7D Autofocus Issue

Discussion in 'Canon EOS' started by george_prescott, May 9, 2015.

  1. I've owned my 7D for about 5 years and generally love it, but I've noticed that on the few times I've shot at very large apertures the focus sometimes is off. I attributed this to user error, but finally today I decided to rigorously test the autofocus. Attached are two examples, using a tripod, cable release and a Canon 80mm 1.8 wide open. The center focus point was used with no repositioning of the camera. Focus setting was on One Shot. The intended focus point is the "8" on the remote control. I tried the same thing on my venerable 20D and it focused perfectly every time. If the results were consistent I could try to adjust the microfocus (although I have never tried this) but sometimes it is OK, sometimes not, so I think this would not help. Time to send it in to Canon?
    George
     
  2. Trouble attaching the example...
    00dHUo-556693384.jpg
     
  3. Even with the best phase detection AF system and properly adjusted cameras and lenses there can be a small amount of focus variation, usually only detectable when looking at 100% crops.
    There is no 80mm f1.8 so you either have the 50mm f1.8 or the 85mm f1.8. My guess is that the low cost, regular AF motor 50/1.8 will likely show more focus variation than the USM driven 85/1.8.
    Why on the 20D AF seems perfect while on the 7D it seems variable I don't know. If anything I'd have thought things should be the other way around. Since the 20D is only 8MP small amounts of defocus might not be as obvious as they are withe the higher resolution 7D.
    In a case like this I'm not sure a trip back to Canon is going to help much.
    This might give you some food for thought - http://www.lensrentals.com/blog/2012/07/autofocus-reality-part-1-center-point-single-shot-accuracy
     
  4. Bob,
    Thanks for the reply. It was actually a 85 mm. I repeated the test today, but outside with a tape measure. This time the 7D had no problems. It looks like maybe it backfocuses around .25-.5 inches, but that may be because the depth of field extends further back than forward. Really hard to tell.
    Actually, one of the reasons I initiated this study was also to look at the effects of recomposing after focusing. Lately I have been reading a lot that the point you focus on will go slightly out of focus after recomposing due to changing the distance from the sensor plane to the focus point due to the camera angle change. Since I was getting good results this morning, I also tested for this and saw no difference in sharpness on my original point with and without recomposing. Looking at the math, if you tilt your camera 10 degrees (which is a lot), on a full frame camera using the entire 35 mm length of the frame, the maximum change in distance is only 6 mm (tan 10 degrees = x/35 solving, x=6). The center of the frame would only change by half of that, or 3 mm. 6 mm is not enough to affect focus quality, at least on portrait shooting, which is what this advice was aimed at.
    George
     
  5. Looking at the math, if you tilt your camera 10 degrees (which is a lot), on a full frame camera using the entire 35 mm length of the frame, the maximum change in distance is only 6 mm (tan 10 degrees = x/35 solving, x=6). The center of the frame would only change by half of that, or 3 mm. 6 mm is not enough to affect focus quality, at least on portrait shooting, which is what this advice was aimed at.​
    That is true if your stance is like the Rock of Gibraltar! With my best efforts, I sway forward and backwards a couple inches after a few seconds and can't keep the plane of focus consistent when I recompose at portrait distances with a fast lens. Focus-recompose at F1.2 to 2.0 works great when I use a pan head!
     
  6. The focus and recompose "problem" is mostly a myth. There are a limited number of situations where you will lose sharpness, but 99% of the time you'll not see a significant difference.
    I wrote an article - http://bobatkins.com/photography/technical/focus_recompose.html - which goes into the math and shows some examples. I think I used a 20mm lens at f2.8 focuses at 1m as an example where focusing in the center and then recomposing to put that center point at the edge would certainly result in poor focus. However with longer focal length lenses, less extreme recomposing and more reasonable distances there's no issue.
    Unless I'm shooting macro with a wideangle lens, I don't worry about focusing and recomposing, in fact I do it all the time.
     
  7. The focus-recompose technique assumes you have a lens with a flat field of focus - if the field of focus is curved, focus and recompose will certainly lead to poorly-focused photos. Just sayin'...
     
  8. Bob,
    I agree with your conclusions, but I think I must respectfully disagree with your trig. The distance from point C to the lens can't physically change that much when pivoting the camera except maybe by the distance the ends of the lens move. I attached a diagram to show what I am talking about. The original camera position is represented by the solid lines to L, C, and R; the new position the dashed lines to L1, C1, and R1. If I understand you correctly, the new distance to C is what was previously the distance to L, but if you rotate the camera essentially around the axis of the centerline of the lens as I did in the diagram, the distance to C is essentially unchanged. Of course, in real life the camera may be rotated more around the sensor plane, which could swing the lens around a little. But in your example that distance changes by about 9 inches, which seems like too much.
    George
    00dHb7-556710284.JPG
     
  9. I think if you look at my article again you'll see it's correct. You have to remember that if a wideangle lens (e.g. 20mm on full frame) is used to image a wall from a distance of 1m, at the center of the field the wall is 1m from the camera, but the edges of the field are further away, about 1.5m from the camera, but they are still in focus because they are off axis and that's the way a rectilinear lens sees things. Fixing focus on one and moving it to the other position will result in focus being off by 0.5m.
    Field curvature can be a factor which may help or hinder, depending on which way the field is curved and how strong the curvature is across the field.
    Independent of the math, my tests indicated that, at least for the test conditions, focusing and recomposing didn't result in any sharpness loss.
     
  10. Ok, I did not know that! Then what you say makes sense.
    So, for example if I am taking a full length photo of
    someone standing and I want the image centered in the
    frame I would NOT want to focus on their face and
    recompose, right?
     

Share This Page