This is my first published test using the official ISO 12233 test chart.
Test procedure:
ISO 12233 chart printed on A2 and shot to fill the screen.
Well, I fail to fill the screen and a fill-factor is meant to define to which degree I utilized the available image height (as the ISO scale is applicable only when the frame is filled 100%).
Fill-factor with the DA40 test shots: 84.3%
I embedded smaller (1/4 size) ISO 12233 charts
into the large one. Therefore, the scale figures for the small charts must be multiplied by 4 to be comparable to the figures of the big chart.
Therefore, the resolution figure in
LW/PH (line width per picture height = 24mm * 2 * lp/mm) is as follows (if imprinted "number" is taken from the small charts):
LW/PH = 400 * number / fill-factor
The printed chart has clear lines up to ~4000 LW/PH (the number "10" in the small charts) where a light printing Moiré pattern starts to set in.
Film (Adox 20 CMS Pro) and scanning procedure should be appropriate for up to ~4000 LW/PH as well (cf. above).
Test shots (red line marks the APS-C image circle):
--
f/2.8:
--
f/5.6:
--
f/11: (click onto images for full scale 5062 dpi images)
Numerical evaluation (first figures: as read from small charts) (second figures: in LW/PH):
Center (f2.8 / f5.6 / f11): (6.2/6.2/7.0) (
2900/2900/3300)
APS-C corner (f2.8 / f5.6 / f11): (4.3/6.0/6.0) (
2000/2800/2800)
FF corner (f2.8 / f5.6 / f11): (<2/<6/6) (
<900/<2800/2800)
FF vignetting (f2.8 / f5.6 / f11): (
+/++/++)
Vignetting scale:
+++: not visible
++: visible but not distracting
+: visible and distracting, but can be compensated
o: corners too dark to compensate, but no clipping
-: clipping
The vignetting in the test shots looks more pronounced than it actually is. This is due to the very steep contrast curve of the Adox B&W film.
I will shortly compare figures to photozone.de figures:
Center (f2.8 / f5.6 / f11): (3050/3480/3080)
APS-C corner (f2.8 / f5.6 / f11): (3010/3280/2850)
(Note that figures have been multiplied by 1.53=24.0mm/15.7mm)
There is no reason why both figures should be directly comparable. We have seen already that with a K20D we expect figures to be better. With a K10D, we have to take into account that the finite sensor resolution sets a hard limit well below 4000 (at 3966 exactly).
Nevertheless, we fail to see the better center resolution at f/5.6. Which lets me to believe that accurate focus wasn't reached (always difficult w/o chimming). Also, there is considerable room for interpretation of resolution achieved. Still, for our overall verdict this should be no issue.
Overall rating scale:
+++: no significant difference between APS-C and FF corners
++: difference but usable at full aperture with a small loss in corner IQ only
+: usable staring at f/4~f/5.6, with no or a small loss in corner IQ only
o: usable at f/11 or with big loss in corner IQ only
-: not usable
Final verdict:
+
Usable at f/2.8 with a big loss in corner IQ. Usable starting at f/4~f/5.6, with a small loss in corner IQ.