In a previous technical brief, we looked at spread spectrum modulation in the context of RF systems. We saw that spreading the spectrum of the transmitted signal can make the communication link more robust and reliable. It turns out that the spread-spectrum concept can be applied to digital signals as well, though with a different goal in mind.
Digital circuitry is noisy; there’s no way around it. The rapid transitions that characterize digital signals are sources of intense EMI (electromagnetic interference). To ensure that electronic devices do not cause other nearby electronic devices to malfunction, engineers must do what they can to reduce radiated interference and to make circuits more resistant to received interference.
There are various techniques for dealing with EMI, and as you may have guessed by now, spectrum spreading is one of them.
Reducing the Peak
Actually, I would go further and say that spectrum spreading is a particularly clever technique for mitigating EMI.
Imagine a digital signal operating at a certain frequency, say 25 MHz. The frequency-domain representation of this signal would contain a peak at 25 MHz (i.e., the fundamental frequency) as well as peaks (of lower amplitude) at the harmonic frequencies.
In this article we’ll focus on the fundamental frequency, because that corresponds to the interfering signal that is strongest and therefore most likely to be problematic. But all the harmonics will be affected in the same way as the fundamental, so spreading the spectrum is a valuable technique even if the primary interference concern involves one of the harmonics.
Here is a frequency-domain plot showing the spike at the fundamental.
Now let’s say that your system can function perfectly well even if the frequency of this signal is not always exactly 25 MHz. Maybe all the performance requirements will be met as long as the signal frequency is not lower than 23 MHz and not higher than 27 MHz.
This introduces an interesting option. What if we intentionally vary the frequency of this signal between 23 MHz and 27 MHz? The time-averaged spectrum will now have the general shape shown below.
As you can see, we have just spread the spectrum. The current and voltage are the same, so the total power has not changed—rather, the original power is spread over a larger bandwidth, and thus the peak has been reduced.
This is a significant accomplishment if your goal is to reduce your maximum radiated EMI at a certain frequency (and this is exactly what you need to do when seeking compliance with FCC peak-emission requirements). You have just rendered your device FCC-compliant while maintaining the required performance and avoiding other (perhaps more expensive or complicated) EMI solutions.
The Spread-Spectrum Clock
So far we have been discussing a generic digital signal, but in practice there is one signal that is an excellent candidate for spectrum spreading: the clock. By spreading the spectrum of your clock, you are automatically spreading the spectrum of all the downstream signals that are governed by this clock.
Of course, spreading the spectrum is easier said than done, right? How do we go about continuously varying the frequency of a clock signal? Easy: buy a spread-spectrum oscillator! Here are a few options:
As you can see in the above comparison (taken from the LTC6908 datasheet), spread-spectrum clocking is a powerful yet easy-to-implement tool for reducing radiated EMI.