After my struggles with ISO and noise this week, I decided to try an experiment... camera set on tripod, in AV mode using the speedlight... the only thing i changed from one shot to the next was the ISO...
Not sure how much you can tell from this collage... prolly looking at it large, on black, might give you a better picture, as it were...
One thing i found kind of odd was that the higher the ISO, the larger the file size... any idea why that might be the case?
Thank you so much for the comments on yesterday's bench shot... yes, the same bench as was featured on November 1... everybody needs a bench series, right?
Love your.....my kids think I'm nuts tag!! Mine always seem to catch me when I'm trying something new.....that's the only problem with this site....I'm taking pics of things I never would have thought to before! I love it though!
Both JPEG and RAW files are compressed file formats (RAW files are commonly referred to as being uncompressed, but that's not true -- they're lossless, which is a different thing).
File compression is a very complex subject, but taking it very simply, an uncompressed image file would contain information about the colour of every different pixel in the image. When you compress the file, algorithms try to figure out ways to store the image data in a smaller space.
If you take the simplest case, a totally black image, an uncompressed file format would say that the pixel in the very top left was black, the next pixel across is black, the next pixel across is black, and so on, for every single pixel. A file like this can be very efficiently compressed -- you can simply store information that says that the file is x pixels wide, y pixels tall, and all the pixels are black.
That's a very simplistic case, but that's what compression is trying to achieve -- it's looking for patterns in the image that it can represent mathematically in less space than it takes to explicitly say what colour each and every pixel is.
Even when you take a very detailed image, it's possible to do this surprisingly efficiently. However, when you introduce a random element, as camera noise is, the ability for the algorithms to work degrades significantly. Instead of areas or patterns which are nicely compressible, such as smooth areas of sky, skin tone, etc., you have areas where every pixel is a slightly different colour. The more noise there is, the harder it is to compress, to the point where an image that consisted entirely of 'white noise' ends up being impossible to compress at all.
Hopefully this explains why your higher ISO images are larger than a corresponding lower ISO image of the same subject. The good news is that the amount of compression loss is surprisingly predictable if the characteristics of the camera are known. This is why, when you increase the ISO, the camera will automatically reduce the number of shots it shows it can store on your card -- it knows that the higher-ISO images will take more space.
You can also see the way compression alters depending on your subject -- if you take a nice low-ISO shot with your lens cap on, so that you get a totally black image, you'll find the file is tiny. Take a photo at low ISO of a very complex subject, such as branches and leaves of a tree, and you'll find the file is larger than average.
@abirkill wow! thank you Alexis... you've explained it perfectly!!! (and yeah, amazingly enough, a photo of the inside of the lens cap results in a very small file ;p)
Interesting experiment.
I tried something similar a while back to prove to myself I wasn't going mad & was getting vertical banding. One of the things I found was there are real ISO sweet spots for noise (best seems to be 160 on my 7D).
The comparison pics are here if you're interested. ;) http://humphreyhippo.wordpress.com/2012/06/29/canon-7d-banding-noise/
@humphreyhippo tx! good experiment to have done... one thing from this is that i am now a little less afraid to push the iso... for the longest time i was slavishly sticking to iso100 and missing all sorts of shots because i couldn't get the shutter speed (and light)... i've found now for some things i can go to 3200 and still get marginally good quality if i don't plan to crop (much)...
You read my mind! I've been meaning to do this forever and keep not getting around to it. I am always afraid of upping my ISO but I have no good sense of how bad the situation can get.
@victorypuzzle yeah... i was quite impressed... altho' i'm not sure this was the best way to test it... i expect it would have been better to have tired to zoom in on a part of the image... but i wasn't sure how to do that and keep them all the same...
I agree with Lynn; ISO 1600 seems the noisiest of the shots.
One point Alexis passed by, but that you can sort of read between the lines, is that the higher the ISO is set, the more light information the camera is trying to read/record into each pixel, which leads to larger file sizes. Lower ISO gives you seemingly smoother pictures, but leaves you with the risk of loosing details. Too high an ISO setting, and you get noise. Good idea of yours, to experiment like this. ;-)
Northy, thank you for sharing this.. I learned so much just fromyour shots and Alexis explanation :)
So interested as to me ther eis very little differene between the shots.. So like you, I will be braver to use the iso more now.
glad i came here. i learned a $75 lesson on ISO. for free. and i actually understood all the explanations. now, hopefully that gets retained in the old coconut. thanks for sharing!
File compression is a very complex subject, but taking it very simply, an uncompressed image file would contain information about the colour of every different pixel in the image. When you compress the file, algorithms try to figure out ways to store the image data in a smaller space.
If you take the simplest case, a totally black image, an uncompressed file format would say that the pixel in the very top left was black, the next pixel across is black, the next pixel across is black, and so on, for every single pixel. A file like this can be very efficiently compressed -- you can simply store information that says that the file is x pixels wide, y pixels tall, and all the pixels are black.
That's a very simplistic case, but that's what compression is trying to achieve -- it's looking for patterns in the image that it can represent mathematically in less space than it takes to explicitly say what colour each and every pixel is.
Even when you take a very detailed image, it's possible to do this surprisingly efficiently. However, when you introduce a random element, as camera noise is, the ability for the algorithms to work degrades significantly. Instead of areas or patterns which are nicely compressible, such as smooth areas of sky, skin tone, etc., you have areas where every pixel is a slightly different colour. The more noise there is, the harder it is to compress, to the point where an image that consisted entirely of 'white noise' ends up being impossible to compress at all.
Hopefully this explains why your higher ISO images are larger than a corresponding lower ISO image of the same subject. The good news is that the amount of compression loss is surprisingly predictable if the characteristics of the camera are known. This is why, when you increase the ISO, the camera will automatically reduce the number of shots it shows it can store on your card -- it knows that the higher-ISO images will take more space.
You can also see the way compression alters depending on your subject -- if you take a nice low-ISO shot with your lens cap on, so that you get a totally black image, you'll find the file is tiny. Take a photo at low ISO of a very complex subject, such as branches and leaves of a tree, and you'll find the file is larger than average.
Let me know if you have any questions!
@northy Nice idea!
@northy Nice experiment.
I tried something similar a while back to prove to myself I wasn't going mad & was getting vertical banding. One of the things I found was there are real ISO sweet spots for noise (best seems to be 160 on my 7D).
The comparison pics are here if you're interested. ;) http://humphreyhippo.wordpress.com/2012/06/29/canon-7d-banding-noise/
Also, you have a very cool peeler!
I agree with Lynn; ISO 1600 seems the noisiest of the shots.
One point Alexis passed by, but that you can sort of read between the lines, is that the higher the ISO is set, the more light information the camera is trying to read/record into each pixel, which leads to larger file sizes. Lower ISO gives you seemingly smoother pictures, but leaves you with the risk of loosing details. Too high an ISO setting, and you get noise. Good idea of yours, to experiment like this. ;-)
Very good explanation, by the way, Alexis!
Northy, thank you for sharing this.. I learned so much just fromyour shots and Alexis explanation :)
So interested as to me ther eis very little differene between the shots.. So like you, I will be braver to use the iso more now.