Analog filters using inductors, capacitors, and resistors are critical to most circuits and systems. Whether as passive or active designs, it sometimes seems that they have been studied “to infinity and beyond” via theoretical constructs with very intense math, with practical “application note” designs and BOMs, and even with hands-on circuits with actual physical-construction details. This makes sense because filters play multiple essential roles in the signal path, regardless of application area. Whether low-pass, high-pass, bandpass, or notch, they are necessary even if they don’t appear to add value to the signal as they improve SNR, minimize interference from nearby channels, and attenuate 50/60 Hz pickup, to cite just a few of their many uses.

Still, classic filter theory is a subject which can drive students and engineers to weird states of mind, as they come in mind-boggling and mind-numbing versions and realizations There are so many distinct topologies, such as pi-filter (Figure 1), Chebyshev, Sallen-Key, Butterworth, Cauer (elliptical), and Gaussian, to cite a few. Then there are their attributes: first order, second order, roll-off, passband ripple, stopband ripple, phase performance, balanced (differential) – that’s another long list.

20190412_Filter_P1

Figure 1: Pi Filter (Quora)

20190412_Filter_P2

Figure 2: Rose-Hulman Sallen-Key Filter

(Of course, those are just your classic all-analog filters. There are also quasi-analog switched-capacitor filters which use charge-balancing and clocked switching among multiple capacitors to implement filter functions. These are valuable additions to the filter roster as they are compatible with IC processes and can eliminate need for discrete-component filters in many cases.)

Classic analog filters can be used into the hundreds of MHz and even GHz ranges. However, these lumped-element filters are increasingly difficult to design and then fabricate successfully for these higher frequencies. Parasitics along with component tolerance and drift make them a real challenge, and these filters often need to be individually trimmed to counter their difficult-to-model realities.

If there was no alternative to discrete-component filters, many of the devices with which we are surrounded would be impractical due to size, performance, consistency, and cost. Obviously, these products are very practical, and it’s largely due to using a radically different approach to analog filters: the surface acoustic wave (SAW) and the bulk acoustic wave (BAW) filters (and their film bulk acoustic resonator – FBAR – sibling). SAW and BAW technologies have matured over the past decades to create low-cost, high-performance devices with no relationship to discrete-component analog filters.

Instead, they exploit the well-known, versatile piezoelectric effect, transforming electrical energy into acoustic waves which travel along the surface (SAW) or within (BAW) an engineered ceramic-crystal material (Reference 1). SAW devices are viable up to about 1 GHz, while BAW devices conveniently overlap in coverage, starting below 1 GHz and going into the multi-GHz regime. One thing they have in common is that they both obsolete the need for study of classic lumped-element analog-filter design theory and practice.

20190412_Filter_P3

Figure 3: Basic SAW Filter (SAW, BAW, and the future of wireless Figure 1)

20190412_Filter_P4

Figure 4: SAW, BAW, and the future of wireless, Figure 2

Reality is that classic analog filters have a much smaller role in much of today’s design activity, yet they are still taught in detail. I did a non-scientific survey of undergraduate EE courses offered at about a dozen major universities, and all but two listed classic filter design as an available course (although it wasn’t clear if this was a required or optional course). Only two had courses on SAW and BAW devices.

Is this because instructors feel that classic design, with all its math and insight, is still needed? Or is it because they are comfortable teaching it? Or because there is so much vetted material out there at so many levels to easily support such a course? Your guess is as good as mine.

My view is that classic analog filter theory should be now taught as a quick overview/survey course: what filters do, why needed, the different types and key characteristics, and the parameters used to quantify their performance – and with minimal math. Those who eventually need to know more, (or are enamored of the topic) for whatever reason, can easily find it. Instead, focus on the filters of the present and future including SAW, BAW, resonant structure, and even waveguide filters for microwaves (yes, still widely used), see References. Even better would be to look at the filters used for >10 GHz applications such as 5G or 77-GHz car radar.

What’s your view on classic analog filters? Do you love them but are ready to leave them? Do they make you squirm, and you’d like to see them get less-rigorous attention? Or are they a critical building-block along the lines of Maxwell’s equations, one in which every EE needs to be well-versed, beyond just merely acquainted? Or perhaps your view is more cynical, and you feel that filter analysis is really more of a qualifying test to see if the student is really a good fit for the analog-design world?