Compact, Reflectionless Band-Pass Filter: Based on GaAs IPD Process for Highly Reliable Communication

Abstract

For highly reliable and compact communication of front-end modules, a miniaturized reflectionless band-pass filter, based on the GaAs integrated passive device (IPD) process, is proposed in this work. The stop-band signal absorption rate of the filter can reach more than 90% and greatly reduce the influence of electromagnetic interference for sensitive devices. First, a circuit topology of reflectionless filter is proposed. Then, the miniaturized reflectionless band-pass filter is designed and fabricated based on GaAs IPD process with a compact size of only 0.85 mm × 1.33 mm × 0.09 mm (0.011λ × 0.018λ × 0.001λ). The filter operates at frequency ranging from 3.3 GHz to 4.5 GHz for 5G communication, the insertion loss (S₂₁) is less than 3 dB, the return loss in the passband (S₁₁) is over 15 dB, the stopband return loss (S₁₁) is over 10 dB, and the out-of-band suppression (S₂₁) reached 19 dB. All the measured results are in good agreement with the simulated results. It shows great potential in the process of designing highly reliable and compact monolithic integrated wireless modules and wearable electronics.

1. Introduction

A filter, as an important component in the front-end modules of wireless communication equipment, selectively pass and filter the signals in the frequency spectrum according to the different working frequency bands. Most filters have the type of reflection for stopband signals, meaning that the useless stopband signal is reflected back to the source port, which is the way that filters normally work [1,2,3]. However, the reflected stopband signal may bring negative impact on the performance of the previous stage in front-end modules or even affect the whole wireless communication system [4,5,6,7,8]. For instance, the filter is always added between the amplifier and analog-to-digital converter (ADC) for filtering processing, and the useless stopband signal, reflected by filter, may be resampled by the ADC, thus reducing the spurious free dynamic range (SFDR) of the ADC [5]. Besides that, in a solid-state transmitter, the homomorphic power amplifier usually has poor suppression of harmonic waves, and the output of the amplifier contains a large number of second and third harmonic waves. These harmonic waves in the stopband are reflected back to the power amplifier by the filter, which may affect the normal work of the power amplifier, or even cause the power amplifier to burn [8]. Attenuators can be used for solving these problems; however, more chip footprint and large insertion loss are introduced, which seriously affects the development of miniaturized monolithic integrated circuits [9].

Morgan firstly proposed the reflectionless filter in 2011, when he deduced the topology prototype of the first reflectionless filter, based on the two-port network [5]. The drawback of the low, out-of-band suppression of the reflectionless filter was also analyzed, which can only reach 14.47 dB. By cascading multiple reflectionless filters, the out-of-band suppression was increased to 42 dB, but the number of components was doubled. An ultra-wideband reflectionless band-pass filter was proposed by combining a cascading, reflectionless low-pass filter and reflectionless high-pass filter, which works at center frequency of 6 GHz, with a return loss of less than 10 dB in both passband and stopband, and an out-of-band suppression that only reached 10 dB [7]. In [10], a reflectionless inverse Chebyshev filter structure with arbitrary attenuation is proposed, and a low-pass filter with a cutoff frequency of 70 MHz is designed to verify their design concept. In [11], a band-pass filter, based on surface acoustic wave (SAW) is designed and measured at 418 MHz using conventional SAW. The reflectionless filters given in [10,11,12] are designed based on lumped elements; however, they are applied or verified at lower frequencies (~hundreds of MHz) and most of them are stop-band filters [13,14,15,16,17,18]. In the past decade, simplifying the topology prototype of reflectionless filter was pursued as an important goal to realize miniatured reflectionless filter [10,11,12,19,20,21,22] and filters applied for high frequency communication are desirable.

In this paper, a highly reliable and compact reflectionless band-pass filter, based on the GaAs integrated passive device (IPD) process, is proposed. A circuit topology of a reflectionless filter is proposed and the miniaturized reflectionless band-pass filter is designed and fabricated, based on GaAs IPD process with compact size of only 0.85 mm × 1.33 mm × 0.09 mm (0.011λ × 0.018λ × 0.001λ). The insertion loss (S₂₁) is less than 3 dB, the return loss in the passband (S₁₁) is over 15 dB, the stopband return loss (S₁₁) is over 10 dB, and the out-of-band suppression (S₂₁) reached 19 dB. The stop-band signal absorption rate of the filter can reach more than 90% and greatly reduce the influence of electromagnetic interference for sensitive devices. All the measured results are in good agreement with the simulated results. It shows great potential in designing of highly reliable and compact monolithic integrated wireless modules and wearable electronics.

2. Circuit Design and Fabrication Process

A circuit topology of reflectionless band-pass filter is proposed by using the method of transforming low-pass filter into band-pass filter. As shown in Figure 1a, the circuit topology of the reflectionless low-pass filter is presented. For the convenience of design and research, the symmetry of the reflectionless filter is retained, and all the inductance and capacitance values maintained the same symmetricity [4]. The circuit topology in Figure 1a (excluding the unit ➀) can be regarded as an odd-mode and even-mode equivalent circuit [5], and it is obtained that:

$${\Gamma }_{\mathrm{even}}=\frac{Z_{\mathrm{even}}-1}{Z_{\mathrm{even}}-1}=\frac{s^2+1}{2s^3+3s^2+2s+1}$$

(1)

$${\Gamma }_{\mathrm{odd}}=\frac{Z_{\mathrm{odd}}-1}{Z_{\mathrm{odd}}+1}=\frac{-s^2-1}{2s^3+3s^2+2s+1}$$

(2)

where S was complex number variable, which transformed microwave transmission signal into complex frequency domain for more flexible computing and derivation. Γ_even and Γ_odd represent the reflection coefficient of even and odd mode equivalent circuits, respectively, Z_even and Z_odd represent the impedance of even and odd mode equivalent circuits, respectively. It can be obtained from Equations (1) and (2), that

$$S_{11}=S_{22}=\frac{1}{2}\left({\Gamma }_{\mathrm{even}}+{\Gamma }_{\mathrm{odd}}\right)=0$$

(3)

From Equation (3), the low-pass filter can show the 100% absorption performance, and the validity of the reflectionless circuit is verified.

The frequency transformation from the reflectionless low-pass filter to the band-pass filter is deduced as follows. The transformed third-order reflectionless band-pass filter is shown in Figure 1b. Frequency transformed from reflectionless low-pass filter to reflectionless band-pass filter is as follows:

$$\frac{\omega }{{\omega }_c}\to \frac{1}{\Delta }\left(\frac{\omega }{{\omega }_0}-\frac{{\omega }_0}{\omega }\right)$$

(4)

$$\Delta =\frac{{\omega }_2-{\omega }_1}{{\omega }_0}$$

(5)

$${\omega }_0=\sqrt{{\omega }_1{\omega }_2}$$

(6)

where ω₁ and ω₂ are the upper and lower limits of the cut-off frequency of the band-pass filter, and ω_c is the cut-off frequency of the low-pass filter. Considerate of normalized coefficient g_k, the low-pass filter with inductance reactance of g_kZ₀ω/ω_c can be transformed in inductance reactance of Z₀ω/ω₀Δ-g_kZ₀ω₀/ωΔ for band-pass filter, which can be considered as series connection of inductance and capacitor. Similarly, the capacitance in a low-pass filter is equivalent to the shunt connection of the inductance and capacitance in a band-pass filter [22].

Table 1. Frequency transformation of the reflectionless filter.

Low-Pass Filter	Band-Pass Filter
$L=\frac{g_k{Z}_0}{{\omega }_c}$	$L_s=\frac{g_k{Z}_0}{{\omega }_0\Delta }$$,C_s=\frac{Y_0\Delta }{g_k{\omega }_0}$
$C=\frac{g_k{Y}_0}{{\omega }_c}$	$L_p=\frac{Z_0\Delta }{g_k{\omega }_0}$$,C_p=\frac{g_k{Y}_0}{{\omega }_0\Delta }$

lists the frequency conversions of the reflectionless low-pass filter to band-pass filter. In Table 1, L_s and C_s, respectively, represent the series inductance and capacitance in reflectionless band-pass filter; L_p and C_p, respectively, represent the shunt inductance and capacitance in reflectionless band-pass filter. Normalized coefficient g_k can be determined by cut-off frequency of the upper and lower limits of the proposed band-pass filter, which is deduced as follows:

The transfer function of the inverse Chebyshev filter is given, as shown in Equation (7), as follows:

$$\left|H\left(j\omega \right)\right|=\frac{1}{\sqrt{1+{\epsilon }^{-2}{T}_N^{-2}\left({\omega }^{-1}\right)}}$$

(7)

where ε is the ripple factor with a value of 0.1925, and T_N is the N-order Chebyshev polynomial [21].

$${\left|H\left(j{\omega }_{IL}\right)\right|}^{-2}=1+{\epsilon }^{-2}{T}_N^{-2}\left({\omega }_{IL}^{-1}\right)={10}^{IL/10}$$

(8)

$$T_N\left({\omega }_{IL}^{-1}\right)=\cosh \left(N{\cosh }^{-1}\left({\omega }_{IL}^{-1}\right)\right)=\frac{1}{\epsilon \sqrt{{10}^{IL/10}-1}}$$

(9)

$${\omega }_{IL}^{-1}=\cosh \left(\frac{1}{N}{\cosh }^{-1}\left(\frac{1}{\epsilon \sqrt{{10}^{IL/10}-1}}\right)\right)$$

(10)

where ω_IL represents the frequency at the specified insertion loss (i.e., IL = 3 dB in this work). For IL = 3 dB, g_k must be multiplied by the appropriate relative offset α = 3^1/2/2 [22], so that:

$$g_k=\alpha {\omega }_{IL}$$

(11)

Figure 2a shows the stacks of IPD process on the GaAs substrate with thickness of 75.75 µm. A series of multi-layer structures such as metal, dielectric, and through holes, etc., are fabricated on the substrate to form MiM (metal–insulator–metal) capacitor, high-Q value spiral inductor, and thin-film resistor. Figure 2b shows major fabrication steps of GaAs IPD process for fabrication of reflectionless band-pass filter. Three layers of electroplated gold M₀, M₁ and M₂ are contained in this structure, where M₀ is used for ground connection and M₁ and M₂ layers are used for upper and lower plates of the MIM capacitor, spiral inductor, and routing connection. N₁ is the first layer of Si₃N₄ with thickness of 0.16 μm, which sets the position of pad and through hole. The thickness of the VM layer is 1.3 μm, which provides the film protection with Si₃N₄ passivation, and provides a protective ring around the capacitor. N₂ is a layer of Si₃N₄ with thickness of 0.1 μm, which acts as the dielectric layer of MIM capacitor. PV is a layer of polyimide with thickness of 1.6 μm and serves as the packed dielectric between the M₁ and M₂ layers to reduce cross capacitance. N₃ is 4.8 μm-thick Si₃N₄ and provides the final Si₃N₄ layer for passivation of the device.

3. Results and Discussion

In this work, a reflectionless band-pass filter, which has upper and lower limits of 3.3 GHz and 4.5 GHz (i.e., 3 dB bandwidth) of the cut-off frequency of the band-pass filter, respectively. Based on the above parameters, derivation, and the description in Table 1, the component values can be calculated as: g_k = 0.6573, L_S = 4.36 nH, C_S = 0.38 pF, L_P = 0.96 pF, C_P = 1.74 nH.

Based on the IPD process schematic, as shown in Figure 2, the electromagnetic field model of the 5G reflectionless band-pass filter is designed with FEM simulated software HFSS. The model of the filter is shown in Figure 3a, the layout of the inductor, capacitor, resistor, and ground–signal–ground (GSG) pads are marked. Figure 3b shows the fabricated reflectionless band-pass filter based on GaAs IPD process. The dimensions of each inductor and capacitor are marked and shown in

Table 2. Component size of reflectionless band-pass filter based on GaAs IPD process (Unit: μm).

LS1	LS2	LS3	LS4	LP1	LP2
210	180	502	180	220	244
LP3	LP4	CS1	CS2	CP1	CP2
460	100	54	50	52	44

. The physical size of the proposed reflectionless band-pass filter is only 0.85 mm × 1.33 mm × 0.09 mm. The filter is measured using a vector network analyzer (Keysight PNA N5247A) with on-wafer GSG probes, as shown in Figure 4.

The simulated results are shown in Figure 5a with dash line and show that the bandwidth of the filter is 1.2 GHz (i.e., 3.3–4.5 GHz) with insertion loss less than 3 dB, the insertion loss of central frequency is 1.8 dB, the stopband return loss is more than 10 dB, and out-of-band suppression is greater than 23 dB. As the stopband return loss S₁₁= −20 log|S11| = −20 log|Γ| > 10 dB, the reflection coefficient |Γ| < 1/3, which suggests that the absorption rate of the reflected signal reaches 90%, and only 10% is reflected. The good absorption characteristics and high suppression characteristics of this band-pass filter are initially verified. Additionally, the squareness factor (SF) was presented for describing the selectivity of filter by following equation:

$$SF=\frac{B{W}_{20\mathrm{dB}}}{B{W}_{3\mathrm{dB}}}$$

(12)

where BW_20dB and BW_3dB were the bandwidths of filters within the range of 3 dB and 20 dB. SF was calculated as 2.08 for the filter in this paper.

As can be seen from Figure 5a, the measured S-parameter result shows that in the 3.3 GHz–4.5 GHz frequency band, the insertion loss of filter is less than 3 dB, the minimum insertion loss is 3.9 GHz and 2.1 dB, the in-band return loss is more than 15 dB, and the out-of-band suppression is more than 19 dB. The relative bandwidth is up to 30.8%, and the out-of-band return loss is more than 10 dB. Except the S-parameter, the measured group delay is also shown in Figure 5b, where the in-band group delay of the proposed filter is 250–280 ps, and the group delay fluctuation is only 30 ps, showing a good phase response. In summary, the measured results of the reflectionless filter are in good agreement with the simulated results, and the slight deviation is mainly due to the error brought by the test environment. Hence, the accuracy of the circuit topology, the GaAs IPD process model, and the good reflectionless characteristics of the band-pass filter are verified.

Table 3. Performance summary of proposed reflectionless band-pass filters.

Reference	Process	Structure	Size (mm)	Component Number	|S11| (dB)	|S21| (dB)	Center Frequency (GHz)
[5]	SMD	Third order BPF	/	18	15	12	0.18
[5]	SMD	Cascaded filter with three third-order BPF	/	54	15	42	0.18
[6]	IPD	Third order BPF	1 × 1	18	16	15	2.6
[7]	IPD	Cascaded filter with three HPFs and three LPFs	2 × 2	20	8	15	5
This work	IPD	Improved third order BPF	0.8 × 1.3	21	10	20	3.9

lists various important performance indicators of proposed reflectionless band-pass filters. However, research of reflectionless band-pass filters are limited nowadays, especially by IPD process. The main forms of third-order, reflectionless band-pass filters were listed in Table 3. Symmetrical structure of band-pass filter was presented in reference [5]. The reflectionless band pass filter in Reference [6] was presented with the same symmetrical structure as that in [5], but the process was different. A cascaded reflectionless band-pass filter was presented in reference [7], which cascaded HPFs (high-pass filters) and LPFs (low-pass filters). The cascaded structure has the advantages of forming UWB (ultra-wide band) filter. In this paper, an absorption suppression unit (unit ➀ marked in Figure 1a) was added to the structure in [5] and [6], which improved the performance with fewer elements and was realized by the IPD process. From the comparison, the proposed reflectionless band-pass filter in this paper has obvious advantages in design method, layout size, component number, and high out-of-band suppression.

4. Conclusions

This paper proposes a circuit topology of the reflectionless band-pass filter, and the miniatured reflectionless band-pass filter is realized based on a GaAs IPD process with compact layout. The measured results show good agreement with the design parameters. The circuit topology of the reflectionless filter is versatile and is not limited to the working frequency mentioned in this paper, and can be used in a series of designs for applications in different signal bands. The research of the reflectionless filter points out a new idea for the design of passive filters with good performances and compact chip sizes for highly reliable communication in front-end modules.