Component Usage in Prototyping: Latch-Up, Filters and Spare Inputs

Most product failures traced to component choice are not caused by the wrong part number. They are caused by a part that was used correctly on paper and left in a condition the circuit could not tolerate: an input floating, an output driving a long cable, a sensitive gate with no protection. The practices below come from prototype work, where these problems surface first.

Inputs That Are Left Floating

A floating input is an undefined input. The pin sits at a potential determined by leakage and by whatever couples into it, which means it can oscillate, draw excess current or change state with temperature. In CMOS logic the condition also raises the risk of damage, because a pin held near the middle of the supply range turns both transistors partly on.

Every unused input should be tied to a defined level, either directly or through a resistor. Spare gate inputs, unused enable pins and unused interrupt inputs are the ones most often overlooked, because they are not connected to anything in the schematic and therefore draw no attention.

Latch-Up and How to Limit It

CMOS structures contain parasitic bipolar transistors formed by the source and drain diffusions and the substrate. Arranged as they are, those transistors form a four-layer structure with a positive feedback path, and if something injects enough current into it the structure can turn on and stay on. That is latch-up, and the current it draws can destroy the device.

The current that triggers it comes from outside the chip: an overvoltage on an input, a transient on the supply, or a cable that carries a discharge. Limiting the current that can enter a pin is the standard defence, either with a series resistor at each output or input, or with clamping diodes that divert the transient to the supply rails before it reaches the device.

Protecting Signals That Leave the Board

Any conductor that leaves the enclosure can bring a discharge back with it. The protection belongs at the connector end, between the connector and the circuit, with a short, direct return to ground. Placing the device close to the chip instead lets the transient travel across the board before it is clamped, which is exactly what the protection was supposed to prevent.

Lengthening the cable increases the coupling. Where a long cable connects a logic circuit to a mechanical contact, the whole interface has to be treated as an antenna rather than as a wire. ESD and PCB edge traces covers the routing that keeps a discharge away from the parts it would damage.

Input Filter Networks

Where a cable runs from a mechanical switch or a sensor to a logic input, a filter at the interface removes the interference that the cable collects before it reaches the device. A series resistor with a capacitor to ground is usually enough, and the values are chosen from the impedance of the cable and the frequency of the noise rather than from habit.

The filter has to be placed at the connector end of the trace, not at the chip, so that the trace between the filter and the device carries a clean signal. A filter placed at the chip protects the chip but leaves the trace itself free to couple into everything it passes.

The component values are not critical as long as the corner frequency is well below the noise the cable will collect and well above the signal the interface has to pass. A slow mechanical switch, for example, can tolerate a filter network with a corner at a few hundred hertz, which removes almost everything the cable picks up. A fast data line cannot, and there the answer is a different interface rather than a heavier filter.

It is worth checking the filter against the drive capability of the source as well. A series resistor that is too large, combined with the capacitance of a long cable, will round the edges of a legitimate signal until the receiver no longer recognises it, and the resulting failure looks like a timing problem rather than a filtering one.

RC Networks on Sensitive Inputs

For bipolar devices with sensitive inputs, a resistor of moderate value in series with the input and a capacitor of at least 100 pF to ground forms an RC network that slows the edge arriving at the device and shunts transient energy. It is a crude measure, but it is effective against electrostatic discharge and it costs two components.

The resistance has to be small enough not to disturb the timing of the signal and large enough to limit the current. Where the input is a high-impedance analogue node, the resistor value also has to be considered against the bias current of the device, or the network will shift the operating point.

Supply Decoupling and Transient Current

Every fast device draws a transient current when it switches, and that current has to come from somewhere close by. A decoupling capacitor at the pin supplies it before the supply can respond, and its absence is one of the most common causes of intermittent behaviour on a prototype.

Place the smallest capacitor closest to the pin, connect it with short wide traces, and keep its loop to ground as small as possible. Devices that switch large currents, such as memory and interface drivers, need decoupling at every package rather than one capacitor for a group.

Unused Pins and Mixed-Signal Parts

Unused outputs can generally be left open, but unused analogue inputs on a converter or an amplifier usually need a defined potential, and the data sheet is the authority on which pins need what. On a mixed-signal device, an unused input left floating can couple noise into the conversion result even though it is not being measured.

Reading the data sheet for the unused pins is unglamorous work that prevents a class of problems which are otherwise very difficult to diagnose. Mixed-signal PCB design guidelines covers the layout measures that support the same goal.

Connectors, Cables and Assembly

The interface between the board and the outside world is where most of these measures are needed, and it is also where layout room is usually tightest. Keep protection and filtering at the connector, keep the return path short, and keep the protected traces away from sensitive analogue nodes.

Where the design has several connectors, apply the same treatment to all of them rather than to the one that seems most exposed. A discharge finds the path with the least protection, not the one the designer was thinking about. EMI suppression design principles covers the layout rules that reduce the energy arriving at those interfaces in the first place.

FAQ

Can unused CMOS inputs really be left unconnected? No. A floating input can oscillate, draw excess current and put the device in a state where both output transistors conduct. Tie it to a defined level.

Is a series resistor enough to prevent latch-up? It limits the current that can enter the pin, which is the main trigger, but it is usually combined with clamping diodes for a complete solution.

Where should an input filter be placed? At the connector, so that the signal is cleaned before it travels across the board. A filter at the chip leaves the trace free to couple into everything it passes.