Electrostatic discharge (ESD) is one of the EMI problems that every signal connector interface has to survive. Shielded cables and shielded connectors are the first line of defence: they keep external static charge away from the circuits behind the interface. A second, complementary approach is to protect the signal lines themselves, using transient voltage suppressors (TVS) or varistors placed where the discharge current can be diverted before it reaches a sensitive device.
Protection at the connector is only part of the story. The same interface is also shaped by the integrated circuits that drive it, by how those circuits are packaged, by how their supply is decoupled, by the clock that times them, and by how the transmission line between source and destination is terminated. This part of the series on component selection and circuit design techniques works through those layers in turn: how CMOS devices load the power distribution network, why package style and pin arrangement change EMC behaviour, why the clock circuit dominates so much of the noise budget, how regulators must be bypassed to stay stable, and how the various termination schemes trade cost, delay, power and signal quality against one another.
Highlights:
- Surface-mount devices have a smaller footprint and a lower mounting height than through-hole parts, so they deliver better EMC performance and should be the first choice; keeping the decoupling capacitor between VCC and GND as close as possible to the IC pin minimises the loop inductance that limits its effectiveness.
- Whether the subject is an IC, a PCB or a whole system, the clock circuit is the dominant factor in EMC performance, because most device noise is tied to the clock frequency and its harmonics; adequate grounding, decoupling and bypass capacitors reduce radiation.
- Termination both suppresses reflection and ringing caused by impedance mismatch and slows the fast rising and falling edges; series termination costs little and consumes little power, but it adds delay and requires the series resistor to sit as close as possible to the source driver.
Protecting Signal Interfaces Against ESD
ESD protection is a system-level concern, not a component-level afterthought. A discharge arriving at a cable or connector will couple into the signal conductors and travel toward the receiver unless it is intercepted. Shielding addresses the coupling path by preventing the disturbance from reaching the conductors at all, which is why shielded cable and shielded connectors are used wherever the interface is exposed to the outside world. Clamping addresses the conductor itself: a TVS diode or a varistor on the signal line presents a low impedance during the transient and shunts the energy away from the downstream IC. In practice the two methods are used together, because shielding reduces the amplitude of what arrives, and clamping limits what still gets through.
Integrated Circuits and Dynamic Supply Current
Modern digital integrated circuits are built predominantly in CMOS processes. The static power consumption of a CMOS device is very low, which is one of the main reasons the technology displaced earlier families. That advantage, however, applies to the quiescent state, not to the switching state. When a CMOS device switches at high speed, it demands instantaneous power from the supply. The dynamic power requirement of a high-speed CMOS device can exceed that of an equivalent bipolar device, because the current draw is concentrated into very short intervals at each edge.
The power distribution network cannot supply that current instantly: the inductance between the device and the supply source limits how fast current can be delivered. Decoupling capacitors have to be added to these devices specifically to satisfy the instantaneous power requirement, providing a local reservoir of charge that responds before the supply network does. Without them, the supply voltage at the die collapses briefly at each switching event, and that disturbance propagates as noise into the rest of the design.
How IC Packaging Affects EMC
Integrated circuits are available in a wide range of package structures, and the choice is not purely a manufacturing decision. For discrete components, the shorter the lead, the smaller the EMI problem, because a short lead presents less inductance and a smaller antenna. Surface-mount devices have both a smaller mounting area and a lower mounting height, which gives them better EMC performance than their through-hole equivalents. They should therefore be the preferred choice. The logical extension of this trend is to mount the bare die directly onto the PCB, eliminating the package and its leads altogether.
Pin arrangement on the IC also influences EMC performance. Consider the path that the supply current follows from the centre of the module out to the IC pin: the shorter that path, the smaller its equivalent inductance and the smaller the disturbance generated when the current changes abruptly. This is the packaging-level reason why the decoupling capacitor between VCC and GND becomes more effective the closer it is placed to the pin it is meant to serve. The capacitor and the supply path form a loop, and the loop inductance is what limits how quickly the capacitor can deliver charge.
The Clock Circuit as the Dominant Noise Source
Whether the subject is an integrated circuit, a PCB or an entire system, the clock circuit is the principal factor affecting EMC performance. Most of the noise generated inside an IC is related to the clock frequency and to its harmonics, since a periodic square wave concentrates energy at the fundamental and at integer multiples of it, and those components extend well into the frequency range where cables and traces radiate efficiently. Appropriate component selection, particularly for clock-related devices, can help control unwanted emissions and improve overall EMC performance.
Both the circuit design and the PCB design therefore have to take the clock circuit into account if the noise is to be reduced. A well-designed ground, appropriate decoupling capacitors and bypass capacitors all reduce radiation. High-impedance buffers used for clock distribution also help: by presenting a controlled impedance to the clock line, they reduce reflection and oscillation of the clock signal, so the clock waveform stays clean instead of ringing and radiating the ringing as harmonic energy.
Mixed TTL and CMOS Logic
When a design mixes TTL and CMOS devices, the different switching and hold times of the two families generate harmonics of the clock, of the wanted signals and of the supply. These harmonics are exactly the components that create EMC problems. The most effective way to avoid the issue is to stay within a single logic family throughout the design.
Because CMOS devices have a wide threshold, most designers now choose CMOS parts. Since the manufacturing process is a CMOS process, the interface circuits of a microprocessor are also best built from the same family. One point deserves particular attention: unused CMOS inputs must be tied to ground or to the supply. A floating CMOS input sits at an undefined level and can switch randomly. In microcontroller circuits, noise coupled into an unconnected or unterminated input can cause the MCU to execute incorrect code, which is far more damaging than the small effort needed to tie the pin to a defined level.
Bypass and Decoupling for Voltage Regulators
A voltage regulator needs its own decoupling, arranged differently from that of a digital IC. In a typical regulator circuit, the appropriate decoupling capacitor should be placed as close as possible to the output of the regulator. The reason is that during the regulation process the distance between the regulator output and the load introduces inductance, and that inductance interacts with the control loop and can provoke internal oscillation in the regulator itself. The capacitor close to the output keeps the loop behaved.
A typical example is to add 0.1uF decoupling capacitors at both the input and the output of the regulator. This suppresses the potential internal oscillation and, at the same time, filters high-frequency noise. In addition, a relatively large bypass capacitor is added to reduce output ripple. The bypass and decoupling arrangement for a regulator is illustrated in Figure 11. In every case, the capacitors must be placed as close as possible to the regulator device.

Figure 11 | Bypass and decoupling regulator
Terminating the Line
When a circuit runs at high speed, impedance matching between the source and the destination becomes important. A mismatch causes signal reflection and damped oscillation, so the waveform arriving at the receiver is no longer a clean replica of what the driver launched. The excess RF energy associated with those reflections is either radiated or coupled into other parts of the circuit, and that is precisely how an EMI problem is created. Terminating the signal helps to reduce these unintended outcomes.
Termination does two jobs. It reduces the reflection and ringing that arise when the impedance between source and destination is not matched, and it also slows the fast rising and falling edges of the signal. Slower edges mean less energy at high frequency, which reduces both radiation and crosstalk.
There are many ways to terminate a signal, and each has its own advantages and drawbacks. Table 3 summarises the main methods and the parameters that matter for each.
Table 3. Summary of signal termination methods
| Termination type | Relative cost | Added delay | Power requirement | Critical parameter | Characteristics |
|---|---|---|---|---|---|
| Series | Low | Yes | Low | RS = Z0 = R0 | Good DC noise margin |
| Parallel | Low | Small | High | R = Z0 | Power consumption is a problem |
| RC | Medium | Yes | Medium | R = Z0, C = 20 to 600 pF | Limits bandwidth and adds capacitance |
| Thevenin | Medium | Yes | High | R = 2 x Z0 | Requires high power for CMOS |
| Diode | High | Yes | Low | - | Clamps overshoot; diode ringing |
Read across the rows, the table makes the trade-offs explicit. Series and parallel termination are both inexpensive in component terms, but parallel termination draws continuous current and therefore becomes a power problem. RC termination moderates the power consumption of the parallel approach at the cost of bandwidth and added capacitance on the line. Thevenin termination, which uses a two-resistor divider, is effective but demands high power with CMOS drivers. Diode termination limits overshoot rather than matching impedance, at a higher component cost, and the diodes themselves can ring.
Series or Source Termination

Figure 12 | Series termination circuit
Figure 12 shows the series, or source, termination method. A source termination resistor RS is inserted between the source impedance ZS and the distributed trace of characteristic impedance Z0 in order to complete the impedance match. RS also absorbs the reflection coming back from the load, so the reflected energy is dissipated at the source instead of bouncing along the line and disturbing the driver.
The placement of RS is critical: it must be as close as possible to the source driver circuit. If it is placed further away, the trace section between the driver and the resistor is left unmatched, and the benefit of the termination is partly lost. The value of RS is given as a real quantity by the relationship RS = (Z0 - ZS). In practice RS is usually chosen somewhere in the range of 15 to 75 ohms.
Parallel Termination

Figure 13 | Parallel termination circuit
Parallel termination adds a resistor RP from the signal trace to ground at the load end so that the parallel combination of RP and the load impedance ZL matches the characteristic impedance Z0. The matching condition is RP // ZL = Z0. The termination resistor therefore works together with the load; it is not an independent value. When ZL alone is higher than Z0, RP pulls the total load down toward Z0, and the line sees a matched load at the end of the trace. For a given Z0 and ZL, RP is selected from the parallel resistance relationship, and the resistor is placed at the load end to reduce the reflection at that end.
This method is not suitable for handheld products. RP is normally small, typically 50 Ω, so the termination draws a steady current and wastes power. The source driver must also deliver a relatively high current; with a 50 Ω termination at 5 V, the driver current can reach 100 mA. The parallel combination ZOL = RP // ZL and the load input shunt capacitance Cd also add a small delay. The product ZOL × Cd sets a time constant at the load, so the termination is not purely resistive in transient behavior. For handheld products, both the static current and the additional driver current are difficult to accept, which limits the practical use of parallel termination when the power budget is tight.
RC Termination

Figure 14 | RC termination
RC termination is similar to parallel termination, but it adds a capacitor C1 in the termination branch. As in the parallel method, R provides the impedance that matches Z0. C1 supplies the current for R and filters RF energy from the trace to ground. Because the capacitor is present, the RC method requires less source drive current than the parallel method. The values of R and C1 are determined by Z0, Tpd (the loop transmission delay), and Cd. The time constant is set as RC = 3 × Tpd, with R // ZL = Z0 and C = C1 // Cd.
The time constant ties the termination response to the transmission delay of the loop. R sets the matched impedance, while C1 and Cd determine the effective capacitance used in the RC product. The capacitor blocks DC, so the resistor does not load the driver with a continuous current. This makes RC termination a way to obtain the matching effect of a parallel resistor without the same DC current demand, while still filtering RF energy from the trace to ground.
Thevenin Termination

Figure 15 | Thevenin termination circuit
Thevenin termination uses a pull-up resistor R1 and a pull-down resistor R2. The divider sets the logic high and logic low levels to match the target load. The resistor values are selected so that R1 // R2 = Z0. In addition, the value of R1 + R2 + ZL must ensure that the maximum current does not exceed the capacity of the source drive circuit. For example, with R1 = 220 Ω, R2 = 330 Ω, and a drive voltage VCC of 5 V, the reference voltage is Vref = R2/(R1 + R2) × VCC = 330/(330 + 220) × 5 = 3 V. Here VCC is the drive voltage.
The Thevenin divider therefore performs two tasks at once: it provides the parallel termination resistance seen by the line and it establishes the DC bias at the load. The divider ratio sets Vref, while the parallel value of R1 and R2 sets the termination impedance. The current limit condition is set by the sum R1 + R2 + ZL, so the divider cannot be made arbitrarily low in resistance. This is the main tradeoff between the desired termination value and the drive capability of the source.
Diode Termination

Figure 16 | Diode termination
Diode termination is similar to Thevenin termination, except that the resistors are replaced by diodes to reduce loss. D1 and D2 limit the amount of excessive signal reflection from the load. Unlike Thevenin termination, the diodes do not affect the linear impedance. For this termination method, Schottky diodes and fast switching diodes are preferred. The advantage of diode termination is that it does not require the value of Z to be known, and it can be combined with other types of termination. This method is commonly used inside an MCU to protect I/O ports.
Because D1 and D2 replace the resistive divider, the termination does not present a fixed linear resistance to the line. The diodes act on excessive signal excursions and limit the reflection energy coming back from the load. This keeps the loss lower than in a resistive Thevenin network. The method can also be used together with other termination types, which makes it useful where the line impedance is not known in advance or where the termination must be placed inside an MCU I/O structure.
Related Reading: