OH Lord_Malone, I did a simple test for you to look at

What timbop said.

jdmoto wrote in post #4095752
...

My question to all the good and sharp 40D's is.. how consistent is the AF I had it on the tripod and live view on with mag x10 and every time i would focus the camera some times 2 out of 5 it would front focus. I guess i never really check this with my old one because it was sharp. Now im so anal about sharpness I want to know how consistent the AF should be.

My 40D is also a little inconsistent on focus - I sometimes get the "2 out of 5" OOF (or outside of the so-called "acceptable range") on my images too, although it's maybe more 2 or 3 out of 10 for me (the 'misses' increase as lighting decreases, down to indoors - tungsten). From what I understand, this is to be expected from AF systems as they cannot be 100% accurate and WILL fail from time to time. I've also read about at least one other forum member who believes he has a sharp 40D, but has experienced random OOF shots too...

Addition: I've just got into the habit now of zooming into my shot in playback mode as far as it will go to pick out focus. If I am not happy, I re-take the shot with Live View 10x magnification. Yes, it isn't ideal - but at least Live View is there to fall back on when AF does fail.

I believe that this information is pertinent to this thread -- feel free to tar and feather me if it isn't.

In LM's thread in which he mentioned the possibility of back focusing on his 40D, I mentioned that the actual focus point might lie outside of the selected focus box by a considerable distance relative to the size of the box.

Since I don't have a 40D, I went with what I have (a Digital Rebel XTi) and made an assumption that the gist of my results can be reasonably applied to the 40D. The point of the illustration shown below is that placing a focus box (the center one in this instance) over a desired point (such as a person's eye) and then seeing that the box flashes red when the shutter button is pressed does not necessarily mean that the actual focus point is inside or touching the box. The autofocus function will favor the sharpest line with the greatest change in brightness in its field of regard. If the contrast around an eye is not as great as something else, you will not know it by any means other than experience in dealing with such situations. Pixel Peeping with ZoomBrowser EX where the boxes are overlaid on the image will not help other than telling you which AF sensor was active -- it won't tell you where the actual focus point was.

Figure 1 below shows the maximum vertical distance from the box that the camera was able to obtain focus with respect to a horizontal line. It is interesting that the sensor's field of regard in the vertical direction appears to be much greater than it is in the horizontal distance. Figures 2 and 3 show maximum horizontal displacement left and right of a vertical line where the camera was able to obtain focus.

Based on this observation, the area covered by the center AF sensor on the XTi is many times larger than the box that is displayed in the viewfinder. The conclusion is that when the AF box flashes red, it means that the camera has managed to focus on "something" and it isn't necessarily in the red box. About the only way that we can determine where the focus point actually occurred is by close inspection of the image during post processing.

We may gripe and wonder why the AF sensor area is so large, but consider the consequences of making it very small -- it could result in many more shots in which the camera is not able to automatically lock on focus. It is a design decision which balances selecting a precise point against fast auto focusing capability.

HOSTED PHOTO
please log in to view hosted photos in full size.

bill boehme wrote in post #4098291
I believe that this information is pertinent to this thread -- feel free to tar and feather me if it isn't.

In LM's thread in which he mentioned the possibility of back focusing on his 40D, I mentioned that the actual focus point might lie outside of the selected focus box by a considerable distance relative to the size of the box.

...

We may gripe and wonder why the AF sensor area is so large, but consider the consequences of making it very small -- it could result in many more shots in which the camera is not able to automatically lock on focus. It is a design decision which balances selecting a precise point against fast auto focusing capability.

I don't think you will be tarred and feathered for this post. I found it very useful, and if I weren't so lazy I'd do something similar with my 40D (but I'll just wait for someone else to go ahead :lol

I think you've illustrated the issue a lot of us are facing, although since it is noticeable on your 400D, I'm surprised as to why this wasn't the case with the 30D (I'm assuming) and why it is so prevalent on the 40D.

I've noticed this especially on some of my first shots with my Sigma 30mm f/1.4. If, for example, I took a photo of someone with their shoulder to the camera and had center focused on their face, at 100% I can clearly see that their shoulder is in focus, not their face. Looking at the shot more closely, I can see that the shoulder would have fallen in the "area of regard", as you call it, around the vertical spaces of the center focus box. I guess this, coupled with the fact that the camera tries to focus on the object closest to you, is why I see "front focus" on those shots.

What I find interesting though is that people who have exchanged their soft-focus 40Ds have had some luck in getting a "copy" which is "sharp" - so, is the "focus bleed" varying per body?!

As to this being a design decision - I personally don't think this is a good decision (at least not the way it was implemented). At the very least, there should be a note in the manual and some sort of indication in the viewfinder (make those boxes bigger!) or perhaps flash the focus box continously to indicate that the focus lies just outside it. There's so many ways "it could have been"!

Still, I hope some of us 40D owners can test this phenomenon out and see whether this is in fact the cause of our collective soft-focus grief...

A few other late night thoughts along these lines:

fz_za wrote in post #4098424
....... If, for example, I took a photo of someone with their shoulder to the camera and had center focused on their face, at 100% I can clearly see that their shoulder is in focus, not their face. Looking at the shot more closely, I can see that the shoulder would have fallen in the "area of regard", as you call it, around the vertical spaces of the center focus box. I guess this, coupled with the fact that the camera tries to focus on the object closest to you, is why I see "front focus" on those shots.

A common misconception is that the camera tries to focus on the closest object. The AF sensor is actually a very simple device that is only able to detect an edge and then send a control signal to the lens focus motor to drive the focus to its sharpest point. It has no way of determining distance to the focal point so it doesn't know what is close or far away. The AF sensor is so simple that it can't actually see objects nor can it see color. It operates like a single row of pixels (or two crossed rows) where the desired focus condition is obtained when there is an abrupt change from dark pixels to light pixels along the row.

fz_za wrote in post #4098424
....... As to this being a design decision - I personally don't think this is a good decision (at least not the way it was implemented). At the very least, there should be a note in the manual and some sort of indication in the viewfinder (make those boxes bigger!) ......

As an engineer, I appreciate the decision from a technical perspective, but there appears to be a disconnect in relating these design ideas to the user community where the interest is to use the camera as a tool of their trade and not be required to develop technical expertise that goes beyond the tools of the photography trade. I can support the idea of a larger box, but I also have a feeling that the shape of the boxes limits is not simply rectangular nor is sensitivity constant within the AF zone.

its it just me or does the first set of images not look focused on 0? more like forward to 1 to 1 & 1/2 depending on which side you read...?

the second example seems better...the posts by bill are very interesting.

Keltab wrote in post #4095517
I sent my 30D to Canon out in California... I had it back in 9 days. They calibrated it and it is doing great now - if only I could do better more consistently!

Where is this? Do you have an address? Also, my 40D is new (only 1 week or so) would it be wiser to send it off to be recalibrated or just to send a replacement?

Robf wrote in post #4098623
its it just me or does the first set of images not look focused on 0? more like forward to 1 to 1 & 1/2 depending on which side you read...?

the second example seems better...the posts by bill are very interesting.

Set one is testing front/back focus @ a 45 degree angle in two different light settings/ Showing 100% crops

Set two is testing how large the focus area is around the focus square @ a 90 degree angle.

d8168055 wrote in post #4098733
Where is this? Do you have an address? Also, my 40D is new (only 1 week or so) would it be wiser to send it off to be recalibrated or just to send a replacement?

Remember these "potential" issues are not the norm. Not even close to the norm.

We are also attempting to make it happen and @ 100% crops.

I wouldn't be too worried unless you are a professional 100% crop focus chart photographer that likes to shoot in very low light with focus points with little or no contrast.

true enough, but the AF sensor operates within a system along with a computer and a lens motor (input-process-output). The system as a whole certainly does seem to focus on the closest high contrast object within its field of view. try it on a 3D object like a shoe.

bill boehme wrote in post #4098574
I can support the idea of a larger box, but I also have a feeling that the shape of the boxes limits is not simply rectangular nor is sensitivity constant within the AF zone.

Which would explain why the focus point moves to the area of more contrast (ligher areas have more) and if that point is not dead center of the square then you could have slight focus shifts with shallow DOFs.

Nikon's new line up is supposed to AF track based on color. Do you think this will be more accurate or is it going to get confused with multiple colors in a confined area?

DavidEB wrote in post #4099158
true enough, but the AF sensor operates within a system along with a computer and a lens motor (input-process-output). The system as a whole certainly does seem to focus on the closest high contrast object within its field of view. try it on a 3D object like a shoe.

The only feedback device in the system is the AF sensor. None of the other components of this control system (meaning the motor, driver, or processor) has any feedback information to null the control loop other than what is provided by the AF sensor. The processor can't create distance data on its own -- all that it does is use the AF feedback data along with some logical processing to create a control loop (probably a PID or variant controller) for the motor. There are several reasons that you may see focus occurring in front of the desired focus point -- one is mentioned in my first post and has to do with the size of the focus box and the resultant ambiguity of where the exact focus would be in the surrounding area. Another reason would be related to a hardware misalignment of the AF sensor WRT the focal plane. The path length from the lens mount to the AF sensor must be precisely the same as the path length from the lens mount to the image sensor plane. If it isn't then one of the mirrors needs alignment. And, finally, I think that we should also consider the visual difference between OOF objects in front of versus behind the actual focus point.

S2000 wrote in post #4099161
Which would explain why the focus point moves to the area of more contrast (ligher areas have more) and if that point is not dead center of the square then you could have slight focus shifts with shallow DOFs.

Nikon's new line up is supposed to AF track based on color. Do you think this will be more accurate or is it going to get confused with multiple colors in a confined area?

An engineering rule-of-thumb is that, "more is better than less". In fact, this rule-of-thumb probably applies to almost everything. I am sure that Nikon has done a lot of research in fine-tuning their AF algorithms and color seems to me to have great potential for tracking moving objects against a cluttered background. I suppose that there are some cases in which color could create a problem, but you could say the same thing about any type of tracking system. I suspect that their use of color works in conjunction with their classical tracking system and is not a replacement for it.

CANON FACTORY SERVICE
15955 ALTON PARKWAY
IRVINE, CA 92618
(949) 753-4200

That is their standard service center - they did a great job and were very detailed with the work.

bill boehme wrote in post #4100743
An engineering rule-of-thumb is that, "more is better than less". In fact, this rule-of-thumb probably applies to almost everything. I am sure that Nikon has done a lot of research in fine-tuning their AF algorithms and color seems to me to have great potential for tracking moving objects against a cluttered background. I suppose that there are some cases in which color could create a problem, but you could say the same thing about any type of tracking system. I suspect that their use of color works in conjunction with their classical tracking system and is not a replacement for it.

I have to figure that they aren't going to introduce a auto focus system that doesn't work...oh wait...

btw...you have to work for TI