High Linearity Microwave Photonic Up-Conversion System Based on Parallel Dual-Drive Mach–Zehnder Modulators

Abstract

A large dynamic frequency up-conversion scheme based on parallel dual-drive Mach–Zehnder modulators (DD-MZM) and balance detection is proposed and demonstrated experimentally. By optimizing the distribution ratio of the optical carrier power and the IF signal power between the two DD-MZMs, the third-order intermodulation components in two sub-links cancel each other upon the balanced photodetector. The measured results show that the large spurious-free dynamic range of 112.3 dB·Hz^4/5 is obtained for an intermediate frequency signal of 2 GHz up-converted to 18 GHz, which is a 14.8 dB enhancement compared with the traditional carrier suppression double-sideband modulation mixer. The frequency up-conversion performance of the established system for the broadband signal is measured with the results demonstrating the feasibility of the proposed optimization scheme.

1. Introduction

A frequency mixer, which realizes the up- or down-conversion of radio frequency (RF) signals, is an indispensable module in modern wireless communication systems [1], radar systems [2], satellite payload [3] and electronic warfare [4]. With the increasing demand for high-quality signal transmission, the operation frequency bands of electronic systems are developing towards a high-frequency domain [5]. Due to the electronic bottleneck of the traditional electrical frequency mixer, it is difficult to achieve frequency conversion over a broad frequency band in one step. Compared with electrical frequency mixers, microwave photonic frequency conversion technology can realize one-step frequency conversion over a large frequency band, featuring high RF isolation and immunity to electromagnetic interference [6]. In the past few years, researchers have proposed various schemes to achieve microwave photonic mixing, such as by using cascaded electro-optic modulators [7], mode-locked lasers [8] and optoelectronic oscillators [9].

With the continuous research on microwave photonic frequency conversion technology, the performance of the frequency conversion link was enhanced, such as improving the conversion efficiency [10], adjusting the phase after mixing [11], cancelling image frequency components [12], suppressing conversion spurious components [13] and so on. Linearity is a significant performance of microwave photonic frequency conversion system and can be characterized by spurious-free dynamic range (SFDR). The SFDR represents the highest and lowest power that the microwave frequency conversion system can handle. Compensating the nonlinear distortion of the system in the digital domain is a direct measure to improve SFDR, and it does not need to modify the original link structure [14,15]. However, the digital domain compensation is only suitable for a down-conversion system with a lower output signal frequency. The method of optimizing the structure of the microwave photonic frequency conversion system in the analog domain is much more universal, such as using low relative intensity noise lasers [16] and low noise optical amplifiers [17]. However, these measures only decrease the noise power of the system and cannot suppress the third-order intermodulation distortion (IMD3) components generated during the frequency conversion process.

The schemes to suppress IMD3 components were widely investigated. In [18], by properly setting the three DC bias voltages of a dual-parallel Mach–Zehnder modulator (DP-MZM), the SFDR enhancement of the frequency conversion link can be achieved. However, the modulation of the local oscillator signal requires cascading an additional intensity modulator. In addition, the three DC biases must be precisely controlled at specific voltages, which increases the complexity of the system and the difficulty for real applications. A phase modulator often has higher linearity than an intensity modulator and does not require bias control. By adjusting the polarization state of the light between the two cascaded phase modulators and the response of the back-end optical filter, the enhancement of SFDR can be achieved [19]. However, it is hard for optical filters to filter out low-frequency signal sidebands near the carrier, which makes it difficult to be utilized for frequency up-conversion. The polarization state adjustment of light in polarization multiplexed links can also improve the SFDR of the system [20]. By adjusting the modulation index of the LO signal in the phase modulation link to a certain value, the IMD3 components can be suppressed [21]. However, the increase in modulation depth brings much attenuation of fundamental signal power.

In this paper, we propose a large dynamic frequency up-conversion scheme, which utilizes two dual-drive Mach–Zehnder modulators (DD-MZMs) and a balanced photodetector (BPD) to form two parallel microwave photonic links. The intermediate frequency (IF) signal and the local oscillator (LO) signal are, respectively, loaded into the two RF input ports of the DD-MZM in the two microwave photonic links, and the DD-MZM works at the carrier-suppressed double-sideband (CS-DSB) modulation state. By optimizing the distribution ratio of the optical carrier power and the IF signal power between the two microwave photonic links, the IMD3 components are cancelled upon the BPD. The experimental results show that the 2 GHz IF signal is up-converted to 18 GHz, and an SFDR of 112.3 dB·Hz^4/5 is obtained. Compared with the traditional CS-DSB modulation scheme, the SFDR is improved by 14.8 dB.

2. System Structure and Operation Principle

The schematic of the proposed high linearity microwave photonic up-conversion system is shown in Figure 1. The optical carrier emitted by the laser diode (LD) is divided into two paths by the variable optical coupler (VOC) and sent to two DD-MZMs, respectively. The upper path is the main frequency conversion path, and the lower path is the IMD3 cancellation path. The IF signal is divided into two paths by the variable electric coupler (VEC) and loaded to the RF input port of the two DD-MZMs together with the LO signal equally divided by a 3-dB coupler. The two DD-MZMs work at the minimum DC bias point and output CS-DSB signals. After being amplified by EDFA, the optically carried signals are input to the BPD for optoelectrical conversion to realize the frequency up-conversion.

In order to analyze the conditions for suppressing the IMD3 components in the process of frequency conversion, a two-tone IF signal is input to the two DD-MZMs. The two modulators work at the CS-DSB modulation state. For the convenience of analysis, supposing the power distribution ratio of the IF signal input to the modulator in the upper path and the modulator in the lower path is a²:1, and the power distribution ratio of the optical carrier is b²:1, the optical signal output from the modulator in the upper path can be expressed as

$$E_1=bA{e}^{j{\omega }_{c}t}\left[e^{j(\frac{\pi V_{DC1}}{V\pi }+\frac{\pi a{V}_{IF1}}{V\pi }\cos {\omega }_{{}_{IF1}}t+\frac{\pi a{V}_{IF2}}{V\pi }\cos {\omega }_{{}_{IF2}}t)}+e^{j(\frac{\pi VLO}{V\pi }\cos {\omega }_{{}_{LO}}t)}\right]$$

(1)

By using the Jacobi–Anger expansions, Equation (1) can be expressed as

$$\begin{array}{cc}\hfill E_1& =bA{e}^{j{\omega }_{c}t}[e^{j\phi }{e}^{j{m}_1\cos {\omega }_{{}_{IF1}}t}{e}^{j{m}_1\cos {\omega }_{{}_{IF2}}t}+e^{j{m}_2\cos {\omega }_{{}_{LO}}t}]\hfill \\ & =bA{e}^{j({\omega }_c+{\omega }_{{}_{IF1}}+{\omega }_{{}_{IF2}})t+j\phi }\cdot {\displaystyle \sum _{n=-\infty }^{+\infty }{\displaystyle \sum _{m=-\infty }^{+\infty }{j}^{m+n}{J}_n(a{m}_1)J_m(a{m}_1)}}\hfill \\ & +bA{e}^{j({\omega }_c+{\omega }_{{}_{LO}})t}\cdot {\displaystyle \sum _{p=-\infty }^{+\infty }{j}^p{J}_p(m_2)}\hfill \end{array}$$

(2)

where $J_n$, $J_m$ and $J_m$ are the Bessel functions of the first kind, $m_1=\pi V_{I{F}_1}/V_{\pi }=\pi V_{I{F}_2}/V_{\pi }$ and $m_2=\pi V_{LO}/V_{\pi }$ are the modulation indexes of IF signals and the LO signal and $A$ is the amplitude of the optical carrier. $\phi =\pi V_{DC}/V_{\pi }$ is the phase shift caused by the DC bias voltage. Similarly, the output optical signal from the modulator in the lower path can be expressed as:

$$\begin{array}{cc}\hfill E_2& =A{e}^{j{\omega }_{c}t}[e^{j\phi }{e}^{j{m}_1\cos {\omega }_{{}_{IF1}}t}{e}^{j{m}_1\cos {\omega }_{{}_{IF2}}t}+e^{j{m}_2\cos {\omega }_{{}_{LO}}t}]\hfill \\ & =A{e}^{j({\omega }_c+{\omega }_{{}_{IF1}}+{\omega }_{{}_{IF2}})t+j\phi }\cdot {\displaystyle \sum _{n=-\infty }^{+\infty }{\displaystyle \sum _{m=-\infty }^{+\infty }{j}^{m+n}{J}_n(m_1)J_m(m_1)}}+A{e}^{j({\omega }_c+{\omega }_{{}_{LO}})t}\cdot {\displaystyle \sum _{p=-\infty }^{+\infty }{j}^p{J}_p(m_2)}\hfill \end{array}$$

(3)

When $m=n=p=1$, the output fundamental frequency signal can be obtained, and when $m=\pm 2,n=-1,p=1$, the IMD3 component can be obtained. Once the two optical signals are sent into the BPD for photoelectric conversion and the frequency-converted fundamental signal and the IMD3 component are obtained, the output signals of the upper link can be expressed as

$$i_{fun-up}=b^2{A}^2{J}_1^4(a{m}_1)-2b^2{A}^2{J}_1(a{m}_1)J_1(m_2)\cos ({\omega }_{I{F}_1}+{\omega }_{I{F}_2}+{\omega }_{LO})t+b^2{A}^2{J}_1^2(m_2)$$

(4)

$$\begin{array}{l}{i}_{IM{D}_3-up}=b^2{A}^2{J}_1^2(a{m}_1)J_2(a{m}_1)-2b^2{A}^2{J}_2(a{m}_1)J_1(a{m}_1)J_1(m_2)\cos (2{\omega }_{I{F}_1}-{\omega }_{I{F}_2}+{\omega }_{LO})t\\ +b^2{A}^2{J}_1^2(m_2)\end{array}$$

(5)

and the output signals of the lower link can be expressed as

$$i_{fun-down}=A^2{J}_1^4(m_1)-2A^2{J}_1(m_1)J_1(m_2)\cos ({\omega }_{I{F}_1}+{\omega }_{I{F}_2}+{\omega }_{LO})t+A^2{J}_1^2(m_2)$$

(6)

$$\begin{array}{l}{i}_{IM{D}_3-down}=A^2{J}_1^2(m_1)J_2(m_1)-2A^2{J}_2(m_1)J_1(m_1)J_1(m_2)\cos (2{\omega }_{I{F}_1}-{\omega }_{I{F}_2}+{\omega }_{LO})t\\ +A^2{J}_1^2(m_2)\end{array}$$

(7)

Ignoring the DC term, the IMD3 component of the BPD output can be expressed as

$$i_{IM{D}_3}=[2b^2{A}^2{J}_2(a{m}_1)J_1(a{m}_1)J_1(m_2)-2A^2{J}_2(m_1)J_1(m_1)J_1(m_2)]\cos (2{\omega }_{I{F}_1}-{\omega }_{I{F}_2}+{\omega }_{LO})t$$

(8)

If the coefficient of the IMD3 is equal to 0, the cancellation of the IMD3 components can be realized,

$$b^2{J}_2(a{m}_1)J_1(a{m}_1)=J_2(m_1)J_1(m_1)$$

(9)

Under the condition of small-signal modulation, $J_0(m)\approx 1$, $J_1(m)\approx m/2$ and $J_2(m)\approx m^2/8$, Equation (9) can be simplified as

$$a^3\cdot b^2=1$$

(10)

It can be seen from Equation (10) that when the power distribution ratio of optical carrier and IF signal between two paths is appropriately allocated to match the condition, the IMD3 component in the frequency conversion process can be effectively suppressed.

3. Experiment Results and Discussion

In order to demonstrate the feasibility of the proposed large dynamic frequency up-conversion scheme, an experiment is carried out. An optical carrier from a distributed feedback LD (Emcore 1754W, EMCORE Corporation, Alhambra, CA, USA) with a wavelength of 1545.42 nm and a power of 15 dBm is sent to an optical coupler with a variable splitting ratio to form the upper and lower paths. Two optical carriers with different optical power are input into DD-MZMs (FTM7937EZ, Shenzhen HLT Optical Technology Co., Ltd., Shenzhen, China). The DD-MZMs have a 3-dB bandwidth of 25 GHz and a half-wave voltage of 3.5 V. Two bias controllers (OZ Optics, MBC-MINI-PD-3A, OZ Optics Ltd., Westbrook, ON, Canada) are used to control DD-MZMs to work at the minimum point. The output CS-DSB signals are amplified by two EDFAs (Maxray Photonics, EDFA-PA-35-M, Hefei, China) and then sent to the BPD (u²t, BPDV2150R, Berlin, Germany) to realize the photoelectric conversion. The up-conversion signal is measured by an electrical spectrum analyzer (ESA, Agilent E4440A, Santa Clara, CA, USA).

Firstly, we analyze the frequency up-conversion performance of the system. A 2 GHz single-tone IF signal with a power of 0 dBm generated from a microwave signal generator (Agilent E8267D, Santa Clara, CA, USA) is divided into two parts: one is sent into DD-MZM₁, and the other is sent into DD-MZM₂. The 16 GHz LO signal with a power of 10 dBm generated from the other microwave signal generator (Agilent E8257D, Santa Clara, CA, USA) is also divided into two parts and sent into two DD-MZMs. All the DD-MZMs work stably at the minimum transmission point. The optical spectrum from DD-MZM₁ is shown in Figure 2. (Here, due to the limited resolution of the optical spectrum analyzer, the frequency of the intermediate frequency signal is set as 5 GHz). It can be found that there is still a little optical carrier remaining, and this is due to the limited extinction ratio of the modulators. The modulated signal light in input to the BPD for photoelectric conversion and the measured electrical spectrum is shown in Figure 3. After photoelectric conversion, an up-converted signal with a frequency of 18 GHz and a power of −26.7 dBm is obtained. By means of the carrier suppression modulation, the LO signal output from the system is effectively suppressed. An LO isolation of 31.4 dB is achieved. If the extinction ratio of the modulator is higher, the residual LO signal can be further suppressed. The 3-dB bandwidths of DD-MZM and BPD are 25 GHz and 43 GHz, respectively. With the frequency of the IF signal being unchanged, the proposed large dynamic frequency up-conversion system can achieve a maximum up-conversion frequency of 27 GHz, namely, up-conversion from S-band to Ka-band.

Then, the nonlinearity optimization is investigated. A two-tone IF signal with frequencies of 2 GHz and 2.01 GHz is produced by two microwave signal generators, and the LO signal remains unchanged. A 3-dB power divider and a tunable attenuator are utilized to realize the function of a VEC. By adjusting the distribution ratio of the optical carrier power and the IF signal power input to the two paths, the IMD3 cancellation condition of Equation (10) is matched. By disconnecting the optical path to the lower port of BPD, the measured electrical spectrum of the upper path is shown in Figure 4a. It can be found that in addition to the up-conversion two-tone signal of 18 GHz and 18.01 GHz, there are also strong IMD3 components at 17.99 GHz and 18.02 GHz, with the signal-to-interference ratio (SIR) of 54.3 dB. Figure 4b shows the measured electrical spectrum of the signal output from the lower path after optimizing the power distribution ratio. By comparing Figure 4a,b, the power of the IMD3 component output from the upper path is basically equal to the power of the IMD3 component output from the lower path. The two optical signals are simultaneously input to the BPD for photoelectric conversion, and the output signals are differentially combined. Figure 4c shows the combined output electrical spectrum. It can be seen that IMD3 components are suppressed greatly with the SIR of 79.3 dB, which is increased by 25 dB. During the up-conversion test of the two-tone signal and the subsequent SFDR test, the bandwidth (BW) of the spectrum analyzer is set to 50 MHz, and the resolution bandwidth (RBW) and video bandwidth (VBW) are both set to 10 kHz.

Then, the SFDR of the system is measured with the IF signal power increase step of 1 dB. The diagram of the measured result is shown in Figure 5a with the SFDR of 112.3 dB·HZ^4/5. The slope of the fitting line of the intermodulation distortion component is 5, which indicates that the IMD3 component in the system is almost suppressed, and the fifth-order intermodulation distortion becomes the main nonlinear component. Due to the reduction in common-mode noise by BPD, the noise floor of the proposed up-conversion scheme can still maintain at a relatively low value when two EDFAs are included. The measured noise floor of the proposed scheme is −147.8 dBm/Hz. At the same time, the SFDR of the traditional frequency up-conversion link based on CS-DSB modulation is also measured. By disconnecting the optical path to the lower input port of BPD and adjusting the VOC, all the optical carrier is sent into the upper path. Directly input the IF signal and LO signal to the DD-MZM in the upper path. The measured results are shown in Figure 5b with the SFDR of 97.5 dB·Hz^2/3 and the measured noise floor of −146.9 dBm/Hz. Under the condition of the optical power into BPD being unchanged, the increase in the optical power reduces the gain of the EDFA, thereby reducing the ASE noise. By comparing Figure 5a,b, the SFDR of the proposed frequency up-conversion system is improved by 14.8 dB, demonstrating the effectiveness of the optimization scheme.

Finally, the frequency up-conversion performance for the broadband signal with the established system is tested. The IF signal with a center frequency of 2 GHz and bandwidth of 50-Mbaud of 16-QAM is applied. The output up-conversion electrical spectra with the input IF signal power of 12 dBm, 15 dBm and 18 dBm are measured and shown in Figure 6. Figure 6a,c,e are the measured electrical spectra of the frequency up-conversion system based on traditional CS-DSB modulation. It can be seen that the intermodulation distortion components grow greatly, as illustrated by the spurious out-of-band components with the increase in the input IF power. By using the proposed scheme, the IMD3 components can be suppressed effectively, as shown in Figure 6b,d,f. Certainly, there are some residual spurious components when the IF signal power is 18 dBm, which are mainly the fifth-order intermodulation distortion. The error vector magnitude (EVM) of the demodulated signal in Figure 6f is 5.12% rms, satisfying the international 3 GPP guidelines with the maximum EVM value of 12.5% rms for 16-QAM modulation [22].

For comparison,

Table 1. Performance comparison of microwave photonic mixing links.

Ref.	Approach to Linearization	SFDR	Noise Floor	Fundamental Penalty	SWaP
[9]	None	102.2 dB·Hz2/3	−162 dBm/Hz	—	fair
[11]	None	101.3 dB·Hz2/3	−146 dBm/Hz	—	good
[14]	Digital compensation	114.5 dB·Hz4/5	−154 dBm/Hz	1.1 dB	poor
[19]	Optical filtering	93 dB·Hz4/5	−140 dBm/Hz	~7 dB	fair
[20]	Adjust polarization	92.3 dB·Hz2/3	−145 dBm/Hz	1.6 dB	fair
[21]	Increase modulation index	114 dB·Hz4/5	−170 dBm/Hz	13.8 dB	good
This work	Direct cancellation	112.3 dB·Hz4/5	−147 dBm/Hz	1.2 dB	fair

lists the recent work in developing microwave photonics mixing links. Compensating nonlinear components in the digital domain is a direct and effective solution to improve SFDR [14]. However, this approach requires additional digital processing modules, degrading the size weight and power (SWaP). In [19], a large SFDR is achieved by adjusting the polarization and optical filtering. Cascading of these devices also increases the size. Adjusting the power of the LO signal to improve the SFDR is a simple way [21]; however, it results in a higher fundamental power penalty. The SWaP of the proposed microwave photonic mixer with no other adjustment devices is relatively good. It is worth noting that with the development of photonic integration technology [23], microwave photonic systems can be integrated on one chip, which will improve the SWaP greatly.

4. Conclusions

In conclusion, a high linearity microwave photonic frequency up-conversion scheme is proposed and demonstrated experimentally. By optimizing the distribution ratio of optical power and IF signal power between the upper and lower paths, the cancellation of IMD3 components can be achieved through a differential combination of the balanced detection. In the proof-of-concept experiment, the IF signal of 2 GHz is up-converted to 18 GHz with the SFDR of 112.3 dB·Hz^4/5. Compared with the frequency up-conversion scheme based on traditional CS-DSB modulation, the SFDR of the proposed system is increased by 14.8 dB. The frequency up-conversion for a broadband signal by the established system is also tested with good performance. The proposed high linearity microwave photonic up-conversion system can be applied in many scenarios, such as satellite payload for transmitting, millimeter-wave emitter for wireless access network by radio-over-fiber, radar transmitter and so on.