In the era of 5G and beyond, massive wireless data traffic is transmitted between user equipment (UE) and network operators, predominantly over optical fiber through the radio access network [1–3]. To meet the demand for millimeter wave and THz band wireless communication, photonics-based radio generation provides a promising solution for establishing high-quality and long-distance base stations (BS) [4–6]. The fronthaul radio over fiber (RoF) technique, which is divided into analog RoF (A-RoF) and digital RoF (D-RoF), focuses on transmitting radio signals from a centralized unit (CU) or distributed units (DUs) to an active antenna unit (AAU) [7–9]. Subsequently, at the AAU side, the optical signals are converted into wireless radio signals for broadcasting, as shown in Fig. 1. A-RoF, which transmits analog radio waveforms without quantization, provides high spectrum efficiency and low system complexity [10–12]. By combining heterodyne reception with A-RoF, MMW and THz signals can be generated at remote antennas without the need for additional up-conversion components or optical signal demodulation.

As network operators ensure high capacity and fidelity in front-haul communication, they also explore the potential benefits of ubiquitous optical fibers, such as optical sensing [13–15]. Indeed, optical sensing can “illuminate” the fiber, making various events visible [16,17], particularly for significant disturbances to the fiber. Some previous research [18,19] successfully implemented vibration sensing with the coexistence of RAN. However, the integration of sensing and communication in these works is not sufficiently seamless, and their cost is relatively high. Therefore, there is an urgent need to monitor the mobile front-haul network in a cost-effective manner.

The phase sensitive optical time domain reflectometry (ϕ-OTDR), commonly known as distributive acoustic sensing (DAS) [20,21], is therefore highly suitable for capturing the fast-changing perturbations along the RAN. Featured without the swept laser [22] and bi-directional setup [23], DAS can demodulate the Rayleigh backscattered (RBS) signal from a standard single-mode fiber (SSMF) to obtain dynamic environmental information, such as events induced vibrations or intrusion detection. However, for classic DAS systems via phase demodulation, large-scale vibration can cause the phase to jump beyond [−π, +π], leading to the issue of phase unwrapping and restricting the dynamic range measurement in practice [24]. To realize large dynamic vibration sensing, Froggatt and Moore [25] utilized the shift of the RBS intensity spectrum, also known as frequency demodulation, to recover the strain, while it requires multiple or swept wavelength lasers to obtain the backscattered spectrum pattern of each scattering point along the fiber. The laser requirement for the frequency demodulation method is then eliminated by the chirped pulse OTDR, whose frequency linearly changes inside the pulse [26]. It should be noted that the dynamic range of frequency demodulation DAS corresponds to the overall spectrum width, which is always very limited due to the narrowband signal and inexpensive equipment of the sensing setup. However, integrating the DAS into the telecommunication system can potentially resolve this dynamic range issue, owing to the shared GHz level telecommunication setup.

Recently, integrated sensing and communication over fiber (ISACoF) technique has emerged as a promising solution for practical implementation of optical sensing [27]. While the distributed fibers form a large communication network, ISACoF has the potential to transform deployed fibers into massive sensors. Optical sensing and optical communication, which exhibit distinct characteristics but share the same optical cable, can be multiplexed using various methods, such as frequency band [28], space and mode [29,30], transceiver [31], or even the demodulation process [32]. There is also a new hybrid ISACoF waveform that modulates the communication signal using a linear frequency modulated (LFM) carrier [33], but it is only applicable to intensity modulated direct detection (IM-DD) systems. If classified by the sensing method, these ISACoF systems can be divided into forward sensing systems [34,35] and backscattered sensing systems [36,37]. Generally, these works set up a simultaneous system for sensing and communication, yet they still fail to achieve seamless integration with the communication signal. In other words, most of the existing schemes only allocate a portion of the channel from the communication system for sensing in either frequency or space. Therefore, exploring the perfect coexistence form between optical sensing and telecommunication systems remains an unsolved challenge.

Here, to the best of our knowledge, we present for the first time a large dynamic vibration sensing method integrated into the RoF communication system by utilizing LFM pilots, without any modifications to communication transmitters. Furthermore, we demonstrate a joint optical-wireless communication experiment, transmitting the designed orthogonal frequency division multiplexing (OFDM) frame to the MMW phased array over a long-distance fiber while simultaneously sensing perturbations around the fiber. The LFM pilots, regularly inserted every few subcarriers of the radio OFDM signal, serve dual purposes. On the one hand, they are used for channel estimation and carrier recovery in communication digital signal processing (DSP). On the other hand, by means of pulse compression, the LFM pilots function as sensing probe pulses, achieving seamless integration of the DAS and RoF systems. Moreover, the dynamic range of the vibration measurement is significantly improved, thanks to the use of a large bandwidth communication setup to generate a sensing probe signal.

The designed OFDM symbol frame structure for analog RoF transmission with LFM pilots is shown in Fig. 2. Each OFDM frame consists of five functional parts. The frame head is a temporal domain sequence with excellent auto-correlation properties and a steep roll-off area for fine frequency offset estimation (FOE) and frame synchronization. Next, a training sequence with a single OFDM symbol is included for channel estimation initialization. LFM pilots are then periodically inserted into the frequency grid at fixed subcarrier intervals for coarse FOE and channel estimation. These pilots are continuous across frames to enable post-processing and sensing functions. The direct current (DC) carrier at zero frequency is left unused to avoid low-frequency noise [38], such as the relative intensity noise (RIN) of the lasers. Finally, the data payload fills the remaining frequency resource grid, completing the OFDM frame.

The communication DSP flow of the transmitter side is shown in Fig. 3(a). The transmitted bits are first mapped into quadrature amplitude modulation (QAM) for each subcarrier, with only one training symbol added in front of the payload to perform the initialization of the channel estimation. The modulation format of the training symbol is binary phase shift keying (BPSK). It should be noted that we periodically vacate one or more subcarriers at a fixed interval of channels to spare the space for LFM pilots. After the inverse fast Fourier transform (IFFT), OFDM symbols in the time domain are added cycle prefixes (CPs) to eliminate the inter-symbol interference (ISI) and maintain the orthogonality between subcarriers.

Then, we insert LFM pilots into these grids by directly adding them to the serial data stream, filling the notches of the spectrum. The LFM pilots can be expressed as $$\begin{gathered} P(t)=∑k=1NpA(t)exp(j·(2π(kfint−12γT0)t+πγt2)) \\ +∑k=−Np−1A(t)exp(j·(2π(kfint+12γT0)t−πγt2)), \end{gathered}$$(1)where P(t), 2Np, A(t), fint, γ, and T0 are inserted LFM pilots, number of pilots, window function, interval of frequency, chirp rate, and duration time, respectively. Channel estimation and coarse FOE can be implemented using LFM pilots, while the time duration of each LFM pilot is across several frames and the frequency is separated from the data subcarriers. After loading the LFM pilots, a sequence of robust frame synchronization and fine FOE is inserted at the beginning of each frame [39]. The synchronization head exhibits excellent autocorrelation properties, characterized by a high peak-to-sidelobe ratio (PSLR), and can also be utilized for fine FOE in the subsequent receiver DSP. Thanks to the peak-to-average power ratio (PAPR) optimization of superimposed LFM pilots using a genetic algorithm (GA) (see Appendix A), the final OFDM frame before the clipping process achieves a relatively low PAPR.

After propagation through the optical fiber, the OFDM signal undergoes pulse evolution governed by the nonlinear Schrödinger equation. Given that fronthaul transmission distances are typically below 50 km and the bandwidth of the wireless OFDM signal is at the GHz level, the effects of chromatic dispersion and fiber nonlinearity can be neglected. Thus, in a typical short-reach optical front-haul system, the main challenges for receiver-side DSP in OFDM demodulation are carrier recovery, including frequency offset estimation (FOE) and phase recovery, as well as channel estimation of the frequency response.

Details of the receiver-side DSP flow are shown in Fig. 3(b). Moreover, these DSP procedures can also be effective after the wireless channel, enabling joint processing of the OFDM signal throughout the entire optical and wireless channel in an analog RoF system. Initially, we need to coarsely estimate the carrier frequency of the OFDM symbol using the LFM pilots before resampling the raw data to an appropriate sample rate for further DSP. The estimated frequency offset can be expressed as fcoarse=12Np(∑k=1Npfk+∑k=−Np−1fk),(2)where fk are the frequencies of the symmetric LFM pilots. Then, we perform OFDM frame synchronization and fine FOE using the frame head, with the synchronization principles and fine FOE metrics detailed in Appendix B. The synchronization sequence used has a very high accuracy of correct detection and is robust to frequency offset and deep fade. After these processes, we obtain a complete OFDM symbol in the baseband without frequency offset.

Figure 4 shows the de-chirp process of the LFM pilots to obtain the frequency response, which assists in the subsequent equalization and removal of the LFM pilots. First, we multiply the signal by the LFM signal with an inverse chirp rate exp(−jπγt2). Then, a multi-band digital bandpass filter is applied to filter out the de-chirped LFM pilots in the positive sideband, which can then be used for frequency response estimation and subtracted from the signal. Similarly, this operation is repeated for the negative sideband. The OFDM signal, now devoid of LFM pilots, can then be processed in the same way as in the standard demodulation stages. After serial-to-parallel (S/P) conversion, CP removal, and FFT, the OFDM symbols are equalized using the obtained frequency response before QAM demapping. It should be noted that the initial phase of each subcarrier is provided by a one-symbol-length training sequence, and a simple phase rotation within ±5deg is required to mitigate the residual phase offset. In addition, the power ratio between the communication payload and the LFM pilots should be carefully chosen to balance both communication and sensing performance, as discussed in the experimental section.

In traditional DAS systems, a pulse is launched into the fiber, and RBS light is received and demodulated at the transmitter side as the pulse propagates along the fiber. Since the sensing resolution is determined by the pulse width, achieving higher sensing resolution requires a narrower pulse width. However, narrow pulses with low power result in shorter sensing lengths, leading to a trade-off between sensing length and resolution in traditional DAS systems. Furthermore, it is challenging to seamlessly integrate traditional pulse-based DAS with deployed communication systems.

Fortunately, pulse compression DAS mitigates the trade-off between sensing length and resolution, as it does not require the probe signal to be a pulse, unlike traditional DAS systems. The scheme of pulse compression DAS is shown in Fig. 5. We first consider a continuous repeated LFM signal as an example. After injecting the repeated LFM signal into the fiber, the received RBS signal motivated by the continuous probe can be expressed as r(t)=(exp(j(2πkfstt+πγt2))rect(t/T))repeated∗hFUT(t),(3)where hFUT(t) denotes the ideal impulse response of a fiber backscattered system, ∗ represents the convolution operation, and fst and T are the start frequency and the repetition period of the continuous LFM probe, respectively. Therefore, by continuously sliding a match filter to correlate with the received RBS signal, we can obtain $$\begin{gathered} s(t)=r(t)⊗exp(j(2πkfstt+πγt2))=p(t)∗hFUT(t) \\ =(sin(πγ(T−|t|)t)πγtexp(j2π(fst+γT2)t))∗hFUT(t), \end{gathered}$$(4)where ⊗ denotes the correlation operation, and p(t) is the equivalent probe pulse, which is also known as the compressed pulse. Therefore, the true ϕ-OTDR traces are acquired using the aforementioned sliding match filter, with the sensing resolution determined by the 3 dB bandwidth of the auto-correlation peak. Moreover, due to the continuous probe signal instead of a pulse, the power of the equivalent pulse can be significantly higher, bypassing the limitations of fiber nonlinearities such as modulation instability (MI) and stimulated Brillouin scattering (SBS), thereby enabling long-haul sensing in pulse compression DAS.

If there is strain on the fiber somewhere, a phase difference δϕ can be detected by interrogating the fiber, which can be analytically expressed as [40] $$\begin{gathered} δϕij=4πfcc(n·Δzij+Δn·zij) \\ =4πfcc(n+Cε)zijΔε, \end{gathered}$$(5)where fc is the carrier frequency, n is the effective refractive index, zij is the distance between ith and jth scattering elements, Cε is strain refractive index parameter, and Δε denotes the magnitude of the strain. In Eq. (5), the time-varying term δϕ can be compensated by a frequency shift fshift shown as δϕij=4πnzijcfshift.(6)

Hence, we establish a unique mapping from the strain intensity to the shift of the RBS spectrum if we can obtain OTDR traces of different wavelengths. In contrast, the output phase shift of the DAS system based on the phase demodulation always falls within the range of −π to +π. Table 1 compares a series of key parameters between the phase demodulation method and the frequency demodulation method. In the case of a probe wavelength around 1550 nm, 1με corresponds to a 150-MHz frequency shift of the RBS spectrum [25].Table 1.

Comparison between Frequency Demodulation and Phase Demodulation in DAS Using the LFM Pilots

According to Eq. (1), we can make use of multiple LFM pilots to reconstruct an RBS spectrum. The frequency demodulation enabled by multiple LFM pilots is shown in Fig. 6. The integrated scheme utilizes multiple LFM pilots in OFDM communication and effectively leverages the large bandwidth characteristics of communication signals to achieve a high dynamic range sensing.

Figure 7 shows the DSP flow after receiving the backscattered sensing signal stimulated by the LFM pilots of forward transmission. A simple synchronization is carried out between segments to mitigate the jitter of the sampling trigger, followed by the frequency demodulation or phase demodulation. The least mean square (LMS) algorithm is adopted to obtain the frequency shift of certain positions between segments, eventually demodulating the vibration [41]. The mean square error (MSE) of the LMS can be expressed as MSE(t,Δf)=1F−|Δf|∑f=−F+|Δf|F−|Δf|(href(t,f+Δf)−hmea(t,f))2(7)where F is the frequency spectrum range, equal to the whole OFDM communication bandwidth. MSE is calculated with each frequency shift Δf between the reference RBS spectrum href and the measurement RBS spectrum hmea, while our goal is to minimum the MSE. Hence, the frequency shift could be derived as fshift=argminΔf(MSEref,mea).(8)

Furthermore, to enhance the spectrum resolution, interpolation can be performed [42] between the LFM pilots, whose frequency is located at the interval of certain subcarriers number. By using the frequency demodulation method, an ultra-large dynamic range for vibration sensing can be achieved with the help of LFM pilots through the proposed ISACoF scheme.

We carry on the experiment to verify the proposed integrated large dynamic sensing and communication scheme over an RoF system, as shown in Fig. 8. The linewidth of the fiber laser (NKT X15) used in the ISAC scheme is around 100 Hz to ensure sufficient coherent length. An intensity modulator (iXblue MX-LN-40) is driven by a microwave signal generator (Keysight E8257D) to generate two sideband carriers, while the frequency of the electrical microwave signal determines the radio band of the OFDM signal. Then, the light is split into two tributaries by a polarization maintained optical coupler (PM-OC) to keep the same polarization for both sidebands. In the upper arm, an arbitrary waveform generator (AWG, Keysight M8190A) with a sampling rate of 12 GSa/s is used to generate the OFDM baseband signal. We first generate the OFDM signal occupying about 1.8-GHz bandwidth and composed of 896 subcarriers, every 7 of which insert an LFM pilot tone to perform sensing and channel estimation. It should be noted that each LFM pilot occupies 2 subcarrier grids to avoid crosstalk between pilots and data channels. Subsequently, the analog OFDM signal is directly modulated by an IQ modulator (Fujitsu 7962EP) without a digital quantization procedure. Meanwhile, the lower sideband carrier is amplified by a polarization maintained EDFA to adjust the carrier-to-signal power ratio. In the following, the upper arm and the lower arm are combined by a PM-OC and then transmitted to a 10-km fiber under test (FUT) spool after an EDFA. There is also a 60-m piezoelectrical transducer (PZT) at the end of the fiber, and a variable optical attenuator (VOA) is used before the photodiode (PD) to keep the received optical power the same. In the AAU side, only one PD is needed to receive the optical radio signal. Therefore, the heterodyne A-RoF scheme can take advantage of the no up-converting process remotely. The radio signal directly generated by the PD can then be sent to the phased array antenna, as detailed in Appendix C. The two MMW phased arrays with a 3-dB bandwidth between 27.2 GHz and 29 GHz can operate in two modes, one is the intermediate frequency (IF) mode and the other is the RF mode, corresponding to two types of wireless transceivers. Due to the low quantum efficiency of the PIN diode, we increase the adjustable transmitter gain up to 55 dB before free space transmission, including gains from a low noise amplifier (LNA), a power amplifier (PA), and an antenna. After 4-m transmission, the radio OFDM signal is detected by the other MMW phased array and then mixed into an IF signal. The IF waveform is directly captured and sampled by an 80-GSa/s oscilloscope (Keysight DSAZ594A), followed by the post communication DSP flow. Meanwhile, we use an integrated coherent receiver (ICR) and a 4-GSa/s oscilloscope to demodulate the sensing results and match the bandwidth of RBS signal.

We first analyze the communication performance of the LFM pilot aided RoF system. The baseband OFDM spectrum is shown in Fig. 9(a), with an inset providing a zoomed-in view of the spectrum. The LFM signal is periodically distributed at an interval of 7 subcarriers, 5 of which are allocated to OFDM payloads and 2 to pilots. This interval was carefully selected to achieve desirable trade-offs among communication performance, sensing sensitivity and range, and overall spectral efficiency under the experimental conditions. To address the PAPR increase caused by multiple LFM pilots, the genetic algorithm (GA), as detailed in Appendix A, is used to adaptively adjust the initial phase of the pilots. Figure 9(b) compares the PAPR values under different initial phase allocation methods for varying numbers of subcarriers. We define the power ratio between the OFDM payloads and the LFM pilots as the signal-to-pilot ratio (STPR). For a fair comparison, simulations are conducted under STPR = 6 dB when pilots are inserted. Without pilots, the PAPR of the OFDM payload is approximately 10–11 dB. When LFM pilots are inserted and their initial phases are optimized using GA, the PAPR increases by about 0.5 dB. In contrast, conventional quadratic phase distribution and random phase distribution increase the PAPR by more than 1 dB. It is worth noting that GA has a more significant effect on PAPR reduction under lower STPRs. For further investigation, the number of subcarriers, subcarrier spacing, STPR, and sweeping bandwidth of the LFM pilot are fixed at 896, 2 MHz, 6 dB, and 2 MHz, respectively.

Since the radio signal is generated by the beat between the single optical tone and the IQ-modulated OFDM signal, the carrier-to-signal power ratio (CSPR) is defined as the power ratio between them. Because the single optical tone is much stronger than the IQ-modulated OFDM signal, its power can be treated as the launch power (LP) of the fiber system. We transmit 64-QAM modulated OFDM symbols and scan the CSPR to evaluate the relationship between CSPR and bit error rate (BER) performance at LP = 7 dBm. Initially, the phased array is switched to the IF mode, centering the PD-received OFDM signal at 2.9 GHz. After evaluating the transmission performance in the IF mode, we switch the phased array to the RF mode, without electrical up-conversion inside it, simulating a practical situation where no local oscillator (LO) is present at the AAU side. Figure 10(a) shows that when CSPR exceeds 11 dB, BER significantly increases as CSPR increases. However, when CSPR is below 10 dB, this trend becomes less apparent, indicating that the OFDM signal, which beats with the high-power single tone, enters the gain saturation region at CSPR<10dB. This also suggests that a CSPR of 10 dB is sufficient to avoid being the dominant noise source in the system. Additionally, the BER in the RF mode outperforms that in IF mode, likely due to circuit losses during up-conversion, pink noise, or distortions caused by the mixer, such as LO leakage and conversion noise [43]. Therefore, RF mode is used for subsequent experiments, corresponding to the photonics-generated MMW heterodyne RoF demonstration.

Next, we fix CSPR = 10 dB and STPR = 6 dB and vary the LP and modulation formats, as shown in Fig. 10(b). We observe that high launch powers are feasible in this radio access network since fiber nonlinearity is negligible within a 10-km distance. It is only when LP exceeds 10-dBm that the demodulated SNR remains unchanged or slightly decreases, likely due to the stimulated Brillouin scattering (SBS) effect of the single optical tone. Additionally, both probabilistic constellation-shaped 16-QAM and 64-QAM (PCS-16-QAM and PCS-64-QAM) are evaluated with entropies of 3 bits/symbol and 5 bits/symbol, respectively.

We further examine the relationship between STPR and transmission performance. With CSPR = 10 dB and LP = 10 dBm, we vary the STPR from 3 dB to 12 dB and the pilot interval from 5 to 21 subcarriers to investigate BER variation for the 16-QAM OFDM radio system. Since the LFM pilots also serve as sensing probes, a lower STPR enhances sensing performance. However, because the OFDM payloads and LFM pilots share the total available optical power and quantization levels, reducing STPR consequently degrades communication performance. Figure 11(a) confirms that decreasing STPR leads to a deterioration in BER. Furthermore, the sweeping bandwidth of the LFM pilots, set here at 2 MHz, occupies 50% of the bandwidth allocated for the two subcarriers. We then adjust the sweeping bandwidth, effectively varying the percentage of pilot occupancy within the 2-subcarrier bandwidth. Figure 11(b) shows that communication performance degrades with an increase in the sweeping bandwidth of the LFM pilot. This degradation occurs due to the overlap between adjacent subcarriers and pilots in the frequency domain, degrading their SNR when the LFM pilots are removed. However, there also exists a trade-off between the sweeping bandwidth of the LFM pilots and communication performance. Specifically, the sensing resolution in pulse-compression DAS systems is directly proportional to the sweeping bandwidth of the LFM pilots. Potential methods for increasing the sweeping bandwidth include expanding the overall communication bandwidth, reducing the total number of subcarriers, or allocating additional subcarriers exclusively for pilot signals. Considering practical simplicity and balancing the trade-off between sensing and communication performance, we selected a moderate subcarrier occupancy rate of 50% (corresponding to a bandwidth of 2 MHz) and an STPR of 6 dB for the subsequent sensing experiments.

After coherently receiving the RBS signal stimulated by the radio OFDM signal, matched filters are used to obtain DAS traces at different frequencies. Thanks to frequency isolation, each DAS trace achieves a high SNR above the noise floor, sufficient for short-reach access networks below 20 km. Figure 12(a) shows the unique RBS spectrum of each point along the 10-km fiber, where the color represents intensity. During the experiment, 128 DAS traces are obtained from 128 LFM pilots with an STPR of 6 dB, and the duration of each pilot is 200 μs. When strain or vibration is applied to a specific section of the fiber, the RBS spectrum shifts in proportion to the intensity of the vibration. Moreover, cubic polynomial interpolation with a factor of 100 is used to fit the RBS spectrum of each point, improving the accuracy of measurement. A 60-m PZT is placed at the end of the fiber and is first driven by an 80-Hz, 1-V peak-to-peak (V_pp) sine waveform. To achieve a 1-kHz interrogation rate for the DAS system, the oscilloscope is triggered with pulses of a 1-kHz repetition period. From the waterfall plot shown in Fig. 12(b), a periodic frequency shift is observed at approximately 9.84 km, indicating vibration at the end of the fiber.

The PZT driving voltage is then varied to verify the linear response between the applied strain and the frequency shift, as shown in Fig. 13(a). Notably, the 1.8-GHz OFDM bandwidth enables the measurement of frequency shifts up to 900 MHz, corresponding to a 6-με vibration. Additionally, the 2-MHz sweeping bandwidth yields a sensing resolution of 50 m, as illustrated in Fig. 13(b). In this trace for a specific segment, a frequency shift is clearly observed, indicating strain at that moment. It should be noted that the communication RoF signal bandwidth is typically fixed, implying that the maximum achievable dynamic range is also fixed. As a result, we can optimize sensing sensitivity to achieve the best possible performance at a defined dynamic range. Moreover, owing to the large bandwidth available from ROF signals (compared to conventional dedicated DAS probes), the ISACoF system inherently offers an improved dynamic range compared with most previously reported DAS systems [40], thus effectively accommodating varying levels of vibration intensity in practical applications.

The frequency response of the vibration measurement is tested by applying a chirp-driven signal ranging from 20 Hz to 100 Hz to the PZT. Figure 14(a) demonstrates a flat frequency response, and the chirp vibration is successfully recovered, validating the system’s ability to capture acoustic signals over a wide frequency range. To compare phase demodulation and frequency demodulation, we also perform phase demodulation by summing the DAS traces using the rotated-vector-sum (RVS) method, as described in Fig. 7. After normalizing the two recovered vibration waveforms at the same point but using different methods, Fig. 14(b) shows that vibration is successfully demodulated through the frequency method, while the phase demodulation waveform wraps due to large vibrations.

To analyze sensing sensitivity at the end of the 10-km fiber, a long-term RBS signal lasting 2 seconds is collected under a 50-Hz sinewave vibration, as shown in Fig. 15(a). After normalization with a window function, the single-sideband (SSB) power spectral density (PSD) of the recovered waveform is derived. The noise floor is approximately −45dB across frequencies, corresponding to a strain sensitivity of 0.81nε/Hz. Finally, to investigate how STPR influences sensing sensitivity, Fig. 15(b) scans the STPR from 3 dB to 10 dB to observe the measured strain sensitivity under different power ratios between communication payloads and pilots, together with 4 different pilot intervals. Referring to Figs. 11(a) and 15(b), an STPR of 6 dB represents an effective trade-off for the experimental performance, as at this point the BER is well below the FEC threshold, and sensing sensitivity has not yet decreased significantly.

By designing a combined waveform simultaneously utilized for both communication and sensing, the proposed ISACoF scheme, employing multiple LFM pilots, achieves more seamless integration compared to the LFM-carrier-based ISACoF system previously demonstrated for 5G RoF fronthaul scenarios [44]. Furthermore, unlike wavelength-division multiplexing (WDM)-based ISACoF systems [45–47], this approach of directly reusing communication pilots as DAS probe signals significantly enhances spectral efficiency, thereby realizing a highly cost-effective ISACoF solution.

In this paper, we have designed and demonstrated a novel LFM pilot aided OFDM signal that is compatible with existing communication setups in practice. The continuous LFM pilots not only assist in the demodulation process at the communication receiver-side DSP but also serve as sensing probes to detect strain or vibration along the fiber using pulse compression techniques. Furthermore, leveraging the large bandwidth of the communication OFDM symbol, an ultra-large dynamic range for frequency-demodulated vibration sensing can be achieved, thereby overcoming the limitations of traditional phase-demodulation-based DAS in large-scale vibration scenarios.

We have experimentally validated the proposed ISACoF method by transmitting 128-LFM-pilot-aided OFDM symbols through a 10-km fiber and 4-m free-space channel. With a 1.8-GHz MMW radio signal operating within a frequency range of 27.2 GHz to 29 GHz, dynamic vibration measurements of up to 6με have been achieved. Through optimization of the power ratio between OFDM payloads and LFM pilots, a sensing sensitivity of 0.81nε/Hz has been attained, along with a demodulated SNR exceeding 20 dB for 64-QAM-OFDM at an STPR of 6 dB. Additionally, the system parameters can be comprehensively designed and optimized in accordance with the specific requirements of practical applications, with trade-offs between sensing and communication dynamically and adaptively adjusted over time in response to evolving operational scenarios and practical needs.