Hi guys,
I'm doing some research for one article I'm writing. I'm curious to find out what was for you the thing that you learned that made you say "From now on I'm not just playing with my DSLR".
For me it was finding about the ISO, Aperture and Shutter Speed triangle.
For me it was finding out that the typical Aperture/Shutter/ISO triangle is BS.
That triangle isn't about exposure at all, but just image brightness. It is based on a very widespread misunderstanding of how digital cameras actually work. It leaves out scene luminance as an important parameter of exposure. Increasing ISO level alone does not increase noisiness - in fact, on some cameras it can reduce noisiness. Increasing ISO does not increase the sensitivity of the sensor. The noisiness of an image is determined primarily by exposure. Exposure is light per unit area on the sensor, not how bright or dark the image is.
All of this leads to further understanding of such things as ETTR, how setting ISO before aperture and shutter is usually sub-optimal, when it makes sense to deliberately underexpose, when to use Auto-ISO, and when increasing ISO does and doesn't help.
To be honest, I find your response the most intriguing.
That's because the "exposure triangle" is a very widely distributed conceptual model. It just happens to be wrong in several aspects, but those aspects don't make a practical difference in many standard shooting situations, so people don't often notice it is wrong in the cases where its mistakes matter. Generally these involve low light and/or shooting in fully manual mode.
For me the triangle made perfect sense and gave me solutions for correcting the exposure.
In many cases the triangle does give perfectly usable solutions. It is in the edge cases and use of terminology that it gets things wrong. As a result people who have learned the exposure triangle tend to make mistakes in low light or when shooting manually. The errors of the exposure triangle do not matter too much for people who only shoot SOOC JPEGs and are not really interested in noise performance.
ETTR is good practice especially if your camera has a high dynamic range.
It matters more if the subject has low dynamic range because you have more highlight headroom that allows you to expose even further to the right. The possible range of dynamic ranges between different scenes tends to be greater than the difference in dynamic range between different cameras.
I'm not sure I agree with the fact that a higher ISO can reduce noise.
Well, you've been repeatedly told by many versions of the triangle and statements derived from its presentation that increasing ISO increases noise, so you will naturally be resistant to the counter-claim. Also, you may think you have observed instances where an increase in ISO resulted in an increase in noise. Let's start by addressing the myth that increasing ISO increases noise. Then we can deal with cases when increasing ISO might actually reduce noisiness.
Many people will use their camera in S mode, with Auto-ISO not set. Taking a series of photos of an approaching jogger running in heavy shade, with an f/2.8 prime lens and ISO set to 100, they find they can get a "good exposure" shooting at {1/125, f/2.8, ISO 100}. But 1/125 will result in subject motion blur as the jogger gets closer and begins to fill the frame. The camera won't let them set 1/250 or faster because the lens is already wide open and the ISO is fixed. So they increase the ISO to 400 and are able to shoot at 1/250 and 1/500, and acceptably freeze motion as long as the jogger doesn't quite fill the frame in landscape orientation. As the jogger gets even closer, they switch to portrait orientation. but by the time the jogger is close enough that they can only frame the subject from waist up, there is noticeable motion blur in the hands. So they change ISO to 1600, and are able to shoot at 1/2000. Upon examining the resultant photos, they find the ones shot at ISO 1600 are noticeably noisier than the ones shot at ISO 100.
Haven't we just proven that higher ISO causes more noise? No, we haven't. We have correlated higher noise and higher ISO, but we haven't shown a causal link. As you might have learned back in school, correlation is not causation. There are at least three other obvious factors here that are also correlated with image noise: shutter speed, subject size in the frame, and subject distance. Which one (if any of them) caused the higher noisiness?
When trying to identify which of multiple correlated factors is causal, we have to try to eliminate non-causal factors. We could do that by holding one of the factors constant while allowing the others to continue to vary. If the noise also continues to vary in the same way, then the factor we held constant wasn't the cause. It is trivially easy to eliminate subject size in frame and subject distance as causal variables.
It is less easy to isolate ISO from shutter speed. As long as we use an Auto-exposure mode (P, A or S, or even M with Auto-ISO) and keep the lens wide open and the lighting constant, any change in either one of shutter or ISO setting will result in a reciprocal change in the other factor. If you want to isolate the casual factor, you need to shoot in full manual without Auto-ISO. If you do that, you will find that increasing ISO while holding shutter (and aperture and lighting) constant does not result in noisier photos, just brighter ones. However, increasing shutter speed while holding ISO (and aperture and lighting) constant results not only in darker images, but in noisier images.
The correct conclusion to draw from this is that increasing ISO does not increase noise, but increasing shutter speed does increase the noisiness of images. In fact a little more experimentation will reveal that any change to exposure* on a given camera results in a corresponding change to image noisiness. If you add different sensor sizes and sensor efficiencies to the mix of variables, you will find that the noisiness of an image depends primarily upon the amount of light captured in the image.
*At this point it may be useful to review what exposure is. The term "exposure" has had a technical definition for almost as long as there has been photography. Exposure is the amount of light falling on the sensitive medium per unit area. Exposure is related to how bright a final image appears, but it is not the only factor affecting image brightness.
During the second half of the 20th century, a lot of people lost sight of that definition, and it became common to conflate "exposure" with image brightness. For people who shot roll film and had it developed by a lab, this was an easy mistake to make. Once they loaded their roll of film, the only way they had of controlling image brightness was to change exposure. Photography became a widespread hobby during the latter half of the 20th century, and the errors in use of terminology that developed at that time persist to this day. E.g. most photographers who see an image that is too bright will call it "overexposed". In fact, it is possible that it is
underexposed, but over-
brightened. This mis-use of terms continues with the so-called "exposure triangle". In fact, the triangle partly models image brightness, not exposure.
While the distinction between exposure and image brightness was unimportant to roll-film photographers who used standard development, digital cameras work differently. The distinction between exposure and brightness can be important if you shoot RAW or want to produce optimal JPEGs in less than ideal light.
Exposure is the amount of light falling on the sensor per unit area. It is controlled by three parameters:
- T-Stop (which is approximated by f-stop)
- Length of time of exposure (shutter speed)
- The amount of light in the scene (scene luminance)
The noisiness of an image depends primarily on the amount of light captured by an image. This depends on exposure, sensor surface area and efficiency of the sensor at converting incident photons to an electrical charge. Since sensor size and efficiency are generally constant for a given camera, exposure is usually the only significant variable affecting image noisiness.
Exposure is one of two variables that affect image brightness. The other is the ISO setting. Increase either by a stop, and image brightness doubles. This relationship between exposure and ISO means that usually a high ISO is used with a low exposure and a low ISO is used with a high exposure. Since a low exposure causes more noisiness and a low exposure usually comes with a high ISO, people associate a high ISO with increased noisiness. Where the "exposure triangle" goes wrong is to claim that the higher ISO actually results in or causes the higher noisiness. The confusion is enhannced when most versions of the triangle fail to distinguish what happens when you change a camera setting in fully manual mode from what happens when you change a setting in an auto-exposure mode.
The phenomenon of images getting less noisy as you increase ISO (but leave exposure unchanged) is not universal. That's why I said "
on some cameras it can reduce noisiness". It only happens to a noticeable degree if a camera implements an ISO increase using analog gain and introduces a significant amount of camera-added noise after the gain stage. Not all cameras do both of these, Most Canon cameras exhibit this behaviour, especially ones produced more than a couple of years ago.
To better understand, it helps to know that the apparent noisiness of an image depends not on the absolute amount of noise in the image but rather on the ratio of the signal to the noise. This Signal to Noise Ration (SNR) has as its signal component the amount of light captured in the image, and as the noise component several different noise sources. The most significant noise source at most exposures is shot nose. Noise is variation. Shot noise is variation naturally occurring in light due to the way light is created.
The other noise components are pretty well all variation that is added unintentionally but unavoidably by the camera. At high exposures, their magnitude is small relative to shot noise so they have little impact on SNR.
When a camera implements an ISO increase using analog gain, it increases the signal and any noise already present by the same amount. As a result the SNR does not change (Multiply the numerator and the denominator of a ratio by the same factor and the value is unchanged.) However, when you add some more noise after the gain stage, that noise is not multiplied, so the final SNR is greater than it was when there was no gain applied.
E.g. Let's say that at base ISO the signal was 100 and the noise was 12 before the gain stage. After the gain stage the camera adds 2 more units of noise. The final SNR is 100/14 = 7.14. When we apply four stops of gain, the Signal is multiplied by 16, the pre-gain noise is multiplied by 16 and then the post gain noise is added. The SNR becomes (16 x 100)/ ((16 x 12) + 2) = 8.25. The SNR is higher so the image looks less noisy. (This is a simplified example. The actual math is different from this because noise adds in quadrature, but the principle of the interaction remains the same.)
In general, the improvement in noisiness from an increase in ISO is not as great as the improvement from the same number of stops of increase in exposure.
Do you have some photo examples to show this?
None of my own, but I hope bobn2 won't mind if I share a couple of his:
As you can see, the ISO 1600 image is less noisy than the ISO 100 image. We'll have to take Bob's word for it that the lighting was the same (very low) for both images, and the reason they are the same brightness is that he normalized brightness during development. There are other examples to be found on this site showing more noise in lower ISO images than in images of the same scene taken with the same exposure but higher ISO.
