Role of Filters within Any Communication System
Filters are an indispensable part of any communication system, whether wired or wireless. The main attributes of a filter are traditionally specified as follows: low insertion loss within the passband to avoid thermal issues, brick-wall transition from the passband to the stopband, and high rejection in the stopband as far away from the corner frequency as possible. These three attributes are not adequate in today’s terms because of the complexity introduced by overcrowding of the frequency spectrum by the voracious demands for sophisticated communication systems.
Given this, and the need for more bandwidth (BW), the pressure for higher dynamic range and better protection against adjacent channel interference has increased significantly, even compared to the very recent past. As more systems are simultaneously deployed, the need for clean transmission is paramount and rigorously enforced by the Federal Communications Commission (FCC). The system’s proprietor must assume the responsibility to ensure compliance with this stringent spectral management specification.
Under such constraints, the practice of using filters to remove intermodulation distortion (IMD) products after they are generated is inadequate and inefficient. To keep up with today’s requirements, filters must be used to prevent the generation of IMD, as well as perform the traditional filtering duties. Prevention is always better than a cure.
Figure 1 shows an ideal lumped circuit of a reflective lowpass filter, within which only reactive elements are present. A reflective filter sends stopband signals back to the source, which usually is a nonlinear active device such as an amplifier or a mixer. The elements responsible for reflecting the stopband signals are the shunt capacitors to ground in Figure 1. The junction comprised of the two series inductors and the shunt capacitor is that of a rudimentary diplexer circuit, separating and routing the passband and stopband signals to different paths. The passband signals follow a straight path from left to right because they see a favorable impedance along that route. The stopband signals follow the shunt capacitors to see a grossly mismatched impedance of the ground plane and reflect back to the source, without much reaching the other side.
The deployment of reflective filters in today’s complex systems causes serious, yet preventable, issues such as instability, dynamic range degradation, efficiency reduction and thermal management upheavals [1, 2, 3, 4]. These issues, if ignored for short term and individual profits, can lead to more serious, long-term and collective, problems such as poor mean-time-between-failure (MTBF) and costly maintenance cycles.
An absorptive filter captures stopband signals such as harmonics and IMD to eliminate unceasing reflections and standing waves at all out-of-band frequencies . Also, absorptive filters can be easily cascaded together if needed, to obtain sharper cut-off properties, without passband distortion. Figure 2 shows an ideal lumped circuit of an absorptive lowpass filter in which resistive elements are deployed to achieve the reflectionless characteristic. Note that the shunt capacitors in Figure 2 are augmented with resistive networks allowing the absorption of stopband energies . In this case, the stopband signals are navigating through the capacitors and then absorbed by the resistors, causing negligible reflection back to the source.
Why Do Stopband Reflections Adversely Affect a System?
Stopband reflections can, and do, cause abnormal system behaviors that are intermittent or steady in nature, with the former being most difficult to troubleshoot and to remedy.
In fact, Werlatone developed our line of absorptive filter products starting in 2010 at the request of various customers who were puzzled by the bizarre system behaviors they could not resolve. These plagued systems worked sometimes but not all the time and were influenced oddly by cable lengths, locations of transmitter mount and the presence or absence of nearby objects. These issues were largely suppressed when absorptive filters were deployed. As it turned out, the stopband standing waves were the main cause of high-frequency radiations through bias ports, inter-component nefarious interactions and inconsistent cable length dependencies.
In a mixer circuit, the reflections of stopband signals back to the nonlinear active element, be it a diode or a transistor, can cause any or all the following effects: occasional oscillations, the generation of higher-order IMD as well as degradation in dynamic range and subsequently, electromagnetic interference (EMI) issues. Instability occurs because the reflected waves alter the instantaneous impedance seen by the active device at one or numerous out-of-band frequencies. The change in impedance can steer a stable operation toward a region of instability enough to initiate oscillation. Worse yet, the oscillation can sometimes be strong enough to self-bias the active device to cause a secondary wave of ill effects, such as a reduction in conversion efficiency and operational hyper-sensitivity. Stopband reflected signals can also interact with the existing IMD of a mixer to generate more spurious signals. The net result is a reduction in dynamic range. Furthermore, stopband-frequency standing waves leak through bias ports and radiative openings, as a means of relieving bottled up energies, to cause EMI issues. These ill effects are also observed in similar nonlinear circuits such as frequency multiplier and divider . Figure 3 shows a block diagram of a mixer circuit in which reflective filters are deployed, illustrating the IMD infested spectral responses. Figure 4 shows a block diagram of a mixer circuit in which absorptive filters are deployed, illustrating the cleaned up spectral responses.
In an amplifier circuit, stopband reflections can also cause instability and high-order IMD as seen in a mixer circuit. But more seriously, they can also trigger a drastic reduction in power-added efficiency (PAE) due to an impedance load-pull effect on the transistors at the harmonic frequencies.
Why would reflected harmonic energy affect power-added efficiency? Recall that one of the techniques to improve efficiency in a power amplifier is to utilize harmonic injection for shaping the voltage and current waveforms. The waveforms are shaped to minimize the duration of time-domain overlapping regions in which both the voltage and current magnitudes are simultaneously not zero. In a proper waveform-shaping amplifier, the voltage is non-zero only when the current is zero, and vice versa, hence minimizing the dissipated power. This technique is well known and utilized in class E and F switching amplifiers . Figure 5a shows ideal voltage and current waveforms of a class A amplifier in which no waveform shaping is attempted. As can be seen in Figure 5a, there are many regions where both the voltage and current are concurrently non-zero, leading to high dissipative loss. Figure 5b, in contrast, shows ideal voltage and current waveforms of a class F amplifier in which waveform shaping is performed. As shown in Figure 5b, there are no instances of non-zero overlap between voltage and current, leading to the idealized efficiency of 100%.
Conversely, when the useful harmonic injection technique is being exercised without permission or awareness, it can do more harm than good. Uncontrolled reflected harmonic signals from reflective filters may degrade efficiency by modifying the waveform shapes in a menacing manner. In some cases, the modification of the waveform shapes is erratic in nature, making system integration and testing difficult and costly. In a marginally efficient amplifier, any degradation in PAE can lead to serious, if not catastrophic, thermal issues. In turn, these issues can directly contribute to ballooning costs due to size and weight inflations of the heatsinks and housing. Poor efficiency directly impacts the MTBF and maintenance cost of the system under consideration.
In other words, a power amplifier designer takes great care to shape the voltage and current waveforms within his/her circuit to achieve the utmost efficiency possible. Reflective filters precariously inserted after or before the power amplifier in the system chain may allow uncontrolled harmonic reflections to degrade his/her masterpiece.
Figure 6 shows the simulated results of two wideband power amplifier assemblies, one employing reflective and the other absorptive filters. The class AB amplifier modules in both assemblies are identical and designed using Cree’s GaN transistor models, inside AWR Microwave Office. There is no special technique applied in this distributed design to peak the output power or PAE over the 30-512 MHz band, since the circuit is meant to reveal the roles of the filters. The harmonic lowpass filters are placed immediately after the amplifier modules, inside the assemblies. The reflective assembly responses are blue (output power, triangle symbol, left axis) and pink (PAE, square symbol, right axis). The absorptive assembly responses are brown (output power, diamond symbol, left axis) and red (PAE, hourglass symbol, right axis).
As can be seen in Figure 6, the absorptive responses show consistently higher output power and PAE and contain less ripples. Specifically, the PAE of the absorptive assembly is 12% at around 380 MHz whereas the reflective is 5%, which points to a serious thermal issue in the reflective choice. The curves shown in Figure 6 illustrate the need for absorptive filters in critical circuit locations where stopband reflections can and do cause issues.
In summary, in case of uncertainty about what to do with harmonic signals, terminating them in absorptive loads is the most predictable and sensible course of action, particularly over a wide bandwidth.
How Does an Absorptive (Reflectionless) Filter Work?
A reflective filter provides a complete reflection of stopband signals back to the source. Figure 7 shows an ideal S-parameter response of a reflective lowpass filter. Note the full reflection (S11 at 0dB) in the stopband.
An absorptive filter, on the other hand, possesses multiple diplexer-like junctions that separate the low-frequency signals from their high-frequency counterparts. The high-frequency signals are routed to a different part of the circuit and then resistively terminated to keep them from wreaking havoc. Figure 8 shows an ideal response of an absorptive lowpass filter. Note the good match at all frequencies (S11 well below -20dB), both in the passband and stopband.
Typically, an absorptive filter may require more discrete elements than its reflective counterpart, perhaps twice as many in some designs. At first glance, the increase in parts seems to be a poor choice. This increase in component-level cost, however, is well justified because such an investment more than pays for itself at the system level. It allows for the reduction, if not the elimination, of countless hours of system-level troubleshooting and tuning. Changes in logistics such as shape of mounting platform, antenna relocation, and component line-up no longer require a re-trimming of the power components. The roll-out plans are no longer dependent on last minute logistical changes, speeding up deployment.
If not properly designed, an absorptive filter can incur higher insertion loss than its reflective counterpart because of the additional lossy (resistive) elements. However, Werlatone has successfully produced absorptive filters that can handle as much as 5000 W in CW operations in the HF frequency range and hundreds of watts at higher frequencies.
In summary, utilizing the wrong type of filters inside a system can lead to the deterioration of system stability, efficiency and spectral purity that are unacceptable in today’s electromagnetic environment and standards.
Be a Team Player
1. Component vs system engineering: The intrinsic see-saw dynamics between a component engineer negotiating the specifications and a system engineer are a fact of life. This occurs because these two engineers are driven by two distinct objectives. The system engineer over-specifies a component to add margins for unforeseen issues inherent in new installations. The component engineer asks for specification relief to ensure lower cost and to improve manufacturing yields. The result is always a poor compromise, with initial system cost taking the highest priority, given a tight budget. The aggregate cost of all the parts is minimized to lower the final system price.
However, inexpensive individual components do not always play well together within a system. For example, in a system in which reflective filters are used, performance in and out of band and the physical location of the parts may cause serious and persistent problems.Changes in physical layout and packaging are inevitable as a system evolves. Mitigating the effects of these “growing pains” fully justifies the cost of the more capable type of filters, namely absorptive. Time to market can be reduced by avoiding internal conflicts where adjacent components work against one another, lowering efficiency and exacerbating IMD performance in the process. The real price, at the system level, must include the careful consideration for consistency and reproducibility. Look to absorptive filters for providing this consistency.
2. Supplier vs customer: Similarly, given the complexity and substantial number of systems being deployed, a supplier can no longer deliver to his customer a spec-compliant component and relinquish responsibilities when system issues arise. If the customer fails in his system endeavour, the component supplier does not receive the high volume follow up order. Thus, a plug-and-play component, a filter in the context of this paper, for both in-band and out-of-band performance is crucial. As traditionally practiced, when a supplier saves the component cost by reducing the parts count inside a filter and delivers a reflective unit, he may look like a hero at the contract signing ceremony. However, if the said reflective filter in turn generates system issues, such a supplier may quickly find himself in a much less enviable position. In fact, that supplier causes at least a delay in system deployment and at worst a contract cancellation.
3. The absorptive filter solution: Absorptive filters, beside the aforementioned technical benefits, can augment the shortcomings and improve the accuracy of today’s harmonic-balance nonlinear simulation tools. These filters provide explicitly the resistive terminations of all electrical signals, be they fundamental, intermodulation, or harmonic in nature. The active-circuit designers can have faith in their simulation results and thus have meaningful correlation between simulated & measured data. The system engineer can reap maximum benefits as integration issues evaporate. With the absorptive filter solution, it is truly a win-win scenario because the component and system engineers work together seamlessly within a predictable and cohesive workflow. The system works as intended when all the pieces of the puzzle are plug-and-play, not plug-and-pray.
What Werlatone Brings to the Game
Werlatone’s mantra is “We deliver power and bandwidth.” We thrive where others shy away, because we work hard to account for every 0.1 dB of insertion loss gone astray. Less loss means more power and smaller size. Smaller size means wider bandwidth. Thus, our attention to details pushes back on the laws, or limitations, of physics.
Furthermore, as an agile company, Werlatone has engaged in many customization projects throughout our years in business. We have produced non-50-Ohm absorptive filters used at the transistor level or at the antenna level to achieve optimal system benefits. Our kW-class of absorptive filter products serves as an industry first, given our ability to minimize insertion loss even within an absorptive environment. Our unique experience in high power designs covering multiple octaves in BW contributes directly to the success of our absorptive filter product line.
2. “Reflectionless Filters,” Morgan, Artech House, ISBN 9781630813277, 2017.
3. “Improving Linearity by Using Absorptive Filters,” Mayott, Analog Devices Technical Article.
5. “Planar Constant-Resistance Hybrid Filter,” Werlatone Inc., Podell et al., Patent US8704611, 2014.
6. “Sub-Network Enhanced Reflectionless Filter Topology”, Associated Universities Inc, Morgan, Patent US9705467, 2017.
7. “Thinking Outside the Band: Design and Miniaturization of Absorptive Filters,” Morgan, IEEE Microwave Magazine, vol. 19, no. 7, pp 54-62, November 2018.
8. “Novel Wide Band High-Efficiency Active Harmonic Injection Power Amplifier Concept”, AlMuhaisen et al., MTT-S IMS 2010, pp 664-667, May 2010.