[Mister Guy]

Just to completely clarify, if I'm reading this correctly, you are shooting fully manually in natural light with the same lens and different bodies and getting different results. We'd really like you to remove that last variable of natural light, and shoot completely in artificial steady lighting. If possible, it would also be helpful if you would put both bodies on a firmly anchored tripod as well, as well as putting both white balance options. Once you've done that, please upload the original JPGs or even the raw files somewhere so we can verify all the associated information.

Particularly exposure compensation.

[nanok]

a lot of confusion here.

please repeat the test with the two bodies, same lens, same scene, same light (artificial light if needed, to be sure it doesn't change), and same settings (iso, aperture, shutter speed), do it in raw, and process the two raw files with the same settings.

if you were shooting jpeg for this test, i have a hunch your settings for in camera processing might be different on the two bodyes.

if i understand your questions correctly, the above should shed some light on this. oh, and don't strip the exif, please.

[jimby]

Tell me how is the camera doing that if the ISO, aperture and shutter speed are locked by the user ? And they are the same on two cameras ?

From my knowledge the camera is "doing a good job protecting the higlights" only by changing the ISO or/and the aperture or/and the shutter speed - and here is not the case.

If these 3 elements are locked, camera cannot do anything - maybe just use it's hdri function (post process in the camera after the picture is taken).

So how do you think the camera protects highlights? By metering bias, of course. How does one set exposure manually? By using the same meter with the same bias.

The OP was initially complaining about the "bad exposure" of his K20D. We pointed out that the exposure wasn't bad, but was perfectly acceptable. THEN we descending into the complaints about the K20D not shooting with a high enough shutter speed, and not having the same exposure as the K10D, blah blah blah.

Now the fact that the user can't recognize a well-exposed image might lead me to believe that there are other factors contributing to unpredictable results here.

We're on page 4 and still no EXIF data.

The point is that the K20D exposure is perfectly acceptable, and the highlights aren't blown. We don't have nearly enough information, and the user doesn't have nearly enough control over the variables, to draw any serious conclusion about what's going on.