Wavelength-swept lasers (WSLs) are critical components in optical sensing systems, such as fiber Bragg grating (FBG) sensing, optical coherence tomography (OCT), tunable diode laser absorption spectroscopy (TDLAS), and light detection and ranging (LiDAR), as they significantly affect the performance of these systems [1–12]. For example, regarding the FBG sensing technology, the wider the wavelength sweep range of the WSL, the more sensors with different central wavelengths that can be connected simultaneously. This also increases the sensor measurement range. The higher the output power of the WSL, the more channels that can be simultaneously monitored based on the wavelength-division multiplexing (WDM) principle. This further increases the number of sensing points. The higher the sweeping rate of the WSL, the higher the measurement speed and data acquisition frequency of the sensing system.

One common wavelength-swept light source uses an amplified spontaneous emission (ASE) broadband light source or a semiconductor optical amplifier (SOA) to provide broadband light, with a fiber Fabry–Perot (F-P) tunable filter for wavelength selection. The F-P filter relies on injecting a periodically varying current into piezoelectric ceramics, which drives the coated cavity and causes the transmission spectrum to vary periodically. The mechanical components in this design compromise its stability and vibration resistance, leading to a complex design and high costs [13–15]. Recently, various new types of WSLs have been proposed, including Fourier domain mode-locked (FDML) lasers [16,17] and actively mode-locked (AML) lasers [18]. These light sources have significantly improved sweeping speed. However, they are often complex in structure, expensive, and unstable, which limits their practical applications. Another technique uses tunable distributed Bragg reflector (DBR) semiconductor lasers. This approach utilizes the Vernier effect, which enables the laser to rapidly lase at specific wavelength steps to achieve wavelength sweeping [19–22]. The WSLs based on DBR semiconductor lasers have been commercially applied in certain fields. However, they have wavelength gaps during sweeping. Additionally, their wavelength control is relatively complex, and mode hopping tends to occur during fast wavelength sweeping. Their robustness is also poor, making them susceptible to external environmental changes.

Among these, distributed feedback (DFB) lasers are attractive because of their stable single-longitudinal mode (SLM) properties and simple wavelength-tuning mechanisms [23–25]. These lasers achieve fast, continuous (gap-free) wavelength sweeping by injecting a rapidly varying current. Previous studies applied DFB-based WSLs in optical frequency-domain reflectometry (OFDR) and FBG sensor interrogation systems [25,26]. However, the wavelength sweeping range of these DFB lasers is limited to 1–4 nm, and no fast wideband-sweeping WSL has been developed. This imposes significant limitations on optical sensing systems that require a broadband wavelength sweeping range. Especially for FBG sensor interrogation systems and OCT technologies, traditional DFB laser arrays are unable to meet the demand for a large wavelength sweeping range. This is primarily owing to the limited wavelength tuning range of a single DFB laser. Theoretically, a broad wavelength tuning range, such as 60 nm, requires an integration of dozens of DFB laser units to cover the wavelengths of interest [27,28]. Such integrations significantly increase the fabrication complexity and reduce the yield of the chips. In addition, owing to the influence of thermal effects, the wavelength of the DFB laser changes nonlinearly with the injection current. This prevents the achievement of a linear sweep across the entire wavelength range, acquisition of the wavelength–time relationship, and achievement of precise mapping from time to frequency as it introduces significant difficulties in accurate wavelength interrogation.

Thus, with the aim of overcoming the above-mentioned drawbacks and expanding the application range of WSLs, we developed an ultra-wideband and high-speed wavelength-swept DFB laser array in this study and proposed a high-precision measurement system for nonlinear frequency variation. Our proposed WSL contains 24 monolithically integrated lasers in an $8\times 3$ interleaving matrix structure, with a wavelength spacing of 2.5 nm. In addition, we used the reconstruction-equivalent-chirp (REC) technique to simplify grating fabrication and enhance precise control of the wavelength spacing. A cascaded $\mathrm{Y}$-branch combiner, which shares an identical active layer with the lasers, was used to realize a single waveguide output, which only required a small current to compensate for material absorption and reduce the fabrication complexity compared with the butt-coupling technique [29–31]. A semiconductor optical amplifier (SOA) was integrated in front of the laser array to amplify and adjust the optical power. To achieve high-precision mapping from the time to frequency domain, a nonlinear frequency variation measurement system based on dual F-P etalons with different free spectral ranges (FSRs) was designed. The system accurately measures the dynamic relationship of frequency variations over time, enabling precise parameters interrogation. The WSL integrates an all-solid-state structural design with no movable mechanical parts, thus exhibiting high robustness. As a result, the wavelength of the laser array was precisely controlled, with the wavelength deviations of all 24 lasers maintained within $\pm 0.2\mathrm{nm}$. Additionally, its side-mode suppression ratio (SMSR) was $>45\mathrm{dB}$, and relative intensity noise (RIN) was $<-130\mathrm{dB}/\mathrm{Hz}$. By sequentially injecting pre-distorted driving currents, a high-speed, wideband wavelength continuous (gap-free) sweep over the 60 nm range with a sweep rate of 82.7 kHz was achieved, and an ultrafast wavelength tuning speed of 4.96 nm/μs was achieved. The output power was stably maintained at 10 mW through a real-time dynamic adjustment of the SOA current.

The proposed WSL is applied to the FBG sensor interrogation system. In the high-low temperature and strain experiments conducted to verify the effectiveness of the proposed system, the system dynamically interrogated FBGs in real-time for up to 3 h. The demodulation results indicated the interrogation linearity of these FBGs to be $>0.9983$, demonstrating a good relative accuracy and excellent interrogation performance of the system. Continuous measuring of FBG spectra showed a maximum wavelength deviation of $<2\mathrm{pm}$ (fluctuating within $\pm 1\mathrm{pm}$), which demonstrated its remarkable wavelength sweep repeatability. The interrogation system achieved precise measurements of high-frequency vibrations in the frequency range of 2–8 kHz, with a time resolution of 12.09 μs. Furthermore, a stable working state was maintained under challenging vibration and shock conditions.

Compared to other typical DFB laser arrays, such as the multi-channel interference (MCI) laser with a complex tuning mechanism [32], lasers integrated with star couplers and optical amplifiers [29], matrix-grating strongly gain-coupled (MG-SGC) DFB laser arrays [33], lasers using multi-mode interference (MMI) couplers [34,35], and micro–electro–mechanical systems (MEMS)-based tunable DFB lasers [36], our proposed device offers significant advantages. These include a simpler design, a straightforward tuning mechanism, lower cost, better vibration and shock resistance, and a larger wavelength sweeping range without the need for redundant components or complex coupling structures. Table 1 presents a comparison of our results with those from other research groups. Notably, this represents the broadest wavelength sweeping range reported to date for high-speed WSLs utilizing DFB laser arrays. The proposed WSL is considered highly advantageous for high-performance optical sensing systems, such as FBG sensor interrogation systems and OCT, that require fast, broadband, stable, and continuous wavelength sweeping light sources.Table 1.

Parameters of Wavelength-Swept Lasers Based on DFB Laser Arrays Reported in Recent Years

The remainder of this paper is structured as follows. In Section 2, we have detailed the design and fabrication of the proposed WSL. In Section 3, the static characteristics of the proposed WSL are detailed. In Section 4, we have discussed the pre-distortion of the driving currents injected into the device to maintain constant output power and achieve more uniform frequency variation over time. In the same section, we have also provided a detailed explanation of the precise measurement method for nonlinear frequency variation of the WSL. In Section 5, the dynamic sweeping characteristics of the WSL and real-time FBG interrogation experiments have been detailed. Finally, the paper is concluded in Section 6.

To achieve a continuous wavelength-tunable range of 60 nm ultra-wideband, we integrated 24 DFB lasers based on the REC technique. However, as the number of lasers on the same waveguide increases, the lasers become more susceptible to grating crosstalk and reflections from other lasers. This can cause unpredictable mode problems and deteriorate SLM properties [37]. For DFB laser arrays designed in parallel, the larger number of parallel connections requires on-chip integration with multilevel optical combiners. This can result in significant optical power loss and an increased chip size [35].

Matrix laser arrays require precise grating phase control. A phase error can lead to undesirable reflections and deteriorate the SLM properties of all channels. DFB lasers typically use phase-shift gratings to ensure excellent SLM properties. However, conventional phase-shift gratings require electron beam lithography (EBL) for fabrication. While EBL enables nanoscale precision, it introduces a series of issues, including high cost, low yield, and complex fabrication methods. The REC technique was used in grating designs to ensure excellent SLM properties and high laser array yields. It also helps improve the accuracy of the gratings [40–42].

Gratings can be easily created using the REC technique with only one step each for holographic exposure and microscale-resolution lithography. This eliminates the need for time-consuming and expensive EBL and ensures consistent wavelength spacing between adjacent channels. Devices based on the REC technique require special sampling of the uniform Bragg grating, which is then integrated into the device. The grating structure is illustrated in Fig. 2. For the sampled grating without the introduction of a $\pi$ phase shift, two degenerate modes exist, causing lasing instability. After introducing the $\pi$ phase shift, the number of degenerate modes reduces to one. When the equivalent $\pi$ phase shift is located at the halfway point of the cavity length, the stability of the laser’s SLM property is optimal [38]. In this case, the longitudinal mode corresponds to the Bragg wavelength. To ensure the SLM property of the laser, an equivalent $\pi$ phase shift was introduced at 1/2 of the cavity length of each DFB laser, which helps to prevent mode hopping. The equivalent $\pi$ phase shift was designed using the REC technique. The principle of REC technology is described as follows [41].

Initially, a specially designed sampling process was used for uniform gratings. The index modulation change $\mathrm{\Delta }n(z)$ of a sampled Bragg grating can be described by $$\mathrm{\Delta }n(z)=s(z)\exp \left(j\frac{2\pi z}{{\mathrm{\Lambda }}_0}\right)+\mathrm{c.c}\mathrm{.}$$(1)where $s(z)$ is the periodic function of the sampling modulation, with $s(z)=1$ at the positions where the grating is retained, and $s(z)=0$ at the positions where the grating is not retained. ${\mathrm{\Lambda }}_0$ is the period of the uniform seed grating. According to the Fourier series expansion, $s(z)$ can be expressed as $$s(z)=\sum _m{F}_m\exp \left(j\frac{2m\pi z}{P}\right),$$(2)where $P$ is the period of the sampling grating and $F_m$ is the Fourier coefficient corresponding to the $m^{\mathrm{th}}$ order channel of the sampled grating. From Eqs. (1) and (2), we can obtain $$\mathrm{\Delta }n(z)=\sum _m{F}_m\exp (j\frac{2m\pi z}{P}+j\frac{2\pi z}{{\mathrm{\Lambda }}_0})+\mathrm{c.c}.$$(3)

The grating period of the $m^{\mathrm{th}}$ order sub-grating can be calculated as $${\mathrm{\Lambda }}_m=\frac{P{\mathrm{\Lambda }}_0}{P+m{\mathrm{\Lambda }}_0}.$$(4)Equation (3) can be rewritten as $$\mathrm{\Delta }n(z)=\sum _m{F}_m\exp \left(j\frac{2\pi z}{{\mathrm{\Lambda }}_m}\right)+\mathrm{c.c}.$$(5)According to Eq. (5), the sampled grating is a superposition of different orders of sub-gratings with different grating periods ${\mathrm{\Lambda }}_m$. To achieve SLM properties, we introduce an equivalent $\pi$-phase shift to the sampled Bragg grating. The equivalent phase shift corresponds to introducing a change $\mathrm{\Delta }P$ in the sampling period at $z_0$. The index modulation of the $m^{\mathrm{th}}$ order sub-grating changes to $$\mathrm{\Delta }{n}_m(z)=\{\begin{array}{cc}\sum _m{F}_m\exp (j\frac{2\pi z}{{\mathrm{\Lambda }}_0}+j\frac{2m\pi z}{P})+\mathrm{c.c}\mathrm{.}& z\le z_0\\ \sum _m{F}_m\exp (j\frac{2\pi z}{{\mathrm{\Lambda }}_0}+j\frac{2m\pi z}{P}-j{\theta }_m)+\mathrm{c.c}\mathrm{.}& z>z_0\end{array},$$(6)where the equivalent phase shift ${\theta }_m$ is expressed as follows: $${\theta }_m=2m\pi \frac{\mathrm{\Delta }P}{P}.$$(7)

In general, the $+1^{\mathrm{st}}$ order sub-grating is used as the laser resonant cavity. When $\mathrm{\Delta }P=P/2$, an equivalent $\pi$ phase shift can be achieved in the $+1^{\mathrm{st}}$ order sub-grating. The grating period of the $+1^{\mathrm{st}}$ order sub-grating is derived as $$\frac{1}{{\mathrm{\Lambda }}_{+1}}=\frac{1}{{\mathrm{\Lambda }}_0}+\frac{1}{P},$$(8)where ${\mathrm{\Lambda }}_{+1}$ is the grating period of the $+1^{\mathrm{st}}$ order sub-grating, ${\mathrm{\Lambda }}_0$ is the uniform seed grating period, and $P$ is the sampling period. The Bragg wavelengths of the uniform seed grating and the $+1^{\mathrm{st}}$ order sub-grating are designed to be distant from the gain region to prevent lasing and at the center of the gain region, respectively. Using the same uniform seed grating, different laser wavelengths were obtained for different sampling periods. This is the main concept of the REC technique [41]. The core idea is that when $m$ is not equal to zero, a smooth change in $\mathrm{\Delta }P$ results in an equivalent chirp. Correspondingly, when $\mathrm{\Delta }P$ is changed discretely, an equivalent phase shift can be obtained.

In this case, the period variation of the sub-grating ΔΛ${}_{+1}$ was derived from the variation of the sampling period $\mathrm{\Delta }P$ [Eq. (9)]. Notably, the tolerance to grating phase errors can be relaxed by a factor of ${(P/{\mathrm{\Lambda }}_0+1)}^2$, which is typically several hundred [42]. Therefore, the REC technique achieves an improvement in the precision of grating fabrication: $${\mathrm{\Delta \Lambda }}_{+1}=\frac{1}{{(\frac{P}{{\mathrm{\Lambda }}_0}+1)}^2}\mathrm{\Delta }P.$$(9)

The basic principle for fabricating the required grating in this work is to first perform a holographic exposure to obtain a uniform grating, and then use a mask for a standard photolithography step. Compared to the time-consuming and expensive EBL, the gratings produced using the REC technique not only benefit from the low cost, short fabrication time, and strong grating uniformity of holographic exposure, but also maintain the flexibility of electron beam exposure. Additionally, as analyzed earlier, the resulting error in the lasing wavelength is small, and precise wavelength control can be easily achieved. Therefore, this technique has the potential for large-scale industrial applications. The period of the uniform grating ${\mathrm{\Lambda }}_0$ was set to 257.886 nm. To precisely control the laser wavelength and compensate for the wavelength deviations, a chromatic dispersion coefficient $\mathit{\alpha }(=-0.0002)$ was introduced. The Bragg grating center wavelengths of LD1–LD24 were set to cover a wavelength tuning range of over 60 nm (Table 2). Using Bragg’s law and Eq. (7), the sampling period of each channel [Fig. 3(a)] was obtained. This period was distributed in the range of 2.998–6.063 μm, requiring only μm-level lithography to accurately control the lasing wavelengths of each channel.

The transmission spectrum of the entire laser grating provides a method for predicting the lasing characteristics. Therefore, the transmission spectrum of the grating in this structure was calculated using the transfer matrix method [43]. For instance, the grating transmission spectra of three randomly selected in-series laser units (LD5, LD13, and LD21) located in the same waveguide are shown in Fig. 3(b). The three marked peaks in the figure correspond to the lasing modes of the three DFB lasers, which were caused by the equivalent $\pi$ phase shifts. To avoid lasing, the Bragg wavelength of the uniform grating was set to 1650 nm, which was distant from the gain region.

Our device employed a ridged waveguide structure with a width of 2 μm. The epitaxial layers were grown using a two-stage metal–organic chemical vapor deposition (MOCVD) process. Initially, an $\mathrm{n}$-InP buffer layer, $\mathrm{n}$-InAlGaAs lower optical confinement layer, InAlGaAs multiple-quantum-well (MQW) structure, $\mathrm{p}$-InGaAsP upper optical confinement layer, $\mathrm{p}$-InGaAsP etch-stop layer, and $\mathrm{p}$-InGaAsP grating layer were sequentially grown on an $\mathrm{n}$-InP substrate. The MQW layer, which is shared by the laser, the cascaded $\mathrm{Y}$-branch, and the SOA, consists of five compressively strained InGaAlAs quantum wells and six tensilely strained InGaAlAs quantum barriers, sandwiched between two graded-index (GRIN) InGaAlAs separate confinement heterostructure (SCH) layers [44]. The etch-stop layer is designed to control the depth of the subsequent grating etching and prevent etching into deeper layers. It is separated by a $\mathrm{p}$-InP spacer layer. Subsequently, the sampling grating based on the REC technique design was fabricated using holographic exposure combined with micrometer-scale photolithography and etching techniques. The specific steps are as follows. Holographic exposure is first used to create a periodic stripe pattern. At this stage, no development step is performed. Instead, another contact exposure using a mask (with μm-level structures) is carried out. The pattern obtained from this exposure corresponds to the structure of the sampled grating. The development process is then performed to obtain the desired grating pattern. The pattern is etched, transferring the design to the grating layer on the substrate, thus fabricating the grating structure. The above are the processing steps for the sampling grating structure based on the REC technique design. A second epitaxial growth is then performed, where $\mathrm{p}$-InP cladding and $\mathrm{p}$-InGaAs contact layers were grown on the grating structure. The cladding layer is used for ridge waveguide etching and other processes, while the contact layer is used for contact with the metal electrode later. The ridged waveguides were formed by etching two 23-μm-wide grooves on either side. To ensure that the lasers on the same waveguide do not emit light simultaneously, adjacent sections were electrically isolated via shallow grooves etched on the ridged waveguides. Typically, the etching depth reaches about half of the waveguide height. The subsequent fabrication process included opening $\mathrm{p}$-metal contact windows, metallization, and wafer cleaving to form lasers. A microscopic top-view of the laser array chip is shown in Fig. 4. A 550 μm SOA was integrated at the front end of the laser array to adjust and equalize the optical power. The SOA was designed with a 7° bend and a tapered shape to minimize reflection. The waveguide width was tapered from 2 μm at the start to 6 μm at the end of the SOA. The $\mathrm{Y}$-branch combiner used in this work is a three-level combiner. The first level combines eight waveguides with a spacing of 25 μm into four waveguides with a spacing of 50 μm, with a longitudinal length of 230 μm. The second level combines four waveguides with a spacing of 50 μm into two waveguides with a spacing of 100 μm, with a longitudinal length of 350 μm. The third level combines two waveguides with a spacing of 100 μm into a single waveguide, with a longitudinal length of 500 μm. Each level of the waveguide combining is formed by two smoothly connected arc segments with opposite directions. The total length of the three-level combiner is 1080 μm. The cavity length of the laser is 400 μm. A 50-μm-long tail absorption region (TAR) was designed at the tail end of the laser array to absorb the light output from the rear facet, which was also designed with a 7° bend to minimize the reflection. The length and width of the chip were 3000 and 500 μm, respectively.

To stabilize the DFB laser array for fast and broadband continuous wavelength sweeping, a stable butterfly package was designed along with a high-speed driver circuit based on a field-programmable gate array (FPGA). For efficient heat dissipation and power supply, the DFB laser array was soldered onto an AlN submount using an Au-Sn solder to create a chip-on-submount (COS) assembly [Fig. 5(a)]. The electrodes of the DFB laser array were bonded to the electrodes of the submount using gold wire bonding, enabling a connection to the butterfly package housing [Fig. 5(b)]. An optical isolator was placed inside the package housing to prevent external reflection. The light emitted from the DFB laser array was focused at the center of the optical window using a coupling lens. The housing contained a thermoelectric cooler (TEC) to control the operating temperature of the laser chip. The packaged device was mounted on a drive circuit board for the power supply and temperature control. We introduced three $8\times 1$ electric switch chips, which, under FPGA control, enable multi-channel current source switching. As shown in Fig. 1, the eight lasers in the same row, such as LD1–LD8, LD9–LD16, and LD17–LD24, share a common current source within each group and are sequentially switched and multiplexed through electric switch chips. This allows for the periodic and sequential current injection for the 24 DFB lasers in the $8\times 3$ array with just a three-channel current source design. This design simplifies the circuit, significantly reduces the number of required electronic components, further lowers the cost, and improves yield.

For the drive circuit, a high-frequency FPGA served as the central control module and was connected to a 14-bit high-speed digital-to-analog converter (DAC). Then, the DAC output of analog voltage signals was converted into current signals using voltage-controlled current sources (VCCs). These VCCs powered the $8\times 3$ laser array, a $\mathrm{Y}$-branch combiner, and an SOA. A negative temperature coefficient (NTC) thermistor and a TEC, controlled by a proportional–integral–differential (PID) circuit, were used to maintain the laser chip temperature at 35°C for consistent wavelength sweeping.

The static lasing performance and wavelength accuracy of the proposed WSL, fabricated using the REC technique, were investigated. For LD1–LD24, the SOA injection current ($I_{\mathrm{SOA}}$), $\mathrm{Y}$-branch waveguide injection current ($I_Y$), small current for compensating for the material absorption of the front row laser ($I_{\mathrm{com}}$), and laser diode injection current ($I_{\mathrm{LD}}$) were set to 180, 180, 30, and 100 mA, respectively. The static lasing spectra for each channel [Fig. 6(a)] were recorded using a Yokogawa AQ6370 optical spectrum analyzer (OSA). All channels exhibited normal lasing with SMSR $>45\mathrm{dB}$ across all channels, indicating excellent SLM properties. The lasing wavelength of each channel was fitted linearly. The results are shown in Fig. 6(b). The fitted slope was 2.525 nm/channel, representing the average channel spacing, which deviated by only 0.025 nm from the design value. Figure 6(c) presents the wavelength deviation of the lasing wavelength from the fitted values for all 24 channels, with all deviations within the $\pm 0.2\mathrm{nm}$ range. This demonstrates the remarkable wavelength-control accuracy of the REC technique.

The differential resistance, threshold current, and optical power performance of the proposed WSL were investigated by testing its power (current) diagram; and $I_{\mathrm{SOA}}=100\mathrm{mA}$, $I_Y=100\mathrm{mA}$, and $I_{\mathrm{com}}=30\mathrm{mA}$. The $I_{\mathrm{LD}}$ was increased from 0 to 200 mA, and the voltage and output power of the laser were recorded. The P-I and V-I diagrams of the three randomly selected laser units (LD4, LD12, and LD20) located in the same waveguide are shown in Figs. 7(a) and 7(b), respectively. As shown in the P-I diagram, the threshold current of the proposed WSL ranged from 25 to 30 mA, and the V-I diagram indicated the differential resistance of LD4, LD12, and LD20 as 5.4, 5.8, and $5.2\mathrm{\Omega }$, respectively.

An SOA was monolithically integrated at the output port of the WSL to amplify and equalize the output power. The performance of SOA was also studied. The amplification characteristics of the SOA were tested for LD4, LD12, and LD20. $I_{\mathrm{SOA}}$ increased from 0 to 200 mA, and $I_Y=180\mathrm{mA}$, $I_{\mathrm{LD}}=100\mathrm{mA}$, and $I_{\mathrm{com}}=30\mathrm{mA}$. The output power with respect to $I_{\mathrm{SOA}}$ was recorded using a Thorlabs-PM100A optical power meter. As shown in Fig. 7(c), the output power of all three lasers on the same waveguide exceeded 22 mW at $I_{\mathrm{SOA}}=200\mathrm{mA}$.

To investigate the wideband current tuning capability of the proposed WSL, the chip temperature was maintained constant at 35°C using a TEC, and $I_{\mathrm{SOA}}=180\mathrm{mA}$, $I_Y=180\mathrm{mA}$, and $I_{\mathrm{com}}=30\mathrm{mA}$. Injection currents ranging from 50 to 200 mA in steps of 5 mA were sequentially applied to LD1–LD24. The lasing spectra and SMSRs for each current were recorded and analyzed using an optical spectrum analyzer (Yokogawa AQ6370, resolution 0.02 nm). As shown in Fig. 8, each laser achieved a wavelength tuning range of over 2.5 nm, ensuring full coverage of the wavelength interval between the channels. The 24 lasers achieved a continuous (gap-free) wideband wavelength current-tuning range of 1515–1575 nm, totaling 60 nm. The wavelength-tuning coefficient with respect to the current was $\sim 0.0187\mathrm{nm}/\mathrm{mA}$. Each laser exhibited a wavelength shift of 2.7–2.9 nm within the 150 mA current tuning range, demonstrating excellent broadband current tuning capability.

The fluctuations in the light beam intensity emitted by a laser diode are quantified by the RIN, which imposes a fundamental limit on optical communication systems [45,46]. Elevated RIN levels degrade the signal-to-noise ratio (SNR) and increase the bit error rate (BER) [47], underscoring the importance of minimizing RIN in lasers. The RIN measurements were conducted for all 24 channels of the WSL with $I_{\mathrm{SOA}}=100\mathrm{mA}$, $I_Y=100\mathrm{mA}$, $I_{\mathrm{LD}}=100\mathrm{mA}$, and $I_{\mathrm{com}}=30\mathrm{mA}$. The electrical-spectrum-analysis-based method outlined in Ref. [48] was used. The superimposed RIN spectra in the frequency range of 50 MHz to 15 GHz are shown in Fig. 9. As shown in Fig. 9(a), the RINs for LD1–LD24 were all below $-139\mathrm{dB}/\mathrm{Hz}$. The impact of varying the $I_{\mathrm{LD}}$ on RIN was extensively investigated. As depicted in Fig. 9(b), LD12 was configured with $I_{\mathrm{SOA}}=100\mathrm{mA}$, $I_Y=100\mathrm{mA}$, and $I_{\mathrm{com}}=30\mathrm{mA}$. $I_{\mathrm{LD}}$ was increased from 50 to 100 mA in 10 mA increments, and the RIN was measured at each current level. The highest RIN near the relaxation oscillation frequency ($f_r$) decreased rapidly from $-133$ to $-155\mathrm{dB}/\mathrm{Hz}$ with increasing $I_{\mathrm{LD}}$. Additionally, the influence of $I_{\mathrm{SOA}}$ on RIN was studied by setting LD12 with $I_{\mathrm{LD}}=100\mathrm{mA}$, $I_Y=100\mathrm{mA}$, and $I_{\mathrm{com}}=30\mathrm{mA}$. $I_{\mathrm{SOA}}$ was varied from 20 to 160 mA in 20 mA increments. As shown in Fig. 9(c), when $I_{\mathrm{SOA}}$ varied from 20 to 160 mA, the RIN of LD12 remained almost unchanged. Thus, the RIN did not degrade when the output power was amplified. Thus, the low RIN of the proposed DFB laser array is beneficial for applications in optical sensing and communication systems.

The linewidth of the proposed DFB laser array was measured using the delayed self-heterodyne (DSH) method [48–50]. The laser emitted from the proposed WSL was split into two arms, one of which passed through a 25 km delay fiber, while the other was frequency-shifted by 80 MHz using an acousto-optic modulator. Both arms were recombined to generate a beat frequency on the photodiode, which was then measured by the signal and spectrum analyzer (Rohde & Schwarz). Half of the full width at half maximum (FWHM) of the beat frequency signal represents the actual linewidth of the laser. During the measurement, the DFB laser array temperature was set to 35°C, with $I_{\mathrm{SOA}}=180\mathrm{mA}$, $I_Y=180\mathrm{mA}$, $I_{\mathrm{LD}}=100\mathrm{mA}$, and $I_{\mathrm{com}}=30\mathrm{mA}$. A typical laser unit was selected for testing, as shown in Fig. 10, with a 3 dB linewidth of approximately 3.48 MHz. The DFB laser array proposed in this work, although no special linewidth optimization design was made, still achieves a narrow linewidth performance of several MHz, which is comparable to the performance of traditional DFB lasers. Compared to vertical-cavity surface-emitting lasers (VCSELs) and some wavelength sweeping sources based on tunable F-P filters, it has a significant linewidth advantage [51,52]. This laser array is well-suited for applications such as FBG sensor interrogation, OCT, and high-resolution spatial imaging technologies [53–56].

For a constant $I_{\mathrm{SOA}}$, the SOA exhibited different amplification capabilities for light of different wavelengths owing to the influence of the SOA bandwidth and material gain spectrum range. The output optical power of the WSL was converted into an electrical signal using a photodetector (PD) and recorded using an oscilloscope (RIGOL MSO5354) (Fig. 10). To distinguish between each sweeping cycle accurately, a trigger signal (${\mathrm{Trigger}}_1$) was provided, which outputted high levels at the beginning and end of each sweeping cycle.

As shown in Fig. 11(a), when the $I_{\mathrm{SOA}}$ was maintained constant, the flatness of the output power was poor, which can severely affect the interrogation results in applications, such as FBG sensing. Therefore, to achieve a more balanced output power, the $I_{\mathrm{SOA}}$ was pre-distorted to continuously vary during wavelength sweeping, maintaining the output power at 10 mW. An automatic calibration program was developed using Python to improve efficiency and accuracy. As shown in Fig. 11(b), after real-time adjustment of $I_{\mathrm{SOA}}$, the maximum fluctuation of the optical signal collected by the oscilloscope decreased from 1.413 to 0.071 V. The output power of the WSL remained stable, demonstrating a significant power equalization effect.

To accurately measure the frequency variations of the WSL over time, a real-time measurement system (Fig. 12) was developed. F-P ${\mathrm{etalon}}_1$ and F-P ${\mathrm{etalon}}_2$ were selected with FSRs of 130 and 12.5 GHz, respectively. An F-P etalon consists of two parallel flat plates coated with a highly reflective layer on the inner side, allowing multiple reflections of incident light, which interferes with the creation of bright and dark interference patterns. F-P etalons are widely used in fiber-optic sensing and communication [57,58]. Each transmission peak in the transmission spectrum has a distinct fixed wavelength and frequency. The center frequencies of any two adjacent transmission peaks are equally spaced (equal to the FSR) and are minimally affected by temperature changes. In this study, the optical frequency variations over time were precisely calculated from the changes in the transmission spectrum, enabling assessment of the frequency-sweeping process of the WSL.

To facilitate understanding of the detailed pre-distortion process of WSL frequency-sweeping, LD12 and LD13 were randomly selected as examples to clarify. Figure 13 shows the dynamic variation in the frequencies of LD12 and LD13 with $I_{\mathrm{LD}}$. As shown in Fig. 13(a), before performing the $I_{\mathrm{LD}}$ pre-distortion, the rate of frequency change over time varied significantly. Near the end of the channel sweeping, the frequency changed too rapidly; the acquisition time intervals for this part of the data will be relatively short, which imposes higher requirements on the hardware acquisition speed and brings challenges to the hardware design of the acquisition system. Thus, pre-distortion needs to be applied to the $I_{\mathrm{LD}}$ to ensure that the frequency change becomes more uniform with time. Firstly, fitting and peak-searching operations are performed on all transmission peaks of F-P ${\mathrm{etalon}}_2$ within a single channel, calculating the peak positions along the horizontal axis ($t_1\sim t_{N+1}$, where $N$ is the total number of peaks). The time intervals $\mathrm{\Delta }{T}_i$ between each adjacent peak are then calculated, representing the time required for a 12.5 GHz change in optical frequency. The average time interval $\mathrm{\Delta }{T}_{\mathrm{ave}}$ is computed and each $\mathrm{\Delta }{T}_i$ is compared to $\mathrm{\Delta }{T}_{\mathrm{ave}}$. The laser injection current is adjusted for each pair of peaks based on their ratio to $\mathrm{\Delta }{T}_{\mathrm{ave}}$. After completing the adjustment for all $N$ segments, the new current values are fitted and interpolated to generate the current curve. This process is iteratively optimized until the $\mathrm{\Delta }{T}_i$ values are within a 15% error range of $\mathrm{\Delta }{T}_{\mathrm{ave}}$, completing the current pre-distortion optimization. After performing the $I_{\mathrm{LD}}$ pre-distortion, as shown in Fig. 13(b), the frequency variations with time became more uniform, which is beneficial to the design of the acquisition system. It is worth mentioning that strict linearization of the frequency or wavelength was not performed for the proposed WSL for fast pre-distortion accomplishment. The optical signal demodulation method proposed in this work does not rely on frequency or wavelength linearization sweeps.

When a WSL is used in the field of optical sensing, determining the relationship between the frequency of the sweeping light and time is necessary to map the sensing signal from the time domain to the frequency domain for accurate interrogation. The DFB laser array achieved ultra-wideband and high-speed wavelength sweeping by sequentially applying a rapidly varying current to each laser unit; however, its linearity was poor. To fully cover the wavelength spacing between channels, the wavelength spacing was set to be smaller than the tuning range of a single channel, resulting in a “frequency overlap” between adjacent channels. These issues represent the primary challenges in the use of DFB laser arrays for optical sensing. To address these issues, a high-precision measurement system (Fig. 12) for nonlinear variations in frequency over time based on dual F-P etalons was developed. The continuous wavelength sweeping light emitted by the WSL passed through the F-P etalons, forming a transmission spectrum with equal frequency intervals. The spectrum was converted into two electrical signals using two PDs. The signals were recorded using an oscilloscope (RIGOL MSO5354). In Fig. 14, the blue and red signals represent the transmission spectrum formed by the light passing through F-P ${\text{etalon}}_1$ and F-P ${\text{etalon}}_2$, respectively. The corresponding frequency intervals between adjacent peaks were ${\mathrm{FSR}}_1$ and ${\mathrm{FSR}}_2$, respectively.

In this study, ${\mathrm{FSR}}_1=130\mathrm{GHz}$ and ${\mathrm{FSR}}_2=12.5\mathrm{GHz}$. The intervals between the transmission peaks in the time domain were not uniform, indicating a nonlinear change in the frequency of light over time. Regions with faster frequency variations exhibited denser transmission peaks, whereas those with slower variations exhibited sparser peaks. Each peak had a specific frequency. This enabled the calculation of the relationship between the frequency variation and time based on the time-domain transmission peaks. As shown in Fig. 14, a trigger signal (${\mathrm{Trigger}}_2$) was also introduced. The signal level was inverted when switching between the unit lasers. This marked the start and end of each sweeping cycle and the time points for switching between different channels.

In Fig. 14(b), each blue peak has a red peak on both the left and right sides. The distance between the blue and first red peaks on the left is denoted as $a$ and that between the first red peaks on the left and right sides is denoted by $b$. The peak distance ratio ($R$) is defined as $R=a/b$. As ${\mathrm{FSR}}_1$ is more than 10 times ${\mathrm{FSR}}_2$, ${\mathrm{FSR}}_1$ is not an integer multiple of ${\mathrm{FSR}}_2$. Therefore, the value of $R$ for the blue peaks does not repeat within the range of two adjacent channels. In addition, the $R$ values of adjacent blue peaks differed significantly. Therefore, by calculating the $R$ of the blue peaks, we determined the existence of a repeated frequency sweeping band.

To facilitate the understanding of how to identify the overlap frequency sweeping section between two adjacent unit lasers, we use the example of LD7 and LD8. Figure 14(d) shows the enlarged view during the switching from LD7 to LD8. In the figure, the $R$ values of the two blue peaks before and after 3.38 μs were equal, indicating the same frequency value represented by the two blue peaks. Thus, the period from 3.33 to 3.42 μs was a repeated sweeping band.

To accurately calculate the relationship between the frequency variations and time, a fitting and peak-finding process was performed for each blue and red peak, as shown in Fig. 14(a). The $R$ value for each blue peak was calculated to identify the repeated frequency sweeping bands. The blue peaks were numbered sequentially from 1 to $i$, starting with the first blue peak at the beginning of the sweeping cycle, as shown in Figs. 14(c) and 14(d). This leads to the derivation of Eq. (10): $$j=\frac{(m-1)·{\mathrm{FSR}}_2+R_1·{\mathrm{FSR}}_2+(i-1)·{\mathrm{FSR}}_1-R_i·{\mathrm{FSR}}_2}{{\mathrm{FSR}}_2},$$(10)where $m$ represents the number of red peaks to the left of the first blue peak at the start of the sweeping cycle, and $R_1$ and $R_i$ are the peak distance ratios of the first and $i$^th blue peaks, respectively. This leads to the derivation of Eq. (11): $$f_j=f_0-j·{\mathrm{FSR}}_2,$$(11)where $f_0$ is the frequency of the first red peak at the start of the sweeping cycle, and $f_j$ is the frequency of the ($j+1$)^th red peak. Based on Eqs. (10) and (11), Eq. (12) can be obtained as $$f_j=f_0-(m-1+R_1-R_i)·{\mathrm{FSR}}_2-(i-1)·{\mathrm{FSR}}_1.$$(12)

Based on Eq. (12), the center frequency of each red peak was calculated. This enabled the precise determination of the WSL output frequency variation over time through further fitting and interpolation. This measurement method is simple and efficient, and has been proven to have high measurement accuracy in subsequent interrogation experiments. The proposed method provides a new measurement approach for nonlinear continuously sweeping light sources. By reducing the value of ${\mathrm{FSR}}_2$, the measurement accuracy of the system can be further enhanced. This flexibility renders the system suitable for various applications. In addition, the effectiveness of this method is not affected by the frequency sweeping speed. Thus, it is applicable to the calibration of ultrahigh-speed WSLs.

An experimental system (Fig. 15) was established to investigate the dynamic sweeping characteristics of the WSL. The light from the WSL was split into two channels using a $1\times 2$ optical splitter (OS). One channel was transmitted to the OSA, and the dynamic sweeping spectra of the WSL were measured (Fig. 16). Peak-max-hold spectra were recorded using an AQ6370 OSA instrument.

According to the measured spectrum, a wavelength tuning range of 1515–1575.4 nm, corresponding to a total continuous (gap-free) wavelength sweeping of 60.4 nm, was achieved. Another channel of light was transmitted to the three-port optical circulator (CIR) and output to the optical fiber connected to ${\mathrm{FBG}}_1$ through port-2 of the circulator. The ${\mathrm{FBG}}_1$ sensor had a central wavelength of 1559.547 nm, reflectivity of 92.51%, and 3 dB bandwidth of 0.25 nm. The FBG signal was then reflected through the CIR and transmitted to the PD. The PD converted this reflected optical signal into an electrical signal, which was subsequently captured using a RIGOL MSO5354 digital oscilloscope. The reflection spectra of ${\mathrm{FBG}}_1$ were accurately recorded during each sweeping cycle. The sweeping period is 12.09 μs, and the sweeping frequency is 82.7 kHz, achieving a high-speed wavelength sweeping rate of 4.96 nm/μs, as shown in Figs. 17(a) and 17(b).

To verify the continuous sweeping stability of the proposed WSL and to evaluate the wavelength repeatability between different cycles, a high-speed sensor interrogation system (Fig. 18) was designed. The laser frequency was accurately measured using the previously mentioned nonlinear frequency variation measurement system, enabling precise demodulation. The principle of the sensor interrogation system is illustrated in Fig. 18(a). The continuous wavelength sweeping light emitted by the WSL was divided into multiple channels using a $1\times N$ OS based on the principle of WDM. Each light channel passed through a CIR to the FBG array, generating multiple reflected light signals that matched the central wavelengths of the FBGs. The reflected light signals, which carry information on the measured physical quantities, were passed through the CIR and entered the PD. The PD converted the reflected light signals into current signals, which were then sent to a high-speed data acquisition board (HS-DAB). The HS-DAB is responsible for signal acquisition and data processing. The hardware employs a multi-stage transimpedance amplifier and a high-speed analog-to-digital converter (ADC) for acquisition, and a Xilinx Kintex-7 Family FPGA for data processing. In the design, a field effect transistor (FET) input operational amplifier with a gain–bandwidth product of 4 GHz, input bias current of 3 fA, and slew rate of 1500 V/μs was selected. The amplifier has high input impedance, low noise figure, and low input bias current, meeting the requirements for high-precision conversion of tiny currents. Through the multi-stage transimpedance amplifier design, the amplified voltage signal is sampled by a 14-bit, single-channel ADC chip with a sampling rate of up to 250 MSa/s. The FPGA on the high-speed acquisition board generates the sampling clock for the ADC chip, and the ADC, under the control of the FPGA, collects data at time points corresponding to equal frequency intervals.

Although the frequency sweeping of the proposed WSL is continuous, the data acquisition is selectively sampled at certain time intervals. To achieve fast and accurate demodulation, the acquisition system collects data points with equal frequency intervals in sequence, allowing the frequency of each point to be determined by simple counting. The frequency interval between data points can be set arbitrarily based on the requirements, which also determines the resolution of the sensor interrogation system. The frequency interval can be very small, allowing for finely sampled data points, but this creates challenges for the data acquisition and processing systems. Therefore, the sampling frequency interval should be set according to the specific requirements. In this FBG sensor interrogation system, a 5 GHz frequency interval is selected for data collection. Peak location searching, fitting, and interpolation calculations are performed on the transmission spectrum of F-P ${\text{etalon}}_2$ to obtain the sampling time points corresponding to the 5 GHz frequency interval. The wavelength sweeping range of the proposed WSL exceeds 60 nm, so approximately 1500 data points are collected per sweep cycle. The ADC has a data interface to FPGA via low-voltage differential signaling (LVDS) output. The 14-bit differential data output, along with the differential output clock signal, is transmitted to the FPGA for further processing. To achieve high-speed real-time data processing, peak detection, fitting, and frequency calculation algorithms are embedded in the FPGA. The collected data can be directly processed in the FPGA without the need to transfer large amounts of data to the computer for computation. This greatly reduces the challenge of data transmission bandwidth and enhances the ability to process data in real-time. In this system, the computer only plays a role in data display and storage.

At the same time, it should be noted that an F-P etalon (F-P ${\text{etalon}}_3$) with a designed defect transmission peak is used for absolute wavelength and frequency calibration, which has an FSR of 100 GHz. The F-P ${\text{etalon}}_3$ was connected to the proposed sensor interrogation system, and the corresponding transmission spectrum was recorded. The first transmission peak (${\mathrm{Peak}}_1$) on the left side of the defect (marking wavelength) was selected for absolute wavelength and frequency calibration. It should be noted that to exclude the influence of the surrounding environment on the center wavelength of ${\mathrm{Peak}}_1$, the etalon was placed in a temperature-controlled chamber set to 25°C to ensure it was not affected by the surrounding environment. Meanwhile, an ASE broadband light source, along with a high-precision multi-wavelength meter, was used to accurately measure the center frequency $v_0$ of ${\mathrm{Peak}}_1$. The fitting and peak search operations were performed on the sampling points near ${\mathrm{Peak}}_1$ using Gaussian fitting and the least squares algorithm. Based on the relative positional relationship between ${\mathrm{Peak}}_1$ and the sampling points A and B on the left and right sides, the absolute frequencies $v_A$ and $v_B$ of points A and B were calculated. Since the sampling points have equal frequency intervals (5 GHz), the absolute frequencies and absolute wavelengths of all 1500 sampling points can be calculated from $v_A$ and $v_B$.

An integrated design was implemented to enhance the stability and portability of the interrogation system. SolidWorks software was employed to design the instrument housing, which was made of an aluminum alloy with dimensions of $25\mathrm{cm}\times 18\mathrm{cm}\times 10\mathrm{cm}$. The WSL, OS, CIRs, PDs, and HS-DAB were integrated into the aluminum alloy housing [Fig. 18(b)]. The integrated FBG interrogation system was compact and lightweight, with a fully solid-state structure and no movable mechanical parts. This design enhanced the resistance to shock and vibration, thereby significantly improving the working stability. It is well-suited for challenging environments such as aerospace and rail transport. A graphical interactive control program for the sensor interrogation system was created using LabVIEW software. This program enables integrated control of the interrogation system and real-time data processing, visualization, and storage. As shown in Fig. 18(c), the FBG reflection signal was displayed in real time within the control interface. The reflection wavelength of each FBG sensor was continuously monitored, and the corresponding physical parameter data were presented using this program.

The real-time interrogation capabilities of an FBG sensor interrogation system based on the proposed WSL were also investigated. To investigate the long-term interrogation stability, two sensors were selected: a temperature sensor (${\mathrm{FBG}}_{\mathrm{A}}$) and a strain sensor (${\mathrm{FBG}}_{\mathrm{B}}$). The central wavelengths of the two sensors, at an ambient temperature of 30°C, were 1543.812 and 1531.426 nm, respectively.

In the high- and low-temperature sensing experiments, the ${\mathrm{FBG}}_{\mathrm{A}}$ was connected to the channel of the FBG sensing system. It was then placed inside an HOTC-100 high- and low-temperature test chamber. This chamber can perform temperature-cycling experiments according to a preset program. The experiment lasted for 300 min, during which the interrogation system interrogated the ${\mathrm{FBG}}_{\mathrm{A}}$ in real time. After maintaining a temperature of 30°C for 40 min, the chamber was gradually heated up to 90°C. After maintaining 90°C for 45 min, the chamber temperature was reduced to $-30°\mathrm{C}$. After maintaining $-30°\mathrm{C}$ for 45 min, the chamber temperature was increased to 30°C, where it was maintained for 40 min.

As shown in Fig. 19(a), the sensor interrogation system accurately recorded the central wavelength of the ${\mathrm{FBG}}_{\mathrm{A}}$ as it changed with the ambient temperature. Based on the test results of the sensor interrogation system, the central wavelengths of the ${\mathrm{FBG}}_{\mathrm{A}}$ at different temperatures are shown in Fig. 19(b). The relationship between the central wavelength of the ${\mathrm{FBG}}_{\mathrm{A}}$ and the temperature exhibited good linearity. After linear fitting, the slope of the wavelength change with temperature was $\sim 10.91\mathrm{pm}/°\mathrm{C}$, with $R^2=0.9983$.

In the strain-sensing experiments, the ${\mathrm{FBG}}_{\mathrm{B}}$ was fixed to a strain-testing platform. At the 70th minute of the test, strain values of 400, 800, $-800$, and $-400\mathrm{\mu \epsilon }$ were sequentially applied to ${\mathrm{FBG}}_{\mathrm{B}}$, with each strain state maintained for 40 min. Finally, the test ended after 70 min at $0\mathrm{\mu \epsilon }$. As shown in Fig. 19(c), the sensor interrogation system accurately recorded the changes in the central wavelength of ${\mathrm{FBG}}_{\mathrm{B}}$ with the applied strain. Based on the test results of the sensor interrogation system, the central wavelengths of the ${\mathrm{FBG}}_{\mathrm{B}}$ under different strain levels are shown in Fig. 19(c). The relationship between the central wavelength and strain exhibited good linearity. After linear fitting, the slope of the wavelength change with strain was $\sim 1.26\mathrm{pm}/\mathrm{\mu \epsilon }$, with $R^2=0.9985$.

Under the influence of temperature and strain, the central wavelengths of the FBG sensors changed in real time. The interrogation system precisely interrogated and recorded the changes. According to the demodulation results, the interrogation linearity of these FBGs was $>0.9983$, which demonstrated good relative accuracy and excellent interrogation performance of the system. The data from the initial and final stages of Figs. 19(a) and 19(c) were magnified for observation; the maximum deviation in the continuously interrogated central wavelengths of ${\mathrm{FBG}}_{\mathrm{A}}$ and ${\mathrm{FBG}}_{\mathrm{B}}$ was $<2\mathrm{pm}$ (within a fluctuation range of $\pm 1\mathrm{pm}$). The interrogation results at the beginning and end of the tests were consistent. The results indicate that the FBG sensor interrogation system can be used for real-time sensing and monitoring of multiple physical parameters with high interrogation accuracy and long-term stability. This demonstrates high wavelength sweeping stability and repeatability of the proposed WSL and high accuracy of the proposed nonlinear frequency measurement system.

To verify the interrogation speed of 82.7 kHz, a high-speed vibration experiment was performed. The HEV-50 vibration exciter was selected, and a corresponding cantilever structure was designed. A bare fiber Bragg grating (${\mathrm{FBG}}_{\mathrm{C}}$) with a grating length of 3 mm and a reflectivity of 87% was chosen. Both fiber pigtails of the ${\mathrm{FBG}}_{\mathrm{C}}$ were fixed to the cantilever beam and the longitudinal displacement stage. A signal generator and a power amplifier applied sinusoidal signals of 2, 4, and 8 kHz to the vibration exciter. The vibration exciter, driven by the sinusoidal signals, generated corresponding frequency sinusoidal vibrations, which were transmitted to the ${\mathrm{FBG}}_{\mathrm{C}}$. The changes of the FBG reflection wavelength owing to vibration are measured for each sweep. In the continuous interrogation of the ${\mathrm{FBG}}_{\mathrm{C}}$, a period of 0.75 ms interrogation result is randomly selected to plot the waveform, as shown in Fig. 20. Within 0.75 ms, a total of 62 sampling points were collected, resulting in a time resolution of 12.09 μs. This indicates that the system’s interrogation speed has reached 82.7 kHz. Using the Fourier method, the interrogation wavelengths were also analyzed, providing a complete frequency analysis within the 40 kHz range, as shown in Fig. 20. We selected 16,384 data points for fast Fourier transform (FFT) calculation, and the resulting frequency resolution in the FFT domain is approximately 5 Hz. We selected the horizontal coordinates of the extreme points in the FFT plot as the peak positions. Using this method, the peak positions measured in Figs. 20(a), 20(b), and 20(c) were 1.99869, 3.99837, and 8.00112 kHz, respectively. The experimental results demonstrate that the FBG sensor interrogation system based on the proposed WSL successfully demodulated high-frequency vibration signals of up to 8 kHz [59,60].

The proposed WSL, along with its FBG sensor interrogation system, features a fully solid-state package structure without movable mechanical parts or fragile optical components. Therefore, the system exhibits remarkable vibration and shock resistance, making it suitable for stable operation in high-vibration environments. To verify this hypothesis, vibration and shock tests were conducted using a sensor interrogation system.

The interrogation system was subjected to continuous vibration and shock tests along the $X$-, $Y$-, and $Z$-axes in both forward and reverse directions to evaluate its vibration and shock resistance performance. The system was mounted on the horizontal table of a DC-20000-200/ST-1515 electrodynamic vibration test system, which applied vibrations and shocks to the interrogation system in the $X$- and $Y$-directions. The test lasted for 20 min, during which the vibration table induced sustained sine wave vibrations at a frequency of 70 Hz with an amplitude of 0.31 mm. Three consecutive forward half-sine pulse shocks were applied at the 4th and 8th minutes of the test. At the 12th and 16th minutes, three consecutive reverse half-sine pulse shocks were applied. Each pulse shock had a peak acceleration of 30g, with a duration of 10 ms per pulse. An FBG sensor with a central wavelength of 1538.416 nm was connected to the sensor-interrogation system at room temperature. During the experiment, the interrogation system monitored the FBG continuously to determine whether it was affected by vibrations and shocks. After the vibration and shock tests in the $X$- and $Y$-directions, the sensor interrogation system was mounted on a vertical table of the DL-6000-60 electrodynamic vibration test system for the vibration and shock tests in the $Z$-direction. The $Z$-axis vibration and shock tests followed the same procedure as those for the $X$- and $Y$-axes. The test results are shown in Fig. 21.

As shown in Fig. 21, the interrogation results remained consistently stable at $\sim 1538.416\mathrm{nm}$, with a maximum fluctuation of no more than 2 pm. This indicates that the interrogation system maintained a stable and accurate performance throughout the vibration and shock tests in all six directions. Therefore, the FBG sensor interrogation system based on the proposed WSL demonstrates the potential for applications in environmentally challenging fields such as the aerospace, rail transportation engineering, and petrochemical industries.

The maximum sweeping speed achieved in this work, 82.7 kHz, is limited by the bandwidth of the DAC, operational amplifiers, and switching chip selected in this study. By selecting electronic components with larger bandwidths and optimizing the driver circuit design, the sweeping speed can be further improved. We will further enhance the sweeping speed of the WSL in our future research work. In addition to improving the sweeping speed, we can further expand the wavelength sweep range of the DFB laser array by designing more integrated DFB lasers and increasing the single-channel wavelength sweeping range. However, the maximum wavelength tuning range of the proposed DFB laser array scheme is mainly constrained by the electroluminescence (EL) gain spectrum of the epitaxial material structure. Generally, the wavelengths at the edges of the gain spectrum are difficult to lase. Based on the measured wavelength width of the EL spectrum, the maximum designed wavelength range of the laser array proposed in this work is approximately $70-80\mathrm{nm}$. To overcome this limitation, it is necessary to enhance the gain bandwidth through the design of quantum wells to cover a broader wavelength range. In our future work, we will further design and fabricate a WSL with a larger wavelength tuning range for more detailed and specific validations.

In this study, we developed an ultra-wideband and high-speed wavelength-swept DFB laser array with an $8\times 3$ interleaving matrix design and experimentally verified its performance. To achieve the high-precision mapping from the time domain to the frequency domain, a nonlinear frequency variation measurement system based on dual F-P etalons is designed. This system demonstrated the capability of accurately measuring dynamic frequency changes over time, ensuring the precise interrogation of the measured parameters. In the all-solid-state design of a WSL, we avoided movable mechanical parts or fragile optical components to offer high robustness and reliability to the device. The maximum wavelength deviations across all 24 lasers remain within $\pm 0.2\mathrm{nm}$, with $\mathrm{SMSR}>45\mathrm{dB}$ and $\mathrm{RIN}<-130\mathrm{dB}/\mathrm{Hz}$.

The WSL demonstrated remarkable performance during continuous dynamic operation, achieving an ultra-wideband and high-speed continuous (gap-free) wavelength sweeping coverage of 60 nm at a sweep rate of up to 82.7 kHz, reaching a high wavelength tuning rate of 4.96 nm/μs. The output power remained stable at 10 mW owing to the real-time dynamic adjustment of the SOA injection current. The proposed WSL was applied to an FBG sensor interrogation system; the system performed real-time dynamic interrogation of FBGs for up to 3 h. The demodulation results demonstrated the good relative accuracy and excellent interrogation performance of the system. The maximum wavelength deviation during continuous interrogation was $<2\mathrm{pm}$, indicating exceptional repeatability of wavelength sweeping. The interrogation system achieved precise measurements of high-speed vibrations in the frequency range of 2–8 kHz, with a time resolution of 12.09 μs. Furthermore, stable operation was maintained under challenging conditions of vibration and shock. Thus, the proposed WSL can be applied in diverse optical sensing systems, such as FBG sensing and optical coherence tomography.