Effect of IR on focus

Discussion in 'Canon EOS' started by bobatkins, Feb 28, 2016.

  1. While making some focus micro-adjustment measurements I came across something I hadn't expected. It seems that the amount of IR light present during autofocus operation can affect the focus position. More IR shifts the focus forward, less IR shifts the focus backward. So if you setup your focus micro-adjustment indoors using LED or Fluorescent lighting, It won't be accurate for tungsten or daylight.
    Why the IR sensors don't have a strong IR blocking filter over them I don't know, but they clearly don't. In fact AF works quite well with a Hoya R72 IR-pass filter over the lens, as long as the light source emits some IR. You'd think AF sensors would have filters over them that made them "blind" to IR (and UV) light, just like the imaging sensor is. However that doesn't seem to be the case. They probably do have IR blocking filters of some sort, just not very strong ones.
    So if trying to get consistent focus micro-microadjustments is driving you crazy, make sure you're doing the calibration in the same light that you will be shooting in.
    I tested this with a 7D, 70D and 6D and with 50/1.8 and 85/1.8 lenses, and all showed the same effect.
    I've documented the technical details here - http://bobatkins.com/photography/technical/focus_vs_light_source.html
  2. Are you sure that the AF system works to spec with an IR blocking filter in place? ie. Can the 6D can achieve focus properly at -3EV? Likely in plenty of light they work fine, but at the limits (-0.5EV, and -3EV), do the systems rely on the IR component to achieve focus?
  3. All the tests were one in good light. The B+W UV/IR cut filter has a filter factor of less then 1/3 stop. Transmission in the visible is >90% over most of the visible light region.
    The change in focus is repeatable and happens in both direct sunlight (PLENTY of light) and at lower light levels. It doesn't correlate with light level. At lower light levels I'm guessing that focus error on a shot to shot basis probably gets larget, but there's no overall shift in focus on average.
  4. I noticed some color temperature dependent focus error in the Nikon D800; if I set the fine tuning on a 200/2 for daylight, it would not be correct when photographing in coloured tungsten lights for example, and depending on the colour of the gel, the focus offset would change. This has been fixed in newer camera models such as the D810. I haven't heard of Canon users with similar problems but I guess it may depend on the exact body type. Filtration and sensitivity in low light can be conflicting requirements, if you put in strong filters with sharp transitions, the passband is also attenuated more, typically. So if the aim is to detect as much light as possible in low light (when implementing the D800 focus system increasing the sensitivity in low light was one of the new features) then there may be a tradeoff in accuracy due to sensitivity to the colour of the light. But I guess this is just something that the manufacturers will have to figure out.
  5. Silicon goes to about 1100nm. Sharp IR block filters are hard to make, while IR pass filters are much easier.
    That said, I don't know much about how they actually build AF systems.
    I never use the AF assist light on any of my cameras. I believe that some use an IR or red/IR lamp for the AF assist. Again, with a Si sensor that is pretty easy to do.
  6. Sharp IR blocking filters are complex, but the technology for making them is well established. Every imaging sensor has one in front of it. The sharp cutoff is obtained by using a multilayer dielectric filter, with perhaps as many as 30 layers of different thickness. They work on an interference principle rather than bulk material absorption.
    However, you make a good point about the IR AF assist feature of some speedlites. That would explain why Canon allow the AF sensors to be IR sensitive. Of course focus for IR is different than focus for visible light (which is why many lenses have a separate IR focusing mark - usually a red dot), but I guess any focus is better than no focus and most people would never notice that that focus using IR AF assist was slightly off.
  7. That would certainly be an excellent reason to not filter out all IR light.
    ...and, if the camera knew that IR assist was being used, it would be child's play to offset the focal distance based on both the subject distance and the intensity of the IR emission (as a percentage of the visible light). In most cases where I use the IR assist, it is the dominant source of light for the PD AF. I have never found focus achieved with IR assist to be dramatically off, and I usually am shooting WO or near when I use it. Of course IR assist still uses visible light (red light) Though obviously I can't see the IR portion of the emmission, and don't know what portion is NIR, FIR, and visible...
  8. It would indeed be possible to offset IR focus effects (if present) as long as the camera knew the IR characteristics of the lens in use. That information might well be present in the lens's ROM which can be read by the camera, at least for Canon branded lenses.
    Overall the shift would be small. Probably observable as a sharpness issue on flat targets when a fast lens was used wide open - though that depends to some extent on how much IR is present in the flash output.
    I think the place where this IR effect on focus is most likely to be and issue (and where I noticed it) is in setting focus microadjustment. If you set it using a light source rich in IR and shoot (or retest focus microadjectment) under a light source that doesn't have any IR, you can clearly see sharpness suffer.
    In most other situations either the use of a smaller aperture will expand the DOF to over small focus issues or the focus offset on 3-D subjects may make the be lost because some part of the subject will indeed be in focus.

Share This Page