I am looking at RGB to YCbCr circuits at the moment, and the input characteristics of monitors. I've got a few (possibly naive) questions though: - What is the standard for AC coupling? From an article I read TVs/monitors AC couple their input and then reconstruct a DC offset, on the other hand some scart cables that I've got and my GameCube component cable (US) have AC coupled outputs. And I heard of some Japanese RGB cables that don't have AC coupling. Is that one of those regional PAL/NTSC standard differences? Does it hurt to have both the output and the input AC coupled? - Cb and Cr are centred on 0V, with positive and negative amplitude. Does this damage the elcos that are usually used for AC coupling? As the DC offset gets reconstructed anyway by the monitor, would it be a good idea to shift it up so that the whole signal is positive? - most of the circuits I've seen are variations of the same idea to sum RGB together to get Y and then use that to get the differential signals. Most variations are in using different opamps or slight changes in the expression. All of them use some kind of negative voltage. One of them uses a MAX764 voltage inverter which sounds promising, or I found the datasheet for a MAX7660 for a similar function. Has anyone experimented with devices like this? Are there any disadvantages/advantages/things to watch out for? - Going back to the AC coupling, if the actual DC offset doesn't matter, is it possible to do the whole thing with single-supply opamps? Does anyone have a circuit or some experience?
Most consumer video is AC coupled or assumed to be AC coupled for a more robust transmission and recovery system which traditionally can come from many sources like radio signals. AC coupling can be made from an AC signal into a split supply (AC) amplifier or a DC amplifier with capacitive coupling to remove the DC offset. The standard is to use a capacitive coupling because it's simpler and cheaper than an AC amp and split supply. Generally the capacitors are in the device/console so that cables are no more than wires. Capacitors cables seems to be a weird Nintendo thing starting with the PAL SNES. PAL SNES gamepads are also incompatible with NTSC ones so maybe that's related. Some of this I said in the other thread already but I'll switch it up: The required capacitance depends on the time constant made with the output impedance, transmission line and display load. This also depends whether your output amplifier has gain or not. If it doesn't then you're outputting 100 IRE into a 37.5 ohm load (this is probably common in pro equipment because a unity amplifier is twice as fast), if it does then it's 200 IRE into a 150 ohm load (this is standard in consumer equipment because it uses less current and requires smaller capacitors to pass 50 Hz sync components -- 220uF). Most displays do AC couple their input, again for robust recovery (in order to clamp to black level and do AGC). They do this AFTER the (75 ohm) load so that the capacitors aren't in series, worsening the filter situation. Since the next stage is a hi-Z amplifier it shouldn't act as a load, so the display input capacitor can be small. Once you AC couple a signal the capacitor creates a (wobbly) virtual ground, that's normal. There's nothing dangerous about putting two capacitors in series, unless the reduced capacitance allows a high voltage across them which exceeds their tolerance. Assuming you're using a DC amplifier your output is already going to have a DC offset. Whether you want to AC couple it is up to you, but you're supposed to. If you use an AC amplifier then you don't need to capacitively couple the circuit because it's already AC coupled. If you added a DC offset then there's no point in using a more expensive AC amplifier since the negative half of it isn't being used. DC amplifier = easier, single supply, no distortion but capacitive filtering (low power efficiency) AC amplifier = harder, split supply, distortion but no capacitive filtering (high power efficiency) In theory you should use an ideal AC amp for everything but that's impractical. With small signals power efficiency isn't really important because it's a matter of milliamps vs cost and complexity so DC amps are used extensively. Op-amps are often used because it's a whole lot easier than building linear differential amplifiers with transistors. A discrete circuit will probably have better performance at a much lower cost though. Yes it's possible to use single supply circuits for any application, but in some cases (such as this one) it's more complicated. When you're transcoding R'G'B' to Y'PbPr the circuit's complexity is highly dependent on whether the input is already AC coupled or not. If it isn't then obviously it's much easier to process with a single supply since you don't need to clamp the signal at sync's back porch (to recover the DC offset).
Never rely on the input's DC offset in your circuits. First is that it varies with device (some have 2V, others 2.5, others 1...), second is that it will screw you over on your own VRefs.. How? Picture this: You have a weak video signal which is supposed to have a DC offset of ~2.5V and you want to amp it 6dB (x2) for example to drive a 75Ohm load (typical scenario with a source resistor = 75Ohm) So you take your favourite high bandwidth, low distortion, JFET input opamp because we don't want to load our weak signal, etc etc... and do this easy circuit with 2 500Ohm resistors closing the feedback loop on the opamp and to a virtual ground. Now this virtual ground should be 2.5V, but you must generate it locally... How? Well using a regulator, or a resistor divider or a voltage reference, whatever suits your bill.. Now a problem here is that this voltage won't be equal to the DC offset the signal originally has (and can never be: they are generated differently), and the opamp will amp this difference, obviously. In this case 2 times. A problem that may happen here is if the signal amped + the DC offset amped may go outside the opamp linear area (specially with 5V single supplies), which will distort the video when it's going to bright or too dark... Tl,dr: Use series caps to remove the inputs' DC offsets and dual supplies for the best.. Even if you have to generate the negative supply locally onboard with some ripple typical of chargepumps (like that 7660 IC and others), opamps usually have decent noise rejection from the power rails so it's not a problem.
Thank you, that is helpful. Now that I have an idea what it does I understand why it was bad in the first version of the converter. I should really have learned more analogue stuff.
