Strange amplitude fluctuation in video signal

Many video devices are AC coupled on the inputs and outputs; it might be nice if that weren't the case, but in practice unless one knows one will only be accepting signals from devices with DC-coupled inputs, one has to accept AC-coupled signals. Further, in many cases, if one will have to deal with AC-coupled signals, it's often helpful to AC-couple one's own input stage.

If one has positive and negative supply rails available, the easiest approach is probably to have the input (past the coupling capacitor) pulled weakly to the negative supply and have a clamp diode to a point about 0.35 volts above ground. The effect will be a signal where the lowest parts of the signal are at -0.35 volts. This will make it easy to detect sync signals, and will provide a usable (though not wonderful) black reference for superimposing opaque video. If negative supply isn't available, one could use a pulldown to VSS and clamp to 3 volts or so and expect all levels to be shifted. The sync detection from this approach is fine; the biggest weaknesses are:

• If the pulldown sinks much current, the end of each line will be made a little darker than it should be; if it doesn't, the circuit may be slow to deal with shifts in the input DC bias;
• One has to 'guess' the difference between the input signal's sync level and black level. Ideally the sync voltage will be 0.34 volts below the black level, but not all video output devices are perfectly "to spec".

If one wants broadcast-quality video, one should decode the sync signals in a way that won't disrupt the original video (e.g. by passing it through a buffer and then AC coupling it to the described clamp circuit), and then ground an AC coupled output briefly after each rising edge of the sync signal (about 500ns if one doesn't have a 3.579MHz notch-reject filter, or 4.5us if one does).

Although the screen of an analog color monitor is divided into many individual picture elements (RGB phosphor triplets), there is no fixed relationship between those and the pixels generated by the computer. Instead, an analog monitor will project a slightly blurred image of the computer's pixels onto the back of the phosphor tube; if the beam passes closer to one phosphor pixel than to another, more energy will be received by the former, making it appear brighter. When two adjacent phosphor pixels are illuminated by different amounts, the eye will perceive a bright spot whose center is closer to the more-brightly-illuminated pixel.

It would be possible for an LCD monitor to use simulate that behavior electronically, and indeed many LCD television sets do precisely that. To avoid weird interactions between computer pixels and LCD pixels, however, it is necessary to have a minimum of roughly two LCD pixels' worth of "blur". When watching videos, such blur is generally not objectionable, but when displaying things like text, it's generally undesirable. To generate a clear display, it's necessary that each computer pixel light up exactly one monitor pixel. Because VGA standards were designed at a time before anyone expected pixel-accurate displays, the VGA standard does not provide any mechanism to indicate precisely where the picture data for each discrete pixel should appear in the video stream. Thus, a monitor must try to "guess", based upon the display's image content; when it guesses correctly, it will sample each pixel's value in the middle of the pixel's display time. If a screen shows a lot of detail, then there will be only one way the monitor can line up computer pixels with LCD pixels without obvious (detectable by the monitor) problems. If, however, the screen consists of nothing but large solid areas of color, then many possible mappings between computer pixels and LCD pixels would appear to "work", but only one would be correct. If the monitor guesses wrong, then when the computer shows more detail the monitor would be able to detect that and react accordingly.