- Email: [email protected]
Mach-Zehnder modulator modulated radio-over-fiber transmission system using dual wavelength linearization
Optics Communications 385 (2017) 229–237
Contents lists available at ScienceDirect
Optics Communications journal homepage: www.elsevier.com/locate/optcom
crossmark
Mach-Zehnder modulator modulated radio-over-fiber transmission system using dual wavelength linearization ⁎
Ran Zhua,b, Ming Huic, Dongya Shena, , Xiupu Zhanga,b a b c
iCom Laboratories, School of Information Science and Engineering, Yunnan University, Kunming, Yunnan 650091, China iPhotonics Laboratories, Department of Electrical and Computer Engineering, Concordia University, Montreal, Quebec, Canada H3G 1M8 School of Physics and Electronic Engineering, Nanyang Normal University, Nanyang, Henan 473061, China
A R T I C L E I N F O
A BS T RAC T
Keywords: Microwave photonics Radio-over-fiber systems Linearization Fiber optic communications
In this paper, dual wavelength linearization (DWL) technique is studied to suppress odd and even order nonlinearities simultaneously in a Mach-Zehnder modulator (MZM) modulated radio-over-fiber (RoF) transmission system. A theoretical model is given to analyze the DWL employed for MZM. In a single-tone test, the suppressions of the second order harmonic distortion (HD2) and third order harmonic distortion (HD3) at the same time are experimentally verified at different bias voltages of the MZM. The measured spurious-free dynamic ranges (SFDRs) with respect to the HD2 and HD3 are improved simultaneously compared to using a single laser. The output P1 dB is also improved by the DWL technique. Moreover, a WiFi signal is transmitted in the RoF system to test the linearization for broadband signal. The result shows that more than 1 dB improvement of the error vector magnitude (EVM) is obtained by the DWL technique.
1. Introduction Distributed radio access network (RAN) based structure is mainly employed in present wireless networks. In distributed RAN, backhaul connects core network and baseband units (BBUs) through Internet protocol (IP) network. BBUs are connected to remote radio units (RRUs) which up-convert baseband signals to radio frequency (RF) signals and the RF signals are transmitted to antennas. Both BBU and RRUs are located in a cell site cabinet. However, data capacity increases continuously and seamless coverage is required for various situations, so more small and micro cells are deployed. The increasing cell sites raise the consumption of power and space. Therefore, the movement from distributed RAN to Cloud-RAN (C-RAN) architecture is a tendency [1–3]. C-RAN consists of two parts, back-haul and fronthaul network. The back-haul is used to connect BBUs and the fronthaul is introduced to connect BBUs and RRUs. The centralization can reduce the cost of deployment, maintenance, and power consumption, and increase the network flexibility. But C-RAN has strict requirements of latency and synchronization which are the main challenge. Moreover, the Common Public Radio Interface (CPRI) protocol runs between BBUs and RRUs demands the supports of multiple formats. Radio-over-fiber (RoF) transmission system can be a solution for the front-haul network. RF signals are transmitted through optical fibers in RoF systems because optical fiber has tremendous capacity. Some
⁎
functions of RRU can be moved to BBU so that RRU can be simplified. Thus, the network is further centralized and the total cost is reduced. Furthermore, it is easier to meet the requirements of CPRI because the signal processing functions are centralized in central office. Although RoF systems provide several advantages, they are susceptible to nonlinearities due to the high peak-to-average power ratio (PAPR). The nonlinearities need to be suppressed to improve the spurious-free dynamic range (SFDR). Many linearization techniques were studied and they can be divided into three major methods, i.e., digital linearization using digital signal processing [4–7], analog predistortion circuit (PDC) [8–13], and optical linearization [14–21] using optical components. Digital linearization has potential to suppress nonlinearities substantially, but it is inadequate in dealing with multiband signals. Although two concurrent bands linearization was studied, the complexity will be increased dramatically when more signals are in consideration [5]. Analog PDC is broadband and lowcost, but it suppresses only odd order intermodulations (IMDs) for broadband. Optical linearization provides huge bandwidth and can be used to suppress odd and even order nonlinearities. In this paper, dual wavelength linearization (DWL) technique is studied to suppress even and odd order nonlinearities simultaneously in a Mach-Zehnder modulator (MZM) modulated RoF transmission system. The requirements of the DWL for MZMs are theoretically analyzed and experimentally evaluated. Single-tone test is given to
Corresponding author. E-mail address: [email protected] (D. Shen).
http://dx.doi.org/10.1016/j.optcom.2016.09.066 Received 7 September 2016; Received in revised form 28 September 2016; Accepted 29 September 2016 Available online 05 November 2016 0030-4018/ © 2016 Elsevier B.V. All rights reserved.
Optics Communications 385 (2017) 229–237
R. Zhu et al.
Fig. 1. Schematic diagram and experimental setup of the DWL for MZM in an analog optical link.
measure the spurious-free dynamic ranges (SFDRs) and WiFi test is also given to verify the linearization of broadband signal. The rest of the work includes as follows. The principle and theory are presented in Section 2. The experiments of single-tone test and WiFi test are given in Section 3. The conclusion is given in Section 4. 2. Principle The schematic diagram of the DWL for MZM is given in Fig. 1. The main nonlinear source is assumed to be the MZM and all other components are assumed to be linear. Two lasers emit lights with different wavelengths which are coupled and transmitted to an MZM which is used for optical subcarrier modulation. The two lasers are incoherent with each other. A 180° hybrid is used for push-pull modulation of MZM. MZM is wavelength-dependent so the two lights experience different modulation characteristics. So different signals and nonlinearities are generated and carried by the two optical carriers. In photodiode (PD), the two optical carriers are demodulated incoherently and thus the sums of the signals and nonlinearities are obtained. The total signal and nonlinearities can be adjusted by tuning the power
Fig. 2. Measured optical transmission characteristics of the MZM.
Fig. 3. (a) Measured optical power of the 1640 nm laser for the maximum suppression of the HD2. (b) RF power of the RF output signals, (c) the HD2s, (d) the HD3s for DWL, 1553 nm and 1640 nm, respectively. The HD2 is maximally suppressed.
230
Optics Communications 385 (2017) 229–237
R. Zhu et al.
Fig. 4. (a) Enhancements of the RF output signal and the HD3 compared to using single 1553 nm laser. (b) Suppression of the HD2 compared to using single 1553 nm laser.
Fig. 5. (a) RF power of the HD2 and (b) RF power of the RF output signal and the HD3 using the DWL technique versus the optical power of the 1640 nm laser.
Fig. 6. Measured SFDRs with respect to HD2 and HD3 for (a) DWL, (b) 1553 nm, and (c) 1640 nm. The bias voltage is 0.1 V.
Two incident lasers are assumed to be laser A and laser B. The halfwave voltages for laser A and laser B are different because MZMs are wavelength-dependent. We assume the half-wave voltages for the two lasers are Vπ_A and Vπ_B and Vπ_A < Vπ_B. The two lasers are incoherent, so they can be demodulated separately in the PD. After photodetection, the total photocurrent can be given as follows:
ratio of the two lasers. If the nonlinearities carried by the two lights are antiphase, the nonlinear distortion can be suppressed. When the MZM is set to push-pull modulation, the optical transmission characteristic of the MZM can be described as
T=
⎡π ⎤⎫ Po 1 ⎧ = ⎨1 + cos ⎢ (V +Vb ) ⎥ ⎬ ⎣ Vπ ⎦⎭ Pi 2L ⎩
(1)
where T is the optical transmission of the MZM, Po and Pi are the output and input optical power of the MZM respectively, L is the insertion loss, Vπ is the half-wave voltage, V is the RF signal, and Vb is the bias voltage applied to the MZM. Using Taylor series, (1) can be expanded as
T=
1 1 + 2L 2L
∞
⎡ ⎛ π ⎞n ⎛ nπ πVb ⎞ V n ⎤ ⎟ cos ⎜ + ⎟ ⎥ ⎝ 2 Vπ ⎠ n! ⎦ ⎣ Vπ ⎠
∑ ⎢ ⎜⎝
n =0
I=
RPA RPB + + 2L 2L Vn ⎫ ⎬ n! ⎭
∞
⎧⎡
⎛
⎞
⎩ ⎣ 2L
⎝ 2
⎠
⎛
⎞⎤
⎝ 2
⎠⎦
∑ ⎨ ⎢ RPA mAn cos ⎜ nπ +mA Vb⎟+ RPB mBn cos ⎜ nπ +mB Vb⎟ ⎥ ⎪ ⎪
n =0
2L
⎪ ⎪
(3)
where PA and PB are the optical input power for laser A and laser B respectively, and R is the responsivity of the PD. mA and mB are the modulation indexes of the MZM for laser A and laser B respectively,
(2) 231
Optics Communications 385 (2017) 229–237
R. Zhu et al.
PD with the responsivity of 0.6 A/W from Discovery is employed. A wideband power amplifier (PA) ZVA-213+ is used to boost the photocurrent. It has the gain of 26 dB, output power at 1 dB compression point (P1 dB) of 24 dBm, noise figure (NF) of 3 dB, and bandwidth from 0.8 to 21 GHz. A spectrum analyzer (SA) is used to measure the output powers. The two wavelengths of the lasers are 1553 nm and 1640 nm. Measured optical transmission characteristics of the MZM are shown in Fig. 2. The maximum transmission points of the MZM are −2.7 V for 1553 nm and −2.5 V for 1640 nm. The minimum transmission points are 2.3 V for 1553 nm and 2.9 V for 1640 nm, respectively. So the half-wave voltages are 5 V for 1553 nm and 5.4 V for 1640 nm. Referring to the requirements of (4), the even order components can be suppressed when −0.2 V < Vb < 0.2 V and it is marked as Area 1 in Fig. 2. The odd order components can be suppressed when 2.3 V < Vb < 2.9 V and it is marked as Area 2. Fig. 7. Measured P1 dBs for DWL, 1553 nm, and 1640 nm at the bias voltage of 0.1 V.
i.e., mA= V
π
π_A
and mB= V
π
π_B
3.1. Single-tone test
. The nth order component in the photocurrent
Generally speaking, odd order IMDs have almost the same phase with the signal. But the phases of harmonics and even order IMDs fluctuate depending on the transmission distance. So the requirements of the suppression of harmonic distortions (HDs) are stricter than that for IMD. Therefore, single-tone test is conducted to suppress HDs. The RF signal generator is set to 7 GHz and 10 dBm. The bias voltage is swept to suppress the second order HD (HD2). The optical power of the 1553 nm laser is set to 0 dBm and the optical power of the 1640 nm laser is adjusted to obtain the maximum suppression of HD2. Back-toback is employed in the experiments. It is found that the HD2 can be suppressed for the bias voltage from −0.2 to 0.1 V. The measurement results agree with the theories in Section 2. The required optical power of the 1640 nm laser for the maximum suppression of the HD2 is given in Fig. 3(a). When the bias voltage of the MZM is increased, the required optical power of the 1640 nm laser is also increased. From −0.2 to 0.1 V, the required optical power is increased from −10 to 5.4 dBm. The reason is that the HD2 is the minimum when the MZM is biased at Vπ/2. When the MZM is biased at −0.2 V, the HD2 for 1553 nm is its minimum. And the HD2 for 1640 nm is its minimum when the MZM is biased at 0.2 V. When the bias voltage is increased from −0.2 to 0.2 V, the HD2 for 1553 nm is increased and the HD2 for 1640 nm is decreased. So the optical power of the 1640 nm laser has to be increased to balance the power of the two HD2s. Fig. 3(b) shows the measured RF output signals for DWL, 1553 nm, and 1640 nm, respectively. The power of the RF output signal for DWL is higher than that for 1553 nm and 1640 nm from −0.2 to 0.1 V. This is because odd order components for the two wavelengths are inphase in this bias range. So the RF signal is improved by the DWL technique. Fig. 3(c) gives the HD2s. The HD2 for DWL is suppressed when the powers of the HD2s for 1553 nm and 1640 nm are the same. It illustrates that the HD2s carried by the two optical carriers are
can be eliminated when the following condition is met. nπ
⎛ m ⎞n cos( 2 + mB Vb ) PA =−⎜ B ⎟ ⎝ mA ⎠ cos( nπ PB + mA Vb ) 2
(4)
The power and modulation indexes are positive and the power ratio of the two lasers is adjustable. For the even order components, Eq. (4) can be met when Vπ_A /2 < Vb < Vπ_B /2 . At this time, the even order components carried by the two optical carriers can cancel each other if their RF power is identical. Moreover, it can be found that the odd order components are increased in the meantime. For the odd order components, Eq. (4) can be met when Vπ_A < Vb < Vπ_B . So the odd and even order components cannot be eliminated simultaneously. However, if the power of the signal and nonlinearities are enhanced by the same level, the nonlinearities are suppressed relatively. This will be proved in later experiments. For a practical single mode fiber link, dispersion compensating fiber is required in the proposed method to compensate for the dispersion induced delay between the two wavelengths. 3. Experiments The RF output signals, second and third order nonlinearities of an RoF system are measured to verify the performance of DWL technique. Experimental setup is given in Fig. 1. A Sumitomo Osaka Cement 40 Gb/s C-band MZM is used. Two polarization controllers are used and adjusted to get maximum output optical power of MZM for each laser. The two lasers are combined and transmitted to the MZM through a 3 dB coupler. An RF signal generator generates RF signal. The MZM is set to push-pull modulation using a 180° hybrid. A 40 GHz
Fig. 8. Measured SFDRs with respect to (a) HD2 and (b) HD3 versus the bias voltage. The HD2 is maximally suppressed.
232
Optics Communications 385 (2017) 229–237
R. Zhu et al.
Fig. 9. Improvements of the SFDRs with respect to HD2 and HD3 using the DWL compared to using (a) the 1553 nm laser and (b) the 1640 nm laser.
Fig. 10. (a) Measured output P1 dBs for DWL, 1553 nm, and 1640 nm. (b) Measured improvements of the output P1 dB using the DWL compared to using the single laser.
antiphase from −0.2 to 0.1 V. It also shows that the HD2 for 1553 nm is increased with the bias voltage. Fig. 3(d) gives the measured third order HDs (HD3s). It can be seen that it is increased by the DWL. Moreover, when the bias voltage is increased, the optical power of the 1640 nm laser has to be increased as shown in Fig. 3(a). So the odd order components such as first and third order components are also increased with the bias voltage. The improvements of the RF output signal and HD3 are given in Fig. 4.(a). It shows that the power improvements of the RF output signal and the HD3 by the DWL are similar. This illustrates that the SFDR with respect to the HD3 can be improved. When the bias voltage is 0.1 V, the improvement of the RF output signal is 17.6 dB. Fig. 4(b) shows that the suppression of the HD2 is higher than 17 dB for the bias voltage from −0.2 to 0.1 V. And the highest suppression is 38.3 dB at the bias voltage of 0.1 V. To verify the dependence of the optical power ratio, the power of the 1640 nm laser is swept. The power of 1553 nm laser is fixed at 0 dBm. The RF input power is 10 dBm and the bias voltage is set to 0.1 V for the maximum suppression of the HD2. Fig. 5 shows the power of the HD2, RF output signal, and the HD3 using the DWL technique versus the optical power of the 1640 nm laser. In Fig. 5(a), the maximum suppression of the HD2 is achieved at 5.4 dBm. When the optical power is shifted to 4.6 dBm or 5.9 dBm, the suppression degrades by 10 dB. Fig. 5(b) shows that the power of the RF output signal and HD3 are increased with the power of the 1640 nm laser. This is because the odd order components carried by the two optical carriers are inphase with each other. And it can be seen 1 dB increase of optical power causes 2 dB increase of RF power. Then the SFDRs at 0.1 V bias voltage with respect to HD2 and HD3 are measured and given in Fig. 6 with respect to the HD2 and HD3 for DWL, 1553 nm, and 1640 nm, respectively. To simplify the explana-
tion, we define the SFDR with respect to HD2 as SFDR2 and the SFDR with respect to HD3 as SFDR3. Fig. 6(a) shows the HD2 is fourth order limited that means the HD2 is eliminated by the DWL. It also can be seen that RF output signal and the HD3 are both increased in power by the DWL. Fig. 6(b) and (c) show the SFDRs for 1553 nm and 1640 nm, respectively. Compared to Fig. 6(b), the SFDR2 and SFDR3 are improved by 38.4 dB and 12.1 dB, respectively. Despite the increase of the HD3, the SFDR3 is still improved. The reason is that the RF output signal and the HD3 are increased by the same level. So the SFDR is improved by the increase of the RF output signal. The measurement results agree with the analysis in Section 2. Compared to Fig. 6(c), the SFDR2 and SFDR3 are improved by 11.3 dB and 0.7 dB, respectively. It can be seen that the HD2 in Fig. 6(c) is third order limited. This is because the even order nonlinearities for 1640 nm are suppressed when the MZM is biased at its quadrature point. So the measured HD2 in Fig. 6(c) is a combination of the suppressed even order nonlinearities and optical beating of the odd order nonlinearities in the PD. Measured input and output P1 dBs for DWL, 1553 nm, and 1640 nm at the bias voltage of 0.1 V are given in Fig. 7. The input P1 dBs for DWL, 1553 nm, and 1640 nm are 16 dBm, 15.5 dBm, and 15.4 dBm, respectively. So the input P1 dB is improved by 0.5 dB and 0.6 dB. The output P1 dBs for DWL, 1553 nm, and 1640 nm are −26 dBm, −43.9 dBm, and −27.5 dBm, respectively. It can be seen that the RF output signal is improved by the 1640 nm laser. The output P1 dB is improved by 17.9 dB and 1.5 dB, respectively. This illustrates that the system is linearized by the DWL technique. Fig. 4(b) shows that the DWL can suppress the HD2 by more than 17 dB for the bias voltage from −0.2 to 0.1 V. To verify the linearization performance of the DWL, the SFDRs and P1 dBs are measured and given in Fig. 8 for bias voltage from −0.2 to 0.1 V. 233
Optics Communications 385 (2017) 229–237
R. Zhu et al.
Fig. 11. (a) Measured optical power of the 1640 nm laser for the maximum suppression of HD3. (b) RF power of the RF output signals, (c) the HD2, (d) the HD3 for DWL, 1553 nm and 1640 nm, respectively. The HD3 is maximally suppressed.
Fig. 12. (a) Suppressions of the RF output signal and the HD3 compared to using the 1553 nm laser. (b) Enhancement of the HD2 compared to the using the 1553 nm laser.
the optical power of the 1640 nm laser is enhanced so that the SFDR3 for 1640 nm is improved. The RF output signal is improved much more by the DWL at 0.1 V as shown in Fig. 3(b). So the SFDR3 achieves the maximum at 0.1 V. The improvements of SFDRs with respect to HD2 and HD3 using the DWL compared to using the single laser are given in Fig. 9. The comparison between using the DWL and using the 1553 nm laser is given in Fig. 9(a). The SFDR2 and SFDR3 are improved by more than 20.6 dB and 0.7 dB, respectively. The maximum improvements which are 38.5 dB for SFDR2 and 12.1 dB for SFDR3 are obtained at 0.1 V. In Fig. 9(b), compared to using the 1640 nm laser, the SFDR2 and SFDR3 are improved maximally by 36.9 dB and 10.8 dB respectively at −0.2 V. And the minimum improvements are 11.3 dB and 0.8 dB respectively at 0.1 V.
Measured SFDRs with respect to HD2 and HD3 versus the bias voltage of the MZM are given in Fig. 8(a) and (b), respectively. The optical power of the 1553 nm laser is fixed at 0 dBm. In Fig. 8(a), the SFDR2 for 1553 nm drops with the increase of the bias voltage because the HD2 for 1553 nm achieves the minimum at the quadrature point of −0.2 V. For the same reason, the HD2 for 1640 nm achieves the minimum at its quadrature point of 0.1 V. And the optical power for the 1640 nm laser is increased for the maximum suppression of HD2. So the SFDR2 for 1640 nm increases from −0.2 to 0.1 V. Compared to using the single laser, the SFDR2 is improved for the bias voltage from −0.2 to 0.1 V by the DWL. And the SFDR2 achieves the maximum at 0.1 V. Fig. 8(b) presents the measured SFDR3s. The HD3 is increased by the DWL for the range of bias voltage, so the SFDR3 is improved by the increase of the RF output signal. When the bias voltage is increased, 234
Optics Communications 385 (2017) 229–237
R. Zhu et al.
Fig. 13. Measured SFDRs with respect to HD2 and HD3 for (a) DWL, (b) 1553 nm, and (c) 1640 nm. The bias voltage is 2.6 V.
Fig. 14. Measured SFDRs with respect to (a) HD2 and (b) HD3 versus the bias voltage. The HD3 is maximally suppressed.
Fig. 15. Experimental setup of the WiFi test.
Fig. 2. The optical power of the 1553 nm laser is 0 dBm and the optical power of 1640 nm is swept to achieve the maximum suppression of the HD3. The HD3s carried by the two optical carriers need to be same in power to cancel each other referring to (3). The required optical power of the 1640 nm laser for the maximum suppression of the HD3 is given in Fig. 11(a). The required optical power is increased from −9 to 6.5 dBm while the bias voltage is increased from 2.3 to 2.8 V. The reason is that the HD3 carried by the 1640 nm laser drops to the minimum at its minimum transmission point of 2.9 V. And the HD3 carried by the 1553 nm laser increases from its minimum transmission point of 2.3 V. So the optical power of the 1640 nm laser has to be enhanced to make the RF power of the two HD3s carried by the two optical carriers same. Fig. 11(b) shows that the RF output signal is reduced by the DWL because the odd order components are suppressed in this range of bias voltage. Because of the same reason, Fig. 11(d)
Fig. 10(a) gives the measured output P1 dBs for DWL, 1553 nm, and 1640 nm. The output P1 dB using the DWL is increased with the bias voltage. It is enhanced by the increase of the optical power of the 1640 nm laser. For the system using the 1553 nm laser, the output P1dB changed little while the bias voltage is swept. And it can be seen that the output P1 dB using the DWL is higher than that using the single laser for the bias voltage range. Fig. 10(b) shows the improvements of the output P1 dB compared to using the single laser. When the bias voltage is 0.1 V, the output P1 dB is improved by 17.9 dB compared to using the 1553 nm laser. And when the bias voltage is −0.2 V, it is improved by 16.4 dB compared to using the 1640 nm laser. We sweep the bias voltage and the optical power of the 1640 nm laser to suppress the HD3. It is found that the suppression can be obtained for the bias voltage from 2.3 to 2.8 V. The range of the bias voltage agrees with the calculation and prediction given in (4) and
235
Optics Communications 385 (2017) 229–237
R. Zhu et al.
Fig. 16. Measured EVMs versus the received RF power for DWL, 1553 nm, and 1640 nm. (a) BTB. (b) 8 km SMF. The bias voltage is 0 V.
two lasers is obtained for the maximum suppression of the HD2 as measured in Fig. 3(a). At the receiver, a digital storage oscilloscope (DSO) 81204B from Keysight is employed to demodulate and verify the signal. The WiFi signal is tested in back-to-back (BTB) and 8 km single mode fiber (SMF) transmission. Fig. 16 shows the measurement results. Error vector magnitudes (EVMs) are measured versus the received RF power of the DSO. Fig. 16(a) depicts that the EVM is improved by more than 1.1 dB by the DWL compared to using the single 1553 nm laser for BTB. It also shows that the minimum EVM point is extended to higher received RF power level. The minimum EVM for DWL technique is achieved at the received RF power of −26 dBm while that for 1553 nm is achieved at −29 dBm. Fig. 16(b) shows the EVM is improved by 1 dB by the DWL compared by using the single 1553 nm laser for 8 km SMF transmission. Referring to [19], the delay caused by the dispersion in SMF may degrade the performance of DWL technique, so the length of the SMF needs to be adjusted slightly. In the measurement, the 8 km SMF is connected with 80 m SMF to maximize the linearization. The measurement results illustrate that the DWL technique can enhance the RF signal power and suppress the nonlinearities.
shows that the HD3 is suppressed too. It also shows that the two HD3s carried by the two optical carriers are almost same. However, it can be seen that the HD2 is increased by the DWL in Fig. 11(c). The suppressions of the RF output signal and the HD3 by the DWL compared to using the 1553 nm laser is given in Fig. 12(a). The RF output signal is suppressed by more than 9.8 dB for the bias voltage from 2.3 to 2.7 V and 3.3 dB at 2.8 V. The HD3 is suppressed by more than 11.7 dB and the maximum suppression is 21.8 dB at 2.7 V. However, Fig. 12(b) shows that the HD2 is increased by using the DWL technique. The enhancement is increased with the bias voltage because the optical power of the 1640 nm laser is increased with the bias voltage as shown in Fig. 11(a). The minimum enhancement is 1.3 dB at 2.3 V and the maximum enhancement is 16.8 dB at 2.8 V. The measured SFDRs when the MZM is biased at 2.6 V are given in Fig. 13. The HD3 is maximally suppressed by adjusting the optical power of the 1640 nm laser. It can be seen that the SFDR3 is improved by the DWL compared to using the single laser although the RF output signal is also suppressed by the DWL. But the SFDR2 is reduced because the HD2 is increased by the DWL. And Fig. 13 shows that the power of the RF output signals are low because the MZM is biased near its minimum transmission point. Fig. 14 gives the measured SFDRs with respect to HD2 and HD3 versus the bias voltage. In Fig. 14(a), the SFDR degrades when the DWL is used. The reason is that the HD2 is increased in this bias voltage range. And it can be seen that the SFDR2s are lower than 52 dB in this bias voltage range. This is because the bias voltage is near the minimum transmission point of the MZM. So the RF output signal is weak and the HD2 is strong. Fig. 14(b) shows that the SFDR3 is improved by using the DWL compared to using the single laser. And more than 6.3 dB improvement is achieved for the bias voltage from 2.3 to 2.8 V. The maximal SFDR3 using the DWL is obtained at 2.8 V. Moreover, it should be noted that the P1 dB degrades by using the DWL for this bias voltage range because the RF output signal is reduced.
4. Conclusion In this work, the DWL technique is studied to suppress odd and even order nonlinearities simultaneously in an MZM modulated RoF transmission system. The theoretical requirement of the linearization is derived. In the single-tone test, the concurrent suppressions of the HD2 and HD3 are experimentally verified for a range of the bias voltage of the MZM. The experimental results agree with the calculation. The measured SFDRs with respect to the HD2 and HD3 are improved simultaneously compared to using the single laser. The output P1 dB is also improved by the DWL technique. Moreover, a WiFi test is given to evaluate the linearization of broadband signals. More than 1 dB improvement of the EVM is obtained by using the DWL technique.
3.2. WiFi test
Acknowledgments
To verify the DWL technique for broadband signal, WiFi signal at 2.4 GHz is transmitted in the MZM modulated RoF transmission system. The experimental setup is shown in Fig. 15. At the transmitter, an arbitrary waveform generator (AWG) 7122B from Tektronix generates a WiFi signal at 2.4 GHz compliant with 802.11a. The OFDM signal consists of 64 subcarriers with 16-QAM occupying 20 MHz bandwidth. And the signal rate is 36 Mbit/s. The incident power from the AWG is −13.2 dBm. A PA ZVA-213+ and a broadband variable attenuator are injected between the AWG and the hybrid. The variable attenuator is adjusted to change the RF input power to the MZM. The MZM is biased near the quadrature point at 0 V. The optical power is 6 dBm for 1553 nm and 4.5 nm for 1640 nm. 1.5 dB power ratio of the
This work was supported in part by a Quebec Fonds de recherche du Québec – Nature et technologies (FRQNT) project in broadband photonic devices, an open project from State Key Laboratory of Millimeter Waves (K201615), a project from Henan Science and Technology Foundation (162102310479), and a project from the Science and Technology Foundation of Henan Educational Committee (16A510009), Yunnan high-tech top-talents recruitment project (2012HA005), a key project from Yunnan Science Foundation (2013FA027),and two projects from National Natural Science Foundation of China (61561051 and 61561052). 236
Optics Communications 385 (2017) 229–237
R. Zhu et al.
US patent 4,992,754, Feburary, 1991. [12] C. Han, S.-H. Cho. H.S. Chung, J.H. Lee, Linearity improvement of directlymodulated multi-IF-over-fibre LTE-A mobile fronthaul link using shunt diode predistorter, in: Proceedings of the European Conference on Optical Communication, Valencia, Paper We4-4-4, September, 2015. [13] R.ZhuZ.XuanY.ZhangX.ZhangD.Shen, Novel broadband analog predistortion circuit for radio-over-fiber systems, in: Proceedings of the IEEE International Microwave Symposium (IMS), Phoenix, Arizona, May, 2015, pp. 1–4. [14] B.M. Haas, T.E. Murphy, A simple, linearized, phase-modulated analog optical transmission system, IEEE Photonics Technol. Lett. 19 (10) (2007) 729–731. [15] B. Hraimel, X. Zhang, Performance improvement of radio-over fiber links using mixed-polarization electro-absorption modulators, IEEE Trans. Microw. Theory Tech. 59 (12) (2011) 3239–3248. [16] K.K. Loi, J.H. Hodiak, X.B. Mei, C.W. Tu, W.S.C. Chang, Linearization of 1.3 μm MQW electroabsorption modulators using an all-optical frequency-insensitive technique, IEEE Photonics Technol. Lett. 10 (7) (1998) 964–966. [17] B.M. Haas, V.J. Urick, J.D. McKinney, T.E. Murphy, Dual-wavelength linearization of optically phase-modulated analog microwave signals, J. Lightwave Technol. 26 (15) (2008) 2748–2753. [18] R.ZhuX.Zhang, Linearization of radio-over-fiber systems by using two lasers with different wavelengths, in: Proceedings of the IEEE International Microwave Symposium (IMS), Tampa Bay, Florida, June, 2014, pp. 1–3. [19] R. Zhu, X. Zhang, D. Shen, T. Liu, Analysis of dual wavelengths linearization technique for radio-over-fiber system with electro-absorption modulator, IEEE Trans. Microw. Theory Tech. 99 (2015) 1–11. [20] H.JungS.LeeS.Han, Gain saturated SOA cascaded EML analog optical transmitter, in: Proceedings of the IEEE International Topical Meeting Microwave Photonics (MWP), Awaji-Yumebutai, November, 2002, pp. 205–208. [21] D.H. Jeon, H.D. Jung, S.K. Han, Mitigation of dispersion-induced effects using SOA in analog optical transmission, IEEE Photonics Technol. Lett. 14 (8) (2002) 1166–1168.
References [1] A. Pizzinat, P. Chanclou, T. Diallo, F. Saliou, Things you should know about fronthaul, in: Proceedings of the European Conference on Optical Communication, 2014. [2] T. Pfeiffer, Next generation mobile fronthaul and midhual architectures, IEEE J. Opt. Commun. Netw. 7 (11) (2015) B38–B45. [3] J. Bartelt, P. Rost, D. Wubben, J. Lessmann, B. Melis, G. Fettweis, Fronthaul and backhaul requirements of flexibly centralized radio access networks, IEEE Wirel. Commun. 22 (5) (2015) 105–111. [4] R.DuanK.XuJ.DaiY.CuiJ.WuY.LiY.DaiJ.Lin, Linearity improvement based on digital signal processing in intensity-modulated analog optical links incorporating photonic frequency downconversion, in: Proceedings of the Optical Fiber Communication Conference, Los Angles, Calofonia, 2012, pp. 1–3. [5] Y. Pei, K. Xu, J. Li, A. Zhang, Y. Dai, Y. Ji, J. Lin, Complexity-reduced digital predistortion for subcarrier multiplexed radio over fiber systems transmitting sparse multi-band RF signals, Opt. Express 21 (3) (2013) 3708–3714. [6] Y. Zhang, J. Li, H. Chen, C. Yin, Y. Dai, F. Yin, K. Xu, Clip-and-filter-based crest factor reduction and digital predistortion for WLAN-over-fiber links, IEEE Photonics Technol. Lett. 26 (23) (2014) 2315–2318. [7] X.N. Fernando, A.B. Sesay, Adaptive asymmetric linearization of microwave fiber optic links for wireless access, IEEE Trans. Veh. Technol. 51 (6) (2002) 1576–1596. [8] R. Zhu, X. Zhang, B. Hraimel, D. Shen, T. Liu, Broadband predistortion circuit using zero bias diodes for radio over fiber systems, IEEE Photonics Technol. Lett. 25 (21) (2013) 2101–2104. [9] R.ZhuX.Zhang, Broadband predistortion circuit design for electro-absorption modulator in radio over fiber system, in: proceedings of the Optical Fiber Communication Conference, San Francisco CA, March, 2014, pp. 1–3. [10] B.-H. Son, K.-J. Kim, Y. Li, C.,-I. Park, Y.-W. Choi, Simple electrical predistortion method using Schottky diode for radio-over-fiber systems, IEEE Photonics Technol. Lett. 27 (8) (2015) 907–910. [11] H.BlauveltH.Loboda, Predistorter for linearization of electronic and optical signals,
237
Contents lists available at ScienceDirect
Optics Communications journal homepage: www.elsevier.com/locate/optcom
crossmark
Mach-Zehnder modulator modulated radio-over-fiber transmission system using dual wavelength linearization ⁎
Ran Zhua,b, Ming Huic, Dongya Shena, , Xiupu Zhanga,b a b c
iCom Laboratories, School of Information Science and Engineering, Yunnan University, Kunming, Yunnan 650091, China iPhotonics Laboratories, Department of Electrical and Computer Engineering, Concordia University, Montreal, Quebec, Canada H3G 1M8 School of Physics and Electronic Engineering, Nanyang Normal University, Nanyang, Henan 473061, China
A R T I C L E I N F O
A BS T RAC T
Keywords: Microwave photonics Radio-over-fiber systems Linearization Fiber optic communications
In this paper, dual wavelength linearization (DWL) technique is studied to suppress odd and even order nonlinearities simultaneously in a Mach-Zehnder modulator (MZM) modulated radio-over-fiber (RoF) transmission system. A theoretical model is given to analyze the DWL employed for MZM. In a single-tone test, the suppressions of the second order harmonic distortion (HD2) and third order harmonic distortion (HD3) at the same time are experimentally verified at different bias voltages of the MZM. The measured spurious-free dynamic ranges (SFDRs) with respect to the HD2 and HD3 are improved simultaneously compared to using a single laser. The output P1 dB is also improved by the DWL technique. Moreover, a WiFi signal is transmitted in the RoF system to test the linearization for broadband signal. The result shows that more than 1 dB improvement of the error vector magnitude (EVM) is obtained by the DWL technique.
1. Introduction Distributed radio access network (RAN) based structure is mainly employed in present wireless networks. In distributed RAN, backhaul connects core network and baseband units (BBUs) through Internet protocol (IP) network. BBUs are connected to remote radio units (RRUs) which up-convert baseband signals to radio frequency (RF) signals and the RF signals are transmitted to antennas. Both BBU and RRUs are located in a cell site cabinet. However, data capacity increases continuously and seamless coverage is required for various situations, so more small and micro cells are deployed. The increasing cell sites raise the consumption of power and space. Therefore, the movement from distributed RAN to Cloud-RAN (C-RAN) architecture is a tendency [1–3]. C-RAN consists of two parts, back-haul and fronthaul network. The back-haul is used to connect BBUs and the fronthaul is introduced to connect BBUs and RRUs. The centralization can reduce the cost of deployment, maintenance, and power consumption, and increase the network flexibility. But C-RAN has strict requirements of latency and synchronization which are the main challenge. Moreover, the Common Public Radio Interface (CPRI) protocol runs between BBUs and RRUs demands the supports of multiple formats. Radio-over-fiber (RoF) transmission system can be a solution for the front-haul network. RF signals are transmitted through optical fibers in RoF systems because optical fiber has tremendous capacity. Some
⁎
functions of RRU can be moved to BBU so that RRU can be simplified. Thus, the network is further centralized and the total cost is reduced. Furthermore, it is easier to meet the requirements of CPRI because the signal processing functions are centralized in central office. Although RoF systems provide several advantages, they are susceptible to nonlinearities due to the high peak-to-average power ratio (PAPR). The nonlinearities need to be suppressed to improve the spurious-free dynamic range (SFDR). Many linearization techniques were studied and they can be divided into three major methods, i.e., digital linearization using digital signal processing [4–7], analog predistortion circuit (PDC) [8–13], and optical linearization [14–21] using optical components. Digital linearization has potential to suppress nonlinearities substantially, but it is inadequate in dealing with multiband signals. Although two concurrent bands linearization was studied, the complexity will be increased dramatically when more signals are in consideration [5]. Analog PDC is broadband and lowcost, but it suppresses only odd order intermodulations (IMDs) for broadband. Optical linearization provides huge bandwidth and can be used to suppress odd and even order nonlinearities. In this paper, dual wavelength linearization (DWL) technique is studied to suppress even and odd order nonlinearities simultaneously in a Mach-Zehnder modulator (MZM) modulated RoF transmission system. The requirements of the DWL for MZMs are theoretically analyzed and experimentally evaluated. Single-tone test is given to
Corresponding author. E-mail address: [email protected] (D. Shen).
http://dx.doi.org/10.1016/j.optcom.2016.09.066 Received 7 September 2016; Received in revised form 28 September 2016; Accepted 29 September 2016 Available online 05 November 2016 0030-4018/ © 2016 Elsevier B.V. All rights reserved.
Optics Communications 385 (2017) 229–237
R. Zhu et al.
Fig. 1. Schematic diagram and experimental setup of the DWL for MZM in an analog optical link.
measure the spurious-free dynamic ranges (SFDRs) and WiFi test is also given to verify the linearization of broadband signal. The rest of the work includes as follows. The principle and theory are presented in Section 2. The experiments of single-tone test and WiFi test are given in Section 3. The conclusion is given in Section 4. 2. Principle The schematic diagram of the DWL for MZM is given in Fig. 1. The main nonlinear source is assumed to be the MZM and all other components are assumed to be linear. Two lasers emit lights with different wavelengths which are coupled and transmitted to an MZM which is used for optical subcarrier modulation. The two lasers are incoherent with each other. A 180° hybrid is used for push-pull modulation of MZM. MZM is wavelength-dependent so the two lights experience different modulation characteristics. So different signals and nonlinearities are generated and carried by the two optical carriers. In photodiode (PD), the two optical carriers are demodulated incoherently and thus the sums of the signals and nonlinearities are obtained. The total signal and nonlinearities can be adjusted by tuning the power
Fig. 2. Measured optical transmission characteristics of the MZM.
Fig. 3. (a) Measured optical power of the 1640 nm laser for the maximum suppression of the HD2. (b) RF power of the RF output signals, (c) the HD2s, (d) the HD3s for DWL, 1553 nm and 1640 nm, respectively. The HD2 is maximally suppressed.
230
Optics Communications 385 (2017) 229–237
R. Zhu et al.
Fig. 4. (a) Enhancements of the RF output signal and the HD3 compared to using single 1553 nm laser. (b) Suppression of the HD2 compared to using single 1553 nm laser.
Fig. 5. (a) RF power of the HD2 and (b) RF power of the RF output signal and the HD3 using the DWL technique versus the optical power of the 1640 nm laser.
Fig. 6. Measured SFDRs with respect to HD2 and HD3 for (a) DWL, (b) 1553 nm, and (c) 1640 nm. The bias voltage is 0.1 V.
Two incident lasers are assumed to be laser A and laser B. The halfwave voltages for laser A and laser B are different because MZMs are wavelength-dependent. We assume the half-wave voltages for the two lasers are Vπ_A and Vπ_B and Vπ_A < Vπ_B. The two lasers are incoherent, so they can be demodulated separately in the PD. After photodetection, the total photocurrent can be given as follows:
ratio of the two lasers. If the nonlinearities carried by the two lights are antiphase, the nonlinear distortion can be suppressed. When the MZM is set to push-pull modulation, the optical transmission characteristic of the MZM can be described as
T=
⎡π ⎤⎫ Po 1 ⎧ = ⎨1 + cos ⎢ (V +Vb ) ⎥ ⎬ ⎣ Vπ ⎦⎭ Pi 2L ⎩
(1)
where T is the optical transmission of the MZM, Po and Pi are the output and input optical power of the MZM respectively, L is the insertion loss, Vπ is the half-wave voltage, V is the RF signal, and Vb is the bias voltage applied to the MZM. Using Taylor series, (1) can be expanded as
T=
1 1 + 2L 2L
∞
⎡ ⎛ π ⎞n ⎛ nπ πVb ⎞ V n ⎤ ⎟ cos ⎜ + ⎟ ⎥ ⎝ 2 Vπ ⎠ n! ⎦ ⎣ Vπ ⎠
∑ ⎢ ⎜⎝
n =0
I=
RPA RPB + + 2L 2L Vn ⎫ ⎬ n! ⎭
∞
⎧⎡
⎛
⎞
⎩ ⎣ 2L
⎝ 2
⎠
⎛
⎞⎤
⎝ 2
⎠⎦
∑ ⎨ ⎢ RPA mAn cos ⎜ nπ +mA Vb⎟+ RPB mBn cos ⎜ nπ +mB Vb⎟ ⎥ ⎪ ⎪
n =0
2L
⎪ ⎪
(3)
where PA and PB are the optical input power for laser A and laser B respectively, and R is the responsivity of the PD. mA and mB are the modulation indexes of the MZM for laser A and laser B respectively,
(2) 231
Optics Communications 385 (2017) 229–237
R. Zhu et al.
PD with the responsivity of 0.6 A/W from Discovery is employed. A wideband power amplifier (PA) ZVA-213+ is used to boost the photocurrent. It has the gain of 26 dB, output power at 1 dB compression point (P1 dB) of 24 dBm, noise figure (NF) of 3 dB, and bandwidth from 0.8 to 21 GHz. A spectrum analyzer (SA) is used to measure the output powers. The two wavelengths of the lasers are 1553 nm and 1640 nm. Measured optical transmission characteristics of the MZM are shown in Fig. 2. The maximum transmission points of the MZM are −2.7 V for 1553 nm and −2.5 V for 1640 nm. The minimum transmission points are 2.3 V for 1553 nm and 2.9 V for 1640 nm, respectively. So the half-wave voltages are 5 V for 1553 nm and 5.4 V for 1640 nm. Referring to the requirements of (4), the even order components can be suppressed when −0.2 V < Vb < 0.2 V and it is marked as Area 1 in Fig. 2. The odd order components can be suppressed when 2.3 V < Vb < 2.9 V and it is marked as Area 2. Fig. 7. Measured P1 dBs for DWL, 1553 nm, and 1640 nm at the bias voltage of 0.1 V.
i.e., mA= V
π
π_A
and mB= V
π
π_B
3.1. Single-tone test
. The nth order component in the photocurrent
Generally speaking, odd order IMDs have almost the same phase with the signal. But the phases of harmonics and even order IMDs fluctuate depending on the transmission distance. So the requirements of the suppression of harmonic distortions (HDs) are stricter than that for IMD. Therefore, single-tone test is conducted to suppress HDs. The RF signal generator is set to 7 GHz and 10 dBm. The bias voltage is swept to suppress the second order HD (HD2). The optical power of the 1553 nm laser is set to 0 dBm and the optical power of the 1640 nm laser is adjusted to obtain the maximum suppression of HD2. Back-toback is employed in the experiments. It is found that the HD2 can be suppressed for the bias voltage from −0.2 to 0.1 V. The measurement results agree with the theories in Section 2. The required optical power of the 1640 nm laser for the maximum suppression of the HD2 is given in Fig. 3(a). When the bias voltage of the MZM is increased, the required optical power of the 1640 nm laser is also increased. From −0.2 to 0.1 V, the required optical power is increased from −10 to 5.4 dBm. The reason is that the HD2 is the minimum when the MZM is biased at Vπ/2. When the MZM is biased at −0.2 V, the HD2 for 1553 nm is its minimum. And the HD2 for 1640 nm is its minimum when the MZM is biased at 0.2 V. When the bias voltage is increased from −0.2 to 0.2 V, the HD2 for 1553 nm is increased and the HD2 for 1640 nm is decreased. So the optical power of the 1640 nm laser has to be increased to balance the power of the two HD2s. Fig. 3(b) shows the measured RF output signals for DWL, 1553 nm, and 1640 nm, respectively. The power of the RF output signal for DWL is higher than that for 1553 nm and 1640 nm from −0.2 to 0.1 V. This is because odd order components for the two wavelengths are inphase in this bias range. So the RF signal is improved by the DWL technique. Fig. 3(c) gives the HD2s. The HD2 for DWL is suppressed when the powers of the HD2s for 1553 nm and 1640 nm are the same. It illustrates that the HD2s carried by the two optical carriers are
can be eliminated when the following condition is met. nπ
⎛ m ⎞n cos( 2 + mB Vb ) PA =−⎜ B ⎟ ⎝ mA ⎠ cos( nπ PB + mA Vb ) 2
(4)
The power and modulation indexes are positive and the power ratio of the two lasers is adjustable. For the even order components, Eq. (4) can be met when Vπ_A /2 < Vb < Vπ_B /2 . At this time, the even order components carried by the two optical carriers can cancel each other if their RF power is identical. Moreover, it can be found that the odd order components are increased in the meantime. For the odd order components, Eq. (4) can be met when Vπ_A < Vb < Vπ_B . So the odd and even order components cannot be eliminated simultaneously. However, if the power of the signal and nonlinearities are enhanced by the same level, the nonlinearities are suppressed relatively. This will be proved in later experiments. For a practical single mode fiber link, dispersion compensating fiber is required in the proposed method to compensate for the dispersion induced delay between the two wavelengths. 3. Experiments The RF output signals, second and third order nonlinearities of an RoF system are measured to verify the performance of DWL technique. Experimental setup is given in Fig. 1. A Sumitomo Osaka Cement 40 Gb/s C-band MZM is used. Two polarization controllers are used and adjusted to get maximum output optical power of MZM for each laser. The two lasers are combined and transmitted to the MZM through a 3 dB coupler. An RF signal generator generates RF signal. The MZM is set to push-pull modulation using a 180° hybrid. A 40 GHz
Fig. 8. Measured SFDRs with respect to (a) HD2 and (b) HD3 versus the bias voltage. The HD2 is maximally suppressed.
232
Optics Communications 385 (2017) 229–237
R. Zhu et al.
Fig. 9. Improvements of the SFDRs with respect to HD2 and HD3 using the DWL compared to using (a) the 1553 nm laser and (b) the 1640 nm laser.
Fig. 10. (a) Measured output P1 dBs for DWL, 1553 nm, and 1640 nm. (b) Measured improvements of the output P1 dB using the DWL compared to using the single laser.
antiphase from −0.2 to 0.1 V. It also shows that the HD2 for 1553 nm is increased with the bias voltage. Fig. 3(d) gives the measured third order HDs (HD3s). It can be seen that it is increased by the DWL. Moreover, when the bias voltage is increased, the optical power of the 1640 nm laser has to be increased as shown in Fig. 3(a). So the odd order components such as first and third order components are also increased with the bias voltage. The improvements of the RF output signal and HD3 are given in Fig. 4.(a). It shows that the power improvements of the RF output signal and the HD3 by the DWL are similar. This illustrates that the SFDR with respect to the HD3 can be improved. When the bias voltage is 0.1 V, the improvement of the RF output signal is 17.6 dB. Fig. 4(b) shows that the suppression of the HD2 is higher than 17 dB for the bias voltage from −0.2 to 0.1 V. And the highest suppression is 38.3 dB at the bias voltage of 0.1 V. To verify the dependence of the optical power ratio, the power of the 1640 nm laser is swept. The power of 1553 nm laser is fixed at 0 dBm. The RF input power is 10 dBm and the bias voltage is set to 0.1 V for the maximum suppression of the HD2. Fig. 5 shows the power of the HD2, RF output signal, and the HD3 using the DWL technique versus the optical power of the 1640 nm laser. In Fig. 5(a), the maximum suppression of the HD2 is achieved at 5.4 dBm. When the optical power is shifted to 4.6 dBm or 5.9 dBm, the suppression degrades by 10 dB. Fig. 5(b) shows that the power of the RF output signal and HD3 are increased with the power of the 1640 nm laser. This is because the odd order components carried by the two optical carriers are inphase with each other. And it can be seen 1 dB increase of optical power causes 2 dB increase of RF power. Then the SFDRs at 0.1 V bias voltage with respect to HD2 and HD3 are measured and given in Fig. 6 with respect to the HD2 and HD3 for DWL, 1553 nm, and 1640 nm, respectively. To simplify the explana-
tion, we define the SFDR with respect to HD2 as SFDR2 and the SFDR with respect to HD3 as SFDR3. Fig. 6(a) shows the HD2 is fourth order limited that means the HD2 is eliminated by the DWL. It also can be seen that RF output signal and the HD3 are both increased in power by the DWL. Fig. 6(b) and (c) show the SFDRs for 1553 nm and 1640 nm, respectively. Compared to Fig. 6(b), the SFDR2 and SFDR3 are improved by 38.4 dB and 12.1 dB, respectively. Despite the increase of the HD3, the SFDR3 is still improved. The reason is that the RF output signal and the HD3 are increased by the same level. So the SFDR is improved by the increase of the RF output signal. The measurement results agree with the analysis in Section 2. Compared to Fig. 6(c), the SFDR2 and SFDR3 are improved by 11.3 dB and 0.7 dB, respectively. It can be seen that the HD2 in Fig. 6(c) is third order limited. This is because the even order nonlinearities for 1640 nm are suppressed when the MZM is biased at its quadrature point. So the measured HD2 in Fig. 6(c) is a combination of the suppressed even order nonlinearities and optical beating of the odd order nonlinearities in the PD. Measured input and output P1 dBs for DWL, 1553 nm, and 1640 nm at the bias voltage of 0.1 V are given in Fig. 7. The input P1 dBs for DWL, 1553 nm, and 1640 nm are 16 dBm, 15.5 dBm, and 15.4 dBm, respectively. So the input P1 dB is improved by 0.5 dB and 0.6 dB. The output P1 dBs for DWL, 1553 nm, and 1640 nm are −26 dBm, −43.9 dBm, and −27.5 dBm, respectively. It can be seen that the RF output signal is improved by the 1640 nm laser. The output P1 dB is improved by 17.9 dB and 1.5 dB, respectively. This illustrates that the system is linearized by the DWL technique. Fig. 4(b) shows that the DWL can suppress the HD2 by more than 17 dB for the bias voltage from −0.2 to 0.1 V. To verify the linearization performance of the DWL, the SFDRs and P1 dBs are measured and given in Fig. 8 for bias voltage from −0.2 to 0.1 V. 233
Optics Communications 385 (2017) 229–237
R. Zhu et al.
Fig. 11. (a) Measured optical power of the 1640 nm laser for the maximum suppression of HD3. (b) RF power of the RF output signals, (c) the HD2, (d) the HD3 for DWL, 1553 nm and 1640 nm, respectively. The HD3 is maximally suppressed.
Fig. 12. (a) Suppressions of the RF output signal and the HD3 compared to using the 1553 nm laser. (b) Enhancement of the HD2 compared to the using the 1553 nm laser.
the optical power of the 1640 nm laser is enhanced so that the SFDR3 for 1640 nm is improved. The RF output signal is improved much more by the DWL at 0.1 V as shown in Fig. 3(b). So the SFDR3 achieves the maximum at 0.1 V. The improvements of SFDRs with respect to HD2 and HD3 using the DWL compared to using the single laser are given in Fig. 9. The comparison between using the DWL and using the 1553 nm laser is given in Fig. 9(a). The SFDR2 and SFDR3 are improved by more than 20.6 dB and 0.7 dB, respectively. The maximum improvements which are 38.5 dB for SFDR2 and 12.1 dB for SFDR3 are obtained at 0.1 V. In Fig. 9(b), compared to using the 1640 nm laser, the SFDR2 and SFDR3 are improved maximally by 36.9 dB and 10.8 dB respectively at −0.2 V. And the minimum improvements are 11.3 dB and 0.8 dB respectively at 0.1 V.
Measured SFDRs with respect to HD2 and HD3 versus the bias voltage of the MZM are given in Fig. 8(a) and (b), respectively. The optical power of the 1553 nm laser is fixed at 0 dBm. In Fig. 8(a), the SFDR2 for 1553 nm drops with the increase of the bias voltage because the HD2 for 1553 nm achieves the minimum at the quadrature point of −0.2 V. For the same reason, the HD2 for 1640 nm achieves the minimum at its quadrature point of 0.1 V. And the optical power for the 1640 nm laser is increased for the maximum suppression of HD2. So the SFDR2 for 1640 nm increases from −0.2 to 0.1 V. Compared to using the single laser, the SFDR2 is improved for the bias voltage from −0.2 to 0.1 V by the DWL. And the SFDR2 achieves the maximum at 0.1 V. Fig. 8(b) presents the measured SFDR3s. The HD3 is increased by the DWL for the range of bias voltage, so the SFDR3 is improved by the increase of the RF output signal. When the bias voltage is increased, 234
Optics Communications 385 (2017) 229–237
R. Zhu et al.
Fig. 13. Measured SFDRs with respect to HD2 and HD3 for (a) DWL, (b) 1553 nm, and (c) 1640 nm. The bias voltage is 2.6 V.
Fig. 14. Measured SFDRs with respect to (a) HD2 and (b) HD3 versus the bias voltage. The HD3 is maximally suppressed.
Fig. 15. Experimental setup of the WiFi test.
Fig. 2. The optical power of the 1553 nm laser is 0 dBm and the optical power of 1640 nm is swept to achieve the maximum suppression of the HD3. The HD3s carried by the two optical carriers need to be same in power to cancel each other referring to (3). The required optical power of the 1640 nm laser for the maximum suppression of the HD3 is given in Fig. 11(a). The required optical power is increased from −9 to 6.5 dBm while the bias voltage is increased from 2.3 to 2.8 V. The reason is that the HD3 carried by the 1640 nm laser drops to the minimum at its minimum transmission point of 2.9 V. And the HD3 carried by the 1553 nm laser increases from its minimum transmission point of 2.3 V. So the optical power of the 1640 nm laser has to be enhanced to make the RF power of the two HD3s carried by the two optical carriers same. Fig. 11(b) shows that the RF output signal is reduced by the DWL because the odd order components are suppressed in this range of bias voltage. Because of the same reason, Fig. 11(d)
Fig. 10(a) gives the measured output P1 dBs for DWL, 1553 nm, and 1640 nm. The output P1 dB using the DWL is increased with the bias voltage. It is enhanced by the increase of the optical power of the 1640 nm laser. For the system using the 1553 nm laser, the output P1dB changed little while the bias voltage is swept. And it can be seen that the output P1 dB using the DWL is higher than that using the single laser for the bias voltage range. Fig. 10(b) shows the improvements of the output P1 dB compared to using the single laser. When the bias voltage is 0.1 V, the output P1 dB is improved by 17.9 dB compared to using the 1553 nm laser. And when the bias voltage is −0.2 V, it is improved by 16.4 dB compared to using the 1640 nm laser. We sweep the bias voltage and the optical power of the 1640 nm laser to suppress the HD3. It is found that the suppression can be obtained for the bias voltage from 2.3 to 2.8 V. The range of the bias voltage agrees with the calculation and prediction given in (4) and
235
Optics Communications 385 (2017) 229–237
R. Zhu et al.
Fig. 16. Measured EVMs versus the received RF power for DWL, 1553 nm, and 1640 nm. (a) BTB. (b) 8 km SMF. The bias voltage is 0 V.
two lasers is obtained for the maximum suppression of the HD2 as measured in Fig. 3(a). At the receiver, a digital storage oscilloscope (DSO) 81204B from Keysight is employed to demodulate and verify the signal. The WiFi signal is tested in back-to-back (BTB) and 8 km single mode fiber (SMF) transmission. Fig. 16 shows the measurement results. Error vector magnitudes (EVMs) are measured versus the received RF power of the DSO. Fig. 16(a) depicts that the EVM is improved by more than 1.1 dB by the DWL compared to using the single 1553 nm laser for BTB. It also shows that the minimum EVM point is extended to higher received RF power level. The minimum EVM for DWL technique is achieved at the received RF power of −26 dBm while that for 1553 nm is achieved at −29 dBm. Fig. 16(b) shows the EVM is improved by 1 dB by the DWL compared by using the single 1553 nm laser for 8 km SMF transmission. Referring to [19], the delay caused by the dispersion in SMF may degrade the performance of DWL technique, so the length of the SMF needs to be adjusted slightly. In the measurement, the 8 km SMF is connected with 80 m SMF to maximize the linearization. The measurement results illustrate that the DWL technique can enhance the RF signal power and suppress the nonlinearities.
shows that the HD3 is suppressed too. It also shows that the two HD3s carried by the two optical carriers are almost same. However, it can be seen that the HD2 is increased by the DWL in Fig. 11(c). The suppressions of the RF output signal and the HD3 by the DWL compared to using the 1553 nm laser is given in Fig. 12(a). The RF output signal is suppressed by more than 9.8 dB for the bias voltage from 2.3 to 2.7 V and 3.3 dB at 2.8 V. The HD3 is suppressed by more than 11.7 dB and the maximum suppression is 21.8 dB at 2.7 V. However, Fig. 12(b) shows that the HD2 is increased by using the DWL technique. The enhancement is increased with the bias voltage because the optical power of the 1640 nm laser is increased with the bias voltage as shown in Fig. 11(a). The minimum enhancement is 1.3 dB at 2.3 V and the maximum enhancement is 16.8 dB at 2.8 V. The measured SFDRs when the MZM is biased at 2.6 V are given in Fig. 13. The HD3 is maximally suppressed by adjusting the optical power of the 1640 nm laser. It can be seen that the SFDR3 is improved by the DWL compared to using the single laser although the RF output signal is also suppressed by the DWL. But the SFDR2 is reduced because the HD2 is increased by the DWL. And Fig. 13 shows that the power of the RF output signals are low because the MZM is biased near its minimum transmission point. Fig. 14 gives the measured SFDRs with respect to HD2 and HD3 versus the bias voltage. In Fig. 14(a), the SFDR degrades when the DWL is used. The reason is that the HD2 is increased in this bias voltage range. And it can be seen that the SFDR2s are lower than 52 dB in this bias voltage range. This is because the bias voltage is near the minimum transmission point of the MZM. So the RF output signal is weak and the HD2 is strong. Fig. 14(b) shows that the SFDR3 is improved by using the DWL compared to using the single laser. And more than 6.3 dB improvement is achieved for the bias voltage from 2.3 to 2.8 V. The maximal SFDR3 using the DWL is obtained at 2.8 V. Moreover, it should be noted that the P1 dB degrades by using the DWL for this bias voltage range because the RF output signal is reduced.
4. Conclusion In this work, the DWL technique is studied to suppress odd and even order nonlinearities simultaneously in an MZM modulated RoF transmission system. The theoretical requirement of the linearization is derived. In the single-tone test, the concurrent suppressions of the HD2 and HD3 are experimentally verified for a range of the bias voltage of the MZM. The experimental results agree with the calculation. The measured SFDRs with respect to the HD2 and HD3 are improved simultaneously compared to using the single laser. The output P1 dB is also improved by the DWL technique. Moreover, a WiFi test is given to evaluate the linearization of broadband signals. More than 1 dB improvement of the EVM is obtained by using the DWL technique.
3.2. WiFi test
Acknowledgments
To verify the DWL technique for broadband signal, WiFi signal at 2.4 GHz is transmitted in the MZM modulated RoF transmission system. The experimental setup is shown in Fig. 15. At the transmitter, an arbitrary waveform generator (AWG) 7122B from Tektronix generates a WiFi signal at 2.4 GHz compliant with 802.11a. The OFDM signal consists of 64 subcarriers with 16-QAM occupying 20 MHz bandwidth. And the signal rate is 36 Mbit/s. The incident power from the AWG is −13.2 dBm. A PA ZVA-213+ and a broadband variable attenuator are injected between the AWG and the hybrid. The variable attenuator is adjusted to change the RF input power to the MZM. The MZM is biased near the quadrature point at 0 V. The optical power is 6 dBm for 1553 nm and 4.5 nm for 1640 nm. 1.5 dB power ratio of the
This work was supported in part by a Quebec Fonds de recherche du Québec – Nature et technologies (FRQNT) project in broadband photonic devices, an open project from State Key Laboratory of Millimeter Waves (K201615), a project from Henan Science and Technology Foundation (162102310479), and a project from the Science and Technology Foundation of Henan Educational Committee (16A510009), Yunnan high-tech top-talents recruitment project (2012HA005), a key project from Yunnan Science Foundation (2013FA027),and two projects from National Natural Science Foundation of China (61561051 and 61561052). 236
Optics Communications 385 (2017) 229–237
R. Zhu et al.
US patent 4,992,754, Feburary, 1991. [12] C. Han, S.-H. Cho. H.S. Chung, J.H. Lee, Linearity improvement of directlymodulated multi-IF-over-fibre LTE-A mobile fronthaul link using shunt diode predistorter, in: Proceedings of the European Conference on Optical Communication, Valencia, Paper We4-4-4, September, 2015. [13] R.ZhuZ.XuanY.ZhangX.ZhangD.Shen, Novel broadband analog predistortion circuit for radio-over-fiber systems, in: Proceedings of the IEEE International Microwave Symposium (IMS), Phoenix, Arizona, May, 2015, pp. 1–4. [14] B.M. Haas, T.E. Murphy, A simple, linearized, phase-modulated analog optical transmission system, IEEE Photonics Technol. Lett. 19 (10) (2007) 729–731. [15] B. Hraimel, X. Zhang, Performance improvement of radio-over fiber links using mixed-polarization electro-absorption modulators, IEEE Trans. Microw. Theory Tech. 59 (12) (2011) 3239–3248. [16] K.K. Loi, J.H. Hodiak, X.B. Mei, C.W. Tu, W.S.C. Chang, Linearization of 1.3 μm MQW electroabsorption modulators using an all-optical frequency-insensitive technique, IEEE Photonics Technol. Lett. 10 (7) (1998) 964–966. [17] B.M. Haas, V.J. Urick, J.D. McKinney, T.E. Murphy, Dual-wavelength linearization of optically phase-modulated analog microwave signals, J. Lightwave Technol. 26 (15) (2008) 2748–2753. [18] R.ZhuX.Zhang, Linearization of radio-over-fiber systems by using two lasers with different wavelengths, in: Proceedings of the IEEE International Microwave Symposium (IMS), Tampa Bay, Florida, June, 2014, pp. 1–3. [19] R. Zhu, X. Zhang, D. Shen, T. Liu, Analysis of dual wavelengths linearization technique for radio-over-fiber system with electro-absorption modulator, IEEE Trans. Microw. Theory Tech. 99 (2015) 1–11. [20] H.JungS.LeeS.Han, Gain saturated SOA cascaded EML analog optical transmitter, in: Proceedings of the IEEE International Topical Meeting Microwave Photonics (MWP), Awaji-Yumebutai, November, 2002, pp. 205–208. [21] D.H. Jeon, H.D. Jung, S.K. Han, Mitigation of dispersion-induced effects using SOA in analog optical transmission, IEEE Photonics Technol. Lett. 14 (8) (2002) 1166–1168.
References [1] A. Pizzinat, P. Chanclou, T. Diallo, F. Saliou, Things you should know about fronthaul, in: Proceedings of the European Conference on Optical Communication, 2014. [2] T. Pfeiffer, Next generation mobile fronthaul and midhual architectures, IEEE J. Opt. Commun. Netw. 7 (11) (2015) B38–B45. [3] J. Bartelt, P. Rost, D. Wubben, J. Lessmann, B. Melis, G. Fettweis, Fronthaul and backhaul requirements of flexibly centralized radio access networks, IEEE Wirel. Commun. 22 (5) (2015) 105–111. [4] R.DuanK.XuJ.DaiY.CuiJ.WuY.LiY.DaiJ.Lin, Linearity improvement based on digital signal processing in intensity-modulated analog optical links incorporating photonic frequency downconversion, in: Proceedings of the Optical Fiber Communication Conference, Los Angles, Calofonia, 2012, pp. 1–3. [5] Y. Pei, K. Xu, J. Li, A. Zhang, Y. Dai, Y. Ji, J. Lin, Complexity-reduced digital predistortion for subcarrier multiplexed radio over fiber systems transmitting sparse multi-band RF signals, Opt. Express 21 (3) (2013) 3708–3714. [6] Y. Zhang, J. Li, H. Chen, C. Yin, Y. Dai, F. Yin, K. Xu, Clip-and-filter-based crest factor reduction and digital predistortion for WLAN-over-fiber links, IEEE Photonics Technol. Lett. 26 (23) (2014) 2315–2318. [7] X.N. Fernando, A.B. Sesay, Adaptive asymmetric linearization of microwave fiber optic links for wireless access, IEEE Trans. Veh. Technol. 51 (6) (2002) 1576–1596. [8] R. Zhu, X. Zhang, B. Hraimel, D. Shen, T. Liu, Broadband predistortion circuit using zero bias diodes for radio over fiber systems, IEEE Photonics Technol. Lett. 25 (21) (2013) 2101–2104. [9] R.ZhuX.Zhang, Broadband predistortion circuit design for electro-absorption modulator in radio over fiber system, in: proceedings of the Optical Fiber Communication Conference, San Francisco CA, March, 2014, pp. 1–3. [10] B.-H. Son, K.-J. Kim, Y. Li, C.,-I. Park, Y.-W. Choi, Simple electrical predistortion method using Schottky diode for radio-over-fiber systems, IEEE Photonics Technol. Lett. 27 (8) (2015) 907–910. [11] H.BlauveltH.Loboda, Predistorter for linearization of electronic and optical signals,
237