Evanescent Mode Ridge Waveguide Bandpass Filter for High Frequency Applications

Background

A band pass filter is a key component of various applications that allows signals with certain frequency bands to pass through and blocks signals with other frequencies. This component is important in wireless technology, such as in communication or radar systems, to extract the desired signals and remove unwanted signals, such as interference or noise.

To implement such a component at microwave frequencies, several well-known technologies can be employed, e.g. lumped components (with inductors/capacitors or LC filter) or distributed elements implemented in planar (e.g. microstrip or stripline) and non-planar (e.g. coaxial combline or waveguide) topology. Despite their bulky structure, non-planar filters offer a high quality (Ǫ_u) factor (typically above 1000) among those technologies, resulting in low insertion loss in the passband region. For narrowband filters, which offer high selectivity, a high Ǫ-factor is important as the filter curve is greatly distorted at the corners of the passband, as shown in Fig. 1. High-quality factor structures are also interesting for high-frequency applications, where most of the structures become more lossy compared to the same structures at low frequencies. In addition, non-planar filters typically offer better power handling compared to LC or planar filters.

Fig. 1. Effect of the quality factor (Q_u) on a filter. A circuit model with three different values of Q_u (100, 1000 and 10000) for a bandpass filter at 25 GHz with a bandwidth of 500 MHz is simulated. When the Q_u is too low, for example, 100, the S-Parameter curves are heavily distorted, especially at the edges of the passband. The rounded edges, along with the increase in insertion loss, reduce the filter selectivity. Therefore, a filter with a high-quality factor is often required. Note: the red curves depict S₁₁ and the blue curves depict S₂₁.

Fig. 2. Examples of non-planar microwave filters: (a) rectangular waveguide bandpass filter, (b) combline filter with iris coupling and (c) combline filter with distance coupling.

A Rectangular waveguide is an example of a popular topology used to implement a non-planar microwave filter, especially at high frequencies (mostly above 20 GHz). As the frequency increases, the size of the waveguide becomes smaller. As can be seen in Fig. 2a, the filter consists of several waveguide resonators with a length around half of the guide wavelength coupled to each other through irises. The fabrication can be realized by using a CNC milling machine. However, while waveguide filters can support various propagation modes, the operating frequency range of a waveguide structure is limited. For example, waveguide-based bandpass filters designed with the fundamental resonant mode (so-called TE101-mode) can only reject signals commonly up to 1.5 times the center frequency (f₀) of the filter. For many applications such as 5G/6G, this is not sufficient.

On the other hand, coaxial combline filters have a better out-of-band rejection, typically up to 3 times the center frequency of the filter. A post inside a cavity, which has a length less than a quarter of the guide wavelength, can be utilized as the resonator (see Fig. 2b). On the opposite side, tuning screws can be added for post-fabrication tuning. An iris can be opened between two resonators to form the coupling structure between them. In the case of broadband filters, the wall between the resonators can be eliminated so that the coupling depends on the distance between the post resonators. For high-frequency applications, the fabrication is, unfortunately, more difficult than for waveguide filters, because the dimensions of a combline filter are much smaller than that of a waveguide, especially when it comes to the coupling between the first/last resonator to the input/output connector.

Therefore, the goal is to develop a filter technology that combines the excellent rejection capability of a combline filter with the easy fabrication of a waveguide filter.

Ridge Waveguide Bandpass Filter: Basic Structure and Design Principle

One of the possible ways to increase the rejection frequency range is to insert a ridge structure inside a waveguide as depicted in Fig. 3. A rectangular waveguide without ridges has a certain cutoff frequency that is inversely proportional to the width of the waveguide. Every signal with frequencies below the cutoff will be attenuated exponentially. This condition is called evanescent mode. The introduction of capacitive loading from the ridge inside a waveguide will lower the cutoff frequency. Employing a ridge as a resonator inside a waveguide, the separation between the evanescent operating frequency and the fundamental mode of the original waveguide is greater, resulting in a wider operating or rejection frequency range. This separation can be controlled by designing the cross-sectional variables a (waveguide width), b (waveguide height), s (ridge width) and d (gap between the ridge and the top of the housing), as shown in Fig. 3.

Fig. 3. Ridge waveguide structure (left) with cross-section (right).

There exists a tradeoff in the design. For example, a smaller gap d between the ridge and the waveguide housing increases the separation between modes and makes the overall waveguide size smaller. The drawback is a reduction of the quality factor and the power handling capability of the filter. To optimize the cross-sectional dimensions of the ridge for a certain out-of-band rejection range, a mode-matching-based simulation can be performed. In this case, Wainwright Instruments has been involved in developing a calculation tool in cooperation with the Chair of High-Frequency Engineering, Technical University of Munich, Germany.

Fig. 4. Arrangement of ridge resonators inside the filter.

The arrangement of ridge resonators inside a bandpass filter can be seen in Fig. 4. Between the ridges, there are small gaps operating in evanescent mode, which enables coupling between those resonators. The strength of the coupling is determined by the gap length (l_g1, l_g2, l_g3, …). On the other hand, the length of the ridge (l_r1, l_r2, l_r3, …) determines the resonant frequency of the filter.

Those lengths must be designed according to the required specifications, e.g. the center frequency and bandwidth, which define the passband region of the filter, as well as the required rejection at specific frequencies – the stopband region. The value of the rejection determines how many resonators will be required, i.e. the filter order. More resonators result in greater rejection in the stopband region, or in other words, in a steeper transition (slope) between the passband and the stopband. The filter responses can be represented by certain mathematical functions or polynomials, for example, Chebyshev, which provides an optimal tradeoff between the transition slope and the variation of magnitude or ripple in the passband. The Chebyshev filter design usually starts from a lowpass filter prototype, which is designed according to the filter specifications. The lowpass component values g_i are obtained and are related to the coupling coefficient value (k_i,i+1) between resonators. For the interested reader, a general overview of filter design and calculation from a lowpass prototype can be found in classical literature, e.g. “Microwave Filters, Impedance Matching Networks and Coupling Structures” by Matthaei, et. al.

Now that the needed couplings are known, the gap lengths can be calculated using 3D electromagnetic full-wave simulation, which can map a certain gap length to a certain coupling coefficient. Theoretically, the ridge resonators should have lengths equal to half the wavelength of the ridge waveguide. However, the evanescent coupling structures also affect the resonant frequency. Therefore, a correction of the ridge length should be carried out during simulation or optimization.

As the feeding structure, either a waveguide or other transmission lines such as coaxial lines can be used. In several cases, a transition line between the feeder and the first ridge resonator is necessary, especially when direct feeding is too difficult to realize. A quarter wavelength impedance transformer with several sections can be utilized for this purpose, as shown in Fig. 5.

Fig. 5. Ǫuarter wavelength impedance transformer between the ridge and waveguide adapter.

Fabrication and Measurement

The filter can be fabricated by using a CNC milling machine. A split block housing can be milled and screwed together to form the filter, as shown in Fig. 6. If possible, tuning screws can be placed on the opposite side of the ridge resonator. This screw will change the capacitance and thus the resonant frequency, in order to compensate for the frequency shift due to fabrication tolerances.

Fig. 6. Typical structure of a fabricated split block housing: (a) top part and (b) bottom part. On the top part, some holes for the tuning screws are visible.

Several filters have been fabricated and measured. Three design examples are shown below.

Case I: Filter with Center Frequency of 25 GHz, Bandwidth of 1 GHz

A Chebyshev bandpass filter with 5 resonators is demonstrated at a center frequency (f₀) of 25 GHz with 1 GHz bandwidth (4%), 30 dB rejection at f₀ ± 1.250 GHz, and a return loss greater than 14 dB. An impedance transformer is used between the coaxial pin/waveguide transition and the ridge waveguide feeder for the filter. As can be seen from Fig. 7a, the filter is silver-plated. According to the measurement results in Fig. 7b and 7c, the filter has an insertion loss of around 1 dB (equivalent Ǫ-factor of around 1000) with an excellent out-of-band response (rejection of more than 30 dB) up to 52 GHz (more than 2f₀). It can be inferred here that this filter has a better spurious response than a standard waveguide filter and a loss or Ǫ-factor that is comparable to a combline filter, which is difficult to fabricate at this frequency.

Fig 7. An evanescent ridge waveguide bandpass filter example at 25 GHz with 5 resonators and 1 GHz bandwidth (4%): (a) fabricated filter, (b) measurement results in the vicinity of passband and (c) out-of-band measurement results.

Case II: Interdigital Filter with Center Frequency of 20 GHz, Bandwidth of 2 GHz

In cases where the filter bandwidth is too large, the gap between the ridge resonators is too small to be fabricated with a milling machine, especially around the first resonator. Coupling coefficient enlargement can be achieved by inverting the neighboring resonator so that the base is on the top of the housing instead of the bottom, as illustrated in Fig. 8. This so-called interdigital configuration is well-known in the design of combline filters when large bandwidths are required.

Fig. 8. Interdigital ridge waveguide bandpass filter.

A 5-resonator interdigital filter with a center frequency of 20 GHz, bandwidth 2 GHz (10%) and 30 dB rejection at f₀ ± 2.5 GHz has been fabricated, as can be seen in Fig. 9a. Although the metallization material here is brass, a better material can be used in future designs,

e.g. silver as in case I. The S-Parameter of the filter has also been measured and plotted in Fig. 9b and 9c, for the passband and out-of-band response, respectively. The insertion loss is almost the same as in case I, around 1 dB. Although brass has an inferior conductivity compared to silver, the fractional bandwidth of this filter is more than twice that of the filter in case I, which helps in reducing the insertion loss.

Fig. 9. An interdigital ridge waveguide bandpass filter example at 20 GHz with 5 resonators and 2 GHz bandwidth (10%): (a) fabricated filter, (b) measurement results in the vicinity of passband and (c) out-of-band measurement results.

Case III: mmWave Applications, Filter with Center Frequency of 47.7 GHz, Bandwidth 1 GHz

The evanescent ridge waveguide concept can also be utilized at millimeter wave frequencies. Wainwright Instruments GmbH is conducting research and development in this area. For example, a 5-resonators filter, which has a passband of 47.2 – 48.2 GHz (f₀= 47.7 GHz, bandwidth 1 GHz or ca. 2,1%), has been designed, fabricated and measured, as shown in Fig. 10. The filter can be employed for 5G NR channel n262.

Fig. 10. An interdigital ridge waveguide bandpass filter example at 47.7 GHz with 5 resonators and 1 GHz bandwidth (2,1%): (a) fabricated filter, (b) measurement results in the vicinity of passband and (c) out-of-band measurement results.

Due to the brass metallization and smaller bandwidth, the filter shows a higher insertion loss (ca. 4 – 5 dB). Nevertheless, it can be improved in the future with silver plating. In addition, it has the advantage that the out-of-band rejection range can reach 100 GHz (ca. 2.1f₀).

Conclusion

In this article, an evanescent ridge waveguide bandpass filter has been demonstrated from 20 GHz to 47.7 GHz with various bandwidths. In general, the performance of these filters lies in between that of waveguide and coaxial combline filters, which means that these filters are suitable when a wide out-of-band rejection range is needed, but a reasonably low insertion loss is also required.