In recent years, microwave photonic technology has caught extensive attention in electronic information equipment due to its low transmission loss, ultra-wideband, and high amplitude-phase consistency. Since Mach-Zehnder (MZ) modulators have inherent nonlinear characteristics, the electro-optic modulation of radio frequency (RF) signals will induce nonlinear distortions. For applications such as optical beamforming, optical analog-to-digital conversion, and optical frequency conversion which integrate electro-optic modulation into RF analog transceivers, the nonlinear distortion caused by electro-optic modulation will reduce the spurious-free dynamic range (SFDR) of RF transceivers. This in turn limits the application and advancement of microwave photonic technology. Thus, we address the nonlinear distortion in microwave photonic links with blind identification and digital compensation methods. By adopting these methods, the SFDR of microwave photonic links can be improved without modifying the original microwave photonic systems.

By converting the output of microwave photonic link into digital signals by the high-speed analog-to-digital converter, digital signal processing techniques that employ blind identification compensation can be utilized to suppress the link’s nonlinear distortion (Fig. 2). The digital signal is first converted to an intermediate frequency (IF) signal via digital down-conversion, and then a memory polynomial model is leveraged to fit the nonlinear intermodulation distortions within the signal. This fitted distortion is subtracted from the original IF signal to achieve nonlinear distortion suppression. We propose employing a spectrum reduction algorithm based on time-frequency transformation to blindly identify the high power signal and distortions within the IF signal. The processing enables the parameter extraction for the memory polynomial model. The IF signal is first transformed into the frequency domain using the fast Fourier transform (FFT). By setting a power threshold in the frequency domain, the separation of high-power signals from low-power signals is achieved. Subsequently, by applying the inverse FFT, the separated high-power and low-power signals are converted back to the time domain, thus yielding a high-power signal that approximates an undistorted ideal signal, and a low-power signal containing nonlinear distortions. By adopting the high-power signal as an input, a first memory polynomial model is employed to fit the components of nonlinear distortion in the low-power signal, and the parameters of the nonlinear model are extracted using the least squares method. By fitting the nonlinear distortion with the first memory polynomial model and adding it to the high-power signal as input for a second memory polynomial model, the nonlinear distortion components in the low-power signal are fitted again to yield the final parameters of the nonlinear model. Additionally, we obtain the nonlinear model parameters by this two-stage fitting process, which can enhance the digital nonlinear compensation effectiveness in microwave photonic links with strong nonlinear distortion.

We employ measured data from the microwave photonic link transmission of a two-tone signal centered at 13.8 GHz to perform offline processing, validating the proposed digital nonlinear compensation method (Fig. 4). By adopting single nonlinear fitting for digital compensation, the third-order intermodulation (IMD3) suppression of the compensated signal is 41.7 dB, an improvement of approximately 18 dB compared to the original signal [Fig. 4(a)]. By utilizing the proposed twice nonlinear fitting for digital compensation, the IMD3 suppression of the compensated signal is 60.7 dB, an increase of 37.1 dB compared to the original signal [Fig. 4(b)]. Meanwhile, there is also a significant improvement in the fifth-order intermodulation suppression. We conduct digital compensation processing using the proposed twice nonlinear fitting on 72 sets of two-tone signals with center frequencies ranging from 2.6 to 16.8 GHz (Fig. 5). After digital compensation, the two-tone signals show an IMD3 suppression of approximately 48 to 62 dB, which shows an improvement of 22 to 46 dB over the uncompensated signals.

We introduce a blind separation method for nonlinear distortion compensation of microwave photonic links based on the spectrum power threshold of the optical link output. Furthermore, we propose a digital compensation technique that employs twice nonlinear fitting to suppress intermodulation distortions in microwave photonic link with significant nonlinearity. Additionally, a high-speed oscilloscope is adopted to sample the output signal of a microwave photonic link, and then offline digital compensation is performed. The nonlinearity of the optical link is fitted and digitally compensated using a memory polynomial model with a nonlinear order of 5 and a memory depth of 16. Finally, this approach improves the IMD3 suppression of the microwave photonic link by more than 20 dB across the frequency range from 2.6 to 16.8 GHz.