tesch: I don't come to this website very often but when I do I find all of the talk is about why Sony isn't a real camera. Very strange!
This is a good article about a lot of things that really have no effect on the images taken by people on this site. If the photographers on this site spent as much time researching composition and color theory they wouldn't have to worry about this nonsense. But that would mean they would have to think which seems to be an issue..........
Get over it!
@HowaboutRAW A better comparison would have been with a composer thinking about music theory when he's composing. Which I imagine he does.
But the issue isn't as moot as tesch seems to think. That "moon and Space Needle" shot doesn't seem like an especially tricky exposure, but it's way too artifacty (which is a word, I've decided) to rank as stock-ready.
Frank_BR: Why doesn't Sony have so far the option of a lossless RAW format? Most critics say that this is just a Sony error, but the question is not so simple. To try to understand better, I did an experiment. I took the RAW files from DPR Studio Scene for Sony 7RII and Nikon D810, and compressed them with WinRAR. The results were as follows:
Camera.....RAW original....WinRAR compr7RII .........41.4MB...........38.3MBD810 ........74.3MB..........43.9MB
WinRAR is a lossless compression, so the RARs files contain the same information as the RAWs from camera.
Surprisingly, WinRAR managed to reduce to almost half the size of NEF file, but failed to appreciably reduce the size of Sony RAW. This shows that NEF is inefficient since it produces much larger files than necessary. There is room for a better lossless RAW coding. I draw the conclusion that a reason for Sony has been reluctant to use a lossless RAW coding is that the current processor technology does not allow high efficiency of coding.
I think all Frank is saying is that NEF is quite inefficient if a general-purpose lossless compression algorithm can reduce the size by a third. Since Nikon has a much better knowledge of the data stream than RAR does, it should be able to do even better than RAR. I assume the redundancy comes from keeping data aligned on byte or word boundaries for faster processing (e.g. they might store a 12-bit sample in two bytes, wasting 4 bits per sample, rather than storing two samples in three bytes without waste.
I agree, though, that the main question is why does Sony think it's acceptable to apply lossy compression to their RAW format, especially when it can lead to such ugly artifacts.
Peter Cockerell: I'm a bit confused by use of 5/6 as the lens equivalence ratio between the IQ250 sensor and a 35mm one. Of course, it's tricky to come up with a single number when the aspect ratios of the sensors are different, but in this case the relevant numbers are: horizontal: 36/44 = 0.82 (or 9/11 if you prefer fractions), vertical: 24/33 = 0.73 (or 8/11), diagonal: 43.3/55 = 0.79 (or about 393/500, or 8.7/11). Concentrating on the diagonal, since that's what you mention in the article, the fraction is much closer to 4/5 than 5/6, and the focal length equivalents would be: 120mm becomes 94mm, 80mm becomes 63mm, and 45mm becomes 35mm.
True enough. I guess it all comes down how important those incremental differences are to the person taking the photos. I had to make a similar decision last year when I moved from a 5D3 to a GH4, because I was just getting fed up with lugging a around a full-frame and all its associated lenses. For my (purely personal) photography. I'd say for 80% of the photos I take (which are mostly for web display), the loss of resolution and noise performance hasn't made any difference to the end results, but for the other 20% I sorely miss the benefits of the full-frame. Which is why I'm seriously considering buting an A7R II to use with the lenses I couldn't bear to part with (Canon EF 70-200 L II and a couple of Sigma primes).
Thanks for the confirmation, JNR and Androole. Another thing that occurs to me is how close the ratio is to 1. How much of a real-world advantage would a small "medium format" back like this have over, say, an EOS 5DS or a Sony A7R II? Especially when you consider all the other compromises inherent in this model. (Though it's true that the sensor size ratio is more significant if you constrain your 35mm camera to a 4:3 aspect ratio, I guess.)
I'm a bit confused by use of 5/6 as the lens equivalence ratio between the IQ250 sensor and a 35mm one. Of course, it's tricky to come up with a single number when the aspect ratios of the sensors are different, but in this case the relevant numbers are: horizontal: 36/44 = 0.82 (or 9/11 if you prefer fractions), vertical: 24/33 = 0.73 (or 8/11), diagonal: 43.3/55 = 0.79 (or about 393/500, or 8.7/11). Concentrating on the diagonal, since that's what you mention in the article, the fraction is much closer to 4/5 than 5/6, and the focal length equivalents would be: 120mm becomes 94mm, 80mm becomes 63mm, and 45mm becomes 35mm.
Robert Garcia NYC: I would've asked him about the noise reduction can you please turn it completely off. This is what makes the Fuji files looks smudgy.
OK, it's not obvious from either my 5D3 or GH4, and indeed the user guides imply the opposite (except for the dark frame subtraction during long exposure NR, but that's a very different kind of NR). I'm actually having trouble imagining how you'd write an NR algorithm that works on a Bayered (or equivalent) image, but it's been a while since I wrote any image processing code, so I can't claim current knowing knowledge of the field.
Not sure I understand that. If Robert Garcia is asking for zero noise reduction, presumably he's referring to in-camera JPG conversion, since AFAIK manufacturers don't apply NR to RAW files (they wouldn't be very raw, otherwise). In this case, the quality of LR's RAW conversion is irrelevant.
LaFonte: But why do I even have to care about equivalent or not exposure? If I say to two guys don't use your camera metering, pull out my old light meter and tell to two guys with very different camera: set your ISO 100, set 1/60, and aperture 2.2 and you will be fine, they would both get properly exposed picture. Right? Even that one geezer have 7d and the other have e-pm.As I understand, that is the whole point of having equivalent exposure that translates to everybody. So we understand each other without looking what size of sensor you have. Starting recalculating what aperture means in different sensor sizes is good only and only for assuming DOF not for exposure.So maybe call it equivalent DOF.Or is it that I totally don't get it?
@quezra" larger sensors have less signal-to-noise ratio than smaller sensors"
You might want to think about what the expression "signal-to-noise ratio" means and then edit your comment accordingly.
I think what's missing from the ISO discussion is how the image is going to be used. If you're shooting primarily for the web, the size reduction process is going average out so much of the noise that except for at really high ISOs, any FF advantage over a smaller sensor is going to be imperceptible. The same applies to a lesser extent with, say, 4" x 6" 240 DPI prints.
Although it touched on it, the article could have emphasized more that, unlike, say, f-stop equivalency, which is purely mathematical, ISO equivalency is very much affected by physical properties, sensor fabrication technologies etc. and is distinctly nonlinear across the ISO range. After all, if we had "ideal" sensors that generated 0 noise across the whole ISO range, the concept of ISO equivalency wouldn't even exist. There's no similar "ideal" scenario for f-number equivalency.
nigelht: Total light is meaningless to noise. Given the same FOV then the total light will be the same. Given different FOVs with the same intensities (as in the example provided) then for the same sensel size then the same number of photons hit so shot noise is identical since the photon shot noise is the square root of the number of photons on a per sensel basis.
Read noise is more dependent on the sensor design and generation but there is a minor advantage for larger sensor since the base SNR for FF sensors is typically slightly better than m43 or 1" of the same generation. Mostly though the noise is dominated by shot noise.
All you've shown on your example charts is the effects of larger sensel size.
The sensel (pixel) sizes for the cameras listed are:
EOS 1-DX : 41.45 um^2X-A1: 22.66 um^2GH4: 13.99 um^2V3: 6.3 um^2
The area of each EOS 1-DX sensel is 658% larger than a sensel on the Nikon 1 V3.
The 1-DX sensel area is 111% larger than on the X-A1.
"When you crop a 36MP FF sensor to APS-C proportions, you get 16 MP. Are you saying that according to your sensel size only theory, that 16 MP APS-C sensors should be giving equivalent results to FF?"
Well, I'm not, necessarily, but DP Review seems to be:
Of course, if you *scale* a 36Mp image down to 16Mp, its noise figure will improve relative to a 16Mp sensor showing the same scene, but that's not what you seem to be saying.
HoustonPowers: I have a small bug in Lightroom 4. When I mouse over the histogram in the Develop Module, the pop up tips are E-mail labels, such as "Mark as read", "Mark as spam" "delete", "reply"..etc..Anyone else have this happening to them?Just curios if its just my system. Its a peculiar bug.
That IS bizarre, then! It's hard to imagine how those strings even appear in LR (or any Adobe program, come to that). I'd be interested to know if those tool tips are the same as the ones in whatever mail program you use...
Those tool tips most likely come from the app *under* LR. It might still be a bug in LR, but could also be a problem with Windows, or even the application underneath.