A new DC/DC controller incorporates active filtering of switching noise on a DC power rail into its basic design.
Electromagnetic interference (EMI) due to noise created by switching-topology power devices, even low-power ones, is an ongoing problem. Not only does noise affect its direct circuitry load, but that conducted and radiated EMI can corrupt nearby circuits; either case is especially serious if low-level sensors are involved.
There are many standards from various regulatory agencies that define maximum allowable limits of EMI as measured under defined conditions. The most commonly cited regulatory dictates are CISPR 25 for automotive applications and CISPR 32 for multimedia equipment, although there are others (CISPR is short for “Comité International Spécial des Perturbations Radioélectriques”).
Not surprisingly, there is no single “best” technique to minimize EMI and keep it below both regulatory limits as well as what the load can tolerate. The solutions are generally a function of noise frequency and how much attenuation is needed, and a combination of two or more approaches is often needed. Some of these noise-reducing techniques are internal to the offending switcher design itself (Figure 1).
Figure 1 There’s no single, simple solution to suppression of conducted and radiated noise; one or more techniques are needed depending on noise frequency and magnitude. Source: Texas Instruments
Even so, external filtering is often still needed, generally consisting of resistors, inductors, and capacitors (RLC) in a passive-filter topology. But there are limitations on such filter performance as well as the bulk of their passive components.
There are other options beyond the standard passive filter, such as “active filtering,” which Texas Instruments is incorporating into their LM25149 synchronous buck DC/DC controller. Active filtering is somewhat analogous to active noise cancellation used in headphones and even some automobiles, where an equal and opposite signal to the undesired one is added to the signal plus noise, resulting in fairly good cancellation. However, the noise is non-electronic and must be captured via microphones, which brings many complex issues.
In contrast, the switching-supply EMI is already in electronic form, thus making it much easier to capture, invert, and cancel. Active filtering is a case where the more you (or the circuit) know about the specifics of the noise – beyond its general characteristics of magnitude, frequency, and perhaps probability distribution – the easier it is to develop schemes to counter it.
Although this seems like a power-and-noise scenario, it’s under the broader subject of signal theory. Using the classifications analyzed by H.L. Van Trees in his classic equation-intensive, three-volume textbook series “Detection, Estimation, and Modulation Theory,” this is a case of the known signal (here, a flat voltage-output rail) with known noise (the associated EMI noise with its known frequency span and general characteristics). As such, it is the type of signal-and-noise challenge that is most amenable to a successful solution.
The benefit of active filtering in this case is that the overall bulk of the active-filtering DC/DC solution is reduced, primarily since only far-smaller passives are needed for the cancellation circuit rather than the larger ones of brute force passive filters (Figure 2). Texas Instruments provides data and graphs showing the resultant filtering performance (References 2 and 3).
Figure 2 While the commonly-used passive filter (a) can provide adequate noise reduction, active filtering (b) does so with smaller passive components, resulting in greatly-reduced footprint and volume. Source: Texas Instruments
Note that there is some potential for confusion between the phrases active filter and active filtering. Engineers are familiar with passive filters, which, as their name implies, implement an RLC topology. For many EEs, the mere mention of filters – especially passive ones – brings back memories of arcane equations and topologies such as Butterworth, Chebyshev, Bessel, elliptical (Cauer), Gaussian, and Sallen-Key, but that’s another story.
In contrast, the active filter uses amplifiers (op amps) to set gain and phase of the filter response. Active filters offer many other functional advantages over passive filters, such as high input impedance and low output impedance and thus provide good isolation between stages. This greatly eases cascading multiple stages to improve filter characteristics. However, an active filter (the noun) is still just a forward signal-path filter with no reverse-cancellation path. It is not the same as active filtering (the verb).
Have you ever resorted to using advanced approaches such as active filtering? How much did you know about the desired signal and the associated noise? Did the result work as well as you expected, or did “real world” imperfections compromise the performance you achieved?
This article was originally published on EDN.
Bill Schweber is an EE who has written three textbooks, hundreds of technical articles, opinion columns, and product features.