Weekly Watch: Return of the BBC Test Tx

Continued from last week.


The fact that it is almost certainly sourced from a BBC broadcast can be verified with the content of the video. The plate at 3:00 has a small picture in the upper left corner that comes from a famous BBC test card. The little girl is Carole Hersee, the daughter of one of the test engineers, and a photo of her was used for the Beeb's first color test card, transmitted just after the start of color broadcasts in 1967. She was paid £100 for her services, which must have been a fortune to her at the time. Ms. Hersee works as a costume designer these days. She still has the clown doll, and more recently brought it out to recreate her famous photo.

Aside from that, the lettering is all done in a standard teletext font used by UK services like Teletext Ltd and Ceefax. (They all shared the same typeface because they were engineered to use the same decoder, which had an on-chip character generator.) The capital N, with the diagonal bar that spans only the center vertical third of the letter, the high-waisted capital A, and the short horizontal tails on the lowercase e and descenders like g and y are pretty distinctive. It's visually similar to the system font on UK-only 8-bit computer systems like the BBC Micro, and quite different from the fonts used on the equivalent systems used in the US and Japan.

And they spell 'colour' with a u.


The screens of fine black and white lines are a test for false color. Color is transmitted in an analog television signal in a way that is simultaneously remarkably elegant and unbelievably ugly. When color TV was first being developed, one of the major demands on both sides of the Atlantic was that the color signal be compatible with existing black and white sets. (The Brits had already been through a switchover between incompatible units when they changed from broadcasting a 405-line picture to a 625-line picture. It was not fun, and they did not wish to repeat the experience.) A method had to be developed whereby they could wedge color information into the signal without compromising the monochrome picture.

The compromise they came up with involved transmitting the color as patterns of tiny dots. In a monochrome signal, different shades of gray are obtained by lighting up a certain proportion of phosphors in a given area. As long as the individual dots of light are too small to see, it doesn't matter which ones you light, as long as you get the ratio right. Those three tiny grids to the right there are all considered 50% gray, but all of them achieve it with different patterns of dots. A color television has additional circuitry that detects which specific pattern of dots is lit up, and interprets that pattern as a particular color. All of those grids would light up the same shade of gray on a black and white set, but a color set might render the top pattern as 50% luma + red, the middle as 50% luma + blue, and the bottom as 50% luma + green. It gets more complicated than that in practice, obviously, since all sorts of colors occur in combination with all sorts of brightness levels, but that covers the basic idea.

This causes problems on color TVs when some section of the luminance signal has a lot of fine detail. If the pattern is tiny enough, and has a high enough contrast level, the signal starts to look like chroma dots to your television. Johnny Carson used to drive the crew crazy by wearing checked jackets. The pattern was so tiny and so high-contrast that they kept bursting out in false-color rainbows when he moved.

The eye-rending psychedelic plates full of black and white lines in the test transmission are purposely engineered to trigger this. While the problem occurs on both NTSC (US/Japan) and PAL (Europe) signals, the specific dot patterns are different for both, so the test cards are different, and trigger different colors. PAL color is encoded not just in the angle of the pattern but also in the phase of the signal; the hues of the false-color splotches that flicker back and forth are consistent with YUV color space. The cobalt/mustard splotches are inverting with respect to the U axis, and the rose/mint ones are inverting on V. NTSC video (theoretically) uses YIQ instead, where getting those particular colors to alternate would involve reflecting the signal across a line of either slope 1 or slope -1 that passed through the origin, which sounds like something nobody would bother to do for a simple test plate when they could just change one sign and be done with it.


The colo(u)r bars in the signal are the EBU (European Broadcast Union) version. An official test transmission from an NTSC region would have used the SMPTE color bars, which reserve the bottom third of the screen for calibrating brightness, contrast, and black levels. The "75%" bit is standard, and just means that the color bars are engineered to be at 75% of max legal signal level.