Originally posted by Ira I'm too stupid to understand this.
I will try to explain it as it has been a hotly discussed issue over the past 3 years.
When the K10D came out, there were a lot of issues relating to exposure, both with Manual aperture lenses and with teleconverters.
What it all came down to in the end, and is the origin of the chart shown here, is the focusing screen used in the K10D and K20D. These screens seem to have some odd characteristics when it comes to how light reflects off the screne to the metering sensors.
This characteristic is somewhat all over the map, but is consistant and repeatable as a function of the aperture.
This was ultimately proven out by swapping the focusing screen our of an *istD and putting it into a K10D (they are interchangeable) and being able to duplicate the *istD's metering behavior in the K10D.
Since that time, I have carefully checked most of my lenses for exposure accuracy using a uniform surface as the test subject.
a uniformly lit uniform color surface should have a greyscale histogram value of about 110-120 (the histogram ranges from 0-255)
Through other testing, I established that for all camera settings set to neutral, a change of 1 stop (by either stopping down a lens or by changing shutter speed) would result in a change of about 45 greyscale.
So basically, if you consider a line through 120 as the correct exposure, values above 120 are over exposure and those below 120 are under exposure, AND when you get to a value of 165 you are at +1 stop and when you get to a value of 75 you are at -1 stop.
The attached chart shows the exposure accuracy of my SMC P 50mm F1.4 on each of my 3 bodies with the origonal screens, and 2 additional curves for the K10D, one curve showing the K10D with the focusing screen from the *istD, and the other showing the K10D with my Chineese Diagonal split screen installed.
essentually the curve shows that with a K10D and stock screen, and the SMC-P 50mm F1.4 I can have under exposure by a little over a stop at F2, and over exposure by almost 2 stops at F9.6. The behavior is predictable from day tyo day, etc, but unless you know the behavior, the general observation would be to conclude as others in this thread have posted, that exposure is all over the map.
The K7 takes about 1/2 stop off the under exposure, and almost a full stop off the over exposure and therefore could be considered as metering more accurately, although it is still not perfect.
The *istD stays within +/- 1/2 stop over the entire aperture range, which is why I conisder it to be the best metering camera out there for manual lenses.