Low cost Data Acquisition Devices (DAQ) often use a single analog-to-digital converter (ADC) and an analog multiplexer chip to sample many channels. Each channel is one-by-one connected to the ADC for the measurement. Some of devices dedicate an ADC to each channel, but that comes at an added expense. In the future, as chip prices continue to drop, ADCs per channel may be common place, until then DAQ users should understand how a signal’s source impedance affects the signal measurement on multiplexed device.

All multichannel-single ADC DAQs, such as the 3000 series product line from IOtech, have some degree of charge coupling effect (cross-talk) caused by stray capacitance. The degree to which the charge coupling affects a signal is called the Channel-to-Channel Crosstalk and is typically specified in negative dB. Ideally the Crosstalk specification would be -96dB for a 16 bit converter but in practice it is not. A simple example consists of a two channel system with -75dB crosstalk and the first channel is measuring a 6.0 Vdc and second is measuring 0.0 Vdc. The -75dB specification tells us that some of the 6.0 Vdc will get into the 0.0 Vdc and that can be determined by the following equation:

Crosstalk = 20 Log ( X / 6.0Vdc )

X = inv Log ( -75dB/20 ) * 6.0 Vdc

X = 1.06mV

Above is a typical system that pictures the stray capacitance between channels. It also shows the input signal as an ideal voltage source and associated resistance. During operation, the multiplexer switches from channel one to channel two and (for example) back to channel one again. This switching action causes a voltage spike to appear across the stray capacitance. Because the voltage spike is an AC signal, the capacitance acts like a momentary short circuit imparting some voltage onto the next channel. This voltage does not have much energy and quickly dissipates (exponentially) through the applied signal's source impedance (resistance). Hopefully, by now, it is easy to picture what happens when the source impedance is significant. It should be noted that an unconnected channel has infinite impedance and, if measured alongside the actual signal, will indicate erroneous values that should be ignored.

The system needs to optimize settling time and source impedance to minimize crosstalk. If the DAQ unit has circuitry that switches back in forth between channels at a high rate of speed, then it is essential to have a signal with low impedance. On the other hand, if the switching time is in the realm of seconds, the source impedance is less critical. The ideal source impedance should be zero; however, under 100 ohms is reasonable. With many high-speed devices, such as the 3000 series, increasing settling time is not an adjustable parameter as their clocking circuitry is driven by a fixed 1MHz clock. Therefore, overcoming high source impedance issues is impossible by increasing settling time. Thus, source impedance should be minimized to maintain accuracy.

As shown below, buffer circuits are often used to convert high impedance to low impedance. However, I would like to point out that because the switch times are fast, fast op amps like the OP27 should be used. Otherwise, the short pulses of the multiplexer can disturb the slower op-amp output stage, negating any positive effects of the buffer

Depending on the signal throughput, slower op amps like the INA128 can be used with a low pass circuit as shown: