|
You have all hit on contributing elements, I think. The 5-IF designs allowed many tuned elements to shape the IF precisely. Too few stages and you do not reach the noise-limited fringe areas. The 2-stage Muntz and Motorola B&W sets proved that. It was also definitely a matter of economics, to produce sets that would provide service people could afford. The number of IFs also affected selectivity and ultimately how tightly stations could be packed geographically - a huge trade-off off spectrum efficiency vs. an affordable TV broadcast service.
Regarding the IQ demods, I have posted elsewhere that the IFs that had become standard practice by the 70s had trouble with quadrature distortion if you tried IQ demodulation - mostly due to effects of the sound trap and upper end. A technically viable (but economically and logistically impossible) solution would have been to retrofit all NTSC encoders with different Q filters that had traps at 920 kHz (baseband). This would eliminate the residual 2.66 Mhz and lower Q energy at IF, which is what caused the residual quadrature distortion. The NTSC fathers never thought of chopping off the lower Q sideband as sharply as the 4.5 MHz trap would chop off the upper Q sideband.
Besides fewer than 5 IF stages being necessary for gain, designers discovered that a haystack shaped IF would pre-peak the video at 2 Mhz, which is desired for apparent sharpness, and since this could be a phase-leading function producing pre-shoots (not customer adjustable), it would complement a simple R-C video peaking control that produces variable post-shoots. So, with the peaking (customer "sharpness" control) set to mid range, you would have a medium-peaked picture with symmetrical pre and post shoots. Different manufacturers made different decisions from time to time as to how much of this IF prepeaking was desirable. Eventually, Zenith and others developed analog video ICs that could make a symmetrical customer peaking adjustment, but still in conjunction with a haystack IF, which by then was being done with surface wave filters instead of discrete components.
|