CHG_CANON wrote:
Without searching for the video, let's 'pretend' he was talking about the reality that the digital sensor captures at just 1 (one) ISO setting, the 'base' ISO for that model. This 'base' ISO is typically ISO-100, but some cameras it is ISO-200, or somewhere in this range. When you dial any other ISO value, the 'computer' that is the processor inside the digital camera will process the data from the digital sensor to achieve the specified ISO. And yes, those international standards are how each camera manufacturer determines the amount of 'gain' to add to the sensor data to achieve the specified and standardized ISO setting. The signal from the sensor is 'amplified' and it is this applied amplification to the data that is the source of most digital noise in the resulting image file. This is also why the best noise performance and widest Dynamic Range of the camera occurs at each digital camera's 'base ISO'.
Without searching for the video, let's 'pretend' h... (
show quote)
This is correct.
Digital cameras actually only have a single "base" ISO and everything else is derived from that. The base ISO in any digital is the lowest setting within their "normal" or "native" range of ISOs... the lowest setting that can be made without "expanding ISO". In most cases today, it's ISO 100.... A few cameras use ISO 64. In the past a few have used ISO 200 and, I dunno, maybe some still do.
It's important to note that most digital cameras have two types of amplification. It's an oversimplification, but call them "hardware" and "software" amplification.
Hardware amplification is the "native" range of the camera (the "un-expanded" range). The camera has circuitry with components that amplify the image, for any ISO other than the "base".
Software amplification is the "expanded" ISO found on many (most?) cameras... usually these are ultra high ISOs (25600, 32000.... even 51200, 102400 or higher... much, much higher in some cases). Some camera also have a lower "expanded" ISO such as 50 or 32.
Images that are software amplified are actually shot at a lower ISO (probably the highest possible within the native range)... essentially under-exposed... and then "tweaked" by the camera's software to appear as if they were shot at the higher "expanded" ISO that was selected by the user. In a sense, with the "expanded" ISOs you are further amplifying an image that was already hardware amplified! This is why the image quality tends to drop off dramatically with the expanded ISOs.
Another way of looking at this.... "expanded" ISOs are much like you shot the image at the highest possible "native range" ISO, and then in post-processing you used the exposure adjustments to brighten the image. This always greatly increases the appearance of noise in images.
Personally, my rule of thumb is to keep to the native range of any camera... To NOT use the expanded ISOs. About the only exception I make is if I plan to do a black & white conversion of the image. Image noise generally isn't a problem in B&W (or any other type of monochrome conversion).
Everyone talks about the increase of "noise" appearing in images at higher ISOs. But what many people overlook is that higher ISOs also cost you dynamic range and resolution. Sometime do a series of test shots with your camera at various ISOs. If you do so carefully using the exact same subject, only changing the ISO from image to image and offsetting that by changes to the shutter speed, you'll find that the higher the ISO, the smaller the file size. If you look at the histogram of the images, you'll also notice the higher the ISO, the narrower the range of tonalities that graph will display. Because of this, it's not uncommon for higher ISO images to need their black point and white points adjusted, for contrast to be added. But it also represents a drop in resolution (which can be made worse by applying too much noise reduction, which removes fine detail from images).
On the cameras that offer it, a lower expanded ISO is also software derived and will have some image quality loss. I'd really like a digital camera that had ultra low ISOs available... ISO 25, 16, 12, 8 and maybe even lower would be useful at times. However, I have never seen this done digitally and have been told there's just too much loss of IQ to make it worthwhile (a higher "base" ISO along with a Neutral Density filter can do the same job better).
So, yes, in a sense all ISOs are "fake". They're all derived, one way or another, from a single ISO. Over the last 20, 25 years, digital cameras have gotten better and better at this... They now offer usable ISOs that are far higher than we ever dreamed would be possible when we were shooting film.
But it's important to know HOW the higher ISOs are being derived.
Best thing you can do is make a series of test shots with your camera, to learn how it handles various ISOs. Try both RAW (conversion and noise reduction in post-processing) and JPEG (in-camera conversion and NR)... Also try different types and methods of noise reduction. Learn what works and what doesn't.
In my opinion, the biggest mistake people make is evaluating their images for things like noise at waaaaaayyy too high magnification. Viewing an image from a 24MP camera "at 100%" on the typical computer monitor is much like making a 3.5 FOOT by 5 FOOT print (approx. 1 meter by 1.5 meter, for our metric friends), and then viewing it from only 18 or 20" away. OF COURSE it looks like crap! Are you planning to print it that large? And, even if you do, will you ever view it from so close?
Everyone does it. Looks at their images "at 100%". It's even common on photo forums for "100% crops" to be used in discussions.
BACK OFF... while it can be handy to highly magnify an image for retouching work, it makes little sense to evaluate many of their aspects so highly magnified. Zoom to a lower magnification that's closer to the size the image will actually be used... whatever that may be. With a 24MP camera, even 50% is like a very large 20x30" print. Even 25% is similar to a 16x20" print. (All this assumes no cropping and that the user's monitor is set to it's native resolution around 100 pixels per inch). Very often, by the time an image has sized for the "real world" purpose, any noise (or even minor focus error, etc.) will no longer be noticeable. So the only person to ever see the noise (or whatever) is the photographer them self, while they're working with the original image file.