I've been doing some experiments using my legacy M42 Takumars on my K10D, and am confused by my results to date. I am hoping someone here can either explain my results, or tell me what I am doing wrong!
First, the set-up. I'm using a small light box as a "constant" source; I measured the light box brightness using a Pentax 1/21 spotmeter and found it to be a pretty uniform 11-2/3 (11.67) EV. I mount one of the Takumars (35/3.5, 50/1.4, 85/1.8 or 105/2.8) on the K10D using a Pentax K Mount adapter. Using a tripod, and with the lens set to infinity, I position the lens
against the face of the light box. I set the camera to ISO 100, M-mode, center-weighted metering, EV steps set to 1/3, lens set to M (manual aperture). With the eyepiece covered, I step the lens through it's aperture range, pressing the green button 3-4 times for each aperture and record the indicated shutter speed.
Second, data reduction and analysis. For each Av,Tv pair I compute the equivalent Exposure Value (EV) using the standard formula {EV=log2(f*f/time/(ISO/100))}. I then graph EV versus Aperture for each lens.
Looking at the graphs is where things get confusing! My expectation, given a constant EV source, is that I'd get more or less horizontal lines. My results are anything but! Each lens produces a falling EV curve as the aperture is stopped down. The 50/1.4 and 85/1.8 lenses have an extra quirk in that the EV curve initially goes up, before following the downward trend after about f/2.8. The first attachment is a composite graph for the four lenses. Note that it appears from visual inspection that the 50/1.4 does not actually close down to f/16 (no change in the iris when the ring is moved from f/11 to f/16), so the upswing in its curve at f/16 is bogus.
To compound my confusion, I repeated the tests using three KA-mount lenses - a FA 50/1.4, a FA 100/3.5, and an A 50/1.7 - and obtained similar results, as shown in the second attached graph.
Next, I decided to see what I would get from the Takumars if I mounted them on my ZX-5n. Using the same set-up and procedure, I took data on the same four lenses. When graphed, these data were much more in line with my expectations as you can see by examining the third graph...
I was a little surprised that there was a definite spread in the detected EV across the lenses, but examination of the lenses suggested one possibility: the 50/1.4, which recorded the lowest EVs, seems to have a yellowish tint, while the other lenses seemed more neutral.
I made one final graph in an attempt to take out these possible 'transmission' effects. I plotted the difference in indicated EV for a particular lens/aperture between the K10D and the ZX-5n. This should leave only the difference in detected EV between the K10D and the ZX-5n metering systems. This graph is the last of the attachments.
As you can see, now all four lenses show a reasonably consistent behavior. Unfortunately the consistent behavior suggests that the K10D manual metering mode using the green button with legacy lenses, whether KA lenses set to manual apertures, or M42 lenses with manually set apertures, is not particularly accurate. Granted this is not a very real world test, but it is fairly well controlled, or at least I believe it to be.
So, where have I gone wrong? Or, if I've got it right, why would Pentax design the system this way? Inquiring minds want to know!!!