Analog Switch and Multiplexer Circuit Design

An analogue switch is a pair of MOSFETs that connects or disconnects a signal path, and it behaves quite differently from the mechanical switch it replaces. It has a resistance that varies with the signal, it injects a small charge every time it changes state, and it only works within a defined signal range. Understanding those three properties is most of the design work.

What an Analog Switch Is

The switch element is a MOSFET, usually a pair in parallel with one device conducting the positive part of the signal and the other the negative part. The parallel arrangement keeps the resistance more constant across the signal range than a single device would, and it extends the range over which the switch can pass a signal of either polarity.

Control is separate from the signal path. A logic input turns the switch on or off, and the control circuitry is referenced to the switch supply rather than to the signal. Some devices include level shifting so that a low voltage logic signal can control a switch that operates on a higher supply.

Because the device is a semiconductor, the switch has no moving parts and no contact wear, switches in microseconds and can be operated millions of times. It also has no isolation when off, only a very high resistance, which matters when the off state has to block a large voltage or when leakage in the off state has to be extremely small.

On Resistance and Signal Range

on resistance is the resistance of the switch when it is closed, typically a few ohms for a general purpose device and a fraction of an ohm for a low resistance type. It forms a divider with the load, so a switch feeding a high impedance input produces a negligible error, while one feeding a low impedance load produces an attenuation that has to be accounted for.

The resistance is not constant. It rises as the signal approaches the supply rails, because the gate to channel voltage falls, and it changes with temperature and between parts. A ratio measurement that depends on the switch resistance matching should use a device that specifies matching rather than absolute resistance.

signal range is the range of voltages the switch can pass without distortion or leakage. A device operating from a single 5 volt supply passes signals from zero to about 5 volts with an increasing resistance near the extremes, while a device with a bipolar supply can pass signals of both polarities. Signals outside the supply rails will forward bias the internal protection structures and must be prevented.

Charge Injection and Switching Transients

charge injection is the small amount of charge that is transferred from the control gate to the signal path when the switch changes state. It appears as a step in the signal, with a magnitude that depends on the device, on the signal level and on the load impedance. In a sample and hold circuit it is one of the dominant error sources.

The resulting voltage step is the injected charge divided by the total capacitance at the node. A high impedance node therefore shows a larger step than a low impedance one, which is why buffers are often placed between the switch and a high impedance load. Reducing the node capacitance and driving it from a low impedance source both help.

Switching transients are also produced by the control signal coupling into the signal path through the package capacitance. A fast control edge produces a spike on the signal line even when the switch is not changing state, and the effect is worse when the control line runs parallel to the signal. Separating the two in the layout is the practical remedy.

Multiplexer Configuration and Crosstalk

A multiplexer is a set of switches sharing a common output, and its principal extra concern is crosstalk between channels. Two channels are coupled through the capacitance of the switches and through the common output node, and the coupling rises with frequency. An unselected channel carrying a large signal can therefore appear at the output at a level that matters.

crosstalk is specified as the level of an unwanted signal at the output relative to the level applied to the unselected channel, and it generally worsens as the frequency rises. Where the application has a large dynamic range, the specification has to be checked at the highest frequency of interest rather than at the direct current value.

Multiplexers are also used in a differential configuration, where two switches operate together to route a pair. The matching between the switches in a pair sets the common mode rejection, so a device that specifies the on resistance matching between channels is preferable for that application. Where the signals are small, the matching matters more than the absolute resistance.

Supply, Level and Protection

The switch supply sets the signal range and has to be chosen with the signals in mind. A multiplexer fed from a single 3.3 volt rail cannot pass a 5 volt signal without leakage, and a device with a charge pump can handle signals above its own supply at the cost of additional switching noise from the pump.

Digital inputs that control the switch have their own thresholds, which may not match the logic levels available. Many devices accept a wide logic range, while others need the control level to be at least a defined fraction of the switch supply. Checking the control threshold against the driving logic at both temperature extremes avoids an intermittent problem that appears only at the cold end of the range.

Protection matters where the analogue input comes from outside the board. A series resistor and a clamp diode at the connector limit the current that can reach the switch, and the resistor should be placed before any filter capacitor so that the capacitor also helps to absorb a transient. The approach is the same as for any exposed analogue input and the general methods are covered in our article on mixed signal board design.

Layout of the Switch Array

Keep the analogue signal traces short and away from the digital control lines. The three times the trace width rule, described in our guide to the three W rule, is a reasonable minimum separation, and on a multiplexer the control lines should cross the signal lines at right angles where a crossing is unavoidable rather than running alongside them.

Place the decoupling capacitor close to each supply pin and return it to the analogue ground, and keep the common output node short and direct to whatever follows. A long common node adds capacitance that both slows the switch and increases the charge injection error, so the multiplexer should be placed close to the amplifier or converter that receives its output.

Guard the high impedance nodes where leakage matters. Surface contamination and humidity produce a leakage current that appears as an offset, and a guard ring or a coating keeps it stable. The coating practices in conformal coating apply to the nodes that carry small signals, and the measurement techniques are the same as for any precision front end.

Verification and Common Faults

Verify the on resistance by measuring the voltage across the switch with a known current flowing, at several points across the signal range. The variation across the range is what matters for a measurement channel, and it is usually larger than the datasheet typical value suggests at the extremes.

Check the off isolation by applying a signal to a channel that is switched off and measuring what appears at the output. A reading that is higher than the specification suggests leakage through the printed circuit board rather than through the device, particularly if the board has been handled or cleaned inadequately.

A switch that works at low frequency and fails at high frequency is usually a crosstalk or capacitance problem rather than a device failure. Comparing the measured crosstalk with the specification, and checking the layout for a long parallel run between signal and control lines, identifies the cause. The same measurement discipline used to check a multiplexed converter channel applies here, and it is worth performing before the design is committed to production.

FAQ

Can an analog switch pass a signal larger than its supply? No, not without leakage through the internal protection structures. A device with a charge pump can handle signals within its extended range, but the range must still be respected.

How does on resistance affect accuracy? It forms a divider with the load impedance. Feeding a high impedance input gives negligible error, while driving a low impedance load produces an attenuation that must be calibrated.

Why does the output jump when the switch changes state? Charge injection from the control gate. Use a lower capacitance node, a device with lower injection, or let the transient settle before sampling.