Andre's Blog
Perfection is when there is nothing left to take away
Cross type AF points in EOS 7D

My good old film EOS SII has just a single auto-focus (AF) point and can only focus on vertical or diagonal lines because its AF point had only one bar of photosensitive elements. New SLR cameras have cross type AF points that have two lines of photosensitive elements forming a cross, which allows such cameras to focus on vertical or horizontal lines.

All good things come at a price and cross type AF points are expensive, which is why many entry-level SLR's only have one cross type AF point located at the center of the frame. Remaining AF points are usually sensitive only to either vertical or horizontal lines.

The EOS 7D boasts 19 cross type AF points, which makes it simply great for action photography when you want to track the subject continuously. That is, when tracking a person by keeping the center AF point on their face simply wastes the upper half of the frame. If instead one of the upper AF points is used to track the subject, the frame will be filled more, giving you more image to work with.

This is all nice in theory, but soon after I started using non-center AF points more extensively, I discovered that in many shots the subject is out of focus, while the background doesn't have a hint of blur. At first, I thought I was slipping the AF point off the subject while tracking it, but checking the location of the AF point indicated that it was right where I wanted it to be. Here's an example of such shot. Hover your mouse over the titles to see the location of the active AF point and the close-up crop.

Original | AF points | Crop

Eventually, I decided to map out crosses inside each AF point, so I can focus on the subject more accurately. I placed a white sheet of paper with a small black square on the wall and moved each AF point perpendicularly to each edge of the black square until the focus light in the viewfinder started to blink, which indicated that I reached the end on the AF cross in the given direction.

Repeating this exercise once for each of the edges gave me four shots where the inner edge of the black square indicated the end of the AF cross in each direction. I superimposed all four shots onto each other and ended up with this picture for the center AF point.

Center AF Point

So far so good - the cross inside the center AF point was pretty much what I expected. Note that the cross extends well beyond the red square marker in the viewfinder, which means that if there is more detail outside of the AF point marker, the camera will focus on that detail.

After taking 72 shots and merging each four, I had the complete map of all AF points for my 7D. Let's take a look at the second one form the top in the center column, which is the one used in the sample shot above.

Non-center AF point

Now it all makes sense - the cross inside this AF point extends well into the top AF point and the camera focused on the detail in the background rather than on the fuzzy police bear.

Here is the complete map of all 19 AF points. Hover your mouse over the numbers to see the AF cross boundaries inside of each AF point.

1  2  3  4  5 6  7  8 9  10  11 12  13  14 15  16  17 18  19

In order to focus with confidence, you have to memorize the direction of each asymmetric AF point and make sure that the cross inside does not extend beyond the subject, especially if the subject doesn't have much contrast. Note that there is a blind spot between the second and third columns from each side (the ones with the wider gap between them) and the camera will not focus at all if the detail you are trying to focus on happens to be between these columns.

Posted Sat Oct 2 21:22:07 EDT 2010 by yatin parikh

thanks to give an ideas. great expriments.

Posted Mon Oct 11 05:27:54 EDT 2010 by Toni

Very interesting results. The point is: is this a common "feature" of all 7Ds or only yours? I'll try to replicate the experiment with mine when I have some time...

Posted Mon Oct 11 11:05:03 EDT 2010 by Andre

I tested more than one 7D body and they all behave the same way.

Posted Mon Oct 11 13:30:55 EDT 2010 by Toni

OK, Andre. Then I find it even more interesting because it can explain many common focusing "problems" with 7D (and other cameras using the same AF system).

I used a couple of your pictures to illustrate a post about this subject on ( I hope you don't mind... I also included a link to this blog.


Posted Mon Oct 25 10:48:56 EDT 2010 by David

When you did this test the af mode was single point and not single point with expansion?


Posted Mon Oct 25 16:41:51 EDT 2010 by Andre

Just the single AF point

Posted Tue Oct 26 03:20:58 EDT 2010 by David

Thanks Andre, I wonder if spot af makes the af cross a lot smaller.

Posted Tue Oct 26 07:55:24 EDT 2010 by Andre

Yes spot AF makes all AF crosses smaller, so the one in the center barely extends beyond the edges of the AF point square. Non-center AF points are similar in shape to what's described in the post, but equally smaller.

Posted Mon Mar 7 06:35:58 EST 2011 by Ciprian Trofin

Thank you. Very usefull test.

I want to clarify something: your conclusion is that focus points 1/2 and 4/5 overlap ?

Posted Mon Mar 7 07:36:53 EST 2011 by Andre

Yes. Canon reused the vertical bar for these points.

Posted Thu Mar 10 05:56:06 EST 2011 by Ciprian Trofin

Interesting. Did you check if the vertical bar is also reused in Spot AF mode ?

Posted Fri Mar 11 00:56:16 EST 2011 by Andre

I don't have a map for spot AF, but I'm using it a lot now and it seems to follow the described pattern (except the bars are shorter), at least for the AF points in the center, which I use most often.

Posted Sun May 8 23:07:25 EDT 2011 by Matt

I did a similar map with mine on Andres flickr post about this and my maps are nearly identical, any differences I chalk up to my measurement technique and the differences are not significant enough to make a difference.

Canon is not being honest with us about the focus points of this camera.  But, when you're aware of it, you can at least make adjustments in your technique.

Thanks for doing the original tests Andre, brilliant idea.

Posted Sat Dec 31 08:06:55 EST 2011 by Ian

Well done.There are a lot of people out there asking for information  about how good this and that camera is for auto focus ect.I think a lot of he problems people are having with out of focus images can be answered  by the results of your tests. I also think that this is the same with Nikon cameras as well.I have had it where I know for certain I have had the camera on the correct focus point and then found it has focused slightly off on something in the foreground or background.Years ago the photography magazines used to test and review cameras properly and in depth, it is a shame they don't anymore.They don't even test the shutter speeds anymore let alone the accuracy of the auto focus.Thanks again. Ian

Posted Mon Jan 28 16:43:33 EST 2013 by Jason
Am I reading this correct, that it does not actually focus where the read dot is but the four areas around the red focus dot?
Posted Mon Jan 28 17:27:47 EST 2013 by Andre

Draw vertical and horizontal lines right through the middle of the red square, each ending at the edges of the dark rectangles. Anything with sufficient contrast along these lines will be picked up by the AF.

Posted Wed Feb 13 06:29:37 EST 2013 by David

Very interesting! Great test.

Annoying that the differences between the rectangles and actual coverage are so large in some cases.

I'm thinking it would be much easier to understand if the AF points were represented by the actual cross-hairs instead of rectangles.   

Posted Thu Aug 1 02:28:52 EDT 2013 by Steve

 Very helpful test! Thanks!

Do you also know how the 5D Mark3 is behaving?

Posted Sun Aug 4 00:53:19 EDT 2013 by Rod Bartlett

 Hi all,  Was this fixed with the latest firmware upgrade?

Posted Tue Aug 6 08:31:19 EDT 2013 by Andre

This is how the AF hardware works in the 7D. It cannot be "fixed" in firmware.

Posted Wed Aug 7 09:15:23 EDT 2013 by Rod Bartlett

Thanks Andre.......Well at least I know what to do with the focus points. Thanks for article, it certainly has helped me...