The quick and dirty method I used last year was to run a string from my target to the lens. I had a vertical line on the target and when the string aligned to the line, it was square. The hight was adjusted prior and I used a level to make sure it was vertical.
Too much hassle compared to a bore sight style shown here. Thanks for the procedure.
nathanlake wrote:
I strongly suggest you apply for a patent or a copyright or something to protect this idea....and hurry up.
I'm no patent lawyer, but since I described the use of file size as a technique to judge sharpness in a post on DPReview last November I don't think there is much here in the way of new intellectual property.
As for automation of the calibration, Since Canon already has contrast detect AF as well as phase detect AF in the latest cameras I can see the next generation of cameras having self calibration built in - simply AF with contrast detection, AF with phase detection, determine degree of lens movement if any, correct the error.
In the meantime a third party automated solution would be great, but if Canon has any clue at all I imagine it would have a short lived value, beyond the current generation of cameras.
Daan B wrote:
Cool concept... I will give this a try myself to double check my own method (which takes much more time ). I am not sure about keeping the target perpendicular to the axis of the lens though... I wonder how I pull that one off?
Apart from the ideas already included within this thread, having a focus target of exactly 3:2 format and framing it perfectly, corner to corner, should do the trick. If your angles are off the converging verticals or horizontals will make it impossible to align all four sides of the target with the viewfinder.
Alternatively, stick a small flat mirror onto the centre of your focus target and sight from the camera. When you see the camera reflected back in the mirror you know you are aligned.
This has been a very interesting thread. I have learned from it, thanks to all. I do have a question though about having the camera perfectly aligned. As long as nothing changed between shots-camera+target alignment-wouldn't the file size readouts all still have the same relationship? It seems that as long as you were pretty close you'd still be shooting the same "scene" each shot, wouldn't you?
Thanks for the great idea.
I went through all my lenses, range was -5 to +9 with my 200 2.8 + sigma 1.4x being dead on at 0.
After calibrating with your method I shot one of those scale charts on an angle, and what do ya know dead on.
I just got my 50D, so its the first chance for micro focus adjustment, but looking at 100% crops of the images, the difference between 1 degree of adjustment is fairly big when shooting text on an angle. Having such a large range with my lenses makes me wonder, because I don't think they've ever been so far off with other bodies.
egd5 wrote:
This has been a very interesting thread. I have learned from it, thanks to all. I do have a question though about having the camera perfectly aligned. As long as nothing changed between shots-camera+target alignment-wouldn't the file size readouts all still have the same relationship? It seems that as long as you were pretty close you'd still be shooting the same "scene" each shot, wouldn't you?
If you are shooting a detailed target and it is at an angle, some of the target will be in focus, some out of focus. When you vary the MA,. you would be selecting the portion of the target that showed the most detail (largest file size) when in focus, which may not be where you intended, in fact, if the level of detail was the same acorss the target, you would probably get the same size file for any MA, since some portion of the target would be in focus because of the angle.
If you were visually inspecting the target instead of relying on file size, you would look at the part of the image where the focus point is and would see when it is in focus so in this case the angle wouldn't matter. BTW, this visual method has the disadvantage of depending on the AF sensor being lined up to the AF point reticule in the viewfinder. The file size method is not vulnerable to this dependency.
photonic, at the expense of looking foolish, and after proper thanks for your most helpful contribution, I would like to ask why, using your approach, selecting the largest file size of any series of identical shots would not work. Why do we need a target and holes, rather than "just" a brick wall for example? Using a tripod with levels would ensure proper horizontal positioning, it being reasonable to assume that most walls are vertical...:-). Of course those of us with a taste for the finer things of life are welcome to hang a Leonardo Da Vinci or Vermeer on said wall, as long as it is vertical, it should work as well...:-). Liveview magnified 10x would ensure proper focus, lighting and aperture to remain constant and "reasonable".
This would make it even easier to set up. Where am I oversymplifying? Again, with thanks.
If your brick wall is exactly perpendicular to the lens axis, then when the brick wall is in focus, all of it will be in focus and when out of focus (modulo differences in distance to the wall due to the FOV), all of it will be out of focus, and the file sizes will be significantly different as you vary MA. The file size differences will be large and it will be easy to tell where the best focus is.
If the brick wall is inclined to the lens axis, then when you vary MA SOME portion of the wall will always be in focus and some out of focus, and your file sizes will differ less if at all.
With visual inspection you only inspect that portion of the wall you expect to be in focus. With file sizes you loose that discrimination.
In fact, the more I think about this method, because of the FOV of the lens, it will have a tendency to back focus the lens since the focus plane is curved.. As you move the center focus point across and past the target varying the MA, there is an annular ring on the target that expands from a point into an annular ring that expands radially outward until nothing is in focus. The annular ring of focus will give a larger files size than the correct single point in focus at the center. Hence the file size distribution would be skewed favoring a back-focus. It may not be very significant though.
BTW, this file size can be done with RAW files as well as jpeg. The RAW size also varies with detail level.
tdodd wrote:
In the meantime a third party automated solution would be great, but if Canon has any clue at all I imagine it would have a short lived value, beyond the current generation of cameras.
Yes, Canon has a clear record of bringing sensible features to their cameras first.
Heard this one already months ago.
Too many variants IMO. 1 variant changes and the file size changes with it.
Also a lot of steps left out from the other version I read months ago.
What about zooms?
Where do you start infinity?
You need a very controlled envirnment to even consider this right down to having a voltage regulator on your lighting, up one volt or down your lighting changed and the file size with it.
For me I use a different technique and the mark1 mod0 eyeball for image comparison. The longest it took me was 5 minutes on a particular lens.
Sorry for the delayed response. I have been on the road recently.
Gene has given a very nice explanation to your question and has also raised a very good point I had failed to considered.
In an earlier post I recommended:
An ideal target is black on a white background, has something good to autofocus on at the center, and includes plenty of detail in the central area.
Based on Gene's observations I would modify this to suggest that all the detail be constrained to the central area. This should ensure that no back focus bias is introduced.
Justin,
Sorry this method did not work for you and I am happy you have developed your own technique that is fast and reliable. As far as your concerns about a very controlled environment go, I can say that I have had little difficulty achieving repeatable results using the same care I would use to light a portrait sitting.
). I am not sure about keeping the target perpendicular to the axis of the lens though... I wonder how I pull that one off?