While shooting some resolution tests with several cameras I noticed something odd about my VX-2000 which I have not read about elsewhere.

I think that the CCD's are not quite in the 4:3 proportion! Now the difference is not much, but it is measurable, and the result is that things will look slightly taller and thinner than reality (which many of us might actually appreciate ;-) It seems like the image is ~17 pixels too skinny.

I printed up the EIA 1956 resolution chart at 16" X 12" and photographed it with my VX-2000, PDX-10 and Nikon 5700 still camera. The resulting images can be seen here:

The Nikon: http://www.greenmist.com/dv/res/b1.JPG
The PDX-10: http://www.greenmist.com/dv/res/d1.JPG
The VX-2000: http://www.greenmist.com/dv/res/d2.JPG

Now these tests were as accurate as I could make them, and I tried to frame each shot so it filled the frame vertically. The nikon image was cropped to fit the chart and resized to 720x480, so perhaps it isn't as valid a comparison. But if you overlay the PDX-10 image on the VX-2000 you'll see that the height is pretty darn close, within a pixel or two I think. But the width is around 17 pixels different.

Now this got me thinking, so I measured the printed chart closely and it isn't exactly 16x12, more lke 15.97" x 12.06". This still isn't enough to account for the difference I see in the framegrabs, and of course it should be the same from both cameras.

So none of this is earth-shaking news I guess, but it was a bit of a surprise.

You did a frame grab on the footage from each camera and then measured the full image size in something like Photoshop?

I just did a screen grab from an old AVI file and it was not a full 720 wide, it was short by about 6 pixels (black lines on both sides of the image).

I'll have to try that with footage from my PD-150.

Yes, I compared the frames in photoshop, but maybe I wasn't clear. The VX-2000 (and presumably PD-150) do capture the full 720 horizontal pixels, I didn't get any black lines. The problem was that the proportions of the image were not geometrically correct. A 4:3 rectangle should exactly fill the frame. Instead, if you photograph such a rectangle and make the height fit exactly in the frame, you will have some extra image to the right and left. So if you photographed a circle it would be more of an oval, slightly taller than it is wide.

DV pixels are not rectangular. Did you take that into account?

Of course. The point is that the proportions of the VX-2000 image are different from the PDX-10, and it appears that the PDX-10 is correct. A picture is worth a thousand words....

http://greenmist.com/dv/vx2000/aspect.JPG

I have overlaid the images from the two cameras such that the pink quadrants are from the VX-2000 and the green are from the PDX-10. Notice the edge of the chart - it extends all the way to the left side of the frame on the PDX-10 but the VX-2000 can see beyond the edge. On the right side the difference is even more pronounced. Both cameras were put on the same tripod without moving it and were aligned/zoomed so the chart filled the frame vertically. The images are slightly off vertically, I guess the zoom settings don't quite match, but this is only a matter of about 2 pixels. In terms of width they just don't match at all; the difference is about 17 pixels.

Boyd I would advise to test again in tele setting. Small lens distortion differences (barrel and pincushion) can result in that kind of apparent aspect ratio deviations if the basic aspect ratio's are not 1/1. Tele settings seldon show these non linearities. This is a big issue in dual camera 3D calibration.

I was zoomed about halfway in. You may very well be right... although would this sort of distortion be anamorphic (only in the horizontal dimension)? And regardless, my point was that mounted on the same tripod at the same distance from the chart the PDX-10 and VX-2000 show differently proportioned images. It's not a big problem or anything and is probably not noticeable 95% of the time. But it was just sort of a surprise to me.

These distortions are concentric, but 4:3 images use a larger lens angle in the horizontal direction than in the vertical direction this ends up in transforming concentric distortions in aspect ratio distortions

Fascinating Boyd, and good work! I've often bemoaned the fact that the side-screens of many cameras aren't 4:3 rectangles either, and the TRV900's is most noticeably more 'wide-screen'
than the 4:3 it should be.

But your revelations are an eye-opener for me too, though Andre's observation that lens distortions (with possible anamorphic overtones) is an interesting one. I was interested to see your intercut lens testing chart on my pc, but now's as good a time to ask - how do the two camera's compare for overall resolution?

With the VX2k's much more sensitive chips, was the chart shot at differebnt apertures in the same light or did you have the PDX10 in auto and use the VX's NDs?

tom.