Not long after posting these images, I received some constructive feedback with respect to the aspect ratio of the Virtual Televisor images. Two areas of concern were noted:
I will return to the first point after I discuss the second. First, it will not do to deal with aspect ratio in terms of pixel dimensions. Given all the complexities of pixel dimensions, pixel shape in different screen modes, and sampling strategies, the only real test is how the images look. The aspect ratio for NBTVA 32-line images should be 3:2 (V:H), or a ratio of 1.5. I measured numerous VTV images as posted on this website, and the average size was 56 mm (H) by 38 mm (W) - a ratio of 1.47! I thought this was pretty good, but I learned that the 3:2 ratio includes the blanked sync interval, which I typically do not display. Well, since that would add 8 more pixels to the 120 I do display, the height of the images would increase to a hypothetical 59.7 mm, or a ratio of 1.57. Either way, the aspect ratio, based on image display dimensions (not pixels) is as close as anyone (a variance of 2-4%) could ask to the 3:2 standard, no matter how computed.
Just for the record, Gary Millard's NBTVView program also does quite well. The average size of the screen-capture images I have is 54 x 38 mm or a ratio of 1.42. Throw in the sync interval which I typically edit out and we get a ratio of 1.52.
Aspect Ratio Based on Displayed Geometric Fidelity
A second way to look at the fidelity of the aspect ratio is to look at the fidelity of the display with respect to geometric shapes. CD#3 has three test patterns that any reasonable person would suggest should be circles - the count-down clock face, the NBTV test pattern, and the "eye". If these images were properly formatted to begin with, a televisor with the proper aspect ratio would display them as circles. A circle has an eccentricity of 0%, so how do these three circles come out when displayed using the VTV software?
Pattern Eccentricity (%)
Clock
2%
NBTV
3%
"Eye"
8%
Both the count-down clock and the NBTV pattern are very close to circles and certainly better than specifications that one might expect from a modern consumer television set or computer monitor. They are also within the 2-4% variance predicted on the basis of display dimensions as discussed in the previous paragraph. If I had to guess, I would say that both these images were computer generated and properly formatted. The "Eye" is further out of round with a slight "vertical stretch". Given the comparison of the two disks which follows, I would predict that the "Eye" may have been done with a live camera and subjected to the same processing as the other "live" image on CD#3. In any case, by any measure, the displayed aspect ratio of the Virtual Televisor images is well with any reasonable standard you are likely to encounter and probably superior to any mechanical display system without the most exacting fabrication standards. (See the section at the end of this page for the impact of fabrication errors on image aspect ratio).
A Comparison of NBTVA CD #2 and CD #3
Even a superficial look at CD #3 suggests that the individuals portrayed are "stretched" vertically in comparison with the images from CD #2. The effect is even more pronounced if you compare specific individuals from the two disks. Since the same software was used to prepare the images and none of the images were altered in any way, it is clear that the two CDs are not the same. CD #3 had a source aspect ratio less than CD #2 or vice versa. Given the fact that I don't know the individuals personally, I can only say, based on several decades in image communications, that the images on the second disk look a bit more "natural" than those from CD #3. The point is, they are not the same and one or both must be off with respect to aspect ratio. Since both disks were edited by the same individual, it is reasonable to dismiss the idea of some differential in the editing process. It seems most likely that there must be some difference in the way the standards (scan) conversion was done in the two cases.
The important point is that you should not expect the images to look the same, no matter what you use for display and no matter how accurate or inaccurate your displayed aspect ratio. Remember, all things considered, the problems of this sort that we are likely to face today pale into insignificance next to the technical problems experimenters faced during the mechanical television era!
Aspect Ratio and Mechanical Televisors
With respect to aspect ratio, the most problematic mechanical display system is one using the classic Nipkow disc. Let's assume we wish to configure a Nipkow televisor so we will have an image that is 1.2 inches high. Since there are 32 scanning lines and, with the vertical NBTVA format, the image height is equal to the distance between holes, our hypothetical disc would have a circumference, at the outermost hole in the spiral of 1.20 x 32 or 38.4 inches. Take this circumference and divide in by Pi (3.1416) and the have a diameter of 12.22 inches or a radius of 6.11 inches. A disk of approximately 12.5 inches in diameter would do nicely and, the right edge of the picture would have a height of 1.2 inches.
Given the 3:2 aspect ratio of the NBTVA format, the image width should be 0.80 inches. Thus each of the 32 holes should be located 0.026 inches inboard of the previous hole (0.8/31). Lest you think we now have a simple display window with a 3:2 aspect ratio, now look at the track for the left-most or last image line. Since it is located 0.80 inches inboard of the first one, it is located on a radius of 5.31 inches (6.11-0.8) or a diameter of 10.62 inches. The circumference with a diameter of 10.62 inches is 33.36 inches. The height of the inner side must thus be 33.36/32 or 1.043 inches! Instead of the 1.5 ratio we would expect with a 3:2 aspect ratio, relative to the left (inner) side, the ratio is 1.043/0.8 or 1.30 - a full 8.7% error! Of course none of this ignores the additional complexities:
In addition to these issues, which are unique to the Nipkow disc, there are other problems that are especially acute when dealing with small display formats. For example, our 0.8 inch calculation was based on the center to center line spacing. If we assume the scanning holes must have a diameter of about 0.025 inches (remember that they are spaced 0.026 inches), the total width of the image is actually 0.825, for a ratio of 1.45 or about a 3% error! In a similar vein, very small variations in measurements, drilling, or other aspects of fabrication can further increase the errors.