AF accuracy vs Lens Aperture - Please explain!

Ok, I have recently been hearing about AF accuracy with different lens apertures and that some do not work or are much less effective at wider or narrower apertures? Could someone please educate me on this. If it helps you to relate it to me, I have a 70D and SL1.

Thanks!

It is my basic understanding that certain cameras, such as the 5DIII, have a subset of their AF points that have an increased accuracy when the body is used with lenses that support wider apertures. I'm not sure what the science/requirements are behind this, I've always chocked it up to those lenses generally having higher quality components that the AF system can leverage.

I'm not familiar with the 70D or the SL1 so I can't say if the AF systems of your bodies function this way, but it should be clearly laid out in the camera manuals under the AF section. I know for my 5Diii there's got to be at least 5 pages dedicated to laying out the various conditions.

Hopefully someone with a solid grasp on the technical aspects will provide you with more information and this helps direct you in your search for the answer.

I'm not sure where exactly you are reading this, or in what subject it pertains to.

Would this be in reference to the combination of camera and lens that will allow focus? usually this is discussed with Bird/Wildlife/Sports shooters using long lenses and teleconverters - it doesn't relate to a normal lens + camera discussion.

another instance is what focus points are cross type, and which ones are only vertical or horizontal types.. cross types are much more accurate. The 6D only has one cross type in the center, and the out focus points are either vertical or horizontal types.. but depending on lighting conditions, can be pretty good in good light.

Your 70D has quiet a few cross type focus points - 19 of them. A review of the owners manual will show you where they are.

There may be other instances where we are talking about depth of field, with other types of lens/camera combos.. but I'm not sure which subject you are thinking of.

I'm not sure, I tend to read a lot from different site. The best I can find right now is from canon and says "All AF points are cross-type at f/5.6.

Center AF point is cross-type at f/2.8" Does that mean that the f/5.6 points dont work when wider than 5.6 or that they are less accurate? And the same for the center point, is it less accurate wider than 2.8? I just like to have an understanding of how/why things work the way they do. Helps me to know why to do certain things. I'm the type that instead of "believing" I ask why. lol

I started reading some article i found in an internet wormhole of link to link to link to link..... And they started talking about how the calculation happen and the accuracy degrades exponentially, so I thought I'd ask you guys instead

So on your camera, all AF points will function if the lens is f/5.6 or faster (smaller F/stop)

The whole lines/cross/ af sensitivity thing is indeed reliant on the max aperture of the lens in use. ie " faster lenses will allow higher AF precision."

In a nutshell, the more light that the larger aperture lets in, the more precision you can get from many of the AF points in your EOS AF system.

This however is not as important as it often sounds, as it turns out that the wider the aperture you set, also means the more precision you NEED due top reduced depth of field.

Here is a GREAT article to explain pretty much the whole mess;


http://www.the-digital-picture.com …-autofocus-explained.aspx

CyberDyneSystems wrote in post #17469201
So on your camera, all AF points will function if the lens is f/5.6 or faster (smaller F/stop)

The whole lines/cross/ af sensitivity thing is indeed reliant on the max aperture of the lens in use. ie " faster lenses will allow higher AF precision."

In a nutshell, the more light that the larger aperture lets in, the more precision you can get from many of the AF points in your EOS AF system.

This however is not as important as it often sounds, as it turns out that the wider the aperture you set, also means the more precision you NEED due top reduced depth of field.

Here is a GREAT article to explain pretty much the whole mess;


http://www.the-digital-picture.com …-autofocus-explained.aspx

Thanks! I will definitely be giving that a read later!

CyberDyneSystems wrote in post #17469201
So on your camera, all AF points will function if the lens is f/5.6 or faster (smaller F/stop)

The whole lines/cross/ af sensitivity thing is indeed reliant on the max aperture of the lens in use. ie " faster lenses will allow higher AF precision."

In a nutshell, the more light that the larger aperture lets in, the more precision you can get from many of the AF points in your EOS AF system.

This however is not as important as it often sounds, as it turns out that the wider the aperture you set, also means the more precision you NEED due top reduced depth of field.

Here is a GREAT article to explain pretty much the whole mess;


http://www.the-digital-picture.com …-autofocus-explained.aspx

well yes but as the lens AF's with the lens wide open, you still get the advantage of the f/2.8 center spot and other higher accuracy points, even if you are then stopping down to f/11 to take the image. The taking f stop is important as it will determine just how accurate the AF has to be to get the subject reasonably in the middle of the available DoF.

Alan

BlakeC wrote in post #17469191
I just like to have an understanding of how/why things work the way they do. Helps me to know why to do certain things. I'm the type that instead of "believing" I ask why. lol

I started reading some article i found in an internet wormhole of link to link to link to link..... And they started talking about how the calculation happen and the accuracy degrades exponentially, so I thought I'd ask you guys instead

I just point the glass end at whatever, then push the clicker thingy all the way down when the camera beeps at me. Works good.

Long articles like the one at the-digital-picture.com give me a headache.

edit. But then again, I'm at work, so already had one.

Big +1 to Alan's post.

AF performance has nothing to do with the aperture you set. It is the lens' maximum aperture that is the important issue.

This is because, as Alan pointed out, the aperture doesn't actually change when you turn the aperture dial. It only changes at the moment you press the shutter button, which is after the AF has done its work. The aperture is always wide open when the AF is operating so the only issue that can reduce AF performance (apart from bad light) is the maximum aperture of the lens.

Most lenses have a wide enough max aperture for AF to operate happily. A 400mm f/5.6 lens will work fine. The problem arises when you reduce that max aperture such as by adding a tele-converter. These extend the reach of your lens but effectively reduce the maximum aperture by one or two stops. That changes your f/5.6 lens to an f/8 max aperture, which doesn't let enough light in for some camera's AF systems to work reliably. You will see this problem mentioned a lot in birding forums exactly because they are all trying to get maximum reach from their lenses to capture elusive birds, so the use of tele-converters is common.

In addition to Dan's very good post, the main reason that there is an aperture limit on Phase Detect AF systems is not about image brightness. After all the Contrast Detect AF system used during Live View operation will still work at any aperture, as long as there is enough light to form some contrast. Phase Detect AF systems operate by splitting the image, so that light rays from either side of the image (which sides depends on AF spot type, Line/Cross/Double cross and orientation. When the image is out of focus the phase (relative amplitude of the wave signal) is out of alignment on either side. The amount of phase difference actually tells the AF system how far out of focus the image is, and in what direction. This means that the AF system can simply drive the lens x far and know that the image will be in focus. This speeds up the AF a lot. All the AF system has to do is Look Drive and you are ready to take the picture. Sometimes it is possible to configure the system to do a look drive look which adds accuracy but at the cost of a slightly slower system. In order for there to be a useable amount of phase difference for the AF to work with there is a limit on how far apart the two sides of the image are. F numbers are useful here as they scale nicely with focal length and so angle of view. A wider angle of view can use a smaller physical aperture, which is exactly what the f number system gives us. So if you have an aperture of f/2.8 or wider then you can measure the phase difference with more accuracy as it will be a bigger value, as phase difference becomes more the further apart physically the signals are. The actual positioning of the AF sensor in relation to the image also matters. If you position the AF sensor in the center of the image it will again receive the maximum phase difference. This is why the center spot is always the most accurate. So for PDAF systems you have two limitations, they need light that is sufficient to actually form a signal, but they also need a wide enough aperture that the two sides are far enough apart so that a phase difference is actually measurable.

Now like any system the limits of accuracy and even the ability to work are not really that hard and fast, the performance drops off slowly. The manufacturers though place limits on the AF system that means that the system will work reliably for pretty much all of the time that there is a possibility of using the camera to take photos. This seems to be f/5.6 for most sensors, although it is possible to get the center spot to work at f/8 in some cases. This is why some non reporting teleconverters, or pin taping on reporting converters will sometimes works. There may just be enough measurable phase difference left for the system to actually operate. usually though this is pretty hit and miss. It also explains why for Canon cameras other brand lenses will still AF at f/6.3, as long as the lens actually reports the right value for max aperture to the camera.

So that's Phase Detect AF, so why don't we just use contrast detect more if there are no aperture limits, it seems like it would be a much better solution. The problem is speed. Because PDAF usually knows how far and in what direction it is out of focus it only needs to look and drive to obtain focus. This is really fast. Contrast Detect on the other hand looks a line of pixels, although you could look at a 2 dimensional group of pixels, and looks for a difference in brightness across them, it also looks at how quickly that change takes place. So what the system does is have a look and then as it doesn't know which way you will have to move it takes a guess and moves the lens by a certain amount. It then has to look again. If the contrast has improved, so that the change in tone is across less pixels that is great and it will then move in the same direction again. If the contrast has got worse it will then start moving in the opposite direction. These initial movements will be quite large, and eventually the signal will stop gaining contrast, and get worse again. At this point the system will then move back in the opposite direction, but by a smaller jump. It will keep doing this moving back wards and forwards in smaller and smaller jumps until it has reached the point where it can't jump any smaller and has reached the maximum level of contrast. Even if they system is in perfect focus at the start of this process it will always have to move the lens to check on the level of contrast. Remember that PDAF knows that ti is in focus just by looking the once. This makes contrast detect a very slow process, that is really very unsuited to moving subjects. The big advantage that CDAF has over PSAF is accuracy on stationary targets, as it uses the actual imaging sensor to do the looking.

Alan

Back to the OP, I believe that with newer cameras there are two "types" of "cross-type" focusing. one type is available at, say, f/5.6 to whatever limit of narrowness, but often a more "dynamic" type of "cross-type" becomes available at f/2.8 and wider.

I don't recall the details because, well, my gear supplies have in one way or another disappeared. but there should be explanations "out there" that refer to your camera's capabilities and limitations!

tonylong wrote in post #17550129
Back to the OP, I believe that with newer cameras there are two "types" of "cross-type" focusing. one type is available at, say, f/5.6 to whatever limit of narrowness, but often a more "dynamic" type of "cross-type" becomes available at f/2.8 and wider.

I don't recall the details because, well, my gear supplies have in one way or another disappeared. but there should be explanations "out there" that refer to your camera's capabilities and limitations!

Tony I will refer you to my post above. PDAF has an aperture limit because it needs to be able to see light coming at an angle from each side of the aperture. The wider the aperture the wider apart the light can be and so the system gets a larger signal for any particular light level. As we are generally looking at an image that is rectangular the most sensitive (so can work with smaller apertures) sensors will be horizontally oriented, and so measure vertical lines of constant (again remember that the sensor is measuring a line of contrast at right angles to it). These sensors will work reliably at f/5.6 at the minimum level of illumination/contrast. As systems have improved Canon has been able to build a sensor that when placed in the center of the view (so gaining the widest possible view) that it can actually work on vertical detail reliably at f/8. Although the sensor works reliably at f/8 it will become more accurate as you make the aperture wider, simply because the visible angle becomes greater, and so the phase difference becomes larger, giving a bigger signal to work from. This accuracy gain works for every increase in maximum aperture value, but because Canon switch certain functions on and off at fixed aperture values, based on the ability to produce reliable results at the specified minimum brightness/contrast levels, Canon have various features that work above/below certain f/numbers. So as the aperture gets wider we start to be able to measure the phase difference at different angles, and locations within the field of view. Once the aperture is wide enough for Canon to switch a certain sensor on, it will continue to get more accurate as the aperture gets wider.

What has happened is that Canon have improved the overall efficiency of the sensors used by the AF systems, so that effectively they can detect the phase difference at a lower level of signal. As far as the difference in sensitivity, at f/2.8 the angle seen through the aperture is twice that seen at f/5.6, so you have double the signal available. Hence the double accuracy at f/2.8 as it is always easier to measure big signal than small signals. There are limits to how wide of an aperture can provide benefit to this system though, as there are limits on the angle of incident to the sensor that can actually be seen. This is also true of the main sensor, and why the 50 f/1.2L has issues at very wide apertures. So past a certain angle the sensor stops seeing the the extra light, and so maxes out on accuracy.

All of the above happens gradually but as I have said, Canon either switches features on or off at certain apertures, or only specifies the sensitivity at certain fixed aperture values.

Alan

So how does all this relate to the fact that a lens is often sharpest stepped down? Many lenses are sharpest between F8-F11.

Colin Glover wrote in post #17550480
So how does all this relate to the fact that a lens is often sharpest stepped down? Many lenses are sharpest between F8-F11.

It doesn't. The AF sensor in your camera expects a certain amount of light to reach it in order to generate enough contrast, and AF sensors will perform better when more light reaches them. Regardless of which aperture you set on your camera, whether it's f/5.6 or f/22, the aperture will always remain fully open (f/2.8 for instance) until the instant you capture the image, during which the aperture closes down at the same time the mirror is going up. But if the widest aperture of your lens is physically incapable of delivering enough light, the sensor will not work, or with reduced functionality.

The only time the case you describe can be a problem is if you're using an adapted lens and physically stopping down to f/8 would prevent focus confirmation from working, but that's about it.

Again I will say, it is NOT the amount of light/contrast that matters here, although more light/contrast always helps, but the angles from which the light rays arrive at the AF sensor. Because f/No scales the physical aperture diameter with angle of view (focal length), the angle is constant at any f/no. So you can define how well PDAF will work based solely on the f/No. If you were to try doing PDAF with the lens stopped down it will actually be less accurate than at full aperture, even though the lens might have slightly more contrast when stopped down, because the separation seen by the Af sensor is less, so the phase difference for any out of focus condition will be smaller. Although increasing the brightness/contrast level will increase the signal it is not nearly as effective as actually making the aperture wider. The same was true when using an MF camera using a split prism focusing aid, which again like the PDAF is relying on seeing the image coming from either side of the aperture. It is also the same reason that when you mount long telephoto lenses with relatively small apertures compared to short FL lenses that one half of the aid goes dark. It was the introduction of these focusing aids that drove the move to using auto apertures to stop the lens down only when taking the shot. This particularly shows on many M42 mount lenses, which have auto diaphragms, triggered by a pin on the rear of the lens, and also a switch on the lens to manually stop the lens down.

Because the PDAF system also needs a certain level of brightness/contrast to work, as well as the necessary angles of light, it is very easy to think that it is the increase in light/contrast that the wider aperture offers that is important. It helps, but no level of brightness/contrast can make up for the fact that the angles of incident of the light from the two sides of the aperture are to small to form a large enough phase difference signal.

Alan