Apologies to slej, I didn't click on his link (
http://www.luminous-landscape.com/tutori...ight.shtml) and read that article.
Of course I forgot that digital sensors are linear, yet the way the human eye perceives light is exponential (each stop is double/half the light of the previous/next one) - so of course digital cameras *do* capture more detail in the lighter areas than in the darker areas, as the highest stop represents 1/2 the entire sensor's dynamic range, the 2nd highest stop represents 1/4, and so on - the low stops are much more susceptible to posterisation and noise.
So in this sense, Slej is absolutely correct in suggesting that the signal-to-noise ratio tends to be much lower in the darker areas (simply because of the way the signal gets converted from a 12-bit linear representation into an 8-bit exponential representation) - so as far as reducing noise goes, you are better off to over-expose and then tone it down in post-processing - and this result can even produce less noise than a "properly" exposed image with no post-processing.
But the original problem at the start of this thread was with the highlights.
With film, as the light reaches the extreme highlights, the film gradually becomes less sensitive to this change in brightness. The result is that detail remains in the highlights, just with a reduction in contrast.
Digital sensors however have no such "shoulder", and simply represent all levels linearly up to their maximum, at which point they clip the signal abruptly.
the result is increased contrast in the highlights right up to the limit of the sensor, and then they are clipped and suddenly become totally blown out and lose *all* detail.
So blowing out highlights is a big no-no for digital, as once a signal is clipped at the sensor, the details cannot be salvaged.
But there is some help at hand, as it appears that most RAW images actually contain more dynamic range than is represented in the JPG output, so often you can extract up to 1/2 stop more detail out of the highlights if you work with the RAW image instead of the JPG. Half a stop is better than nothing, but certainly not enough to make you throw caution to the wind.
So the "Epose to the Right" lesson in that luminous landscape tutorial that Slej pointed to really would give you better results - in theory anyway.
But... is it worth all that effort for each photo to save a bit of noise? If you could automate the workflow - perhaps re-flash the camera with a reprogrammed exposure algorhythym and script something on the PC to expose the pics properly later - then maybe it is? Sounds like a lot of hard work though...
I love working in Photoshop and can see the benefit in working with good RAW software, but I also love it when good photos come out of the camera without needing anything done to them.
My Olympus C750UZ is pretty ordinary when it comes to noisey low-light and high-ISO photos... I keep it stuck on 50 ISO and switch to 100 when I need to. 200 is for emergencies only and if I need 400 then I'll just walk away shaking my head. My tripod and remote are my best friends in low-light. Even if it had a RAW mode, it would still only improve things slightly.
There is another much simpler approach to solving this problem if you, like me are suffering from noisy pics - buy a better camera!
hahaha.. see, it really does come back to "blame the gear"!
Well, actually it comes down to "take advantage of the technology". Something like a 350D can take photos at 1600 ISO with less noise than my C750 does at 50 ISO in low light...
Things have come a loooong way in the last couple of years, and while you might think its unfair to compare a 350D to an Olympus P&S, the fact is that my C750 retailed at AU$1200 when I bought it, and I just ordered a 350D for AU$1470. Both cameras fill a similar price-bracket, yet they are worlds apart in most other aspects.
...if only I could buy new skills as easily (and cheaply)
![Wink Wink](https://www.shuttertalk.com/forums/images/smilies/wink.gif)
I think I need to stop talking and go out and actually take some photos now.
Cheers
Adrian