where the brain stacks ten or twenty images sent from the eye on top of each other to get a sharp and clear "image" of reality. This isn't possible in a still-image camera (yet...)
What your consciousness "sees" is not what your eyes see. You "see" what your brain deduces from what your eyes see...
p.3 #7 · Digital Medium Format / 35mm equiv difference
snapsy wrote:
It would be constructive to see same-scene examples of how Camera A excels at color rendering vs Camera B. I may try if I find the time.
"Excel"? There are two main directions in "being good" at color. One is exactly human-eye-like (SML), and that's fairly objective. I can put an absolute metameric error on that, and call it a day.
The other is "pleasing photographic". The "subjective" rendering. They're not the same thing, and actually few persons will ever agree on what the "perfect" version of "pleasing photographic" means...
Well, once upon a time a few thousand moons ago we were all at the mercy of the film manufacturer's color calibration. Film is in no way "accurate" regarding color. The result of an activated grain didn't reflect the same color or spectral composition as the one that activated it...
Negative color film is even worse, because then you had two such operations in a row. First you take one color of light (spectral composition) into the camera system to let a grain activate. When activated the grain shows the INVERSE of that color, and not even the exact inverse, but something "about so". Then you take this already skewed color transformation - and do it all over again on to a photographic paper, a paper with conversion defects of its' own. The problem is the same when scanning a negative - who can say exactly what the scanner imperfections did to the film imperfections?
-And trying to work that system backwards is almost impossible. When you have two unknowns in an equation, the "correct" answer is a floating function of its' own. The correct answer can be anything, it's not even necessary for the result to be real - it CAN correctly include imaginary numbers :/
p.3 #8 · Digital Medium Format / 35mm equiv difference
theSuede wrote:
"Excel"? There are two main directions in "being good" at color. One is exactly human-eye-like (SML), and that's fairly objective. I can put an absolute metameric error on that, and call it a day.
The other is "pleasing photographic". The "subjective" rendering. They're not the same thing, and actually few persons will ever agree on what the "perfect" version of "pleasing photographic" means...
I was thinking one of the objective measures, like the hue resolution you mentioned.
p.3 #9 · Digital Medium Format / 35mm equiv difference
Suede: immense amount of information. Much appreciated!!!
This makes a lot of sense.
I was reading last night on the CFA and sensor designs but couldn't find comprehensive comparisons. There was nentioning of Nikon's approach to the data reproduction referred to as D-Lighting, which I still need to look up...
p.3 #10 · Digital Medium Format / 35mm equiv difference
Tonal transitions are mainly built from three components: Color sensitivity type, mid-low gray noise and lastly PRNU.
Looking at the electrical part of MF digital, they tend to excel at all three parts. They may not have the absolute DR of the latest Sony sensors, but they're extremely clean from say -9 to -10Ev and on up. -10Ev is the central EV part of "very deep shadow detail".
Their color sensitivity type is for all cases I've tried "over-sensitive". That is, the sensor is slightly more sensitive to hue shifts than the human eye. This also helps in getting good surface structures, as long as the camera profile is well implemented, and does not create luma differences (between hues) where there should be none.
PRNU, or fixed pattern errors in the readout, are also typically very low. At first this was because CCD based sensors had the edge on this parameter, and lately because of well implemented CDS fuctions (and very slow readouts!) on the the newer CMOS sensors they use today.
When you have all three together you get a good starting point for editing, lots and lots of flexibility. You get smooth, but very accurate medium-high frequency rendition. A lot of times TOO much high frequency contrast, hence the moire problems - though that has lessened with the later 60MP sensors. I still got very noticeable amounts of moire on the H4D-50 though, that I've shot in studio a few times.
..........
Then the lenses have a fairly easy time - sure they need to project an image height of almost 30mm (60mm sensor diagonal), but they're often used at F8-F11, and they "only" need to be "very good" at 40lp/mm where a lens used on the D800 or an APS camera needs to be "very good" at at least 60lp/mm to get critical pixel level detail.
Smaller formats need lenses that have higher contrast at higher lp/mm at lower aperture numbers to get the same resolution and contrast per image-width. That is decidedly NOT an easy task, though last years efforts by Sigma and Olympus has shown the way to some really high quality smaller format lenses.
This is a very good 120mm lens, shot at F4.0. To capture this image on µFT in stead (not impossible but very, very hard!) you'd need to shoot a 45mm lens at F1.5. Taht would give the same FoV and DoF. And actually, no matter how good the 45mm lens gets, I doubt it can do this in the extreme corners. The image is scaled to 16MP size, same as a 1:1 scale OM-D E-M5 raw file.
100% crops, at indicated positions.
I honestly don't think it's economically viable to manufacture a 45mmF1.5 lens this sharp wide open for µFT. Again note that this is wide open, at 100% 16MP scale. The Zeiss Distagon 55 might do it at F1.8 on APS, and a very good 85mm lens might do it at F2.5 on FF/FX format. But I'm not to sure, especially about the corner performance.
Sorry for the "borrowed" image, but I couldn't find any relevant images in my own library, need to shoot some wide open stuff on the HB too...
p.3 #11 · Digital Medium Format / 35mm equiv difference
The newer Sony 24Mp sensors do colour much better than anything bar base ISO in the a900/850 series, both the a900 and RX1. Side by side I occasionally would be surprised by the older sensor's performance but only occasionally, and it can deliver a very organic look at lowish ISO - but it cannot match the a99 due to the higher floor. And the deterioration above ISO 400 was a serious problem, and shadows needed careful watching. Balancing WB was also harder.
Files are much easier to work on the a99/RX1 - in another league actually. Jpegs far superior, and noise pattern is superb and very different from the colour blotch land seen in the a900. No one agonises over what ISO for best results with the modern sensors.
All in all, performance is far superior and a huge step forward (a99 vs a900)- low ISO, high ISO, 1.6 stops more DR, all of which assist colour performance and all round photographic use. One is a low ISO trad DSLR, the other a modern weapon. And that is with the SLT in the way.
p.3 #13 · Digital Medium Format / 35mm equiv difference
From a guy who shoot MF backs for Hollywood promos:
'Nine times out of ten, because of the size of billboards and everything, we shoot a lot on the Hasselblad and the Phase One backs. We had to step back a little on the Phase One backs because the higher resolution ones are so high-resolution that they almost have to retouch the sharpness out of them. I couldn’t tell you which one we’re using, but we’re not using the highest-end ones because people complained that they were too clear.
Film was never this sharp. It’s sharper than real life. You shouldn’t be able to read a hair inside the tear duct of someone’s eye. On one of those high-end backs, though, you can almost read what someone is thinking. It’s kind of terrifying.
I think they need to work less on making backs higher resolution, and more on making them able to shoot in darker areas. Being able to shoot with a medium format camera at ISO 2000 would be really nice. You have to know you can’t have too many dark areas.'
p.3 #14 · Digital Medium Format / 35mm equiv difference
But what has changed hugely with digital Mike is that you can get close to MF in 35mm FF and soon the difference may be vanishingly small - and have significant advantages in outlay, ISO potential, size/bulk, etc. You can go too far in one direction (like res'n), when you need a balance of qualities. See above post.
With film, HCB was never going to use 4x5 and Adams would not want a Leica M. The 24x36 'film sensor' was the exact same as loaded into the 4x5 so real estate was the main variable, but very different today in sensor tech. Less can indeed be more, and this can enable small cameras to do much much more. It's just the start.
p.3 #15 · Digital Medium Format / 35mm equiv difference
I am sceptical on 35mm being side by side with MF.
Perhaps I am wrong, but it seems physically impossible to reproduce glass for equal performance as Suede illustrated above. Get very close - yes, but "very close" is ever so subjective of a concept.
Aside from actual possibilities there are also financial and practical reasons.
p.3 #16 · Digital Medium Format / 35mm equiv difference
^Well, it seems like you'd need something along the lines of the Distagon 55/1.4 to get 'close', at least as far as lens resolution is concerned. Then there are still the differences between MF backs and small format cameras that seem to be bigger because they are optimized for different types of shooting/output.
p.3 #17 · Digital Medium Format / 35mm equiv difference
The development of the thread hasn't been too bad :-)
Especially (but not exclusively) welcome to the insights from theSuede on colors.
Here is a small breakdown of differences commonly associated with format differences between 35 mm and medium format image output. The purpose is to get a sense of what is necessarily related to the format, and what is implementation specific.
With low light effectiveness, I mean for a given perspective, FoV, DoF and shutter speed.
And I included the myths about perspective and diffraction (diffraction in the image scale for the same picture / same DoF), that hopefully are sufficiently debunked at least in here.
Not meant to be definite, just a help to structure.
p.3 #18 · Digital Medium Format / 35mm equiv difference
Interesting. Taking D800E into account - it seem to have higher dynamic range (according to DxO of course) than most current digital backs. Exception to the rule perhaps? Or just sensor data interpolation? As far as I remember Nikon read sensor data in 2 passes and averaged accordingly?
p.3 #19 · Digital Medium Format / 35mm equiv difference
Krosavcheg wrote:
Interesting. Taking D800E into account - it seem to have higher dynamic range (according to DxO of course) than most current digital backs. Exception to the rule perhaps? Or just sensor data interpolation? As far as I remember Nikon read sensor data in 2 passes and averaged accordingly?
In short, DXO measures 'engineering' dynamic range, where the lower limit is a signal to noise ratio of 1, or 0 dB. In that case the D800(E) dynamic range is superior to just about anything. That is because of the exceptional low read noise at low ISO with Exmor
For practical photography, we often define the lower limit with a higher quality requirement, say a signal to noise ratio of 20 dB, that looks good. In that case, the lower end is dominated by photon shot noise, the amount of light collected, and then the digital backs have more DR than the D800(E).
p.3 #20 · Digital Medium Format / 35mm equiv difference
'I am sceptical on 35mm being side by side with MF.'
Assuming you refer to digital, the pro cited in the pop photo article above thinks MF is almost 'too good' for his work...so clearly there is a 'threshold of acceptable quality' at play here, and as MF is at or above that threshold for much work currently done on MF, it makes good sense that higher Mp FFs will approach the threshold also and become acceptable in that market and to those buyers (who can afford more if need be).
And Zeiss will not be alone in the arms race of lens and sensor development.
The DR and ISO issues are prime examples of why many will like what high end FF brings to the table, you would think? From reading, DxO's sensor based data is considered to be high quality. Maybe a better way to put it, is that it accords with personal perceptions - of mine at least. My a900 lost a lot of images to wide scene DR that the a99 captures with tone scale to spare, for example.
