How Do Analog Multiplexers Work for Expanding ADC Channels?

Detailed Explanation

An analog multiplexer IC contains an array of CMOS switches, each connecting one of several input channels to a shared common pin, with a digital address (binary-coded, typically 3 address bits for 8 channels, 4 for 16) selecting which switch is closed at any moment. The classic example is the CD4051-family: an 8-channel single-ended multiplexer (or 4-channel differential, in the related CD4052) that has been a standard part for decades. Modern precision alternatives — Analog Devices' ADG-series, Texas Instruments' MUX-series — offer lower on-resistance, better on-resistance flatness, and lower charge injection for measurement-grade applications.

Because the multiplexer is a simple, passive-looking switch from the signal's perspective (no active amplification or buffering inside most parts), a signal can flow through it in either direction — a mux used to select which of several sensors feeds an ADC can equally be wired the other way, to route one signal to one of several destinations.

On-Resistance and Its Effect on Accuracy

Every CMOS switch has a non-zero on-resistance (Ron), typically tens to a few hundred ohms depending on the part and supply voltage. This resistance forms a voltage divider with whatever the mux output feeds — an ADC's sample-and-hold input impedance, or a downstream amplifier's input impedance. For a high-impedance ADC input, this divider effect is negligible; for a low-impedance load, it introduces a measurable gain error. A related, often more important effect is on-resistance flatness — Ron is not perfectly constant across the input voltage range in most CMOS switches, which introduces a small, signal-dependent nonlinearity rather than a simple fixed gain error. Precision multiplexer parts specify Ron flatness explicitly; general-purpose parts like the CD4051 typically do not, and are less suited to high-resolution precision measurement for this reason.

Settling Time and Bandwidth

Switching the mux to a new channel doesn't produce a valid reading instantly — the mux's on-resistance combines with the downstream load capacitance (including the ADC's own sampling capacitor) to form an RC time constant that the signal must settle through before the reading is accurate. For slowly-changing signals (temperature, most DC measurements) this settling time is rarely a practical constraint; for fast or precision signals, the settling time after each channel switch must be budgeted into the total conversion cycle, and a settling delay inserted in firmware before triggering the ADC conversion.

Charge Injection

When a CMOS switch's control gate transitions, a small amount of charge couples through the switch's parasitic gate capacitance into the signal path, producing a brief voltage glitch on the selected channel at the moment of switching. This is usually small enough to be irrelevant for slow measurements once settling time has been allowed, but matters for fast-sampling or very-high-precision applications, where lower-charge-injection parts (typically the more expensive precision mux ICs) may be justified.

Break-Before-Make Switching

Most analog multiplexers are designed as break-before-make: when switching channels, the previously-selected switch opens before the newly-selected switch closes, guaranteeing that two input channels are never briefly connected together during the transition. This matters when multiplexing signals from different sources that should never be shorted together, even momentarily — always confirm break-before-make behaviour in the datasheet for any application where two source channels must never be connected simultaneously.

Practical Examples

A common use case is a multi-point temperature-monitoring system with many thermocouples or thermistors and only one precision ADC (or one channel of an instrumentation-amplifier-fed ADC) — a 16-channel mux ahead of the ADC lets firmware sequentially select and read each sensor, at the cost of one settling delay per channel switch, instead of requiring 16 separate ADC channels or 16 separate ADC ICs.

Another common pattern uses a mux to route a single, carefully-designed signal-conditioning chain (amplifier, filter, ADC) across several sensors that are read one at a time rather than simultaneously, keeping the cost and board area of the precision analog front end to a single instance rather than duplicating it per channel.

Design Considerations

• Match the mux's on-resistance and flatness spec to the measurement's actual accuracy requirement. A general-purpose CD4051-class part is fine for coarse, low-impedance-tolerant readings; a precision measurement application (load cells, RTDs, calibration-grade sensing) usually justifies a precision mux part with tighter on-resistance flatness and lower charge injection.
• Insert a settling delay after each channel switch, sized to the actual RC time constant (mux Ron × downstream load capacitance), not an arbitrary fixed delay — an undersized delay reads a partially-settled voltage, while an oversized delay wastes conversion cycles unnecessarily on a channel-scanning application.
• Consider whether the application actually needs a mux at all. Some ADCs (particularly in the sensor and MCU space) already integrate multiple input channels internally with their own internal multiplexer — see what is an ADC — so an external mux is only needed when the channel count required exceeds what a chosen ADC natively provides, or when isolating/protecting individual channels with external switching is itself a design goal.
• Protect mux inputs from voltages outside the supply rails. Most CMOS mux inputs are not tolerant of voltages beyond the supply rails by more than a diode drop; a sensor fault condition that drives an input above/below the rails can damage the part or the rest of the signal chain through the mux's parasitic ESD diodes conducting fault current onto the supply rail.

For multi-sensor analog front ends needing careful channel-switching, settling-time, and accuracy trade-offs, Zeus Design's product development team designs and validates precision measurement hardware end to end.

Common Mistakes

• Triggering an ADC conversion immediately after switching the mux channel, without a settling delay — the classic cause of a multiplexed measurement system reading a value contaminated by the previous channel's signal, especially when switching between channels with very different voltages.
• Using a general-purpose mux for a high-resolution precision measurement and attributing the resulting nonlinearity to the ADC or sensor, when the actual cause is the mux's on-resistance flatness across the signal range.
• Ignoring the mux's own leakage and off-isolation specifications in high-impedance, low-signal applications — a "deselected" channel is not a perfect open circuit, and its leakage current or capacitive coupling into the selected channel can matter for very sensitive measurements.
• Exceeding the mux's absolute maximum input voltage rating during a sensor fault condition, damaging the part (and potentially the rest of the signal chain) — add input protection (series resistance, clamping diodes) ahead of the mux for any channel connected to a signal that could fault outside the expected range.