Originally posted by Class A That's because the light meter only cares for candela / sq.mm. It doesn't care for the total amount of light required.
But this is *exactly* the value that represents exposure to both film and digital sensors.
Quote: Note that a lens on the larger sensor camera needs to be able to let the larger amount of sunlight through. That's why it has to be faster.
You cannot simply use the same aperture diameter and blow up the image circle. By doing so you would reduce the light flux (same amount of photons spread over a larger area). Hence, the aperture diameter needs to be larger.
No, actually, you can. It's done in large format lenses all the time. There are 120mm lenses that cover 2x3", ones that cover 4x5", and ones that cover 8x10"; most are f5.6 or so. As you know, a 120mm lens @f5.6 is going to be ~22mm, regardless of the coverage of the image circle. You don't have to make the aperture diameter larger, you have to design the lens for a larger image circle. It then accepts light from a greater angle of incidence - allowing more total light through the SAME opening. Same thing applies to FL on small cameras. A 28mm lens produces exactly the same total light on the sensor as a 100mm lens, because it accepts a greater angle of incidence through the opening.
Quote: AFAIC, the problem lies with the definition of exposure. It assumes that one wants the same candela / sq.mm amount for all formats. If you scale various images from different source sizes to the same output size, however, you not only get the effect that larger sources require less magnification and hence yield sharper images, but they also represent more total light being collected. Remember, they received the same exposure, i.e., same amount of light per unit area, but have a larger area, hence collected more total light. This means that images from the smaller sources will not only be less sharp but also noisier.
See, when I read this previous paragraph, I think we're very close together; the only objection I'd have is that the reason exposure is defined that way is that, for a given EV, you *do* want the same candela/sq.mm regardless of format; otherwise, metering wouldn't work. Other than that, I'm with you.
Quote: The reason for this is not that smaller sensors/film formats are noiser, it is because we assumed that "same exposure" provides a level playground. Well, it doesn't. In order to produce an image with the same SNR, a smaller sensor needs a higher exposure. That's why using the same f-ratio isn't enough, it needs to be smaller (which also brings the DOF into line).
Requiring "higher exposure" sounds as if more light is needed for the smaller sensor, but that's not the case. We just need to supply it with the same total amount of light. Only the, unit area based, definition of exposure causes this fact to mean that a higher exposure is needed.
Ok. I see what you're saying here, but it's important to note that these things are only true when we're describing
"equivalency", *not* when discussing standard photography. ISO100 is ISO100 regardless of the format, and has the same SNR at the source (the sensor); since we have to magnify it more to get the same size print, the
output has a lower SNR. f1.9 is f1.9 regardless of the focal length and sensor size in terms of
actual exposure at the film plane, because sensors and films WORK on the candela/sq.mm principles described above. It's not the definition of exposure that's at fault; it's the concept of equivalency, which is an analogy, supposedly to make it easier to understand the trade-offs as we change formats.
Any time we increase magnification of the source, we reduce output quality in measurable ways; smaller sensors must, then, have lower quality over the whole image area (with an ideal reproduction medium and all other things being equal).
But the thing I wanted to get at is that there's no need to re-write photography books to tell people they are incorrectly exposing their film because of the equivalency numbers you've spoken of. Light flux, or candela/sq.mm, or other measure of light power per area is the way sensors and film see light, and f-stops are the way lenses deal with light, and everyone is happy again. I can safely use the same settings for aperture, shutter speed, and ISO, across every film and digital body with impunity, and only be concerned about equivalency when I feel the need for a cheerful argument.
Some of the other stuff we've argued about is related to context; for instance, the discussion about pixel size and noise. Your statements were true, but only if you make the same size print regardless of MP; if I take a FF sensor of 6MP and one of 24MP, and I make an 8x10 from the 6MP (300 dpi) and make a 16x20 from the 24MP (300dpi), the larger print will show more noise; we can say it's because of the smaller pixels, or we can say it's because of the greater magnification, and *both* are true. But if we make both prints 8x10, we'll see exactly the same noise profile, as you said.
On the definition of "fast", I'm afraid we'll have to agree to disagree. It seems nonsensical to me to suggest that a lens is called 'fast' because it gives low DOF rather than because it allows fast shutter speeds.