Compact Monolithic 3D-Printed Wideband Filters Using Pole-Generating Resonant Irises

The design concept of a pole-generating resonant iris is demonstrated in rectangular waveguide filters in this paper. Different from conventional reactive iris, the resonant iris can generate an extra transmission pole without increasing the number of resonant cavities. As a result, several design advantages can be gained: (i) a more compact filter structure; (ii) an ability to realize strong coupling and therefore wide bandwidth; and (iii) a new polarization rotation capability. Two third-order Chebyshev filters are designed and implemented, demonstrating the miniaturization and polarization rotation feature. A fifth-order Chebyshev filter with 20% fractional bandwidth is presented to show the capability of realizing wideband. This also demonstrates the realization of asymmetric coupling between the resonant iris and the cavity resonator on either side. An approach to control and extract the coupling between the iris and the cavity resonator has also been presented. To manufacture the intricate asymmetric iris structure, all the presented filters are printed monolithically using selective laser melting technique. Excellent agreement between the measurements and simulations has been achieved, verifying the design concept as well as the additive manufacturing capability in microwave waveguide devices.


I. INTRODUCTION
Waveguide cavity filters have been widely used in satellite and base station applications because of their excellent performance in terms of low loss and high-power handling capacity. The rapid development of wideband communication and carrier aggregation has spurred the growing research activity in wideband waveguide filters [1], [2]. Typical filter requirements for 5G sub-6 GHz system are wideband (around 10%), compact in size, and low profile. In satellite communication transponders, wideband filters are often an indispensable part of the transmit-receive diplexers.
For waveguide filters, there are two typical approaches to realize wideband filtering performance. The first one requires cascading several coupled single-mode resonators [3], [4], [5], [6]. However, the increasing size and mass is a wellknown concern as the order of waveguide filter rises. Wider bandwidth usually needs stronger coupling. The irises that enable strong coupling strength are often associated with the degradation of out-of-band performance due to the higherorder modes and iris resonances. To address this drawback, one method is to push the iris resonance to higher frequency by dividing the single iris to multiple smaller apertures [7], but the coupling strength will decrease. Another approach to realizing wideband filter is to use multi-mode resonators [1], [8], [9], where multiple resonance modes were excited in a single resonator. The overall filter size can hence be reduced while the filter bandwidth can be increased. However, it is often difficult to implement the high-order filters because of the complicated coupling scheme. Filters with multi-mode resonators are also more sensitive to manufacturing tolerances and often suffer from poor temperature stability [10]. In practical filter applications, waveguide filters with explicit This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ coupling schemes and in-line configurations are still highly desirable.
Waveguide cavity filters usually use coupling irises (thin metallic diaphragms containing a concentric opening) to realize the coupling between cavities. From the circuit point of view, the irises can be normally represented as coupling capacitor or inductor. In the past decades, researchers have modified the coupling irises for new or enhanced functionalities in filters. In [11], [12], [13], capacitive and inductive irises were combined to implement a resonator to replace the waveguide cavity resonators, but the coupling between resonators relies on additional quarter wavelength waveguide inverters, which increases the size of the filter. In [14], [15], [16], [17], the frequency-dependent coupling irises were used to generate transmission zeros (TZs). More recently, iris structures have been used to realize multiple passbands or enhance the bandwidth of waveguide filters [18], [19]. However, it should be noted that all above function-enhancing irises rely on complex iris structures. Therefore, a capable and reliable fabrication solution will be the critical enabler to exploit the potential of the resonant iris and new iris structures.
These novel irises structures are traditionally manufactured in separate piece-parts by computer numerical control (CNC) machine and then bolted together [11], [12], [13], [17]. They could also be integrated into the filter, and the entire filter is machined out of multiple parts [15], [16], [19]. However, assembly error is an important source of performance degradation. The passive intermodulation (PIM) product generated at mechanical interconnections could reduce the power handling capacity. Moreover, the low manufacture freedom of conventional subtractive process severely limits the design flexibility of iris structures. Most previous work only look at rectangular resonant irises or non-centered resonant apertures.
We have recently proposed a 3-D printed waveguide filter using pole-generating resonant iris structure in [33]. By replacing the conventional reactive irises with resonant irises, an extra transmission pole (TP) can be implemented within the space previously occupied by the coupling iris. This allows a higher-order filter without excessive size increase. Selective laser melting (SLM) technique was employed to manufacture the unconventional resonant iris structures. Here we have extended the work in [33] in several areas: (1) Asymmetric coupling resonant iris was implemented, only enabled by the manufacturing capability of 3D printing. A new geometrical freedom was introduced to realize the asymmetric coupling on either side of the resonant iris, which is often impractical for traditional machining methods due to the limited lateral space within the iris. (2) Circuit analysis of the pole-generating resonant iris structure was presented. The frequency dependance of the iris coupling and the iris resonance feature were discussed. An intuitive design procedure was introduced. (3) The wideband capability was explored. A filter with 20% fractional bandwidth is demonstrated. (4) The polarization rotation capability, enabled by the symmetry in the resonant iris structure, was demonstrated. This allows a 90°twist to be easily embedded into the filter. Compared with traditional waveguide twists [34], [35], [36], this design feature eases the fabrication and provides a compact configuration. (5) The reproducibility of the SLM manufacture solution was investigated. Several designed filters were reproduced and measured. Measurement results show a good correlation with the tolerance analysis. This work also gives an assessment to the surface finish, by providing surface roughness measurements and showing the impact on insertion losses.

II. POLE-GENERATING RESONATING IRIS A. CONCEPT OF POLE-GENERATING RESONANT IRIS
The initial idea of this design concept comes from a common phenomenon. The inductive coupling irises often accompany by a spurious resonance above the passband, deteriorating the out-of-band rejection, especially for wideband filters. Fig. 1 shows the frequency response of a waveguide filter with inductive irises and the TE 101 -like electric field pattern of the iris when it resonates. In this work, we try to leverage the iris resonance and propose the pole-generating resonant iris structure. Fig. 2(a) shows the initial L-shaped resonant iris proposed in [33]. To form the resonant iris, an extra metallic diaphragm was added next to the symmetric inductive iris. Since the diaphragm introduces discontinuity into both the H-plane and E-plane, an L-shaped resonant aperture is realized. For comparison, the classic rectangular resonant iris is also presented in Fig. 2(b). From the simulated electric field (E-field) distributions, it is clear that both field patterns are like the TE 101 mode, but the pattern of the L-shaped aperture is folded. Hence, the L-shaped resonant iris has a longer  electrical length than the classic rectangular resonant iris. The iris resonance mode can be readily lowered into the passband, while the condensed E-field also enables the stronger coupling. Nevertheless, the above L-shaped iris can only realize basic pole-generating feature. To implement higher-order filters, a design methodology is needed. Therefore, the circuit analysis and electromagnetic optimization will be discussed next.

B. CIRCUIT MODEL AND ANALYSIS
The circuit model will be first illustrated to help understand the operation mechanism of the resonant iris. This circuit analysis will focus on the pole-generating feature. The inductive iris in Fig. 1 works as an impedance inverter (K-inverter) between two resonators and usually is represented as a shunt inductor as shown in the inset. However, the iris is a distributed physical structure. Its intrinsic frequency dependence will lead to unintended transmission feature. Herein, its embodiment is the extra transmission pole (TP) generated by the iris resonance mode on the upper stopband. To investigate the pole-generating feature, a resonant equivalent circuit was used, as presented in Fig. 3. The shunt inductor in the classical K-inverter is replaced by a parallel LC circuit, and its resonance frequency is determined by We will consider the following two cases: In this case, the resonant iris amounts to a frequency-variant coupling element, which can offer the required coupling at the centre frequency while generate an extra transmission pole on the upper stopband. Considering the circuit in Fig. 3, the parallel LC can be regarded as a shunt reactance, X at ω 0 , and the two-port network acts as a K-inverter. Two conditions can be imposed: 1) The parallel LC circuit resonates at ω p (ω p > ω 0 ); 2) At ω 0 , the transmission matrix of the iris must be the same as that of the K-inverter. i.e., The elements of this matrix can be respectively calculated [37]: where Z 0 is the characteristic impedance and X is the equivalent reactance of the parallel LC circuit at ω 0 .
Combining (1)-(4), we can find Therefore, for waveguide iris filters, once the passband specification is specified, we can predict the position of the transmission pole from the iris resonance, which can also provide insight for the achievable bandwidth of the designed waveguide filter. The corresponding shunt capacitor and inductor values can be further calculated using the following equation.
To verify the circuit model, we consider the two-pole Chebyshev waveguide iris filter in Fig. 1. Its centre frequency and bandwidth with 20 dB return loss is 10 GHz and 400 MHz, respectively. Fig. 4 presents two schematic circuits for this filter, respectively based on the ideal K-inverter and the resonant inverter model, where ω 0 = 1/ √ L 0 C 0 . The circuit elements can be calculated to be: C 0 = 0.151 pF, L 0 = 1.671 nH, K 01 = K 23 = 18.748, K 12 = 7.548, ϕ = 16.71°, L p = 0.071 nH, and C p = 1.601 pF. A comparison of their responses (simulated using AWR Design Environment) and the  full-wave EM model response (from CST Microwave Studio) is shown in Fig. 5. It is apparent that the equivalent circuit with the resonant iris better represents the full-wave EM model in terms of the in-band response as well as the position of the transmission pole at 16.6 GHz.
2) CASE II: (ω P = ω 0 ) In this case, the resonant iris can be regarded as a singlet which produce a pole-zero pair in the frequency response. As demonstrated in [33], when the L-shaped resonant iris is sandwiched between two waveguide TE 101 resonators, the TP generated by the resonance iris can be moved into the passband, as shown in Fig. 6 (the black solid curves) from a three-pole filter with one resonant iris. The transmission zero appears at 15 GHz. If an extra coupling aperture (See the inset of Fig. 6) is added next to the L-shaped iris, the TZ can be further controlled as shown in Fig. 6(the red dash curves). Three TPs are generated within the passband, while the TZ can be manipulated using the extra coupling iris. Therefore, the equivalent circuit in Fig. 3 is no longer applicable for the filter in this case. Considering its coupling topology (see Fig. 7), we can find the resonant iris not only provides the additional resonance node, but also allows the cross coupling k 13 . The core pole-zero block (marked out by the red circle) is equivalent to a singlet. From the circuit perspective, the  singlet can be more explicitly represented by the alternative equivalent circuit, as illustrated in Fig. 7 [38]. The shunt branch (series LC circuit) is added to account for the extra transmission zero. The synthesis-based approach to design filters with cascaded singlets has been well established [38], [39]. However, most methods are developed for narrowband filter and rely on circuit extraction process for physical implementation. Herein, we focus on the pole-generating feature of the resonant iris and use it to enhance the wideband capability of waveguide iris filters. In addition, the extra coupling iris allows us to control the cross-coupling. Thus, a more intuitive design approach is used to link the coupling coefficient to the real geometric dimensions of the resonant iris. Fig. 8 shows the EM simulation model and the simulated S 21 response used to extract the coupling coefficient. According to [40], the frequencies of two peaks of S 21 (f p1 and f p2 ) represent the magnetic and electric frequency f m and f e . The corresponding coupling coefficient k can be calculated by where the plus-minus sign represents the different nature of the coupling, whether magnetic or electric. This can be further determined from the phase information of S 21 . In each iteration of the simulation, both resonances should be adjusted to keep the average of the two peaks (f p1 + f p2 )/2 at the specified centre frequency. In the simulation model, two SMA connectors are inserted into the resonators to provide the weak   coupling. The open end of the resonant iris is approximated by a perfect magnetic wall to account for the loading effect from the adjacent resonator. It should be noted the model in Fig. 8 is simplified to illustrate the extraction process. More detail on the design of the resonant iris will be discussed in the next section.

C. OPTIMIZATION OF THE L-SHAPED RESONANT IRIS
It is often desired that the frequency and the coupling in a resonator can be adjusted seperately. However, for the primitive L-shpaed resonant iris (see Fig. 2(a)) proposed in [33], only two design parameters (t and s) can be used. The parameter l is largely fixed to fit the connected waveguide resonators. The degree of freedom in the design is very limited. New degree of freedom is needed for the resonant iris structure. Observing the simulated E-field distribution in Fig. 2, the E-field concentrates in the central area of the resonant structure and decays along to both sides. Therefore, the resonance frequency of the resonant iris is more sensitive to the dimensional changes of the structure in the central area. If we mitre the two end edges of the resonant iris, a taper transformation along the Z-axis can be formed, as shown in Fig. 9. As a result, an additional degree of freedom to control the frequency can be realized. This also facilitates the tuning of inter-resonator coupling with the adjacent resonators and the realization of asymmetric coupling.  It should be noted that such intricate structures would present significant difficulty with conventional machining. In contrast, they can be easily formed by 3-D printing. Fig. 10 illustrates a few crucial dimensions of the modified resonant iris. The dimensions s 1 and s 2 are used to separately control the size of the middle and end cross-sections whereas the parameter r is used to control the angle of the mitred edges. It is important to note that the two mitered end sides can be different from each other to allow for asymmetric coupling. An example will be demonstrated later. Parametric studies were carried out to understand the behavior of the resonant iris. Both end sides are set as a perfect magnetic wall to imitate the loading effect, to some extent, from cascaded resonators. The parameter l is fixed to 10.16 mm to be compatible with the X-band waveguide. Fig. 10(a) shows the variation of the resonance frequency of the resonant iris versus s 1 or s 2 , where the inset shows the air models of resonant iris at three representative cases. When s 1 equals to s 2 , the modified resonant iris degenerates into the initial version. It is apparent that the resonance frequency is more sensitive to s 1 , especially when  s 2 is smaller than s 1 . Further control can be applied by changing the parameter r. Shown in Fig. 10(b) is the variation of the resonance frequency versus s 2 under different ratio of r/t. As the ratio decreases, the frequency curve exhibits less variation, which means the ratio of r/t can mitigate the influence of s 2 on the resonance frequency. In addition, Fig. 11 shows the simulated resonance frequency and the unloaded quality factor Q u versus the thickness of the resonant iris t. As can be observed, while the frequency change is almost negligible, the Q u can be considerably improved with the increased thickness t. Still the Q u of the resonant iris is smaller than the standard X-band rectangular waveguide resonator at 10 GHz (usually on the order of 8000) due to the condensed E-field. The electrical conductivity of silver (6.3 × 10 7 S/m) was assumed for the extraction of the Q u .
Next, we consider the control of the coupling associated with the resonant iris, using the parameters s 2 and t. Fig. 12 shows the coupling coefficient curves, extracted using the method shown in Fig. 8. The coupling increases with the rising s 2 and decreases with t, while both parameters have comparable effect on the coupling strength. Therefore, the effective control of the coupling can be achieved by adjusting the two parameters. Estimated from the achievable coupling strength, if we consider a common 4 th -order Chebyshev filter, the achievable fractional bandwidth can readily reach at least 20%. Moreover, because of the field distribution of the L-shaped folded TE 101 mode is rotationally symmetric (see Fig. 2(a)), an additional polarization rotation feature can be realized. A filter with the rotation feature will be demonstrated later. It is worth noting that the resonant iris concept could work with other waveguide resonators than the rectangular cavity resonators used here. Now that both the resonance frequency and coupling associated with the resonant iris are shown to be adjustable, the design procedure of the filter using resonant irises can be summarized as follows: 1) change the parameter s 1 to obtain the specified resonant frequency. In this step, s 2 equals to s 1 ; 2) adjust parameters s 2 and t to obtain the required coupling while fine-tuning s 1 and r to modify the resonance frequency; 3) construct the initial filter using the obtained dimensional parameters and then perform full-wave optimization.

III. DESIGN EXAMPLES
In this section, three filter prototypes are designed as demonstrators. Fig. 13 illustrates two third-order waveguide filters based on the pole-generating resonant iris, where an L-shaped resonant iris is sandwiched between two conventional TE 101 resonators. To aid visualization, the metallic wall of the filters is rendered with translucent yellow color. The first filter in Fig. 13(a) is with a traditional in-line configuration, whereas the filter in Fig. 13(b) is similar but with a polarization rotation feature where only the first resonator is rotated by 90°. All other dimensions are kept the same.

A. TWO THIRD-ORDER CHEBYSHEV FILTERS
The first filter is specified with a centre frequency of 9 GHz, a fractional bandwidth of 9.5% and in-band return loss of 20 dB. The detailed discussion about this filter has been reported in [33]. Compared to the common rectangular waveguide filters, this filter has a footprint reduction of 20% while showing an improvement on the out-of-band rejection. The second example in Fig. 13(b) is to demonstrate the polarization rotation feature of the resonant iris. The same design specification as the first example is adopted. Fig. 14 compares the response before and after polarization rotation. Little effect on the performance is shown, due to the rotation symmetry of the resonant iris.

B. A FIFTH-ORDER CHEBYSHEV FILTER
The third example is to demonstrate the wideband capability and the implementation of asymmetric coupling using the resonant iris. It is designed to operate at 10 GHz with the equal-ripple passband fractional bandwidth of 20% and a return loss of 20 dB. The design parameters can be calculated to be k S1 = k 5L = 1.01, k 12 = k 45 = 0.86, and k 23 = k 34 = 0.63. Fig. 15(a) shows the filter structure, where a tapered resonant iris was used to realize the asymmetric coupling on either side of the iris (i.e., k 12 ࣔ k 23 ). Note that the mitred angle of the resonant iris in this example has been adjusted to provide a smoother transition between the two end faces. A comparison between the simulated and the ideal response calculated from coupling matrix is presented in Fig. 15(b). The response of the initial design from physical dimensioning (or coupling extraction) provided a reasonably good starting point for the following full-wave optimization. The filter response after optimization agrees well with the ideal response. Fig. 16 compares the simulated wideband responses between the resonant iris filter and the conventional fifth-order waveguide iris filter, with the inset showing the their internal structure and the enlarged view of the in-band S 21 responses. The 28% reduction in the overall length of the filter has been achieved. From Fig. 20(a), we can observe that the attenuation floor of the lower stopband is slightly worse than the rectangular waveguide filter. However, the upper stopband rejection is improved considerably over the waveguide iris filters as the spurious resonance mode is moved into the passband. The simulated insertion loss is 0.06 dB for the traditional and 0.15 dB for the resonant iris filter. The slightly higher loss is a result of the lower Q u of the resonant iris. Another common concern about the low-Q resonant structure is the passband distortion, such as degraded passband flatness or rounded passband shoulder. To recover the sharp passband selectivity, common methods include predistortion [41] and the use of nonuniform Q-values [42]. Connecting the external ports with a high-Q resonator is also an effective approach [43] in the practical application. For the proposed resonant iris filters, the low-Q resonant irises are sandwiched between two higher-Q rectangular waveguide resonators. There is some intrinsic balancing effect. From the expand view of the in-band S 21 response (see Fig. 20(b)), we can see the variation in insertion loss are similar, around 0.2 dB, for the resonant iris filter and the conventional one. Therefore, it is reasonable to believe the resonant iris structure has a small impact on the in-band transmission response, especially for the wideband applications.

IV. MANUFACTURE AND MEASUREMENT A. MANUFACTURING PROCESS
The resonant iris structure increases the manufacture complexity for traditional milling techniques, and even makes it impractical as far as the tapered resonant iris is concerned (Fig. 15). With this in mind, AM technique is chosen to manufacture the intricate tapered iris structure because of its unique capability of allowing monolithic fabrication of complex microwave components. More specifically, the selective laser melting (SLM) technique was employed on the twin-laser SLM500HL system. All three prototype designs were printed  monolithically using A20X. It is an aluminium-copper-based alloy powder, containing 92% aluminium, 5% copper, and 3% other materials. To examine the reproducibility of the manufacturing process, the 2nd and the 3rd design were printed twice. It should be noted they were printed in different batches to test the practical repeatability. Fig. 17 illustrates the printing direction where the bluecoloured material represents the scaffold supporting structure. In printing, all the filters are tilted 45°to avoid any overhang structures inside the filter, while the downskin surface with the worst surface finish can be moved away from the electrical  current concentrating area (top surface of rectangular waveguide). Table 1 summarizes the laser processing parameters and layer thickness. No surface treatment was applied on these samples except for the flange interfaces. Fig. 18 shows the photograph of the three monolithically printed filters. The small difference between the two third-order filters is caused by the length of feeding waveguide and wall thickness.

B. MEASUREMENT RESULTS
S-parameter measurements of all filters were performed on Agilent E8361C PNA network analyser. The analyser was calibrated using TRL (Thru, Reflect, Line) method prior to the measurements.
For 3-D printed devices, the effective electrical conductivity is significantly affected by the surface roughness. The first third-order filter was used to evaluate the effective conductivity of A20X alloy used in 3-D printing. Its nominal electrical conductivity has been reported to be 1.9 × 10 7 S/m [44]. Herein, the Hammerstad-Bekkadal (HB) method [45] is used to calculate the effective conductivity: where σ 0 is the electrical conductivity of the conductor with a smooth surface and K SR is the Hammerstad correction factor, given by Here, δ is the skin depth of the conductor and is the root mean square surface roughness. of the printed filter in this work was measured to be 3.5 μm using Alicona InfiniteFo-cusSL microscope. It is larger than the skin depth (1.2 μm for the A20X alloy at 9 GHz). The effective conductivity can be calculated to be 0.56 × 10 7 S/m.
To verify the calculated effective conductivity, Fig. 19 shows a comparison between the simulated results and the measured filter response, where the inset is the expanded view of S 21 . It can be observed the surface roughness introduces an additional insertion loss of 0.07 dB. The measured insertion loss agrees well with that predicted from the effective conductivity, so the conductivity of 0.56 × 10 7 S/m will be applied in the following simulations to account for the impact of the surface roughness. Fig. 20 shows the measured and simulated results of other two designs. No tuning was applied. As can be observed, all the measured passbands agree with the simulated results very well. For the third-order filter with polarization rotation, two samples were printed and measured (see Fig. 20(a)). The measured centre frequency is 8.91 GHz (Sample 1), shifted down by 1.0% and 9.05 GHZ (Sample 2), shifted up by 0.6%. Within the passband, the measured minimum return losses are around 15 dB whereas the measured average insertion loss is around 0.17 dB. The simulated insertion loss is 0.16 dB. Fig. 20(b) illustrates the in-band and wideband performance of the fifth-order waveguide filter. An excellent agreement is achieved between the simulated and measured in-band response, especially for the second sample. The measured centre frequency is 10.05 GHz and 9.97 GHz, respectively. The average insertion loss is better than 0.18 dB for both samples, which is close to the simulated 0.15 dB. The wideband performance is also as expected, with the rejection of 20 dB up to 15.3 GHz. It should be noted that 'response calibration' is used for the wideband measurement, so the evident ripple can be observed in the transmission response. Overall, very good agreement between the theoretical and measured results have been demonstrated. A reduced-Q resonant structure will set a lower bound on achievable minimum bandwidth. To evaluate this limit, we first extract the Q u using coupling matrix method from the measurement responses of the two samples of the fifth-order filters. All resonators are assumed to be uniform for simplicity. The Q u can be extracted to be 1700 (Sample 1) and 2100 (Sample 2). The average Q value of 1900 was then used to evaluate the insertion loss for filters with different bandwidths. If we consider the 1 dB threshold of insertion loss, the achievable bandwidth can be estimated to be 140 MHz. For typical filters in satellite payload around X-band, the resonant iris filters could be used for e wideband feeding filters (e.g., with a bandwidth of 250 or 500 MHz), but may not be suitable for the narrowband channel filters (e.g., with a bandwidth of 54 or 72 MHz).
To further investigate the reproducibility of the resonant iris structure using AM fabrication solution, a tolerance sensitivity analysis is carried out. Different from conventional CNC techniques where the tolerance is related to the practical milling process, the geometric error from 3-D printing is independent from the specific milling direction. Therefore, a tolerance  sensitivity analysis based on Monte Carlo sampling method [46], [47] was performed on the fifth-order resonant iris filter.
Six key dimensional parameters with ±100 μm (resolution of printing) tolerance were considered. 300 simulation samples with uniformly random distributed dimensions were taken. Fig. 21 shows the results of the sensitivity analysis, where the biggest frequency shift and worst return loss are 50 MHz and 15 dB, respectively. Both measurement results are within the envelope of the simulation samples. For the wideband application, we believe the 3-D printing technique is a reliable manufacture solution for the intricate pole-generating resonant iris structure. Table 2 provides a comparison with published wideband filters. As can be observed, the resonant iris filter shows a wide bandwidth with competitive performance in insertion loss, out-of-band rejection, and miniaturization. The polarization rotation feature gives it more design flexibility.

V. CONCLUSION
In this paper, the design concept of a pole-generating resonant iris has been demonstrated using rectangular waveguide filters. By introducing the capacitive and inductive discontinuity simultaneously, a transmission pole can be generated by the L-shaped iris. A circuit analysis was performed to represent the operating mechanism. The prototype filters have demonstrated the capability of realizing a wider bandwidth of 20% while maintaining a compact footprint due to the miniature resonant iris. Improved out-of-band performance has also been shown. The polarization rotation feature has been shown in a 3 rd -order filter. The major challenge associated with the resonant iris is its intricate structure and the almost inseparable feature from the cavity resonator, which makes it difficult to control the coupling. An approach to control and extract the coupling has been proposed. An intuitive design procedure was presented. The implementation of the filters is facilitated by 3-D printing technique. Very importantly, this allows the asymmetric coupling on either side of the resonant iris by using the tapered iris structure. The Q u of the resonant iris is lower than traditional waveguide resonator, but in the wideband application this has limited impact on the insertion loss. All three filters were printed monolithically using SLM process. Excellent agreements between the simulated and measured results were achieved, which verifies the resonant iris concept, the design and manufacturing approach. The pole-generating resonant iris structure is not limited to rectangular waveguides and may be used for miniaturization or wideband applications. 3-D printing has proved to be a capable fabrication method for the resonant iris structures. Future work is expected on the design of more complex RF components, benefiting from the proposed resonant iris structure and the 3-D printing technique. The circuit synthesis method will be further developed.