Originally posted by mrechte I will not trust the K20D light meter anymore, and I start wondering whether I made the good choice with this camera...
@mrechte, maybe, you find the following interesting.
I've recently read an article about spot metering in particular and that one should always do test shots in conjunction. The reason?
Well, it only looks at one spot (it's purpose) and will adjust the exposure as for a gray target (18% gray?). But you never really know how gray the spot really was (you may know for a caucasian face, but other than this...). So, a spot metering almost always is accompanied by EV compensation.
So, I went back to your photos and studied some of them with PhotoME (if you don't already use it -- get it!).
It will actually show all the matrix exposure readings in LV layered on top of the image. It also computes the effective LV which your image was taken at with.
What I found is that with spot metering, the effective LV the image was taken at was consistently about 1.1 LV higher than the spot reading.
So, probably ok and spot metering with a narrow spot is just tough to handle, actually. Personally, I don't use it as it doesn't work with split screens.
As for matrix metering. There is a common sense here that the Pentax tend to underexpose by about -0.5EV, but people normally consider this to be ok and more in line with the rule of the digital era to expose to the right.
But then, maybe your light metering is faulty, indeed.
To check this out, you could meter two face subjects (dark/bright), hand-meter both to know their LV (or spot meter them from close enough), take pictures, and control the matrix metering values via PhotoME.
UPDATE:
I had a look at your recent shots and the exposures can all be explained by spot measure +1.1EV.
Except the last one, IMGP0698.jpg:
This is a matrix exposure, the image is taken at effective exposure of 13.5LV, the matrix fields are all in between 12.0 and 13.0. So, the decision to take the image at 13.5LV is perfect.
Still, the histogram shows that the image is underexposed by a full 1EV. Like the original light values all ranged from 11.5 to 12.5.
Therefore, I am tend to believe that your unit's 16 exposure meters measure by 0.5 stops too low.