Quote:
Originally Posted by Tomcomm
Did RCA ever go back to wb IQ video processing? ….
|
In the 80s, RCA produced a chassis with I/Q decoding. They developed a circuit that could drive a delay line and sense its output with one analog IC pin and one external transistor, one of my favorite patents of all time.
I had a project at Zenith to investigate possible wideband chroma, and unfortunately found that a combination of circumstances made it impractical (list below). Of course, we did a competitive analysis of the RCA, and found that the I high frequency response was rolled off severely, which I think was probably due to them discovering the same problems I did in my project:
1) The IF amp designs were "haystack" shaped rather than attempting the flat top response in very early color sets. This was for a combination of reasons; lower cost, better transient response when used with a simple analog sharpness control, less sensitive to fine tuning error. But it meant that chroma sidebands were on a slope that had to be compensated in the chroma take-off / bandpass circuits.
2) The original NTSC Q filter specs in the color encoder were not strictly enough specified. Many encoders had sufficient Q sidebands that non-symmetrical upper and lower sidebands would be present due to the sound carrier notch in the transmitter and especially in the receiver. This produced quadrature color distortion on edges - could be OK for some things, but when the edges of faces get wrong color, that's a deal breaker. If only the original NTSC gurus had thought to specify a 920 kHz null in the Q baseband filter, both Q sidebands would have been suppressed, and the I/Q unequal bandwidth would have been guaranteed to work as advertised.
I went through a complete series of experiments, from full wideband RGB to composite baseband with wideband equiband chroma; to composite baseband with double-sideband wide I and narrow Q using a commercial encoder; to the same over RF (where it got bad); to an encoder with a redesigned Q filter over RF (edges fixed). When I say edges fixed, I mean edges of faces were OK; but we could find some objects that showed distortion. For example, a thin yellow box surrounding titles on one test slide turned orange on the vertical strokes, because only the I component had significant amplitude on the thin line. Yellow title letters were not so bad because they were usually thick enough to get some Q response; but you could see some effect.
Speculation on why the original NTSC specs didn't have the notch specified:
1) effects weren't easily visible on small screen
2) analysis of some original schematics indicates there could be an inadvertent notch (or at least strong high frequency attenuation) in the Q path depending on the mutual coupling in the phase-equalized filter design; but this was never specifically required in the NTSC specs.