Requesting Help Evaluating the Focusing of My 24-105

[rlriii13]

I have been shooting with my copy of the 24-105L for the past five months and I have been continually questioning if I should be content with my results. I have taken countless portrait shots with the lens and more often than not, I feel like the focus is slightly off. Instead of the focus being sharply on the eye, it will be on the nose or the model’s collar. There's no feeling like getting a bunch of blurry eyes and sharp nostrils. What’s more, I would often feel like my intro level lenses were sharper.

At first, I assumed I was autofocusing incorrectly. Maybe I was changing the focus plane after focusing the shot. Maybe I was moving when the shutter fired. So I set up some test shots on a tripod and did not get any really alarming results that had me exchanging my lens copy.

After another couple months, I have got some really great shots with this lens, but I am still seeing a lot of poor focusing. So, I set up another test and decided to ask the best resource I have - the TDP community.

In my eyes, the results reflect my suspicions of front focusing, but seem to be much milder than I see when I’m shooting portraits. I have a hard time believing that this small discrepancy is what’s causing my problem. In your opinion, is the focusing of these images below so different that it would cause me to lose sharpness in a portrait or headshot? Is it just user error?


I thank you for any insight you can provide.

Both shots are straight out of camera, focused on the letter H with the 24-105 at f4 on an XSi (so micro adjustment is not an option).

Autofocus using the center point only (slight front focus):



Manual focus using 10x Liveview:

[HDNitehawk]

Anerror in the checking here is taking a picture of a stick laying at an angle. Your senosors form a +, and depeding on your camera the sensor is larger than the little square that you see that highlights. If the left part of sensor focused correctley it would indicate front focus, if the right part focused it would indicate back focus. For instance my 1D IV the sensor is twice the size as the square.


You need to take a picture of somthing in a flat plane, with the ruler set beside it at an angle to get an accurate reading. You could use somthing like lensalign or this [View:http://www.the-digital-picture.com/Reviews/Datacolor-SpyderLensCal-Review.aspx]


You can make a homeade version yourself with a box and ruller.


That said, it does look like it front focused a bit. The real test is in actual use, and you alread have a sense it is front focusing. So if you camera allows, you could adjust it.


EDIT: Just noticed you indicate it does not have an adjustment. Which only leaves sending it off.

[cls]

Hi, I've had similar concerns with a couple of my lenses. I suppose you could say that mainly my concerns were caused by thinking the AF system can do more than it actually can. After learning more about the AF system and more about proper AF testing I've come to the conclusion that all except one of my lenses have been OK out of the box. The one exception was a 50/1.8 that focused really badly - i sent it back without testing and received a new one that was OK.


Some things to consider when judging the AF:


- You know the exact spot that you want in focus, the camera doesn't. I don't know all the details but basically center point AF is based upon focus detection within a small region around the center point - it's bigger than the small dot. Proper AF function will result in focus on something within this region, not necessarily the spot you had in mind. In order to deal with this you should test focus on something flat (a plane perpendicular to the sensor).


- There is also an uncertainty that one must accept. If you re-focus a number of times you will not get the exact same result every time. You should expect though that the average focusing error is zero (for a large number of shots), and that none of the shots are completely off.


- There are differences between camera bodies and lenses. I have a rebel T1i that I've understood is not as good as the higher end bodies regarding AF precision (never tried one myself). Fast lenses (f/2.8 or wider) also help the AF precision. A rebel body with an f/4 lens can not get the same precision as a 7D body with a fast lens. Still, the average focusing error should be zero.


For your specific example, tight thin DOF portraits, I usually aim right below the eye in order to avoid any eyebrow in the AF detection region (AF will typically pick the closest spot within this region, like an eyebrow instead of the iris). I also take a few shots to make sure that at least one of them has focus exactly where I want it.


So, before you send your gear for adjustment, follow HDNitehawk's advice and re-do your test using a flat contrasty target in good lighting - you may find that AF actually does OK.

[rlriii13]

Thanks guys.
<blockquote>You need to take a picture of somthing in a flat plane, with the ruler set beside it at an angle to get an accurate reading.</blockquote>


I'll definitely give the perpendicular test a go and see what comes of it.


CLS, thanks for the reassurance. I know it's not a terrible lens copy. If there is a problem, it's so minor that I wouldn't think it would matter in most cases.
<blockquote>I usually aim right below the eye in order to avoid any eyebrow in the AF detection region (AF will typically pick the closest spot within this region, like an eyebrow instead of the iris).</blockquote>


I'll try to avoid the brows... that's quite possibly my problem (or at least one of them)!

[neuroanatomist]

I'll definitely give the perpendicular test a go and see what comes of it.

Here's a relevant thread to peruse...

[rlriii13]

Thanks John. I completely missed that thread. There

[richm]

Good Timing - I just read this article and found it quite interesting. www.canonrumors.com/.../this-lens-is-soft-and-other-myths

[andnowimbroke]

Which is why I was getting ready to ask how many lenses you have. If you just send in one with the camera, that might mess up the rest.

[rlriii13]

Thanks for the article Rich. If I follow that logic (which I'm inclined to do), I have two options for this specific situation. I should either exchange the copy and hope that the next one is a better match for my body or upgrade to a new body with micro adjustment. I'll have to see which option my wife thinks is best. [] I have a handful of other lenses that I'm happy with, so I would prefer to not have the body calibrated to this one lens.


Either way, I still plan on running a better autofocus test this weekend just to see the results.