An In-depth Discussion of M + Auto-ISO for Canon SLRs
(Updated on 10-Aug-2012)
The reason I am writing this piece is not to convince everyone that they should convert to an Auto-ISO user, but simply to explain why Auto-ISO is a very good feature. Seeing that many people tend to dismiss this feature without really understand how to use it, and discourage others from trying it at the same time, I feel it calls for a proper discussion.
Auto-ISO is a feature that lets the camera change ISO automatically. The first canon SLR to have real Auto-ISO is 7D (This article incorrectly stated 50D was the first, a DPR forumer has advised me that 50D does not have a working auto-iso). Pre-7D Canons had a non-working Auto-ISO that would always be stuck on ISO 400. Since 7D, Auto-ISO would select the ISO that gives you correctly exposed final image based on the calculations from the camera's metering system in combination with the chosen shutter speed and aperture.
Personally I find Auto-ISO to be the most useful in M mode, I'll explain why and then explain its application in Av mode and Tv Mode. But before we do we will first discuss some surrounding issues.
The Nature of ISO
In digital photography, increasing ISO is increasing the level of amplification to the signal that sensor has received; and base ISO is simply leaving that signal free of any additional amplification. Amplification is either done by hardware (or sometimes called analog, or real iso) or by software (sometimes called fake iso). Canon APS-C SLRs are only capable of hardware amplification at full stops, e.g. ISO 200/400/800 etc. 1/3 and 1/2 stop increments are done by in camera software and you can achieve the same effect in post processing.
Software amplification is manipulation to the image file after hardware amplification is finished; it does not achieve the same image quality as hardware amplification. The reduced quality can be demonstrated by a simple comparison. Shoot two images with the same exposure, i.e. same shutter speed and same aperture under the same lighting; one shot at ISO 100 and underexposed by 3 stops, then increase “exposure” by 3 stop in raw processing software; while the other shot at ISO 800 and is correctly exposed. It would be immediately apparent to the viewer that the ISO 800 image has much less noise as well as better contrast and colour accuracy than the other. This is a very important point to remember. It also should be reminded that when manually setting ISO, the photographer is controlling the amount of amplification the camera will apply to the image after it is captured, whereas controlling aperture and shutter speed is controlling the amount of light is to be captured.
Earlier I have said that these in-camera 1/3 stop ISO increments are done by software and that software is not ideal. While this is true in reality 1/3 stop push or pull makes hardly any real impact to image quality, so for all intended purposes it can be overlooked.
On a side note, all current canon APS-C cameras has ISO’s upper limit set at ISO 6400. ISO 12800 is selectable but is software pushed from ISO 6400 and is not available in Auto-ISO.
Regardless what some people would like to claim, the human eye and brain adapts to different lighting conditions way too well so that we are unable to correctly judge the actual amount of light surrounds us. Camera engineers recognised this long ago and invented electronic metering system. The metering system judges the scene for us and calculates the required shutter speed and/or aperture and/or ISO. If we use these calculated values to shoot the picture we will get a balanced looking image, for the purpose of this article we will call it "correctly exposed image".
As of the second decade of the 21st century, metering system from one of the world's largest and best camera manufactures is very advanced and very accurate. Majority of camera users - those who use Av, Tv, P, or greenbox mode, are quite happily relying on the metering system. The only people that don’t use it are those who use M mode in its classic fashion (i.e. manual ISO). They rely on their "brain and experience" when it comes to metering. As far as I can see, they are not producing "more correctly exposed" images, and if I was forced to make a guess, I would say they produce a lot more unusable images than the rest of us because their need of experimenting different settings to get the correct exposure in a given scene.
Now, if by using Av mode, where the user dictates aperture and ISO value, we trust the metering system to determine the correct shutter speed that would deliver a correctly exposed image, then where the user is to dictate aperture and shutter speed, there really is no reason as to why we would not trust the same metering system, using the same calculation formula, to determine the correct ISO that would deliver a correctly exposed image. That mode is M + Auto-ISO.
The Ideal Image
The goals of a photographer in making an ideal image are usually to: A, create the image as he saw in his head; and B, have the sharpest image; and C, have the best image quality.
To achieve A we need to controls the DOF which directly affects the look of the final image, therefore we must control aperture.
To achieve B and C, it is a mixture of control both aperture and shutter speed. We control aperture because certain aperture gives the lens its optimal performance in sharpness, contrast and CA/flare resistance; we control shutter speed because the slower the shutter speed, the more light hits the sensor and the less amplification required to bring the image to "correct exposure", therefore the better looking the final image. But at the same time it increases the risk of motion blur caused by photographer's own shake or movement of the subject.
There is always a particular shutter speed that is just enough to avoid any motion blur, and that is what we call "minimum safe shutter speed". Since it would give the sensor more light than any other usable shutter speed, the goal is to use this “minimum safe shutter” speed in every shot. Depending on the focal length of the lens, movement of the subject, and surrounding stability support available to the photographer, minimum shutter speed changes in every different situation. Because of the complexity involved, it is a call only the photographer can make correctly. The camera cannot, regardless how advanced it is.
Along Comes M + Auto-ISO
Let’s recap what we know so far:
* To achieve optimal image quality, the photographer must have control over both aperture and shutter speed.
* Because human eyes are not designed to meter light, camera metering are much better at this job.
* It is the combination of shutter speed, aperture and ISO that produce a correctly exposed image.
* Correct hardware amplification produces the best image.
Taking all four points into consideration, it should not be difficult to conclude out that the practice that would ensure best image quality consistently is for the user to choose ideal aperture and the minimum safe shutter, and let the camera's metering system decide on what is the best ISO (hardware amplification) under the given lighting condition - which is in essence, M mode and Auto-ISO.
Comparison to Classic Av Mode Usage
Av mode is where the user controls aperture and ISO, and let the camera select shutter speed. There are several disadvantages when compared to M + Auto-ISO:
1, Because user has no control over shutter speed, you could end up with either shutter speed that is too slow or that is faster than necessary. If it is too slow, the image will have motion blur. If it is faster than necessary, the image would have received less than maximum amount of light and therefore had to receive more amplification than necessary to brighten it to the correct exposure and therefore more noise. Neither scenario is ideal.
2, In the event of the user finding the shutter speed is too fast or too slow and wants to correct it, the user has to go into ISO menu to change ISO in order to maintain aperture and still get shutter speed to change, this is slower than similar the operation in M + Auto-ISO mode, in the latter mode, one only has to turn the shutter speed wheel. Consider a situation where you are shooting a moving subject, you do not know what is the exact minimum safe shutter, you have to test several different shutter speed before you find it. Using M + Auto-ISO, you just shoot, turn wheel by 1 click, shoot again, turn wheel 1 click again, etc, you can easily test 10 different shutter speed inside 10 seconds. Now think how long it would take you to do the same using Av mode and having to change ISO.
3, This is perhaps the most significant disadvantage. With the exception of using flash as primary light source, in all other situations, lighting can be different between every shot. Consider the situation where you shoot pictures of people in an indoor party using 17-55mm F2.8 IS. The slowest you can hold steadily is 1/20s at 55mm (minimum safe shutter speed) and people always pose for you so subject is always static. When you shoot people who are close to windows or lights, they would have more light on their face, and the shots would require less amplification (lower ISO), and conversely shots of these who are not close to windows or lights would require more amplification (higher ISO). Using Av mode means you would:
- leave ISO low and end up with motion blurred images half the time, or
- leave ISO high and end up with half images that could have been much cleaner, or
- constantly go into ISO menu and make change.
With M + Auto-ISO, you just leave the camera on 1/20s and ideal aperture, nothing else, every shot should turn out just right.
Some people question M + Auto-ISO's usefulness when it lacks exposure compensation (EC) so we will take a look at this. Before we begin I would like to iterate an important concept - to maximise image quality, it is best to let more light onto the sensor, it is better than to use hardware amplification, which is in turn better than software amplification.
This brings me to the centre of the EC issue. EC in Av and Tv work very differently to EC in M Auto-ISO. In Av and Tv mode with manual ISO, EC changes either shutter speed or aperture to let more light in. The gain in IQ is therefore substantial, but it also substantially increases the risk of motion blur in Av or weakened lens performance in Tv. In contrast, using EC in M Auto-ISO will only change ISO amplification (you are already letting in the maximum amount of light). The image quality gained compared to using default Auto-ISO then bring it up or down in software by 1 stop is rather minimal, certainly a lot less than that gain by using EC in Av or Tv mode. In fact 1/2 and 1/3 stop EC are complete software based therefore makes zero difference to image quality.
The second significant issue is the rationale behind EC usage. Assuming camera metering is not faulty, generally people want to use EC when the exposure is not what they had in mind. e.g they want certain things in the frame to be brighter than they turned out to be, mostly commonly people's faces. Usual causes are strong contrast or uneven lighting. Rather than use EC, one can use different metering mode to adjust the way exposure is calculated. In my experience, that has worked fairly adequately. In the unusual event where change metering mode still doesn’t give you the right result, You can still easily switch ISO to a particular value that gives you the desired exposure, and then switch back to auto afterwards.
Correctly 'exposure" in post processing works better if you wanted to brighten the image, but not so good if you wanted to darken the image, because after hardware amplification, the information that is gone is gone for good, for example if camera metering chose iso 3200 but you would have preferred iso 1600, the brightest one stop of Dynamic Range is gone forever. To counter this problem you can use Highlight Tone Priority, it will use one stop less ISO.
It should be noted that AE lock does not work on Canon APS-C bodies when using M + Auto-ISO (for no good reason that I can think of), but works on Av + Auto-ISO.
All this does not mean EC should not be implemented in M + Auto-ISO eventually, I can see no reason why we shouldn’t have it. But the point here is that in the absence of EC, M + Auto-ISO still works quite well.
Limiting Auto-ISO Range
We have concluded that hardware amplification is better than software; this is only true to a certain extent. For current line of canon APS-C SLRs with the 18MP sensor, shooting with the same shutter speed and aperture, using ISO 1600 underexposed by 2 stops then increase brightness in post processing should give you the same image quality as ISO 6400. But by going up to ISO 6400, you lose exactly two stops of dynamic range in highlight. This means going to ISO 6400 is actually counter-productive. It would make sense to keep upper limit of Auto-ISO at 1600, and use software to brighten up the image later.
When Not to Use M + Auto-ISO
There are several situation where M + Auto-ISO is not useful.
1, Where light is abundant. Where every shot turned out to be iso 100, it would indicate that there is too much light, sticking with M + Auto-ISO would yield a lot of overexposed images. In this sort of situations switch to Av and use ISO100.
2, Where flash is used. Auto-ISO does not work properly when you use flash.
3, Where lighting is completely controlled. In studios where photographer has complete control of lighting, there is no need to use Auto-ISO. ISO should be left at 100.
That is it, you can assmue that M + Auto-ISO is useful in all other situations
Av + Auto-ISO
Av + Auto-ISO is a completely different animal. How it works is that the camera will first change shutter speed based on changing lighting condition, but as soon as shutter speed would have to drop below *1/(2xFL)* (for example slower than 1/100s for 50mm), it would bump up ISO, in other words it tries to maintain the old rule of thumb *(1/FL)* err on the side of a bit more shutter speed. It does not take IS into consideration (for no good reason that I can think of).
Personally I do not use it because my minimum safe shutter is much slower than 1/(2xFL), especially with IS. However I think for a novice who is still learning how to hold the camera steady, and on his way to find his minimum safe shutter, or someone who is of age and has trouble hold the camera steady, Av + Auto-ISO actually makes a lot of sense. You can be sure that your images will be shake-free and IQ will be quite consistent.
Tv + Auto-ISO
This mode works similarly to M + Auto-ISO except it would always leave aperture at maximum (where there is not enough light to keep ISO at 100). It is not a problem if you have a lens that would give you good image quality at maximum aperture. Most Canon L telephoto lens happen to do just that, making this mode quite useful for Canon white lens owners who shoot action, sports, and wildlife. The advantage of using this mode over M + Auto-ISO is that you can use EC.