Open Access
Add to BibSonomy
Abstract
We investigate the impact of photodetector (PD) aperture size and wire bond length on the performance of analog radio over multi-mode fiber (A-RoMMF) systems using high-frequency bands up to 28GHz. It is realized that transmission characteristics of A-RoMMF is affected by the change of frequency response due to the PD aperture size and the bonding wire length. We also demonstrate A-RoMMF transmission using an optical receiver mounted PD and a low noise amplifier (LNA) on one printed circuit board (PCB) to improve size, link gain, and transmission characteristics.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Mobile data traffic continues to explode. It is expected that mobile data traffic will grow at a compound average growth rate (CAGR) of 46 percent from 2017 to 2022 [1]. The rapid adoption of internet of things (IoT) devices will be one of the factors driving the increase.
The adoption of IoT devices requires a high capacity transmission wireless system [2,3]. To increase wireless system capacity, the super high frequency (SHF) bands, especially 28 GHz band is being reserved for the 5th generation mobile communication system; such bands also enable low latency transmission and massive device connectivity. However, high SHF band signals exhibit strong directivity and high propagation loss. They are also highly attenuated by real world objects such as buildings and the human body. This is likely to create more dead areas in terms of radio frequency (RF) reception for in-vehicle and in-building applications. RF signal relay technology would be useful to reduce these dead areas in such SHF band mobile access systems [4].
Radio over fiber (RoF) was proposed as a promising solution to relay RF signals into dead areas [4–7]. To eliminate as many dead areas as possible, reducing the cost, size, and power consumption are key goals.
An analog radio over fiber (A-RoF) system with simple configuration and no analog to digital converters has been studied to reduce the installation cost [7,8]. It also has the advantage of low latency transmission, which is a basic 5G mobile communication requirement, because it can relay RF signals without complicated digital signal processing. Direct modulation via a vertical cavity surface emitting laser (VCSEL), which can simplify the module structure, has been studied to reduce the cost of RoF [8–11]. For transmission distances less than a few hundred meters (in-vehicle systems being a typical example), RoF systems that use multi-mode fibers (MMFs) of glass or plastic, which can have large cores, are studied [10–15].
However, the link gain of A-RoMMF transmission systems employing VCSELs would not be large enough given the strong attenuation of the received radio-wave during the propagation through the air. An A-RoMMF system that can support weak input RF signals is required. No detailed studies on improving A-RoMMF link gain and its effect to the transmission characteristics have been published as far as we know. This work focuses on link gain improvement in the optical receiver module. Section 2 studies the impact of PD aperture diameter and bond wire length on link gain.
To improve the link gain, adding a LNA is also the solution. Generally, the LNA has narrow bandwidth. Even though the broad bandwidth of digital signal transmission would suffer some degradation, it does not pose a problem for analog signal transmission. To adapt the LNA for use in the A-RoMMF optical receiver module, a bias-tee should be inserted between the PD and the LNA. Mounting them on one PCB without electrical connectors between them yields improved compactness and frequency response. Section 3 demonstrates the effect of mounting a PD, a bias-tee, and a LNA on one PCB for small A-RoMMF optical receiver module.
2. Effect of aperture size of PD and bonding wire length
2.1 Theory of gain
RoF gain is defined as Eq. (1) [11,16]:
where Pin is the input RF power, Pout is the output RF power, ηfd is the optical fiber to detector coupling efficiency, and γd is PD responsivity.For high SHF band transmission, the frequency response must be considered because the gain attenuation is large. Accordingly, a frequency response parameter is added to Eq. (1). The modified equation is Eq. (2):
where FPD is the frequency response of the PD.PD diameter impacts not only the fiber to the PD coupling efficiency, ηlf, but also PD frequency response, FPD, for high frequency transmission due to parasitic capacitance. Larger diameters offer larger ηlf until the PD aperture size exceeds the spot size. However, the parasitic capacitance continues to increase, and FPD could be degraded [17].
Our system sets a lens between the optical fiber (MMF) and the PD as shown in Fig. 1. The optical coupling loss between the MMF and the PD, which impacts coupling efficiency, is related to spot size after the lens and the PD aperture diameter. If the optical profile after the lens has a Gaussian distribution, the relations are Eqs. (3) and (4):
Fig. 2. Calculated plots of (a) optical coupling loss versus the spot size for five PD aperture sizes and (b) the optical coupling loss and the gain change of RoF versus the aperture size for spot size of 9 µm.
Considering the parasitic capacitance is important for high frequency operation as mentioned above. Figure 3(a) and Fig. 3(b) plot the simulated relative gain of the electrical circuit without the optical coupling loss at 28 GHz, and the resonance frequency by the equivalent circuit shown in Fig. 4 and Table 1, respectively. C1 in Fig. 4 was changed from 64 fF to 140 fF to simulate performance at each PD aperture size. The relative gain is normalized using that attained with PD aperture size of 15 µm and spot size of 6 µm. The relative gain of the electrical circuit decreased as PD aperture size increased because of drop in resonance frequency, even though the optical coupling loss is reduced. Thus it is important that the aperture size be considered when assessing the optical coupling loss and the effect of the parasitic capacitance.
Fig. 3. Simulated plots of (a) relative gain versus PD aperture size at 28 GHz and (b) resonance frequency versus PD aperture size.
Table 1. Parameter of equivalent circuit.
Figures 5(a) and 5(b) show the simulated relative gain at 28 GHz including the calculated optical coupling loss in Fig. 2, and the relative gain against aperture size with spot size of 9 µm. The relative gain is normalized using that attained with PD aperture size of 15 µm and spot size of 6 µm. As the results show, PD aperture size φ of 20 µm yields better gain than 15 µm for spot sizes above 7 µm the reverse is true for spot sizes under 7 µm.
Fig. 5. Simulated plots of (a) relative gain including the optical coupling loss at 28 GHz versus spot size for five PD aperture sizes, and (b) relative gain versus PD aperture size at spot size of 9 µm.
2.2 Measurement setup
The measurement setup shown in Fig. 6 was used to assess the frequency response, the error vector magnitude (EVM) and signal to noise ratio (SNR) as a function of PD aperture size. Measured gain includes the optical coupling loss between the optical fiber and the PD. We used a vertical cavity surface emitting laser (VCSEL) whose characteristics are shown in Fig. 7. The VCSEL was driven at 7mA bias current and was directly modulated. The converted optical signal was passed to MMF and attenuated to +2.2dBm at before lens by an optical attenuator not to exceed for absolute maximum rating of input power of PD. The optical power was used for all measurements. The signal was passed to the GaAs-based pin PD chip in Table 2 via a lens. Total MMF length was 2.5 m. This length is so short that the condition can be considered as back to back (BtB) condition. The bonding wire length between the PD and the wire on PCB was 220 µm. The spot size after the lens was 9 µm (FWHM).
Table 2. Characteristics of PD.
2.3 Results and analysis of frequency response
The measured frequency responses for five PD aperture sizes are shown in Fig. 8. They include the VCSEL module’s frequency response. Aperture sizes under 20 µm yield larger gain because the resonant frequency is shifted to higher frequencies as is shown in Fig. 3(b). However, the relative gain with 15µm aperture size is smaller than that with 20 µm aperture size. This is because the optical coupling loss effects on it due to increment of optical loss at 15 µm aperture size as mentioned in subsection 2.2. The measured relative optical coupling loss is almost same with simulated result as shown in Fig. 9(b). As Eq. (2) shows, the change in gain is twice the optical loss considered in units of dB. Thus the increase in optical loss has a larger impact than the gain improvement due to the parasitic capacitance change with aperture size. Figure 10 shows measurement results and the calculated result including optical coupling loss at 28 GHz. Both are for aperture size of 20 µm. As the figure shows, the measurement results closely approach the calculated results. It is considered that the slight difference between them is due to the imperfect Gaussian distribution of the optical profile after the lens and the simulation’s failure to consider the resistance change created by the PD aperture size change.
Fig. 9. (a) Tolerance curve for each PD aperture size and (b) relative optical coupling loss for alignment error of 0 µm.
The result is case of our system. The best aperture diameter of PD would change by kinds of lens and the transceiver module because the spot size could be changed by them. In terms of commercialization, practical levels of fabrication error should also be considered. Aperture size should be decided by the fabrication system and the tolerance curve of lens to PD coupling as shown in Fig. 9(a). The loss at 0 µm is 0.5 dB to 0.7 dB, which is higher than simulated data, because simulated data does not contain material absorption loss dB of lens, loss at mirror, and Fresnel reflection.
Bonding wire length, between the PD and PCB trace, is also important in improving the gain of the optical receiver because it changes the parasitic parameters which yield a change in FPD. We measured the frequency response versus the wire bond length between the PD and the PCB trace using the measurement setup shown in Fig. 6.
Figure 11 shows the effect of wire length on the frequency response with PD aperture size of 20 µm. The shortest wire has the best gain at 28 GHz. When considering wide bandwidth transmission, slope flatness also is important. The wire length of 220 µm offers excellent gain and flatness. Summarizing Fig. 10, gain of -37.4 dB is achieved with PD aperture size of 20 µm and bonding wire length of 220 µm.
2.4 Effect on transmission characteristics
Subsection 2.3 elucidated the influence of the PD aperture size and bonding wire length on basic RF transmission characteristics. This subsection addresses their effect on communication performance. We measured the error vector magnitude (EVM) to evaluate the performance in the BtB condition using the measurement setup of Fig. 6. Input RF signal used Pre-5G signal (64 QAM modulation with the center carrier frequency of 28 GHz, bandwidth of 90 MHz, subcarrier space of 75 kHz, and subcarrier number of 1201). LNA before vector signal analyzer was removed because the module has LNA. Figure 12 plots the measured frequency spectrum of the RF signal inputted to the optical link at -10 dBm amplitude.
Figure 13(a) shows the EVM measurement results for the five PD aperture sizes with bonding wire length of 220 µm. PD aperture size of 20 µm has smallest input amplitude less than 8% EVM in small input amplitude region (region A). The region A demonstrates signal amplitude [8,13]. So the gain improvement leads to the input amplitude improvement if the noise level change is small. On the other hand, the EVM of all samples are almost same at the larger input RF amplitude region (region C) because it is mainly determined by the distortion induced by nonlinear VCSEL behavior [8,13]. As a result, the input RF amplitude range for EVM (dynamic range) values of less than 8% is largest with PD aperture size φ of 20µm.
Figure 14(a) shows the change of SNR (ΔSNR) for each PD aperture size compared to PD aperture size of 35 µm at input RF amplitude of -20 dBm. And Fig. 14(b) shows the change of signal level (ΔS) and noise level (ΔN) for each PD aperture size compared to PD aperture size of 35 µm at input RF amplitude of -20 dBm. The SNR changes with the aperture size, and the PD aperture size of 20 µm has the largest SNR. This is because ΔN is smaller than ΔS as shown in Fig. 14(b). ΔS is almost same with change of the gain in Fig. 10. On the other hand, ΔN is smaller than it. It is assumed that the noise levels are affected by the LNA before vector signal analyzer. The noise level of PD aperture size of 35 µm is more affected by the noise of LNA than that of PD aperture size of 20 µm because the noise level of PD aperture size of 35 µm is smaller than that of PD aperture size of 20 µm. Figure 15 shows calculated result of ΔN for each PD aperture size compared to PD aperture size of 35 µm. It is calculated by Eq. (5). Nout is the noise level, G is the gain of the LNA, Namp is the noise level of the LNA and Nin is input noise level to LNA. We used G of 24 dBm, Namp of -97 dBm and Nin of -115 dBm to -121 dBm for each aperture size. The calculated result is almost similar to result of Fig. 14(b).
Figure 13(b) plots the EVM measurement results for the three bonding wire lengths between the PD and the PCB trace. The bonding wire length of 220 µm sample attains the largest dynamic range.
This result has important meaning for analog RF transmission because demonstrates that not only can the gain be improved but also the input RF amplitudes that offer EVM range of less than 8%, neither of which is considered in digital transmission assessments.
3. Gain improvement by adding LNA after PD
3.1 Effect of adding LNA on frequency response
To improve the link gain, a LNA could be added after the PD. To add LNA to A-RoMMF optical receiver module, a bias-tee is required between the PD and LNA.
The frequency response can exhibit a periodic fluctuation due to a standing wave between the PD and the LNA with the following circuit because their impedance mismatch causes signal reflection. This fluctuation has to be attenuated to realize wide bandwidth RF signal transmission because otherwise the gain would vary too strongly with each subcarrier and signal quality would be degraded.
The static wave frequency is expressed by
where f is frequency, L is length between the PD and the LNA, n is a natural number, c is velocity of light, and ɛr is dielectric constant of the PCB. From these equations, the peak to peak difference in frequencies Δf is given by Figure 16 plots the Δf versus distance between the PD and the LNA as estimated by Eq. (8). To suppress standing waves from 0 - 30 GHz, the PD and LNA should be separated by less than 2.6mm when ɛr is 3.5.3.2 Experimental setup to assess effect of adding LNA
The experimental setup shown in Fig. 17 was used to assess the frequency response, the EVM and SNR as a function of separate distance between PD and LNA. Basic configure is same with setup in subsection 2.2. Input optical power to the samples was set to +2.2dBm not to exceed for absolute maximum rating of input power of PD.
We tested a discrete sample with a PD, a bias-tee (Anritsu V251) and a LNA module connected by connectors (Sample A). The aperture size of the PD was φ20 um. And the bonding wire length was 220 um. The characteristics of the LNA module is shown in Table 3. PD and LNA separation distance of Sample A was 100 mm.
Table 3. Characteristics of LNA module.
We also tested two samples with a PD chip, a bias-tee and a LNA chip mounted on one PCB (Sample B and Sample C). Difference between Sample B and Sample C was separation distance between PD and LNA. PD and LNA separation distances of Sample B and Sample C were 45 mm and 1 mm, respectively. Their PD aperture size and bonding wire length were same with Sample A.
A picture of Sample C is shown in Fig. 18. Its dimensions are length of 35 mm (not including connector), width of 24 mm, and height of 12 mm. Package size is a very important factor in many applications, such as vehicles. The losses at the electrical connector and transmission loss of the PCB are approximately 0.3 dB and 0.3–0.5 dB/cm at 28 GHz, respectively. Mounting the PD, the bias-tee and the LNA on one PCB offers improved gain and fewer electrical connectors and shorter circuit lengths.
3.3 Results and analysis of frequency response
The frequency response of each sample at the BtB condition is shown in Fig. 19(a). This result includes the frequency response of the VCSEL module. Sample A and Sample B exhibited periodic fluctuation in frequency response, the period of which matches the values calculated by Eq. (8) as shown in Table 4. As Fig. 19(b) shows, the frequency responses of a sample without PD and without LNA show no periodicity of frequency. This is evidence that supports the idea that resonance is due to impedance mismatch at the PD and the LNA.
Fig. 19. (a) Frequency response of A-RoMMF with Sample A, Sample B and Sample C and (b) frequency response of receiver module without PD or LNA.
Table 4. Peak to peak frequency difference (Δf).
Sample C showed no periodicity in frequency response from 15 GHz to 30 GHz which matches the calculated result shown in Table 4. The gains for each module at 28 GHz are shown at Table 5. The maximum gain of -12.1 dB is achieved by Sample C.
Table 5. Gain at 28 GHz.
3.4 Transmission characteristics of optical module mounted chips on one PCB
We evaluated the EVM of an A-RoMMF system using Sample A, Sample B and Sample C. All measurements were performed in the BtB condition. The results are shown in Fig. 20. The optical receiver module with the LNA has minimum EVM of 1.9% and dynamic range of more than 34 dBm.
Fig. 20. (a) EVM of receiver module with mounted PD, bias-tee and LNA on one PCB and (b) constellations of Sample C at 0 dBm input RF amplitude.
Figure 21(a) shows the change of SNR (ΔSNR) by distance between the PD and the LNA at input RF amplitude of -20 dBm. The SNR changes by distance between them, and the shortest distance sample (Sample C) has the largest SNR. This is because change of noise level (ΔN) is smaller than that of signal level (ΔS) as shown in Fig. 21(b). ΔS is almost same with change of the gain in Fig. 19(a). On the other hand, ΔN is smaller than it. It is assumed that ΔN is affected by the LNA of the samples as mentioned in subsection 2.4.
Sample C, which is mounted a PD, a bias-tee and a LNA on one PCB with the shortest distance between the PD and the LNA, has the best dynamic range. The lower input amplitude region is shifted because of the gain improvement. That means the Sample C can support smaller RF amplitude compared to Sample A and Sample B.
For Sample C, the EVM of a 100 m MMF was also measured. The frequency response of the tested MMF is shown in Fig. 22. The EVM after 100 m MMF transmission is shown in Fig. 23. Input optical power to the module is adjusted by optical attenuator to +2.2dBm. As the results show, Sample C achieved EVM of less than 8% after 100 m MMF transmission. The minimum EVM and dynamic range were 3.8% and 27 dB, respectively.
Fig. 23. (a) EVM of 100 m MMF A-RoMMF transmission with Sample C and (b) constellations input RF amplitude of 0 dBm after 100 m MMF A-RoMMF transmission with Sample C.
As far as we know, this study has introduced and tested the world’s smallest A-RoMMF optical receiver module with mounted PD, bias-tee and LNA on one PCB.
4. Conclusion
This study is the first to show in detail the effect of PD aperture size on the frequency response of high SHF band signals and the transmission characteristics of A-RoMMF. PD aperture size and wire length between PD and PCB have a large impact on high SHF band RF signal transmission. PD aperture size has to be determined by considering not only the optical coupling but also the parasitic capacitance because it affects frequency response and transmission characteristics (EVM). We also demonstrated link gain improvement by adding LNA and mounted a PD and it on one PCB. We revealed the benefits of it, and the importance of the distance between the PD and LNA with regard to A-RoMMF performance. Not only the frequency response but also the transmission characteristics are affected by the distance. To support further reductions in input RF amplitude, an amplifier could be placed in front of the VCSEL and optical fiber characteristics could be optimized.
Funding
Ministry of Internal Affairs and Communications.
Acknowledgments
This work was conducted as the “Research and development for expansion of radio wave resources,” supported by the Ministry of Internal Affairs and Communications (MIC), Japan.
Disclosures
There are no conflicts to declare.
References
1. Cisco, “Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update, 2017–2022 white paper,” (2019).
2. ARIB 2020 and Beyond Ad Hoc Group, “Mobile Communications Systems for 2020 and beyond,” Oct. 2014
3. T. Kawanishi, A. Kanno, and H. S. C. Freire, “Wired and Wireless Links to Bridge Networks,” IEEE Microwave 19(3), 102–111 (2018). [CrossRef]
4. T. Aiba, A. Kanno, N. Yamamoto, T. Kawanishi, and T. Wakabayashi, “High SHF band RF signal relay employing radio over multi-mode fibers,” The 40th PIERS, Toyama, Japan, August 2018.
5. S. Kaurl, M. Srivastava, and K.S. Bhatia, “Radio over Fiber Technology – A Review,” International Conference of Technology, Management and Social Sciences (ICTMS-15), 85–89 (2015).
6. W. P. Ng, T. Kanesan, Z. Ghassemlooy, and C. Lu, “Theoretical and experimental optimum system design for LTE-RoF over varying transmission span and identification of system nonlinear limit,” IEEE Photonics J. 4(5), 1560–1571 (2012). [CrossRef]
7. H. Kim, “RoF -based Mobile Fronthaul Network Implemented by Using Directly Modulated Laser,” Asia Communications and Photonics Conferences (ACP), China, 10-13 Nov. 2017.
8. P. Hartmann, X. Qian, A. Wonfor, R. V. Penty, and I.H. White, “1-20 GHz Directly Modulated Radio over MMF Link,” in International Topical Meeting on Microwave Photonics (MWP2005), 95–98 (2005).
9. J. V. Kerrebrouck, H. Li, S. Spiga, M. C. Amann, X. Yin, J. Bauwelinck, P. Demeester, and G. Torfs, “10Gb/s Radio-Over-Fiber at 28 GHz Carrier Frequency Link Based on 1550 nm VCSEL Chirp Enhanced Intensity Modulation after 2 km Fiber,” OSA Technical Digest (online) (Optical Society of America, 2018), paper W1F.1 (2018).
10. N. J. Gomes, A. Nkansah, and D. Wake, “Radio-Over-MMF Techniques-Part I: RF to Microwave Frequency System,” J. Lightwave Technol. 26(15), 2388–2395 (2008). [CrossRef]
11. C. Carlsson, A. Larsson, and A. Alping, “RF Transmission Over Multimode Fibers Using VCSELs-Comparing Standard and High-Bandwidth Multimode Fiber,” J. Lightwave Technol. 22(7), 1694–1700 (2004). [CrossRef]
12. H. Kuboki and M. Matsuura, “Modal Dispersion and Feed Light Crosstalk Mitigations by Using Center- and Offset-Launching for Optically-Powered Radio-over-Multimode Fiber Systems,” OFC 2018, San Diego, CA, USA, 11-15 March 2018
13. H. K. Al-Musawi, W. P. Ng, Z. Ghassemlooy, C. Lu, and N. Lalam, “Experimental Analysis of EVM and BER for Indoor Radio-over-Fiber Networks using Polymer Optical Fiber,” 20th European Conference on Networks and Optical Communications (NOC 2015), London, UK, 2015.
14. A. M. J. Koonen and M. García Larrodé, “Radio-Over-MMF Techniques—Part II: Microwave to Millimeter-Wave Systems,” J. Lightwave Technol. 26(15), 2396–2408 (2008). [CrossRef]
15. A. Kanno, N. Yamamoto, and T. Kawanishi, “Radio over fiber-based radio relay link for indoor and in-car applications towards 5G/IoT era,” OSA Technical Digest (online) (Optical Society of America, 2017), paper S3E.2 (2017)
16. C. Cox III, E. Ackerman, R. Helkey, and G. E. Betts, “Technique and Performance of Intensity-Modulation Direct-Detection Analog Optical Links,” IEEE Trans. Microwave Theory Techn. 45(8), 1375–1383 (1997). [CrossRef]
17. N. Dupuis, D. Kuchta, F. E. Doany, A. Rylyakov, J. Proesel, C. W. Baks, C. L. Schow, S. Luong, C. Xie, L. Wang, S. Huang, K. Jackson, and N. Y. Li, “Exploring the limits of high-speed receivers for multimode VCSEL-based optical links,”, Optical Fiber Communication Conference (OFC), San Diego, CA, USA, M3G. 5 2014.
