During my latest venture into lens-test-land (see e.g. here
) I encountered the obvious problem of shooting a standard-target to judge lens performance. I experimented with two targets: The Siemens-star
and patterns on a monitor.
Some first tests with monitor-patterns easily and immediately revealed geometric distortions
, light fall-off
Look at this example (clickable for larger views!):
Let's see whether these results hold under closer scrutinity:
Well, let's kick the first out: Light fall-off
! With a shot from an LCD-monitor you're more likely to record differences in lighting of the minotor than the lens. Especially when you're shooting 30cm away from the monitor with a wide-angle lens, the angle of view from the camera to the edges of the LCD-monitor will produce visible light fall-off. So forget about this
Ok, on to the easy one: Geometric distortions
. That is clear as the sky that an absolutely flat LCD-Monitor with a grid on it is a perfect
target to test the lens. Easy!
Now to the last point: Well, if you see differences in centre- vs corner-sharpness, that is valid. Only thing you have to make absoultely sure is that the camera is aligned parallel to the monitor, as any misalignments will result in one corner (or the other) falling out of the very limited dof
that you have when shooting a 21" monitor "full-screen". That is a magnification of about 1:20 and dof is around 1.2mm (!)
(for f4.0 and 25cm viewing-distance of an A3-print) in this case.
But what about pixel-pitch on the monitor and pixel-pitch of the sensor and projected pixel-pitch of the monitor on the sensor? Well, here's a little math: My monitor has a pixel-pitch (pp) of around 100dpi, my camera's sensor a pp of 4100dpi. So long as you're at a magnification of 1:10 (projected monitor-res = 1000dpi) you end up with 16-18 sensor pixels peeping at one full color monitor-pixel (consisting of red/green/blue subpixels). As you need 4 sensor pixels to analyse all colors (Bayer-matrix) that's a comfortable ratio to say the resolution of the sensor outperforms the projected monitor resolution and thus you should be able to "see" all monitor pixels perfectly sharp - if
you have a perfect lens! So any deterioration in perfect sharpness of a screenshot at 1:10 can be attributed to the IQ of the lens!
At 1:22 you're still at a ratio of 4 sensor-pixels per projected full-color monitor-pixel. That is still pretty safe and gives you a full view of the 24"monitor for a camera with a 1.1" APS-C sensor.
If you shoot at a magnification of 1:100 (camera at 100xfocal length from the monitor) the circumstances turn around. In this case you've got 5-6 full-color monitor-pixels projected on each sensor-pixel. So it is safe to say that the camera cannot see any single monitor-pixels any longer. So tests with single-line-width structures on a monitor are irrelevant from this distance.
Thomas (beware: Nikon-fanboy and moderator!) My Lens Reviews
, My Pictures
, My Photography Blog