Gene51 wrote:
I just re-read this in light of not wanting to be included in my broad generalization.
Well, when you say that "even with fast lenses requires high ISO" seems to imply that you are not at all familar with ISO invariance....
Gene, my statement does not imply that at all. My point was that with an ISO invariant Camera (which encompasses probably the majority of DSLRs in use today), in a low light environment, once you have used the fastest lens available (wide open), and you need a given shutter speed to freeze the action, then your only options are increasing the ISO or raising the brightness in post with the attendant increase in noise. Please give me credit for not making uninformed statements. Not only have I reread the article from which you excerpted the above illustration, I have read everything I can find on the subject, so please don’t presume I am uninformed. In fact, I think I may have a better grasp on the subject than many of its proponents here. Let me start with a quote directly from the article that you drew the illustration from:
“ISO invariance is not a measure of how good a particular camera or sensor is. It simply means that a lot of the exposure is generated with the image processor rather than in the sensor. It's simply a different approach to low light photography.
So while I do think ISO-less shooting is an exciting new technique for low light photographers (doesn't make a lick of difference for daytime photography), it's certainly not worth picking a camera simply based on its ISO-invariance.”
Now some comments and then a challenge:
First, there is nothing magic about ISO invariance. Effective ISO can be changed in-Camera by either amplifying the sensor output prior to the A/D, multiplying the digital output of the A/D by a constant or (more commonly) a combination of the two. It’s been suggested that the A/D reference voltage could also been changed, but to my knowledge, no one has demonstrated that is actually used by camera manufacturers. It should be noted that all the cameras I’m familiar with (including Nikon’s and Canons) use the multiplication by a constant in SW beyound ISO 1000 or so regardless of the method at lower ISOs. Given that, whether the multiplication is done in-Camera or in post, the result is the same - the noise is multiplied along with the signal. There is no difference as to whether this is done by an amplifier or multiplying the output of the A/D - the noise from the sensor and that inherent in the A/D due to digitizing error are multiplied the same amount. If the argument is that the amplifier adds noise, remember that we’re discusing a signal that has a dynamic range of about 12-13 bits (<80 dB) and low level amplifiers with a noise floor below 14bits are common - every $50 CD player has one. While there is an advantage (as pointed out in the referenced article) to doing the operation in post in that the operation can be selectively applied to portions of the image, the result is the same. The point being that the idea that the DR is magically improved by doing the operation in post doesn’t hold up. I think this misunderstanding derives from the well-known fact that DR decreases with increasing ISO as measured prior to post. What it ignores is that those curves never show what happens after post if you choose to underexpose and raise in post. Since the multiplication of the data is the same regardless where it’s done, there is no magic change in the DR. The noise and the signal are multiplied in the same way - just accomplished in a different place. If you read the above-referenced article, where both methods are compared, there is no clear-cut advantage in underexposing and raising the brightness later outside of the selective enhancement in post I mentioned.
Now for the challange. While I have seen your photos that were almost black raised 5 stops in post to a usable image, i’m More interested in some more typical exposures. How about you or Ron (or someone else) post the following: using an “ISO invariant Camera”, take a low light action shot (such as indoor sports) both at ISO 5000 (or whatever is the “correct” exposure) and 5 stops underexposed. Same shutter speed, same lens, same f stop - just a change in ISO. Bring the underexposed shot up in brightness in post to match the brightness of the “correctly exposed” shot and post the results. No noise reduction, no curves, nothing but a change to the gamma to make the brightness the same. I’ll look forward to seeing the results.