My amplifier doesn’t behave properly when multiplexing channels. What could be the cause?
Some people (especially engineers) claim that one way to contest a speeding violation is to request the calibration certificate of the radar unit. They think that if it’s overdue for calibration, it’s impossible to charge the driver. Whether this really works for anyone (the instrument may have been recently calibrated), the safest way to steer clear of a speeding ticket is to stay below the speed limit. But what if you didn’t realize you were going too fast? That’s an excuse that usually doesn’t work.
The same can happen with amplifiers. In some applications, engineers can forget that an amplifier input may be tied to a source with very fast transients. For example, take a multiplexer buffered by a voltage follower (or an instrumentation amplifier) as shown in Figure 1. Say that the input signals are static and filtered by RC networks to reduce the noise bandwidth or RF interference. The amplifier must be fast enough to settle between transitions, so its selection must take slew rate and bandwidth into consideration. In the lab, however, the result isn’t as expected: the amplifier output moves slowly, and it has an unusual waveform with a long settling tail. The settling time is nowhere near the specifications. What could be the problem?
A lot of things could go wrong, but fundamentally, the amplifier inputs are getting overloaded when the channel transitions. If the output can’t move as fast as the input (or at least not before the input does), a large differential voltage will appear across the two inputs. This condition can saturate the input transistors, increase the input bias current, forward bias the internal protection diodes, or cause other undesired effects. The actual reaction to this channel switch depends on the input topology, process technology, and internal protection circuits, and depends on the speed of the transient and the voltage difference between the adjacent channels. In addition to the amplifier’s reaction to the overload condition, the increased input bias current (even if it only flows in the parasitic capacitance between the multiplexer and the op amp) will charge or discharge the capacitor on the input side of the multiplexer. This perturbation, which changes the original voltage level, will only be restored through the resistor connecting the source to the capacitor. If the time constant of the filter is large, a long settling tail will appear at the output of the amplifier.
Several things can be done to remedy the problem, depending on its source. A low-cost solution is to add resistors in series with any inputs that are connected to the fast source—the multiplexer in this case. This can help to avoid saturating the input by increasing the impedance seen by the source during the transient. At a DAC output, or if the amplifier is used to condition a square wave from a microcontroller or FPGA, a simple RC can help slow down the edge to keep the amplifier happy. Each case requires some analysis to determine the best solution, but please pay attention to the transient conditions. And remember, the best way to stay out of trouble is to slow down!