Sony DCR-HC40E CCD Coverage Test
Sunday 4 April 2004

The purpose of this test is to determine the camcorder's CCD coverage under different modes (normal & 16:9 for video, 1152x864 & 640x480 for stills) and get some insight into how much of the CCD is actually used in each case. The test was motivated by a discussion in the Sony HC20/30/40 forum at www.camcorderinfo.com (see the relevant thread for more info - my nick is DrDV).

The equipment & setup:

--- Sony DCR-HC40E (Note: This is the European PAL version, hence the "E" at the end).
--- Custom-made pattern board (made out of four A4 pages printed on a laser printer).
--- Tripod.

The pattern board consists of a number of squares measuring exactly 2cm x 2cm each, arranged in a chess pattern for ease of viewing/counting. I've also added a circle with a diameter of exactly 14cm. This is a perfect circle, not an ellipse (useful for checking geometric distortions, etc.). The camcorder's zoom was set to full wide (W) and all parameters were set to AUTO. The camcorder was placed at a distance of about 60-65cm from the pattern board, and its position was fixed throughout the test. Please note that the room was lit using incandescent lamps and hence the images appear yellowish and a bit dim.

Result #1 --- Normal Video Mode (16:9 WIDE Mode = OFF):

This is an actual frame grab from the resulting DV AVI video file (image size: 720 x 576).

OK, this looks as expected. Now, let's try to determine the CCD area that was used by the HC40. It's difficult to determine the absolute CCD area without knowing the internal structure of the camcorder optics and CCD. However, for comparison purposes, we can determine the relative CCD coverage in terms of number of squares. The good thing about this test is that it doesn't really matter how far away the camcorder is from the pattern board, or how big or small the pattern board actually is. As long as all distances are kept constant throughout the test, we can express the relative CCD coverage in terms of the number of squares by simply counting the squares in the above frame.

We can see that the frame consists of just under 22 full squares horizontally and exactly 16 full squares vertically. So, the total number of squares captured is 22 x 16 = just under 352 squares.

Result #2 --- 16:9 Wide-Screen Video Mode (16:9 WIDE Mode = ON):

This is an actual frame grab from the resulting DV AVI video file (image size: 720 x 576).

In this case we can see that the image is distorted geometrically since a 16:9 frame must be squeezed into the same 720 x 576 image. This process is known as "anamorphic" and will display correctly on a 16:9-capable device, such as a wide-screen TV. Let's count the squares again. Horizontally we have 24 full squares and two halves at the sides, so we count this as 25 squares horizontally. Vertically we have exactly 14 full squares. Comparing with the previous result we see that there is more information horizontally (wider angle), but less vertically. The interesting thing is that the total number of squares in this case is 25 x 14 = 350 squares! Same as the previous result! This means that the HC40's specification of 690000 CCD pixels for video holds true regardless of what mode is being used. In other words, in 16:9 mode there is more information horizontally, less vertically, but the total amount of information remains constant at 690000 light measurement units (CCD pixels).

Some calculations:

I think we can safely assume that the CCD surface is uniformly covered by CCD pixels (i.e. the CCD pixel density remains constant over the entire CCD surface). Given that the HC40 uses 690000 CCD pixels for video, and these are mapped to 350 squares in the above pattern, we can calculate the amount of CCD pixels that correspond to one square:

CCD pixels/square = 690000 / 350 = 1971 CCD pixels per square. This will be useful later on when we analyse the resolution of the still images.

Result #3 --- High Resolution Still Image (1152 x 864):

This is the actual JPG file saved on the camcorder's Memory Stick.

In this case we have just over 26 squares horizontally and just under 20 squares vertically. Total number of squares: 26 x 20 = 520 squares. Expressing this in terms of CCD pixels using our previous calculation, we get 1971 x 520 = 1024920 CCD pixels. This is a very good match to the 1 megapixel advertised resolution in this mode.

Result #4 --- Low Resolution Still Image (640 x 480):

This is the actual JPG file saved on the camcorder's Memory Stick.

This is a very interesting result. If you compare this with the high resolution image, you'll see that they are identical in terms of CCD coverage. They only differ in resolution. Another very interesting observation is that the 640 x 480 image appears sharper and cleaner than the 1152x864 one. It looks like the 640 x 480 image is actually a down-sampled version of the full 1152x864 image. This is good news for the HC40 because it means that the 640 x 480 images taken with the HC40 will be superior to those taken with say the HC20 or HC30. Even though they are all 640 x 480, the HC40 actually takes a 1152x864 image and then down-samples it to 640 x 480, thus resulting in a more detailed image (not in terms of pixel count, but in terms or sharpness and overall quality). This is just a speculation - the only way to verify this is to do a side-by-side comparison between the HC20/30 and HC40.

That's it! I hope these results are useful in understanding some of what's happening inside the HC40. I wrote this page in a hurry and I know it doesn't look very nice. Also, the interpretation of the results given above, although correct to the best of my knowledge, may still be flawed or wrong - please email me if you find any errors, etc.

Many thanks to the authors of WinDV and VirtualDub for creating these wonderful and free tools.

Written by Costas Vlachos --- Last updated: Monday 5 April 2004