Anablog

Electronic Design has a design idea for using video multiplexers to create an attenuator, or digital pot. When I had to design a high-frequency attenuator for an automotives diagnostic tool, I had trouble getting the bandwidth over a few MHz. I called up Jim Williams and he said, “You know when you rotate the vertical knob of a Tek scope you hear little clicks?” I got his implication right away— RF relays were the practical way to make a high-frequency attenuator. That was 10 years ago and there are a lot of video multiplexor chips on the market since then. Now this is a similar setup to what I eventually did, using a Maxim fault-protected multiplexor to switch in various taps off a resistive ladder. My problem was that the input had to have a 1 megaohm input impedance, so I had to use high-value resistors. This meant that the stray capacitance at every ladder node affected the signal attenuation, and that stray capacitance was different and uncontrolled at every ladder node. So the solution was to put little shunt capacitances across the ladder resistors. They had the same value ratios as the resistors. This kept the dc input resistance 1 megaohm but lowered the ac input impedance to a couple hundred ohms at high frequencies. These lower impedances swamped out the stray capacitance of the board layout and the mux chip. So the one thing the design idea does not mention is the value of the resistors needed for the attenuator ladder. I can assure you that you better keep them low if you want good ac performance. That is OK for video signals, you can assume that the input impedance is 50 ohms, I had to make a poor-man’s oscilloscope that might have 200 volt inputs, so I couldn’t buffer the inputs and use a low-impedance ladder. Do be advised that the stray capacitance of each ladder node will affect the video signal in this scheme and might affect the color or even the luminosity of the signal. The lack of real boards or scope traces in any Design Idea should always make you wonder if the folks built the circuit or just whipped it up in SPICE, which will not account for layout stray capacitance and rarely accounts of IC input capacitance. So if you build this circuit do remember that the layout is critical and that you should try to equalize the stray capacitance proportion at each node, in order to all the ac attenuation to match the dc attenuation. One of the beautiful things about NTSC video is that the higher frequencies are phase-sensitive color information; so attenuating them is not too serious if the zero-crossings still are accurate. But when this circuit is used to do RGB as in figure 2, you should be very careful that high-frequency attenuation does not lower the resolution and ruin the color saturation of the video signal. Using a buffer amp before the attenuator ladder and lowering those ladder resistor values is one way to do this. There is a reason an old HP attenuator is still worth a couple hundred bucks on eBay, they are worth every penny. And also realize that if you attenuate NTSC, well you are attenuating the horizontal and vertical sync pulses and you may have to restore those as well, along with the back-porch and other reference values.