Ray, Roland, et al, I have a hypothesis but no K-5 or K-7 so will rely on you to check this out:
Rather than the expense of a CA corrected AF light chain, which likely wouldn't be perfect anyway, it seems that these two cameras use the "+" part of Safox to measure the White Balance (WB) and do a Auto Focus (AF) correction based on that. This works well in bright light where an an accurate measurement of WB can be made, even for bright tungsten; however, it was forgotten that in dim light there isn't really anything of colour to measure: That's why our eye's response in dim light depends on black and white receptors (which are also more sensitive).
In the above case, Pentax could do one of two things, try to continue to use the WB adjustments even though there is a large margin of error but try to design that error to drive the front focus to be no worse than if "+" was not there, or if the errors are larger than what is trying to be corrected for, just give up and don't correct. In either case, even for daylight type of lighting, due to having to allow for the balances between the RGB sensors, it is likely that the green channel is the strongest followed by the red and blue. As the light level drops, especially for the "warrmer" indoor lighting, it is likely the blue level that will get fuzziest to the point of not being able to be discerned followed very slightly later by the red channel. At that point, there is no way of determining WB accurately and it likely occurs about a stop or two above the minimum metering/AF capability for the maximum aperture of the lens used since AF is wide open . For a f/1.4 lens, that would be about LV 0 to 2, for a f/2.8 lens about 2 to 4, and for a f/4 lens about 3 to 5. Just for reference, a f/2.8 lens at ISO 3200 is 1/16 second expsoure for LV = 2.
If I'm right, there isn't much that Pentax can do about accurate AF in low light, as anything they do would be a compromise. They could assume that low light might be tungsten and do a partial correction, but then it might be just low light fluorescent or halogen and the guess would be wrong. The most they could do in this case is use the WB preset as a hint as to how much correction to apply since it can't be measured. However, if the conjecture that low light daylight illumination is also wrong, then they could use a model to predict how AF distance varies with light level and apply that instead of or on top of the normal "+" WB correction.
BTW, my K200D also front focuses for very dim light as described above (f/4 lens at f/4 ISO 1600 1/8 second) which as at about 4000 degrees Kelvin and is fine in brighter light. This also makes sense as the phase detector picks up more and more reddish rather than bluish tones at dimmer light levels, even though it is monotone sensitive.
As a thought, has someone done a similar test for a "Gold AF Standard" D700 or D300(s)?
Regards, GordonBGood
