Spread Spectrum Techniques for Minimizing EMI in Power Supplies

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

This discussion of spread-spectrum topologies for switching regulators and converters continues with a look at some modulation and component details.

Q: What are some of the issues associated with the frequency modulation spread spectrum (FMSS)?
A: Keep in mind that it is a frequency-modulation process, and so it is governed by the many complex equations of FM. Users have three top-tier parameters to consider:

Modulation spreading bandwidth: how far on either side of the nominal frequency the oscillator clock signal should be modulated. For various reasons, this is usually limited to between 5% and 10% of the nominal frequency but may go as high as 20%.
Modulation frequency (rate): the frequency of the modulating signal.
Modulation waveform: the shape of the modulation waveform.

Of course, formal FM analysis shows that these three factors are not fully independent of each other, and the settings of one will have some influence on the others.

Q: What are the modulation-waveform options?
A: Obviously, any waveform can be chosen to modulate the regulator clock oscillator, but some standard shapes are common: sinusoidal (sine), triangle, “Hershey’s kiss”, and digital pseudorandom code (Figure 1).

Figure 1. Four waveform types are often used for spread-spectrum modulation. (Image: Micro Power Systems)

Q: What do the associated spread spectrums look like?
A: While the exact shape depends on the specific values of the modulating waveform, their general shapes are shown in Figure 2.

Figure 2. Each modulation waveform type results in a different spectrum overall shape after modulation. (Image: Micro Power Systems)

Figure 3. This simple example of stepping through the modulation frequency is seen in both time and frequency domains. (Image: Texas Instruments)

Q: How can you visualize these modulations and their effects?
A: You have to look at them simultaneously in both the time and frequency domains; one specific case is seen in Figure 3.

Q: What about the spectrum for pseudorandom modulation?
A: That gets more complicated because it depends on the pseudorandom sequence, and there are many choices. For example, there is a digital spread-spectrum technique called dual random spread spectrum (DRSS), which superimposes two separate modulation sequences and allows tailoring the spectrum to have the least energy in some sensitive bands.

Figure 4. This is the time-domain modulation profile of one specific DRSS pattern. (Image: Texas Instruments)

Q: Can you give an example?
A: Figure 4 shows a DRSS modulation profile in the time domain with a triangular envelope targeting lower RBWs, and a superimposed pseudo-random sequence targeting higher RBWs (RBW is resolution bandwidth, a parameter used when establishing the bandwidth of the spectrum analyzer).

Figure 5 shows the emissions performance with this waveform when used with the Texas Instruments LM5156-Q1, a non-synchronous boost controller, with and without DRSS.

Figure 5. EMI performance of the LM5156-Q1 boost controller before and after spread spectrum, using a printed circuit board (PCB) not specifically designed for lower EMI; note the reduction in the AM-radio broadcast-band spectrum. (Image: Texas Instruments)

There are significantly reduced spectral peaks in both the 150-kHz to 30-MHz band and the 30—to 108-MHz band, which are the two key bands for CISPR 25 in the context of automotive applications. Of special concern is the AM radio band of 550 to 1660 kHz, where converter switching noise is a serious concern (note that most, but not all, cars still have an AM radio).

Q: This issue of modulation is getting very complicated. Do I need to know all about it and make these decisions?
A: You do, and you don’t. If your problem is keeping emissions below various EMI regulatory standards such as CISPR 25, there are many spread-spectrum regulators and support components pre-set to provide that capability. When using these, there are just a few simple decisions to make related to the spread-spectrum functionality.

If, on the other hand, you need to minimize EMI in a specific span or spans of the spectrum due to the particulars of your design and its sensitivities, you may need to tailor the modulation to achieve these goals. These issues can include direct interference with a critical signal or perhaps result in intermodulation between the regulator spectrum and other oscillators.

Q: Can you give an example of an easy-to-use, high-performance spread-spectrum IC?
A: Texas Instruments has the TPS8267x family 600-mA, high-efficiency MicroSiP™ step-down converters, measuring just 2.30 × 2.90 × 1.00 mm high (Figure 6). These ICs are “complete” devices; within the package are the switching regulator, inductor, and input/output capacitors; no additional components are required to finish the design.

Figure 6. The TPS8267x family of devices are complete spread-spectrum regulators in a tiny package and are simple to use; they provide well-defined performance for a specific set of outcomes. (Image: Texas Instruments)

The ICs in this family provide 90% efficiency at their nominal 5.5 MHz operating frequency and are offered with many output-voltage values between 1.2 V and 2.1 V. The spread-spectrum architecture varies the switching frequency by approximately ±10% of the nominal switching frequency, using a triangle profile as the modulating signal that is approximately 1.6% of the carrier frequency.

Q: Are there other ICs that serve primarily as external oscillators (clocks) for regulators?
A: Analog Devices offers the LTC6902, a precision, low-power oscillator that drives up to four multiphase outputs in a single small package (Figure 7). A single external resistor sets the nominal oscillator center frequency from 100 kHz to 20 MHz, and it provides user-selectable spread-spectrum frequency-modulation capability.

Figure 7. The multiphase LTC6902 offers a wide nominal operating-frequency range set by a resistor as well as modulation bandwidth set by another resistor; shown are the before and after spectrums for a 500-kHz nominal output with 20% spreading. (Image: Analog Devices)

The SSFM capability modulates the oscillator’s frequency by a pseudorandom noise signal to spread the oscillator’s energy over a wide frequency band. The percentage of frequency spreading is programmable using a single additional external resistor and can be disabled if desired.

Q: Are there any downsides to the use of spread spectrum for reducing EMI?
A: Yes, there are. Depending on the setting of the three spreading parameters cited earlier, and the specifics of the regulator being managed, the regulator can be less stable, go into undesired oscillator modes, and fail to regular to specification. Also, converter efficiency may be degraded anywhere from a few percentage points to much more. Therefore, the application must be implemented carefully

Of course, these concerns are almost entirely absent when using a self-contained, complete spread-spectrum IC, as compared to a do-it-yourself pairing of a modulation scheme and a regulator.

Q: Is the spread spectrum technique also used for other applications?
A: It is. It was (and still is) used for encoding messages (encryption), so they could not be understood by unintended listeners, even if the signal itself was intercepted. This technique was used for radio conversations between President Franklin Roosevelt in the US and Prime Minister Winston Churchill in the UK during World War II for top-secret conversations.

Hard to believe but true, Roosevelt and Churchill had the same 78-rpm phonograph record, and both were synchronized (doing so was a real challenge). It was used to scramble (frequency modulate) and spread the spectrum of their spoken voices and then to demodulate it when received. This was an awkward system, and the requisite electronics filled a large room, but it was all they could do with the technology at the time – and it worked. Today, spread-spectrum encoding is still used for secure links, but with synchronized digital encoding/decoding “keys” on each side.

Q: Are there other uses for spread-spectrum techniques in addition to encryption of communication links?
A: Yes, in what is called wideband communications, the modulated signal is deliberately spread across a wider bandwidth but at lower spectral energy density to avoid both fixed-frequency and varying-frequency noise that may reside in one or more spectral bands. The receiver must be set to capture and decode this wideband signal.

Q: Will this FAQ address spread spectrum for encryption or for overcoming spectrum noise?
A: No. Although many of the underlying principles and initial equations are the same or similar, they relate to different applications with very different specifics, priorities, and implementations. This FAQ focused only on the use of spread-spectrum technology for managing noise from switching regulators.