FAQ on spread spectrum for switching-supply EMI management Part 1: overview

Spread spectrum minimizes the EMI impact of switching power supplies and helps meet regulatory mandates.

Noise is a concern and consideration for many circuit designs, and there are many types of noise. There is external noise, such as from nearby motors or RF interference sources, and internal noise, which is generated by various sources of thermal motion of electrons in components.

However, noise can also be created by the normal, intentional action of circuit components, such as a “spark,” which occurs when the contacts of an electromechanical relay open. For the circuit designer, this noise is unavoidable and unintentional, but “you can’t blame others; this noise is your own doing.”

This is the situation with the noise generated by the switching action of AC/DC and DC/DC regulators (also called converters, with a modest distinction that is irrelevant here) related to their clock and switching frequency operation. This switching generates system noise and electromagnetic interference (EMI) at the clock-switching frequency and its harmonics.

This FAQ will look at “spread spectrum” clocking, a widely used technique for minimizing the effects of this noise by reducing its peak appearances. Spread spectrum serves two purposes: it lessens the impact of the noise on the circuit itself, and it increases the likelihood that the overall product can meet regulatory mandates and limits on how much EMI is allowed to be radiated by the final product.

Q: What are the basic problems with switching regulators and noise?
A: The clock that drives the regulator’s and the regular’s switching action, such as capacitors transitioning between charged and discharged modes, creates radiated and conducted noise. This noise affects circuit performance and may exceed allowable regulatory limits for noise, especially radiated noise (Figure 1).

EMI — Figure 1. FCC Part 15 and CISPR 22/32 radiated limits for Class A and Class B using QPK and AVG detectors below and above 1 GHz, respectively. (Image: Texas Instruments)

Q: What is an alternative?
A: One alternative is to use a low-dropout regulator (LDO), as these non-switching topologies produce nearly no noise (except thermal noise). However, LDOs are relatively inefficient (approximately 30 to 50% efficiency) compared to switching regulators (80-95% efficiency) and have serious wasted power and thermal issues. In general, an LDO is a possible alternative option only for loads below one-half to one-amp output; above that level, it is simply “no contest.”.

Q: Are all switching regulators noisy?
A: Yes, but the issue is to what degree they are noisy. For example, some switchers from leading vendors such as Analog Devices, Texas Instruments, Micro Power Systems, and others are remarkably low noise despite the inherent switching action. The designers of these parts achieved this extremely low-noise performance by looking at every noise source within the regulator associated with the switching action and then devising clever topologies to minimize or even cancel each one (Figure 2).

EMI — Figure 2. The LT8625S “Silent Switcher” radiated noise performance is well below the CISPR32 Radiated Emission mandated limits, achieved by meticulously identifying various noise sources and design topologies that largely negate them. (Image: Analog Devices)

Q: So, is the switching-regulator noise problem solved?
A: In some cases, yes, but in many cases, no. These ultra-low switchers may not have the current/voltage ratings needed in the target application, and their ratings can’t simply be scaled up. Also, while they create much less noise than a standard switcher, they still create some noise—and that may be too much for the situation.

Q: What’s a way to solve the problem?
A: The approach is called “spread spectrum” modulation of the regulator clock signal. Rather than have a fixed clock frequency, the clock is frequency-modulated around a center frequency,

Q: What does this do?
A: This spreads the EMI and noise energy across a spectrum rather than having it primarily at one frequency and, to a lesser extent, at its harmonics. The noise energy of the previous peak is instead spread across part of the spectrum and with a reduced peak value (Figure 3).

Figure 3. The general scheme of EMI reduction by implementing spread-spectrum techniques via frequency modulation of the switching-regulator clock. (Image: Texas Instruments)

Q: Wait, I thought FM was for communication links and broadcast radio…is this the same FM?
A: Yes, it is the same, and the same complex equations govern it. However, it is a totally different application, and no demodulation is needed (as it is for communication links). Also, users of spread-spectrum devices often do not need to delve into the equations; they just need to select a spread-spectrum regulator with suitable characteristics or user-selectable options to meet their needs.

Q: So, is the overall switching noise energy actually reduced?
A: That’s a trick question. This technique does not reduce the overall noise energy as a filter does. Instead, it manages the noise energy to spread it across a wider spectrum with lower amplitude than it had at a single frequency. So, the answer depends on your perspective and priorities.

Q: Why is this acceptable, and in what ways?
A: First, by spreading the noise spectrum while reducing its magnitude across the spectrum, the EMI generated by the switching action may be attenuated enough to pass relevant regulatory mandates.

Q: Is that all?
A: No. In addition to the regulatory issues, some specific noise frequencies may have a greater adverse effect on circuit performance, such as interfering with a low-level sensor signal. By spreading the noise energy and reducing its peak values, the interference from the switcher may be reduced enough that it is no longer an issue for the circuit.

Q: Is the use of spread spectrum controversial? Is it a “cheat?”
A: Where you stand on the question depends on where you are sitting, as they say. If your primary objective is to meet regulatory mandates and get the noise spectrum away from a sensitive area for your circuit, it may be a good solution with no cost except for a slightly more expensive regulator.

However, it doesn’t address the core problem but only transforms it into another set of noise frequencies, admittedly with lower magnitudes. This is unlike true noise attenuation using filters or shielding. Using the spread spectrum may also have a ripple effect of unintended consequences, and it may instead make your problem into a problem for someone else on the team who is working on another part of the circuit.

In some ways, this is like dealing with excess heat using heat sinks, heat pipes, fans, and other techniques to remove heat from your source. Yes, you are getting that heat away from your area of concern, but moving it to that place called “away” may make it into someone else’s problem!

Q: What are some other implications of using a spread-spectrum clock for the switcher?
A: The answer also depends on the source of the modulating clock signal. The system may source it or be created at the spread-spectrum regulator IC itself. If the applied modulated switcher clock is also used as the system clock – as it is in some designs – then the fact this clock is no longer “precise” but instead is frequency modulated may be a problem, even though the average frequency may be correct. For example, any frequency modulation of the switcher clock may cause issues if that same clock is being used to time a data link associated with the system.

On the other hand, for applications such as appliances, frequency modulation of the system clock may be a non-issue as long as the nominal frequency is correct. Alternatively, if the IC creates its own modulation, then the rest of the system is unaffected by the non-constant frequency of the spread-spectrum clock.

Q: It seems that having a non-constant clock contradicts the general design goal of precise, low-jitter clocking, especially for high-speed data links. Is that so?
A: Yes, there is a contradiction here since a spread-spectrum clock is the opposite of a low-jitter clock. However, these are two different situations, and each must be considered on its own merits and objectives as long as one does not affect the other.

The next part of this article looks at spread-spectrum modulation in more detail and provides some IC-based examples.