Originally posted by Anvh What kind of scale did you use there?
And very strange it's different then what i got with the K10D...
I am somewhat suprised you get different resuilts, however, lets go through step by step, because blende8 was able to reproduce exactly the same behavior with his camera, lenses and measurement of greyscale. the lines with his name are using data he posted in a thread discussing this issue, so I added them to my graph. Note also, I have similar flat line results using y *istD and Sigma APO 70-200 F2.8 EX, and have also measured a slightly increasing over exposure but also a streight line, using my tamron 28-75F2.8 so I know th enethod works.
So for the test, I use either a paved road, sidewalk, or cement block wall. THe surface is uniform shading and illumination for 100% of the frame.
I use spot metering if there is no installed split image finder, but average metering if there is a split image finder, regardless the results seem the same.
I use .5 EV steps because manual lenses have detents in either .5 or 1 stop increments
I set the aperture and let the green button pick shutter speed
I measure the greyscale in the center 10% of the frame, so that lens vignetting does not impact the results (same reason I use spot metering when possible.
I shoot JPEG with neutral contrast and saturation, with highlight and shadow protection off (as these all influence things) Shooting RAW can be misleading unless your raw editor is programmed to use the image jpeg defaults (perhaps a point of error in your test)
the photo editor gives the greyscale in 8 bit resolution, and i did linearity tests by going from a completely black (greyscale 0) to completely white (greyscale 255) in 1/3 stop increments, and from this I know that from about 25-225 greyscale each stop is about a change in greyscale value of about 45. the last 25 or so greyscale at each end of the exposure contains about 3 stops of exposure, but it is very non linear, with the last full stop only having a change of 3, the next 7 and the next 15, before the linear portion of the exposure curve.
I plot greyscale on the vertical, and aperture on the hotizontal, but let the horizontal axis be a log scale, so that there is equal space between full stops.
can you post some data, and make sure that ev comp is zeroed.
also watch out for back light entering the viewfinder and influencing the data