Utilizing Dual-Random Spread Spectrum to improve EMI performance
If you had the option to use a smaller, less expensive EMI filter, would you take it? You might answer quickly yes, but only if you could guarantee avoiding the pitfalls of various modulation schemes. See how Dual-Random Spread Spectrum can minimize filter size and cost, while leveraging the advantages of triangular and pseudo-random modulation to achieve excellent EMI performance at high and low frequencies.
Resources
[CHIME]
[WHOOSH]
[SWISH]
Hello, and welcome to TI's low EMI training series. I'm Sam Jaffe, and in this video, we'll be discussing our advanced spread spectrum feature called dual-random spread spectrum.
First, let's talk a little bit about spread spectrum, and why we need it, and how it works. A switching regulator like a buck regulator will generate EMI at its switching frequency and the harmonics of that frequency. We can see the switch waveform in red, and we can see the period of each switch. The low frequency conducted EMI waveform on the top right shows the corresponding EMI emissions on an EMI plot. The red lines represent the limit lines for a typical EMI test. If the blue line exceeds the red line, you've failed the test.
We can see that the higher harmonics on the high frequency conducted EMI plot on the bottom right exceed the limits at the FM band, around 100 megahertz. Spread spectrum dithers the switching frequency to spread this energy. Instead of switching at a constant frequency with well-aligned harmonics, we switch at multiple frequencies centered at our original frequency in order to attenuate the fundamental and blend the harmonic energies into a smooth, averaged waveform. This gives us more margin, which helps us pass our EMI tests, and can even help reduce the necessary size and cost of our EMI filters.
So how do we implement this in a buck converter IC? The two most common types of dithering are triangular modulation and pseudo-random modulation. Triangular modulation dithers the switching frequency up and down in a triangle shape, where a pseudo-random modulation dithers the frequency randomly for every switch cycle.
Triangular modulation does a great job of spreading the energy evenly at the fundamental frequency. This gives us more margin at the low frequencies if we have a limit that we need to stay below. This allows us to reduce our EMI filter size and cost.
Pseudo-random modulation, depending on its implementation, may not attenuate the fundamental as evenly as triangular modulation. We can see that it is a little bit more peaky, which doesn't allow us to reduce our filter size and cost as much as triangular modulation. At the higher frequencies, however, pseudo-random generally performed better.
Higher frequency EMI scans spend less time measuring the EMI per data point, which means it might not measure long enough to capture the entire triangular ramp up or down. This effectively reduces the spread, which reduces the effectiveness of the modulation. Pseudo-random, on the other hand, can jump between frequencies much faster, maintaining its modulation spread in the high frequency sweeps.
Another downside to triangular modulation is at the very low frequencies in the audible band. Most triangular modulation schemes have a triangular modulation frequency between 4 kilohertz and 15 kilohertz, right in the audible region. This modulation shows up as noise on the output and input, which can create an audible tone at that frequency if the noise couples into nearby audio circuitry or even a ceramic capacitor on a PCB which resonates near that frequency.
Pseudo-random, on the other hand, does not have this tone issue. There's still noise, but it's spread, sounding more like white noise, which is much less noticeable than a high-pitched tone.
So how can we create a better method which uses the best of both schemes? Our optimized scheme is called dual-random spread spectrum, and you'll see why in a minute. First, we start with triangular modulation. To address the audible tone issue at the triangle frequency, we dither the triangle frequency randomly. This will spread that tone into something more like white noise.
Next, to address the suboptimal high frequency performance, we add pseudo-random on top of the modulation. This is the foundation of DRSS, to achieve great performance at the fundamental at higher frequencies and in the audible band. But there's one more trick which sets this scheme apart from the rest.
It's one thing to spread the audible tone into white noise, but we've gone a step further to actively reduce this noise. And to understand this, first we have to discuss where this noise comes from.
As the switching frequency goes up and down, the inductor current ripple amplitude goes up and down. In a peak or valley current-controlled device, the current command will not be able to perfectly compensate for this change in peak to peak ripple. This animation exaggerates the issue by keeping the peak current command fixed in place.
We see that the change in frequency results in a change in inductor ripple amplitude, which results in a change in output current. This rippling output current creates a rippling output voltage, which can couple to other circuitry causing this audible noise issue. We also see a ripple on the input caused by the change in RMS current from the inductor ripple modulation as well as the rippling output current from the interaction with the peak current command.
To address these issues, we can preemptively scale the peak current command with the frequency to eliminate the output current ripple and subsequently eliminate the output voltage ripple and most of the input ripple. When the frequency goes up, the inductor ripple amplitude decreases. The peak current command goes down, and the output current and voltage stay constant. When the frequency goes down, the inductor ripple amplitude increases, the peak current command goes up, and again, output current and voltage stay constant.
Dual-random spread spectrum enables best-in-class EMI performance. It takes advantage of the great fundamental frequency attenuation of triangular modulation, and it utilizes the optimal high frequency performance of pseudo-random modulation. In addition, it actively spreads and reduces the input and output noise at the very low frequency audible band. This allows a power designer to rest easy knowing they don't have to weigh the pros and cons of each modulation scheme. They can simply pick a device with DRSS and spend that extra time working on the rest of the design.
This has been a discussion of spread spectrum, common techniques, and the advantages of dual-random spread spectrum in achieving optimal spread spectrum performance in our buck converters. Check out our other low EMI videos on integrated capacitors, HotRod package, and more. Thanks for watching.
This video is part of a series
-
Designing a low-EMI power supply
video-playlist (19 videos)