Microwave Photonic Filters and Applications

Abstract

Microwave photonics is a promising and rapidly developing interdisciplinary field. It combines microwave and photonic techniques to generate, transmit, process, and manipulate microwave signals by using the advantages of broadband, high frequency, and low loss provided by photonics. As an important branch of microwave photonics, the microwave photonic filter (MPF) can overcome the limitations set by traditional electronic technology and can realize advanced signal processing in modern communication systems due to its higher performance, selectivity, and flexibility. This review provides a comprehensive overview of MPFs, including fundamental principles, typical structures, and key applications. Additionally, the microwave photonic integration is a very important tendence because of its advantages of small size, light weight, low power consumption, and low cost. The recent advances in integrated MPF are also reviewed.

1. Introduction

Microwave photonics (MWP) is an interdisciplinary field that integrates traditional microwave technology with photonic technology [1]. It takes advantage of the broadband, high frequency, and low loss offered by modern photonics to generate, transmit, process, and manipulate microwave signals [2]. Initially focused on defense applications [3], MWP has been extended to address a considerable number of civil applications, including cellular [4], wireless and satellite communications [5], distributed antenna systems [6], and high-resolution optical sensing [7]. The schematic diagram of a typical microwave photonic link is shown in Figure 1 [8]. In the microwave photonic link, we can see that microwave signals are loaded onto the optical carrier by an electro-optical modulator at first. Then, the microwave photonic signals are transmitted and processed by optical devices. Thanks to the ultra-low transmission loss of optical fiber, the microwave photonic signals can be transmitted over a long distance. Meanwhile, compared with electrical devices, the bandwidth of optical devices is much larger, which enables the direct processing of broadband high-frequency microwave signals in the optical domain. Before output, the microwave photonic signals are converted back to electrical signals by using a photodetector (PD). Compared with traditional microwave technology, the major difference of microwave photonic technology is that the microwave signals are processed in the optical domain, which can take the advantages of photonic technology [2], including ultra-low loss, large bandwidth, and high immunity, to electromagnetic interference.

In MPFs, optical devices are used to achieve filtering functions in the microwave frequency band. A comparison of a traditional electronic filter and an MPF is shown in Figure 2 [8]. In the traditional microwave filter, the radio frequency (RF) signal received by the RF antenna is processed in the RF circuit to perform the signal processing task. While, in the MPF, the RF signal received by the RF antenna is firstly modulated onto the optical carrier by an optical modulator, and the RF signal is converted to an optical signal. Then, the optical signal is processed by the optical signal processor. Before output, the optical signal is converted back to the RF signal by the optical receiver. Different from traditional electronic filters, the RF signal is processed by optical devices in the optical domain, and broadband RF signals at high frequency can be directly processed because of the large bandwidth. MPF also shows high tunability [9,10,11] and reconfigurability [12]. This means that the center frequency can be widely tuned, and the bandwidth can be changed to adapt to different applications [13]. Even the frequency response of the MPF can be switched between bandpass and bandstop shapes [14,15]. At the same time, optical signals transmitted in optical fibers or optical waveguides show strong immunity to electromagnetic interference [16,17]. MPF plays an important role in the field of microwave photonics. For example, optical oscillators (OEOs) benefit from MPF to achieve broadband tunability and low phase noise [18,19,20,21,22,23]. The measured accuracy and bandwidth are also improved in microwave frequency measurements through integrated photonic technology [24,25]. Programmable MPF achieves real-time signal processing and facilitates the development of reconfigurable systems [26,27,28,29]. MPF has shown great potential in modern communication systems.

In this review, the first part introduces the theoretical foundation and the classification of MPF. The theoretical part paves the way for the design of MPF. The second part introduces the applications of MPF in various areas, including optoelectronic oscillators and microwave frequency measurements. Since the integrated MWP (IMWP) shows superiorities in size, weight, power consumption, and cost, the third part introduces the integrated MPFs and applications in integrated systems.

2. Fundamental Concepts

Although there are differences between the MPF and the traditional electronic filter in processing microwave signals, their system descriptions are identical, as shown in Figure 3. The input signal x(t) is divided into N parts to generate N taps. The N taps are introduced in turn with a series of delays of T, 2T … (N − 1)T and a series of amplitude weightings of b₁, b₂ … b_N, respectively. Finally, all taps are summed and output to the system. Additionally, the output signal y(t) can be fed back into the system after delaying, weighting, and summing [30]. In the MPF, the functions of splitting, delaying, weighting, and summing can be implemented by using different optical components. For example, the tap sampling can be realized by splitting a single source into N outputs using a 1 × N coupler or a wavelength division multiplexing (WDM) light source [31,32]. The addition unit can be realized by using optical couplers [9,33] or WDM combiners [34,35], the multiplication unit can be realized by using optical attenuators [36,37] or optical amplifiers [33,38], and the delay unit can be realized by using optical fibers [35,39], high Q resonant cavities [40,41,42], or fiber Bragg gratings (FBGs) [43,44,45].

The relationship between the output and the input signals can be expressed as follows:

$${\displaystyle \sum _{n=0}^N{b}_{n}x(t-nT)+{\displaystyle \sum _{m=0}^M{a}_{m}y(t-mT)=y(t)}}$$

(1)

The discrete Fourier transform of Equation (1) yields the transfer function expression of the MPF as follows:

$$H(\omega )=\frac{{\displaystyle \sum _{n=0}^N{b}_n{e}^{-j\omega nT}}}{1-{\displaystyle \sum _{m=0}^M{a}_m{e}^{-j\omega mT}}}$$

(2)

when a_m = 0, the transfer function of the filter can be obtained according to Equation (2), and can be expressed as follows:

$$H(\omega )={\displaystyle \sum _{n=0}^N{b}_n{e}^{-j\omega nT}}$$

(3)

In this case, the filter is called the finite impulse response (FIR) filter because the number of taps is finite. However, when a_m ≠ 0, the filter is called the infinite impulse response (IIR) filter because the tap number is infinite in theory. From Equation (2), it can be seen that the frequency response of the filter is a periodic function, and the period is T.

2.1. Parameters of the MPF

The MPF is a versatile RF signal processing system utilized in a wide range of applications, and different transmission spectra are required in different scenarios. The subsequent discussion presents the pertinent parameters used to assess the performance of the MPF.

The response of a typical FIR-MPF is shown in Figure 4.

(1)

Free spectral range

Free spectral range (FSR) refers to the spectral period of the frequency response. It is determined by the unit time delay T, which is given by 1/T. For some applications, a single-pass band filter in the entire operating range is required. Therefore, the FSR should be larger than the operating range, and a small delay is required for a large FSR. However, the minimum FSR is limited by the coherence of the optical source of the MPF and environmental influences.

(2)

Full width at half maximum bandwidth

In Figure 4, the full width at half maximum (FWHM) bandwidth is denoted as ∆f_FWHM, which refers to the bandwidth when the output power decreases by half from the maximum value. When logarithmic coordinates are used, it is the bandwidth corresponding to a 3-dB drop in magnitude. Therefore, the FWHM bandwidth of the filter passband can also be called the 3-dB bandwidth, which characterizes the spectral width that can be selected by the filter. The FWHM bandwidth of the MPF is mainly determined by the number of taps. When more taps are generated and summed, the quality factor of the MPF is increased, and the FWHM bandwidth is reduced correspondingly.

(3)

Quality factor

The quality factor (Q) is defined as the ratio of FSR to 3-dB bandwidth. It is a parameter to measure the superiority of filter frequency selectivity, and a larger value corresponds to a better filtering selectivity. For the FIR-MPF, Q is closely related to the number of taps N. When the number of taps satisfies N > 10, it can be concluded that Q ≈ N [16]. For the IIR-MPF, Q is mainly related to the loss in the optical loop. A lower loop loss corresponds to a higher Q value.

(4)

Rejection ratio

The rejection ratio is defined as the ratio of the magnitude of the passband to the magnitude of the secondary sidelobe. It describes the suppression degree of out-of-passband signals for the MPF. A larger the rejection ratio indicates that the out-of-band signals are suppressed lower. The rejection ratio of the MPF is also related to the number of taps and the tap coefficients of the MPF.

(5)

Shape factor

The shape factor is defined as the ratio of stopband bandwidth, such as the 20-dB bandwidth to the 3-dB bandwidth. It characterizes the transition edge steepness of the filter passband. The more the shape factor tends to be 1, the more the filter tends to be an ideal rectangular filter.

(6)

Tunability and reconfigurability

The tunability describes to what extent the center frequency of the filter can be tuned. The reconfigurability describes to what extent the filter shape can be changed, such as switching the filter response between bandpass and bandstop responses or adjusting the bandwidth. Both tunability and reconfigurability can be achieved by manipulating the parameters according to Equation (2). For example, the tunability can be achieved by adjusting the phase terms to the weighting complex coefficients. In contrast, changing the magnitudes of the weighting coefficients or the number of taps can change the filter shape, thus achieving reconfigurability.

2.2. Classifications of the MPF

According to the frequency response, the MPF can be classified into different types. In this section, the classifications of the MPF according to different criteria are introduced.

2.2.1. FIR- and IIR-MPFs

According to the number of taps, MPF can be classified into FIR- and IIR-MPFs.

When a_m = 0, only a finite number of taps are summed. Therefore, the corresponding MPF is defined as FIR-MPF. The FIR-MPF is advantageous in reconfigurability, because the coefficient of each tap can be conveniently adjusted. To demonstrate the FIR-MPF, a laser diode (LD) with a single wavelength and a pair of optical couplers are used, as shown in Figure 5a [46]. The modulated optical signal is divided into N paths by the 1 × N coupler. Then, different weightings and delays are applied to the N taps, and finally, they are combined by the N × 1 coupler and converted back to an electrical signal by the detector. Meanwhile, FBGs can reflect the wavelengths that satisfy the Bragg condition and are usually used to obtain finite taps of the MPF, as shown in Figure 5b. The FBGs with different reflection wavelengths are uniformly arranged to reflect different wavelengths of the optical source, generating different taps of the MPF. The coefficient of each tap can be controlled by designing the reflectivity of FBG. Then, the weighted taps experience different time delays and are summed in the PD. Finally, an FIR-MPF is achieved [43,44,45].

The schemes mentioned above are realized based on uniform time delay. Reference [47] uses an array of lasers as the taps of the filter and introduces additional time delays at different taps, resulting in a nonuniformly spaced FIR-MPF. An arbitrary bandpass frequency response with all-positive tap coefficients is obtained. When nonuniformly spaced wavelengths are passing through a dispersive fiber, there will be nonuniform time delays between adjacent taps. The nonuniform time delays provide equivalent phase shifts to the tap coefficients, thus achieving an MPF with arbitrary response shape. Further, the nonuniformly spaced FIR-MPF has been proposed to realize a microwave bandpass differentiator [48]. Compared with the differentiator based on previous negative coefficient MPFs, the complexity of the proposed differentiator based on nonuniformly spaced MPF is reduced.

When a_m ≠ 0, the output signal is fed back to the input port. Consequently, the number of taps is infinite, and the corresponding MPF is defined as IIR-MPF. Figure 6a shows a typical diagram of the IIR-MPF [46]. It can be seen that the number of generated taps tends to be infinite due to the recirculating loop. In IIR-MPF, the frequency response is dependent on the coupling coefficient of the coupler, the loop loss, and the loop length [49]. Since the loop length is fixed and no microwave phase shifter is used in the loop, the corresponding frequency response of the MPF cannot be tuned. Using a tunable delay line can definitely achieve a tunable frequency response, as shown in Figure 6b [50]. However, the FSR of the MPF can be correspondingly changed.

Compared with FIR-MPF, IIR-MPF uses fewer components but obtains higher-Q factors [51,52,53]. Compared with a single IIR-MPF, using the cascading method can significantly enhance the FSR and the Q factor of the MPF. Reference [54] presents two cascaded active recirculating delay line (RDL) loops with a semiconductor optical amplifier (SOA) inserted between the two loops, as shown in Figure 7a. Each active RDL loop acts as a single IIR filter. To eliminate the optical interference caused by cascading two IIR loops [55] and to obtain a stable response, cross-gain modulation (XGM) in the SOA is used to perform wavelength conversion. By doing so, the FSR and Q factor are significantly improved using the Vernier effect technique.

Notably, MPFs can be applied to an instantaneous frequency measurement (IFM). Traditionally, the IFM is performed in the electrical domain, faces challenges in recognizing high-frequency microwave signals, and is susceptible to electromagnetic interference. Microwave photonic techniques have been proposed to solve these problems [6,56,57]. For example, an optical channelizer [58,59,60] can be used to measure microwave frequencies. However, specially designed diffraction gratings and PD arrays are required, which makes the system bulky and costly. In addition, the measurement accuracy is usually low because of the wavelength spacing drift and relative power fluctuations of multiple laser sources. The IFM can also be realized based on frequency to time mapping [61,62,63], which shows the advantage of measuring multiple frequencies simultaneously. To enhance frequency resolution, an IIR-MPF and an FIR-MPF are combined to realize IFM, as shown in Figure 8a [64]. A monotonic frequency response ranging from positive to negative infinity is obtained, and a unique relationship between the output response and the input microwave frequency is established. Based on the calculation, the minimum resolution is 2.67 dB/GHz for a 10 GHz measurement range. Therefore, the microwave frequency measurement resolution is significantly improved, and the complexity of the system is reduced.

Additionally, MPFs can also be applied to OEO, which is an ideal solution for generating high-frequency microwave signals with low phase noise and broadband tunability [65,66]. In the OEO, using a long optical fiber is a common approach to promoting the quality factor of the loop, thus reducing the phase noise. However, a long fiber loop results in ultra-narrow eigenmode frequency spacing. Therefore, an ultra-narrow electrical bandpass filter (EBF) is required for single-mode oscillation [67,68]. The problem is that the ultra-narrowband EBF is usually difficult to tune over a broad band [69,70]. By contrast, MPFs can be widely tuned and thus provide a promising solution to realizing tunable OEOs. Reference [18] proposes a tunable bandpass MPF that successfully replaces the narrow EBF in a traditional single-loop OEO. The MPF was realized by cascading an FIR filter and an IIR filter, as shown in Figure 9a. The FIR filter is realized based on tunable multi-wavelength lasers and a dispersion-compensated fiber (DCF). The IIR filter is realized simply based on a passive RDL loop. By adjusting the wavelength spacing of the multi-wavelength laser, the center frequency of the MPF passband can be tuned, which determines the frequency of the generated microwave signal. As a result, tunability of the OEO can be achieved over a wide frequency range without sacrificing the quality of the produced signal. Thus, a wide oscillation frequency tuning range of 6.88–12.79 GHz is realized.

2.2.2. Positive, Negative and Complex Coefficient MPF

When a_m > 0 and b_n > 0, which indicate that all the coefficients of the MPF are positive, the corresponding MPF is defined as a positive coefficient MPF. The coefficient weighting is usually achieved by using an optical amplifier or attenuator. Since the light power cannot be negative, only positive coefficient MPFs are achieved. However, positive coefficient MPFs always exhibit low-pass transmission and make it impossible to implement some specific response shapes, such as the rectangular filter [71]. On the other hand, positive coefficient MPFs always have a passband at zero frequency, which should be avoided in some application scenarios [72]. When the negative weighting coefficient is obtained, the filter is called the negative coefficient MPF, and the passband at zero frequency can be eliminated by properly setting the positive and negative coefficients [71], as shown in Figure 10. There are several methods that have been proposed to achieve negative coefficient MPFs. The first method is to use differential detection, which means that the taps are divided into two parts, and then, subtraction detection is performed in a microwave hybrid [32,73]. The second method is to use the XGM effect in which the pump light power and probe light power are complementary [74,75,76,77,78]. The third method is to set a pair of Mach–Zehnder modulators (MZMs) at in-phase and counter-phase modulation points, respectively. Therefore, the phase difference between two intensity modulated signals is π, thus achieving negative coefficient taps [79].

Using nonlinear effects in SOA [80,81,82], such as XGM [74,75,76,77,78] and four-wave mixing (FWM) [83], a series of MPFs can be realized. Figure 11a shows a negative coefficient MPF achieved by using XGM in SOA. After XGM, the probe light is counter phase modulated by the pump light, and a negative coefficient is obtained [75,84]. The signal light at wavelength λ₁ is modulated and divided into two paths. One is delayed and another is input into the SOA together with the probe light at wavelength λ₂. Due to the XGM effect, the signal at λ₁ is copied to the probe light at λ₂ in inverted phase. Therefore, a negative coefficient tap is obtained. Meanwhile, reference [85] has proposed a bandpass MPF with high-frequency selectivity based on XGM in SOA. As shown in Figure 11b, the intensity modulated signal by MZM is injected into the RDL loop after the power adjusted by EDFA and the attenuator. After XGM, the amplified spontaneous emission (ASE) spectrum of the SOA is modulated by the intensity modulated optical signal, and negative coefficients are obtained. Then, the converted ASE spectrum circulates in the RDL loop and achieves a high-Q factor frequency response after PD. The residual pump signal, unfiltered by the tunable narrowband optical filter (TNOF), realizes a weak all-pass response which can improve the shape factor.

When b_n is a complex number and the phase of coefficient b_n can be adjusted from 0 to 2π, the MPF is called the complex coefficient MPF, whose transfer function can be expressed as follows:

$$H(\omega )={\displaystyle {\sum }_{n=0}^N\left|b_n\right|}{e}^{j{\phi }_n}{e}^{-j\omega nT}$$

(4)

The biggest advantage of the complex coefficient MPF is its excellent tunability. The passband or the stopband of the complex coefficient MPF can be tuned by adjusting the phases ${\phi }_n$ of the complex coefficients. Notably, by changing the unit time delay T, the MPF can also be tuned, but the bandwidth and FSR will be changed at the same time, which is detrimental for processing discrete signals. However, the complex coefficient MPF can be tuned, while the bandwidth and FSR remain unchanged. Notably, the complex coefficient MPFs are achieved by using microwave photonic phase shifters. Several approaches have been proposed to achieve microwave photonic phase shifters, such as using an all-optical differentiator [36], using nonlinear effects in optical devices [86,87], and using a polarization modulator (PolM) [88,89]. Based on stimulated Brillouin scattering (SBS) in optical fiber, an MPF with tunable complex coefficients is reported in reference [9]. The complex coefficients are realized by combining optical single sideband (SSB) modulation and SBS. It shows that when an optical carrier is aligned with the SBS gain spectrum, the phase of the optical carrier is shifted. After photodetection, a phase shifted microwave signal is generated [90]. However, these approaches are realized based on discrete devices and complex. By contrast, the optical all-pass filter is an ideal candidate to achieve a microwave photonic phase shifter [91,92,93]. The all-pass filter based on a microwave photonic phase shifter is quite simple and compact. A detailed introduction of the all-pass filter is given in Section 3.1.3.

Notably, a pair of MPFs with positive and negative coefficients can be applied to microwave frequency measurements [94]. A low-pass MPF with coefficients of (1, 1) and a band-pass MPF with coefficients of (−1, 1) are implemented using a PolM and two sections of polarization preserving fiber (PMF), as shown in Figure 12a. Due to the complementary frequency responses of the MPF pair, an amplitude comparison function (ACF) can be obtained. The ACF is the ratio of the two transfer functions of the MPF pair and is quasi-linearly monotonically decreasing in a large frequency band range. The measurement of the microwave frequency can be obtained by simply measuring the microwave powers from the two outputs of the MPF pair. Since the MPF pair is realized using a single wavelength, the measurement accuracy is independent of the laser wavelength, which significantly reduces the cost and complexity of the system. As shown in Figure 12c, a measurement accuracy of less than 0.2 GHz over the 36 GHz microwave frequency measurement range is finally realized.

As mentioned above, the OEO requires a narrow passband EBF for mode selection. Since the SBS has a narrow and large gain spectrum, SBS-based MPF can be adopted in OEO to replace the EBF for mode selection. A single-passband MPF with an ultra-narrow passband is realized by cascading an SBS-based MPF and an IIR-MPF based on an RDL loop [18,19,95]. The experimental setup is shown in Figure 13a, and the measured frequency response of the cascaded MPFs is shown in Figure 13b. Results show that the cascaded MPF has an ultra-narrow FWHM bandwidth of 150 kHz, while the mode spacing of the OEO is 200 kHz. Therefore, single mode oscillation can be guaranteed, and the mode hopping and side mode noise are significantly suppressed.

The application of MPF in OEOs offers several advantages. Due to the low-loss characteristics of the optical elements, the MPF achieves a high-Q factor, which enables the OEO to generate high-frequency microwave signals with low phase noise. The wide operation bandwidth and tunable center frequency of the MPF enable the OEO to generate microwave signals at different frequencies over a wide frequency range. Additionally, the MPF can also be reconfigured dynamically to enable the OEO to generate different signals [96].

2.2.3. Coherent and Incoherent MPFs

In the MPF, microwave signals are carried by an optical carrier. The schematic diagram of the MPF in Figure 5a shows that multiple taps are summed together in the detector. Notably, the coherence character of the optical carrier can significantly impact the frequency response of the MPF. According to the coherence of the optical carrier, the MPFs can be classified into coherent MPFs and incoherent MPFs.

In a coherent MPF, the unit time delay T is much smaller than the coherent time of the optical carrier ${\tau }_{coh}$. Therefore, when all the taps are summed in PD, the converted photocurrent is proportional to the square of the summed optical field of all the taps. Therefore, the optical phases of the taps play a predominant role in the frequency response of MPF. Negative and complex coefficients of MPF can be implemented. On the other hand, since the operation of MPF depends on optical interference, the optical phase variation of any tap of the coherent MPF (e.g., delay or refractive index change due to the environmental fluctuations [46]) will change the frequency response of the MPF and cause high instability. To achieve coherent MPFs with enhanced stability, the unit time delay in the MPF should be reduced to be as small as possible. The minimized unit time delay makes the optical phase variations between the arbitrary two taps caused by environmental fluctuations to be identical to each other.

To reduce the unit time delay, phase-shifted fiber Bragg gratings (PS-FBGs) are used [97]. In reference [97], a PS-FBG is used to eliminate one sideband of the phase modulated signal. After phase modulation to intensity modulation (PM-IM) conversion, a microwave bandpass filter is obtained, as shown in Figure 14a. Instead of using a single PS-FBG in transmission, two cascaded FBGs are used in reference [98], in which a single passband is realized without phase-induced intensity noise. The schematic diagram of the coherent MPF is shown in Figure 14b. One FBG is used to select the optical carrier, and the other FBG is used to select a sideband Then, a phase modulated signal is converted to an SSB intensity modulated signal. The center frequency of the microwave bandpass filter can be tuned simply by tuning the center frequency of the notch filter or the wavelength of the laser source. The spectral shape remains constant during the tuning process.

Besides the FBG, the coherent MPF can also be realized by using integrated devices, which have a compact size and minimize optical phase variation among different taps due to environmental fluctuation. Therefore, the frequency response of the MPF tends to be stable. However, the overall impact caused by environmental fluctuations on the coherent MPF, such as the frequency response drifting, still exists. Details of the integrated MPF are presented in Section 3.

When the coherence time ${\tau }_{coh}$ is much shorter than the unit time delay T, the corresponding MPF is called incoherent MPF. The optical phase relationship between any two taps is completely random, and the optical power input into the detector is the sum of the optical power of all taps. In this scenario, the incoherent MPF exhibits little susceptibility to environmental fluctuations and shows exceptional stability [17]. However, a drawback of incoherent MPF lies in its coefficients, which are constrained to positive values, and this leads to a serious limitation of the range of transfer functions. Meanwhile, the FSR of an incoherent MPF is usually very small [72]. Fortunately, there are solutions available for the implementation of incoherent MPF with negative coefficients, as mentioned above. Reference [99] reports a technique for generating negative coefficients of incoherent MPF based on two MZMs biased at complementary transmission slopes. Two MZMs are biased in the linear region in opposite transmission slopes, as shown in Figure 15a, respectively. When microwave signals are applied to different modulators, the envelopes of the modulated optical signals are complementary. At the output of the PD, the counter phase microwave signals are generated, resulting in negative coefficients, as shown in Figure 15b.

3. Microwave Photonic Integration

Integrated optoelectronic devices based on semiconductors have a compact size and excellent reliability, and they can even achieve functions impossible to be achieved based on discrete devices. IMWP has the potential to optimize performance and reduce power, footprint, and cost in the large-capacity and bandwidth systems [100]. IMWP deals with the application of integrated photonic techniques in microwave photonic systems [101,102]. In the past few years, IMWP has become the most active area of microwave photonic research [8], with remarkable advances in integrated photonics based on various material platforms such as indium phosphide (InP), silicon on insulators (SOI), silicon nitride (Si₃N₄), and lithium niobate on insulators (LNOI). The SOI platform is most widely used in IMPF by the feature of COMS compatibility. Several integrated passive elements are available In SOI, such as arrayed waveguide gratings, optical filters, and Ge photodetectors. The InP platform can provide all components for monolithic integration, especially the active optical components, such as the laser diode, semiconductor optical amplifier, and mode-locked laser. The ultra-low loss of Si₃N₄ facilitates the manufacture and application of optical waveguides and tuning elements, such as MRR and MDR. LNOI has a high refractive index and high second-order nonlinearity, and it is widely used in modulators. In this section, the integrated MPF (IMPF) and its applications in integrated systems are reviewed. A detailed performance comparison of the reported IMPFs is presented in

Table 1. Performance comparison of reported IMPFs.

Platform	Filter Type	FWHM Bandwidth (GHz)	Tunable Range (GHz)	Rejection Ratio (dB)	Gain (dB)	Noise Figure (dB)	SFDR * (dB/Hz2/3)
RF [103]	Bandpass	0.025	3–5.6	40	−3.1	3.1	137
As2S3 [104]	Band-stop	3	0–15	>40	−10.1	27.1	96.5
InP [81]	Bandpass	1.9–5.4	0–27	32	N/A	23.2	86.3
InP + SOI [105]	Bandpass/ Band-stop	0.36–0.47/ 0.38–0.45	3–21/ 3–25	>10/ >40	−28.2	51.2	99.7
SOI [106]	Band-stop	0.78	0–40	60	N/A	N/A	N/A
SOI [107]	Bandpass	0.7–5.2	2–35.8	51	8.97	33	90.1
Si3N4 [108]	Band-stop	0.15–0.35	0–12	50	8	15.6	116

* SFDR: Spurious Free Dynamic Range.

3.1. IMPF

In IMPF, the unit time delay can be significantly reduced to as small as 15.4 ps [109]. Therefore, the obtained IMPF is usually coherent MPF, which shows that quite large FSR and complex coefficients can be easily obtained. In IMPF, the performance promotion and flexibility enhancement are two focused problems. Here, we review the advances in promoting the performance and flexibility.

3.1.1. Performance Enhancement

The performance of the MPF includes the rejection ratio, Q factor, shape factor, etc. Traditional FIR-MPFs based on discrete devices have periodic filter transfer functions even within a small operation bandwidth because of rather small FSRs, which block the MPF from being applied in practical RF applications because of periodic passbands in its operation bandwidth. In contrast, IMPFs are advantageous in obtaining very large FSRs, and a single passband filter in a large operation bandwidth can be obtained consequently. The microring resonator (MRR) is usually used to achieve IMPFs with a high-Q factor. However, for an all-pass MRR, there is a trade-off between the rejection ratio and the bandwidth at the output port. It is impossible to obtain the narrowest bandwidth while achieving the maximum rejection ratio. To solve this problem, a technique that optically cancels the microwave signal at the resonant wavelength of the MRR is proposed to increase the rejection ratio of the microwave photonic notch filter [110]. An EOM and a silicon nitride-based MRR are used to operate the phase and amplitude of the two optical sidebands, as shown in Figure 16a. The two sidebands corresponding to the target microwave frequency are placed with equal amplitude and opposite phase. When passing through the PD, the beating between the optical carrier and two sidebands is generated. Notably, the two beating microwave signals are of equal magnitude and opposite phase and, thus, are completely canceled. Therefore, a notch MPF with a peak rejection as high as 62 dB is obtained, as shown in Figure 16b. An improvement of 52 dB in peak rejection is obtained by this technique. The center frequency can be tuned from 1 to 9 GHz, and the peak rejection remains above 55 dB during the tuning process, as shown in Figure 16c.To reduce system complexity, reference [106] proposed and demonstrated an SOI-based widely tunable MPF with an ultra-high rejection ratio. The MPF is realized by using two cascaded tunable Mach–Zehnder Interferometers (MZIs) and an under-coupled MRR, as shown in Figure 16d. Figure 16e shows the principle of the MPF. Two cascaded tunable MZIs are used to achieve destructive interference by controlling the amplitude and phase of the two first order sidebands of the phase modulated signal. When the two first order sidebands are of equal magnitude and counter phase, a microwave photonic bandstop filter with an ultra-high rejection ratio can be obtained. Measured results show that the peak rejection is higher than 60 dB, the FWHM bandwidth is 780 MHz, and the center frequency can be tuned from 0 to 40 GHz. Another advantage of this work is that it simultaneously achieves ultra-high peak rejection and narrow bandwidth without the external electrically orthogonal hybrid coupler.

As mentioned above, the MRR is one of the most important on-chip photonic devices for realizing IMPFs due to its small size, unique amplitude and phase response, and excellent tunability [111,112,113,114]. Many IMPFs using on-chip MRRs [108,111,115,116,117,118] or microdisk resonators (MDRs) [24] have been demonstrated. Most of the MPFs based on silicon MRR exhibit a very coarse resolution in the order of GHz or tens of GHz [24,115,116,119]. This is caused by the high transmission loss due to the rough sidewall scattering of the waveguide. To reduce the bandwidth of the MPF based on MRR, the transmission loss must be reduced. Therefore, several special fabrication processes have been employed, such as thermal reflow [120,121], chemical–mechanical planarization (CMP) [122], and multiple-exposure techniques [123,124]. However, these special fabrication techniques increase complexity. Therefore, approaches that use the design of geometry of the waveguide have been proposed to reduce scattering loss, such as using an ultrathin core waveguide [125,126] and multimode ridge waveguides [127]. The intuitive idea is to reduce the interaction between the optical field confined in the optical waveguide and the rough sidewall. To ensure single-mode propagation, the multimode waveguide can be connected with a single-mode waveguide with a linear adiabatic taper [109] or designed with adiabatic propagation [128]. Figure 17a,b shows the schematic diagram and the microscope image of the fabricated ultrahigh-Q MRR, respectively. The ultrahigh-Q MRR is fabricated based on the standard SOI MPW-foundry processes [129]. The resonator is mainly composed of two multimode straight waveguides (MSWs) connected with two 180° modified Euler multimode waveguide bends (MWBs). Figure 17c shows the simulated light propagation in the Euler MWB and the mode profiles at the input and output ports. The test results show that the transmission loss of the optical waveguide is only 0.065 dB cm⁻¹, which is 20 times smaller than that of traditional SOI strip optical waveguides. Consequently, an intrinsic Q-factor up to 1.02 × 10⁷ is demonstrated for the first time based on the standard 220-nm-SOI MPW-foundry processes. Further, a tunable MPF with an FWHM bandwidth of 20.6 MHz and a tuning range of about 20 GHz is realized based on the ultra-high-Q MRR, as shown in Figure 17e,f, respectively.

Using PM-IM conversion, it is very convenient to achieve a microwave bandpass filter. However, the optical phase of the MRR-based notch filter can also affect the frequency response of the MPF. This is because the detected RF power can be affected by the phase variation of the two sidebands. To solve this problem, it is proposed to use a phase compensator before the PD to overcome the RF distortion induced by the unwanted phase variation [130], as shown in Figure 18a. The measured MPF response is illustrated in Figure 18b. It can be noted that the FWHM bandwidth is only 1.05 GHz, which is 0.8 GHz narrower than the FWHM bandwidth of the MPF without phase compensation.

PM-IM conversion is used to realize that single-passband MPF is effective and works well. The premise is that optical notch filters do not introduce an additional phase outside the notch region. In practice, however, the additional non-zero phase introduced by the filter outside its notch band prevents the complete elimination of microwave interference out of the passband of MPF, resulting in a significant reduction of the rejection ratio of the MPF and the shape factor. In reference [116], a tunable single passband MPF is realized with an improved shape factor and rejection ratio. Figure 19a shows the schematic diagram and illustration of the proposed single passband MPF, which is realized by using an optical double-notch based on two cascaded MRRs. By slightly changing the FSRs of the two MRRs, two optical notches with slightly different bandwidths and center frequencies can be obtained. By locating the optical carrier in the middle of the two microring notch filters, the additional phase introduced by each filter outside the notch regions is effectively eliminated. Meanwhile, the shape factor and the rejection ratio of the single passband MPF are also improved, as shown in Figure 19b. In addition, the bandwidth of the MPF is determined by the bandwidth difference of the two optical notch filters, rather than their respective absolute bandwidths. Therefore, to achieve a narrow-passband MPF, the bandwidth requirement is relaxed. Results show that the shape factor of MPF is 1.78, as shown in Figure 19b. The center frequency can be tuned from 6 to 17 GHz, and the rejection ratio is increased to 20 dB, as shown in Figure 19c.

3.1.2. High Flexibility

Promoting flexibility, including tunability and reconfigurability, is another focus of IMPF. The flexibility of IMPF is a major advantage over traditional electrical filters. To enlarge the bandwidth reconfigurable range, an SOI-based flat-top MPF is demonstrated in reference [131]. A 10th-order MRR based on a coupled resonator optical waveguide (CROW) and a PD is monolithically integrated based on SOI. The schematic structure of the SOI-based reconfigurable MPF is shown in Figure 20a. By adjusting the electrical power applied to the micro heaters, the resonant wavelengths of the 10th-order MRR can be tuned to realize an optical bandpass filter with a flat top, as shown in Figure 20b. By changing the wavelength of the optical carrier, the FWHM bandwidth of the MPF is adjusted from 5.3 to 19.5 GHz, while the rejection ratio remains higher than 30 dB. Notably, although the bandwidth can be reconfigured, the center frequency of the MPF cannot be tuned. Therefore, a tunable and reconfigurable MPF realized by using cascaded MRRs is proposed [107]. The schematic diagram of the proposed device is shown in Figure 20d. Four cascaded MRRs are employed in the device, and each coupling region is designed with a tunable MZI. By adjusting the electrical power applied to the micro heater deposited on either arm of the MZI coupler, the coupling coefficient is adjusted, and the bandwidth of the MRR can be consequently reconfigured. Figure 20e shows that the bandwidth of the MPF can be reconfigured from 0.7 to 2.0 GHz. Figure 20f,g shows that the center frequency is tunable from 5.2 to 35.8 GHz, and the rejection ratio remains above 40 dB in the tuning process.

Similar to the versatile programmability of discrete signal processing algorithms, a variety of advanced and complex signal processing functions can be built flexibly and dynamically based on FIR filter configurations [132]. A silicon-based four-tap FIR programmable MPF is proposed in reference [26], as shown in Figure 21a. The MPF is divided into four taps by using cascaded multimode interference (MMI) couplers. The delay between two adjacent waveguides is 10 ps, which is achieved by using waveguides of different lengths. Both amplitude and the phase modulations are implemented with separate thermal electrodes, realizing adjustable center wavelength, bandwidth, and passband shape, as shown in Figure 21b–e. The solution offers significant advantages such as compactness and ease of integration with electronic devices. This approach also demonstrates that a stable coherent MPF can be achieved by reducing its size. Similar to electronic field-programmable gate arrays, a demonstration of silicon-based reconfigurable two-dimensional photonic waveguide networks is reported in reference [27]. As shown in Figure 21f, the mesh is composed of seven-hexagonal MZI waveguide cells. Each cell allows independent control of amplitude and phase, thus realizing a variety of photonic circuits with different structures. Figure 21h,i illustrates typical amplitude responses of the reconfigured unbalanced Mach–Zehnder Interferometer (UMZI) and a bandpass CROW structure, respectively. In addition, over 20 different configurations of photonic circuits are demonstrated, realizing classical FIR and IIR signal processing functions as well as multi-port linear optical operations. This hexagonal lattice network enhances reconfiguration performance, increases the number of switching elements per unit area, and reduces the loss of per-space resolution.

MPFs using FBGs as the delay unit have been described in discrete devices. Many works based on FBGs have realized incoherent multi-tap MPFs [30,44,133,134]. In contrast, integrated waveguide Bragg grating (WBG) can realize IMPFs due to their compactness and flexibility [28,29,135,136]. Arbitrary amplitude and phase responses can be produced by manipulating the coupling coefficients and perturbations of the grating profile [137]. A tunable and bandwidth-reconfigurable MPF has been proposed in reference [135]. The core structure of the MPF is a high-order distributed feedback resonator (DFBR) based on waveguide Bragg gratings, as shown in Figure 22a. The DFBR is realized by inserting PS segments between Bragg grating mirrors (BGMs) in a corrugated silicon strip waveguide. Therefore, fourth-order multicavity filtering can be achieved. Compared to MRR, DFBR can simplify the design of narrow bandpass transfer functions [138]. Figure 22b illustrates the measured spectral transmission of the DFBR MPF, and Figure 22c shows the details of the passband window. In this case, an FWHM bandwidth of 4.5 GHz is measured. Figure 22d,e shows the tunability from 0 to 70 GHz at reconfigurable bandwidths of 5 GHz and 10 GHz, respectively.

3.1.3. Optical All-Pass Filter

Generally, the MPFs are realized by using the amplitude frequency response of optical devices, including lowpass, highpass, bandpass, and bandstop filters. In contrast, the all-pass filter (APF) has a constant amplitude frequency response, and phases only change with frequency. Therefore, the APF can also be called a phase filter or all-pass network. Because of this unique characteristic, the APF is particularly suitable for situations where only optical phase manipulation is required, such as microwave photonic phase shifters [139,140], MPFs [104,141], the synchronization of optical time division multiplexed (OTDM) systems [142], and dispersion compensation [143,144]. Previously implemented APF does not include full-band all-pass responses, but actually includes flat-top bandpass filters with a large bandwidth [145,146,147,148]. The passband amplitude also undulates with frequency [149,150,151]. Therefore, the APF still produces resonant peaks of a smaller amplitude at resonance [152,153,154,155]. Because of the existence of waveguide loss and coupling loss, the APF directly obtained based on MRR is not truly all-pass. In order to eliminate the amplitude variation caused by loss, an APF in the passive silicon waveguide is proposed and demonstrated in reference [91]. The APF uses optical interference between the outputs of an all-pass MRR and a straight waveguide. Figure 23a shows the schematic diagram of the optical APF. The phase response can be continuously tuned by adjusting the electrical power applied to the micro heaters. The application of APF in a microwave photonic phase shifter is further explored. APF is an ideal microwave photonic phase shifter device because only the phase is changed in the frequency response, while the amplitude remains constant. The measured microwave phase is shifted from 0 to 1.84π, and the corresponding RF power variation is less than 1 dB, as shown in Figure 23b,c. In addition, the APF can also be realized by using a dual-injection MRR with self-loss compensation [92]. The structure shown in Figure 23d can be designed to realize an APF based on the MZI-assisted MRR by using the self-compensation of loss. By adjusting the operating state, the device acts as if the transmission loss from the input to the through port is compensated. When the compensation is precisely equal to the transmission loss, the APF can be realized. The measured microwave phase shift and the corresponding amplitude, with respect to the frequency, are shown in Figure 23e,f, respectively. It can be seen that the power variation is less than 0.8 dB during the microwave phase shifting. However, both of the two schemes realize first-order APFs, which achieve a phase shift range of less than 2π. In order to obtain a larger phase shift range and time delay, it is possible to directly cascade multiple first-order APFs to obtain higher-order APFs. The corresponding insertion loss and system complexity will increase consequently. To solve these problems, it is demonstrated that a second-order APF is successfully achieved by cascading a first-order APF with an all-pass MRR [93]. Results show that the phase shift can be adjusted from 0 to 3.27π. Notably, the topology of the second-order APF can be easily extended to achieve higher-order APFs.

Therefore, APF has a constant amplitude response and varied phase response and is particularly suitable for occasions where pure phase manipulations are required, such as a microwave photonic phase shifter, continuous time delay, and dispersion compensation in devices.

3.1.4. System Integration

IMPF has many unparalleled advantages over discrete-based MPF. To further reduce the size, weight, power, and cost (SWaP-C), it is also desired to realize a fully integrated microwave photonic system, including light sources, electro-optical modulators, filters, delay lines, PDs, and other functional devices. Early demonstrations of tunable and reconfigurable MPFs have been reported by integrating SOA and PM [81,82,156].

A milestone demonstration of the first full IMPF [157] was realized in 2017. The chip contains all of the core components in a complete microwave photonic system, including a tunable distributed Bragg reflection (DBR) laser, a dual-drive Mach–Zehnder modulator (DDMZM), a tunable optical filter based on a ring-assisted MZI (RAMZI), and an InP PD. As shown in Figure 24a, the image of a fabricated die is approximately 6 × 4 mm². The packaged system is shown in Figure 24b. Based on this compact photonic circuit, an RF bandpass filter with a GHz-level bandwidth and moderate tunability is successfully implemented, as shown in Figure 24c. Although the performance of the MPF is constrained by the modulation bandwidth, relatively high waveguide loss, and RF crosstalk, this integration of a microwave photonic system is still the first successful attempt. The performance in the future can be improved based on the high-performance components based on the InP platform [132], including a high-power laser with low relative intensity noise (RIN), MZMs with low half-wave voltage, and high-efficiency PDs with high responsivity [158,159,160,161,162].

As a widely used platform for IMPFs [163,164], silicon-based photonic circuits offer compact size, low loss [120,165], and high-speed modulators [166,167] and PDs [168]. In reference [169], a silicon-based multifunctional integrated MPF is reported, as shown in Figure 25a. A high-speed PM, a thermally tunable high-Q MDR, and a high-speed PD are integrated on the chip. Based on the integrated chip, a bandpass MPF is achieved. The FWHM bandwidth is 1.93 GHz, and the center frequency can be tuned from 3 to 10 GHz. Additionally, the estimated power consumption of tuning the MPF is no more than 0.37 mW. If the microwave signal output from the MPF is fed back to the RF port of the PM, an integrated OEO can be obtained when the loop gain is higher than the loss [170]. The oscillation frequency of OEO is determined by the center frequency of the MPF. The center frequency of the MPF can be tuned by adjusting the electrical power applied to the micro heaters on the MDR. Therefore, the OEO oscillation frequency can be tuned. Notably, silicon is an indirect bandgap material with low luminescence efficiency. Consequently, the implementation of a silicon-based light source is the most difficult constraint for silicon-based fully monolithic integration.

The InP platform enables the monolithic integration of all active and passive devices. However, the large transmission loss severely limits the resolution bandwidth. Due to the low loss, Si and Si₃N₄ platforms are ideal candidates for achieving high-resolution bandwidths. Unfortunately, the optical source integration on these substrates is still a long-standing challenge. To solve this problem, hybrid integration has recently been recognized as a feasible way to build effective bridges between different material platforms [171]. A hybrid integrated MPF that achieved high filtering performance is demonstrated in reference [105]. The MPF is realized through the hybrid integration of an InP chip-based laser and a monolithic silicon photonic circuit consisting of a DDMZM, a high-Q MRR, and a PD. As shown in Figure 26a, the footprint value of the chip is approximately 1.34 × 3.18 mm². In the hybrid integration process, the InP-based LD chip and the Si photonic chip are optically connected based on off-the-shelf micro-optic components. The devices are assembled into a compact package, as shown in Figure 26b. This IMPF features a high-resolution bandwidth of 360 MHz, a wide frequency tuning range from S-band to K-band (3–25 GHz), and a large rejection ratio over 40 dB, as shown in Figure 26c–e. In addition, the amplitude frequency responses of the IMPF can be flexibly switched between bandpass and bandstop functions. The scheme uses a hybrid integration approach to realize miniaturized, high-performance, and low-cost MPF.

3.2. Applications of IMPF

Compared with the MPF based on discrete devices, IMPF can be applied to integrated microwave photonic systems, which is advantageous with its small size, light weight, low power consumption, and low cost.

3.2.1. Integrated OEO

As mentioned in Section 2.2.1, the application of MPFs in OEO offers various advantages, including low phase noise and broadband tunability and high stability. However, most OEOs are implemented based on discrete devices and have a large size, heavy weight, and high power consumption and cost. The silicon-based IMPF in reference [169] utilizes a high-Q MDR, and an integrated OEO can be realized by feeding the MPF output back to the PM [170]. In reference [20], an on-chip tunable PT-symmetric OEO based on an SOI wafer is realized. A high-Q MRR, an adjustable MZI, and two high-speed PDs are monolithically integrated on the chip. The MRR is used to implement a tunable MPF for performing coarse mode selection. To obtain a narrower bandwidth, the optical waveguide scattering loss is reduced by increasing the waveguide width. To ensure single mode propagation, the wide waveguide is connected with a single mode waveguide by an adiabatic taper, as shown in Figure 27a. Figure 27b illustrates the transmission of the MPF. The combination of high-quality MPF and PT symmetry significantly enhances mode selection and enables the OEO oscillation frequency to be widely tuned. Results show that stable single-mode oscillation of the OEO is achieved, and the oscillation frequency can be tuned from 0 to 20 GHz, as shown in Figure 27c,d. Additionally, the on-chip SBS can also generate a narrow-band MPF, which can be used for mode selection in OEO. Reference [21] reports a chip-based highly pure microwave source harnessing SBS. The design and construction of the integrated OEO are shown in Figure 27e. The SBS-based MPF also provides gain to the cavity, and no additional re-amplification is required to compensate for the link loss. Because of the large Brillouin gain employed in a short cavity [172,173], the OEO can achieve single mode oscillation with high stability without detrimental mode hopping. Figure 27f shows the measured electrical spectrum of the OEO, which shows a narrow linewidth. Figure 27g shows that the oscillation frequency of the SBS-based OEO can be tuned from 5 to 40 GHz.

3.2.2. Integrated Microwave Photonic Frequency Measurement

In Section 2.2, several schemes of achieving photonic-assisted microwave IFM were introduced. However, IFMs based on discrete devices are usually large and have low stability against environmental disturbances. In addition to the schemes mentioned above, integrated optoelectronic devices, such as Bragg gratings [25], MRRs [174,175], MZIs [176], and MDRs [24,177], can also be used to realize the frequency-to-power mapping (FTPM). IFM can be achieved by establishing the relationship between the output power and the microwave frequency. To adapt different applications, it is required to achieve an adjustable measurement range and high accuracy. A chip-based microwave photonic notch filter using an integrated silicon MDR can be utilized in microwave IFM with an adjustable measurement range [24]. The schematic diagram of the microwave frequency measurement system is shown in Figure 28a. Figure 28b shows the measured amplitude frequency responses with different laser wavelengths. Therefore, different ACFs can be calculated, and finally, a fixed FTPM can be obtained post-processing. Figure 28e shows that the measurement error is ±0.2 GHz when the measurement range is from 9 to 19 GHz. However, there is a trade-off between the accuracy and measurement range. Namely, a smaller measurement range results in a higher resolution. Reducing the frequency measurement range can increase the steepness of the ACF curve, thus improving the frequency measurement accuracy.

In addition to the FTPM scheme, frequency-to-time mapping (FTTM) can also be used to establish the relationship between microwave frequency and time. Although FTTM does not support IFM, it can be used to obtain the frequencies of microwave signals with different types, including single-frequency signal, multi-frequency signal, frequency-hopping signal, and frequency-chirped signals. It is also essential to achieve higher accuracy. In reference [178], an FTTM scheme is proposed based on an ultra-high Q factor MDR. The schematic diagram of the frequency measurement is shown in Figure 28f, and Figure 28g shows the waveforms corresponding to points A, B, C, D, and E, respectively. The unknown microwave signal is firstly modulated by an intensity modulator (IM) to a periodically sweeping triangular wave and then filtered by a band pass filter to remove one sideband and input to a high-Q MDR. Finally, an oscilloscope (OSC) displays the electrical signals received by the PD. When the input optical frequency is exactly equal to one of the resonant frequencies of the MDR, a depression will appear on the OSC. There will be a time difference ∆t between the appearance of the depression and the beginning of the falling edge of the triangle wave, as shown by the waveform at point E in Figure 28g. Therefore, a mapping of microwave frequency to time can be established, and the unknown microwave frequency can be identified by measuring ∆t. For the magnesium fluoride MDR, Figure 28h shows an FWHM bandwidth of 0.6291 MHz at a wavelength of 1556.77 nm. Figure 28i illustrates the linear relationship between ∆t and microwave frequency, and Figure 28j shows the calculated frequency measurement error from 14.25 to 17.25 GHz. It can be noted that the error remains below 10 MHz over the tested 3 GHz range. This scheme provides a solution to realizing a high-performance microwave frequency measurement system with small size and light weight.

4. Conclusions

The MPF has been emerging as a promising technology to meet the growing demands of modern communication systems, and it offers a wide range of advantages over traditional electronic filters. In this review, we provide a comprehensive introduction of MPFs, discussing their theoretical foundations, various applications, and the evolution of these MPFs from discrete to integrated devices. These concepts are the cornerstone for designing and analyzing MPFs, enabling researchers and engineers to explore new possibilities in signal processing and filtering. Various applications of MPFs are discussed, demonstrating their versatility and effectiveness in different fields. OEOs benefit from MPFs to achieve wide frequency tunability and reconfigurable signals. Advances in microwave frequency measurements have also been made through integrated photonic technology, improving the accuracy and bandwidth of frequency measurements. In addition, novel filter functions, such as APF, are also introduced. APF is a good device to manipulate signal phases. Therefore, microwave photonic phase shifting and adjustable time delay line can be obtained based on APF. The evolution of MPFs from discrete to integrated devices has played a crucial role in expanding their utility and integration into existing systems. These on-chip integrated devices offer compactness, scalability, and higher reliability. However, challenges such as fabrication complexity and device performance limitations need to be addressed to fully realize the potential of IMPFs. For the fast development of MPFs, as well as fabrication techniques and materials, we believe that MPFs have great potential for shaping the future of MWP, enabling new capabilities in various fields and promoting advances in communication and signal processing systems.