When HDTV specifications were developed about twenty years ago, the broadcast and production communities sought to improve the capabilities of television systems. Unfortunately, with every advance comes yet another set of problems, as you will see here.

Aldo Cugnini
Analyst

Central to the story is the change in color space from CCIR-601, which specified 525- and 625-line systems, to ITU-R BT.709, which defines 1125- and 1250-line HDTV systems. The former used SMPTE-C color space, which accounted for the reproduction characteristics of CRT displays; the latter color space was described in SMPTE-177. After some 50 years, the basic equation (Y = 0.29R + 0.59G + 0.11B) changed (to Y = 0.21R + 0.72G + 0.07B) for 1125-line systems. The reason for this is that the CIE x,y coordinates for the primary colors had changed (i.e., the color space changed), and so the equations required modification in order to keep the same D65 white point.

So, why the change in primary colors? Acknowledging the previous deficiencies in video color reproduction, one would think that it was to represent colors better in the new HDTV system. But that isn’t the case. I recently had various communications with several individuals regarding this evolution, and according to one individual, the change in color space was done to accommodate LCD displays. Those of you watching TV since the 60s (I’m dating myself) will immediately recognize the fallacy in that argument. SMPTE-177 and ITU-709 were published in 1993, predating the practical use of LCD monitors as video displays. (It also appears that this story has been propagated by some monitor manufacturers.)

In fact, the change had to do with the problem of constant luminance, wherein the color difference signals contribute to the luminance signal. In an ideal television system, we should mix linear RGB at the encoding side (the "matrix") to generate the luminance and chrominance signals. However, in practical systems, the camera pickup (originally an imaging tube) imposes a non-linear ("gamma") transfer characteristic on the RGB signals before the matrix. Because the color encoding band limits the color-difference signals, this incorrect order results in errors in the displayed brightness of high-frequency signals. This error can usually be noticed on color edges, especially between different colors.

ITU-709 introduces a smaller error than ITU-601, but it is still there. LCD monitors require a different gamma correction, so there is a potential for improving the situation - but this would require a hardware-intensive 3×3 matrix multiplication for each pixel, and most monitor manufacturers would just as soon leave this out. Compounding the situation is the fact that the color space is different for HD and SD video - and that means that a correction should be made when converting from one format to another. If the scan converter takes this into account, then further errors will be minimized, but if it doesn’t, then we just have to live with the results.

As for the color gamuts themselves, there’s very little practical difference in the extent of the colors that can be reproduced by SMPTE-C vs. ITU-709. One would think a better evolution would have been toward a considerably wider gamut, such as that specified in the Digital Cinema Initiative (DCI) - or even the original NTSC color gamut that, ironically, was never used.

Reply to the author