Abstract: In this paper, we experimentally investigate the throughput of IEEE 802.11n 2x2 multiple-input-multiple-output (MIMO) signals in a radio-over-fiber-based distributed antenna system (DAS) with different fiber lengths and power imbalance. Both a MIMO-supported access point (AP) and a spatial-diversity-supported AP were separately employed in the experiments. Throughput measurements were carried out with wireless users at different locations in a typical office environment. For the different fiber length effect, the results indicate that MIMO signals can maintain high throughput when the fiber length difference between the two remote antenna units (RAUs) is under 100 m and falls quickly when the length difference is greater. For the spatial diversity signals, high throughput can be maintained even when the difference is 150m. On the other hand, the separation of the MIMO antennas allows additional freedom in placing the antennas in strategic locations for overall improved system performance, although it may also lead to received power imbalance problems. The results show that the throughput performance drops in specific positions when the received power imbalance is above around 13dB. Hence, there is a trade-off between the extent of the wireless coverage for moderate bit-rates and the area over which peak bit-rates can be achieved.
© 2015 Optical Society of America
Distributed antenna systems (DASs) using radio-over-fiber (RoF) links have been long recognized as a commonly used infrastructure solution in in-building environments [1–4]. Figure 1 depicts a RoF-based DAS [5,6] in a typical in-building environment. Multiple spatially separated remote antenna units (RAUs) are fiber-connected to a central unit (CU) where base station or access point (AP) facilities are placed. The DAS architecture inherently provides dynamic resource allocation and effective management of network elements by routing radio frequency (RF) signals through fiber cabling throughout the building or venue. As Fig. 1 shows, RoF-based DAS can be used to distribute wireless local area network (WLAN) signals and mobile telephony signals (3G, Long Term Evolution (LTE), etc.) . Compared to the conventional wireless communication network where different service systems use their own infrastructure, RoF-based DAS can combine multiple radio streams together, attractive for neutral host providers, due to its transparency, broad bandwidth and low attenuation features. The cellular, public safety radio, and WiFi signals can be transmitted on the same fiber. It can dramatically improve the infrastructure utilization. In addition, direct-modulated laser diode (LD) and PIN photodiode (PD) are used in this kind of application, whose prices have fallen considerably as production scales increased. Also, the cost of a RAU can be reasonably low, hence, it is possible to distribute a number of RAUs over large places of interest. Considering these advantages, the RoF-based DAS is cost efficient. In DAS applications, the base station facility in the CU is usually connected to multiple RAUs to extend the indoor wireless coverage of the base station and to share the hardware and bandwidth resource. By increasing the number of RAUs, the radio signal propagation distance is much shorter and radio signals are presumed to experience better channel quality since there is lower path loss, penetration loss and shadowing, which can significantly improve the system capacity and coverage. Furthermore, upgrading and re-configuration can be more easily performed in this centralized architecture. These features make RoF-based DAS attractive in achieving a cost-effective, scalable and flexible network.
With the shared RoF-based DAS infrastructure, it is preferred that WLAN signals are distributed in the same buildings and public venues as an important complement to telephony signals . IEEE 802.11 is a series of standards for implementing WLAN products . In the last five years, almost every newly-shipped WLAN product supports the IEEE 802.11n standard. More recently, we have seen a dramatic increase in the shipment of WLAN products compliant with IEEE 802.11ac. One distinct characteristic of IEEE 802.11n/ac standards is the use of multiple-input-multiple-output (MIMO) technique. Moreover, the latest mobile telephony standards also include the use of MIMO techniques. The integration of MIMO and RoF in a DAS architecture where any antenna port in CU could connect to any RAU, can improve the system capacity and extend the wireless coverage by increasing the received RF power and the antenna separation [10–13]. Therefore, the spatial streams could be flexibly allocated to any adjacent RAUs depending on the bandwidth demands, as seen in Fig. 1. One major issue with DAS when distributing WLAN signals over a fiber system  is the effect of different fiber lengths on the received OFDM signal. This was shown in a simulcast DAS, when a single AP (BS) feeds multiple RAUs. A similar situation should occur in Distributed-MIMO antenna systems (D-MIMO), where multiple transmitters and receivers are employed in the system. The CU or the clients need to align the OFDM signals and combine the different spatial streams before demodulation. For the design of a practical, re-configurable in-building network, it would be necessary to ensure equivalent fiber lengths for every link, which may prove difficult considering the reuse of the large installed base of fiber links.
Some parameters are helpful in discussing the issue above. OFDM in both IEEE 802.11g and 802.11n has a symbol duration of 4 µs. The cyclic prefix (CP) as the guard interval is 0.8 µs for 802.11g and 0.4 µs for 802.11n in default mode. The CP helps the receiver to recognize and align the signals. In an ideal wireless environment, it theoretically allows a path difference up to 400 feet (122m) for 802.11n , but one might expect the practical limit to be less than this. In a mixed optical-wireless channel, as the speed of light in vacuum is higher than in the fiber, the ideal maximum path-length difference will be less than 400 feet. We first identified and preliminarily analyzed the fiber-length different effect in WLAN-over-fiber DAS in . In this paper, we also analyzed the received power imbalance problem to reinforce the conclusion made easier, and correspond to the fiber-length results.
Another critical feature of MIMO systems is the spatial independence of the wireless channels, which generally requires an antenna separation of at least ¼ the radio signal wavelength . With optical feeding of the antennas, significantly larger antenna spacing can easily be accomplished, which opens up new possibilities for covering wider areas with MIMO [18–20]. Increasing distance between RAUs cannot only extend the wireless coverage, but also improve the throughput performance by reducing the correlation between the different spatial streams . However, large antenna separation may lead to a received power imbalance problem, especially in the positions very close to one of the RAUs. Recent standards employing MIMO techniques, such as LTE, can cope with up to 12-15 dB of power imbalance between different spatial streams at the mobile receiver . Normally in a traditional co-located antenna system (CAS), MIMO antennas are always installed in the same device, hence, the received power in mobile users would not exceed such a figure. However D-MIMO systems cannot guarantee such similar power levels at each position in a real environment with large antenna separation. Hence, both reducing correlation and decreasing the received power imbalance should be considered in the design a D-MIMO system.
In this paper, we experimentally investigate the throughput of an IEEE 802.11n WLAN 2x2 MIMO-over-fiber system with different fiber link lengths and different power levels, respectively. We compare the results with the performance of a spatial-diversity-supported AP. During WLAN operation, the protocol will change the data rate automatically according to the signal-to-noise ratio (S/N) and received power. Hence, the calculated coverage area for different data rates and its dependence on different received power sensitivities is also shown in the paper to explain the trade-off of between the area over which the peak bit-rate is obtained and the total coverage area; it is shown that the larger the distance between the distributed antennas, the larger the overall wireless coverage area, but that the peak-rate area will reduce. In summary, despite these benefits of RoF DAS we stressed above, there are issues with fiber length difference and power imbalance. This paper is about understanding the extent of such limitations and examining trade-offs.
The experimental setup for the proposed system is described in Section 2, whilst the investigation of the effects of fiber length difference is described in Section 3 and the effects of power imbalance described in Section 4. Conclusions follow in Section 5.
2. Experimental setup
Figure 2 shows the Dual-RAU experimental setup. The AP connects to a host PC using a 1 Gb/s Ethernet link. The two antenna ports of the AP are each connected to a RAU by RoF links. In the downlink, the RF signals transmitted from the AP are converted into optical signals by a distributed feedback (DFB) laser diode (LD) and the resulting optical signals are transported over standard single-mode fiber (SMF) links. At the RAUs, the optical signals are detected by photodiodes and electrically amplified for transmission over the wireless path. The same processes occur in the opposite direction in the uplink. The power at the input of the lasers was set at 2.5dBm, achieved by attenuators, at which point, the whole downlink can attain its best performance (measured by Error Vector Magnitude (EVM)). In terms of the RoF link setup, as the typical noise figure of RoF link is around 35dB, suitable RF amplifiers are needed in the RAUs to increase the signal power and hence improve signal to noise ratio. Amplifiers were biased at different voltages, so that they did not operate in their non-linear regions. The AP antennas are used in the experiments and the gain of the antenna is 9.5dBi.
Figure 3 shows the experiment layout: the CU and RAUs were placed in separated rooms to make sure that the signal from the AP cannot wirelessly arrive at the area covered by the RAUs. In the testing room, the distance between the two RAUs was set to either 4.4m or 9m. In Fig. 3, A, B, C indicate a row number and 1, 2, 3 a column number used to identify the mobile client positions at which measurements were carried out. The larger antenna spacing can reduce the spatial correlation and provide higher received RF power for the clients [10–13], which is the specific benefit from the DAS architecture. The testing area is a typical office room with varied furniture (desks, chairs and computers). The RAUs were mounted on trolleys with height extensions, so the signals were generally radiated above all furniture. The TCP throughput measurements were carried out using the commercial software Networx, and achieved by transferring large files between the Host PC and mobile laptops.
3. Fiber length difference in D-MIMO system
In this experiment, the RAU2 is set 4.4 m away from RAU1, and the output power at the antenna terminals (see Fig. 1) was set at 1 dBm. In all cases described in this section, the output power was set at this level, removing the effect of low received power. In order to evaluate the throughput performance with fiber-length difference effects in the RoF DAS, we consider both the MIMO transmission and spatial diversity transmission. Seven configurations were used in our experiment, as shown in Table 1 using different groups of fiber links.
3.1 Single client measurements
In Cases 1-7, we measured the throughput in each location shown in the layout of Fig. 3. At the CU, a Belkin IEEE 802.11n MIMO-supported AP is used in Cases 1-5, and a 3com spatial diversity AP is used in Cases 6-7.
Figure 4 shows the normalized downlink throughput measurement results, which are normalized by the maximum throughput each mode can reach. (It should be noted that this is 96 Mbps for the MIMO AP and 25 Mbps for the spatial diversity AP, and that due to the use of MIMO techniques, the ratio is more than twice the bandwidth ratio.) The measurements were implemented in each location inside the room with the different sets of fiber links for Cases 1-4. In Cases 2-4, the throughput dropped gradually as the fiber-length difference increased. This decrease is mostly because of the severe ISI, as the fixed fiber delay difference increases the interval between arrived signals from the two RAUs. At large intervals between the two spatial streams, the guard interval/CP is insufficient to isolate the current symbol, when the fiber link difference exceeds a certain distance. In Case 5, the throughput does not reduce much due to the extra fiber delay compared to Case 1, as shown in Fig. 5, as the fiber delays are approximately equal; only a small performance decrease is observed due to the protocol overhead (e.g., longer wait times for acknowledgments). This verifies that the throughput decrease observed in Fig. 3 is not caused by the fixed fiber delay, such as due to late arriving acknowledgments, but to the nonalignment between the two signal streams.
Figure 6 shows the result of throughput measurement with the spatial diversity AP, in which only one antenna port is employed at a time. We can observe that the normalized throughput only shows very small changes between Case 6 and Case 7. This is simply because only the antenna which receives higher power or SNR is selected, meaning that the antenna with poor channel conditions does not significantly affect the whole system performance. As the signal streams are not combined, their nonalignment does not cause a deterioration in performance.
3.2 Multiple clients measurements
An additional client was added to determine the effects of multiple clients accessing the AP from both antenna ports. The same model laptops with same Network Interface Cards (NIC) were used to avoid any inequalities of throughput distribution, as different manufacturers often implement different MAC configurations which may cause extremely variable contention performance . Additionally, the MIMO-supported AP was in 802.11n mode only, and no 802.11b&g clients were allowed to access the network. For the 802.11g diversity AP, no 802.11b clients were present within the association radius of the AP, thus, we can ensure the maximum throughput could be obtained.
Furthermore, we use the basic access mode in the measurement, and previous measurement checks around the room have assured us that there was no hidden node problem. The experiments were carried out with RAU1 and RAU2 in different position groups, where the positions are defined according to the representation in Fig. 3. G1 means that both of the laptops are in position A1 (A is the row number and 1 is the column number), G2 means both are in A2, G3 both in B4, and G4 means 1 laptop in A1, the other one in B4. The data recordings were collected at the same time, and after the throughput of the two laptops reached an equilibrium level.
Downlink and uplink normalized throughput performance of the 2-laptop situation are shown in Fig. 7 and Fig. 8. Similarly to the single user situation, we can see an obvious drop of the aggregate throughput both for downlink and uplink with the MIMO AP in the 25m-175m configuration compared to the 25m-25m configuration. However, only a slight decrease in normalized throughput is observed with the diversity AP, as only one antenna is working at one time. Although the results show a reduced effect of the fiber-length difference on the system using the diversity AP, it should be noted that the data throughput (bps) of the MIMO AP system, even at around 0.25 normalized throughput, is not lower than the diversity case.
4. Power imbalance analysis in MIMO-over-fiber distributed antenna system
4.1 Power imbalance effect
As the office in which the tests were carried out was not large enough to create path losses which would validate the power imbalance effect due to different antenna separation, the effect was created artificially by using different transmit power levels at each RAU. The output power levels before the antennas used in the experiment are listed in Table 2. By employing different output power, a larger received power imbalance can be seen in the experiment, contributed to by the initial output power difference and by different path loss. The ITU indoor path loss model is used to predict the path loss in the tested area, given by :24], we use N = 28 for the typical office area. Hence, we predict the two spatial stream’s path loss and calculate the received power imbalance for each position.
The relationship between D-MIMO throughput performance and received power imbalance can be observed in Table 3, Table 4 and Table 5. In the Tables, S represents Throughput, Pr1 and Pr2 represents Received Power (dBm) from RAU1 and RAU2 respectively, and Pi represents Power Imbalance (dB). The experiment is conducted with a single user at locations defined as in Fig. 3. From Sets 1-6, we find the throughput dropped in the positions where the received power imbalance exceeds around 13dB, these occurences being indicated by the gray shading. We normalized the throughput performance by the maximum throughput each mode can reach, as described in Section 3 (96 Mbps for the MIMO AP and 25 Mbps for the spatial diversity AP). For example, the received power imbalance in Position C4 is 9.43 dB in Set 4 (Table 4), and 14.03 dB in Set 6 (Table 5), and a normalized throughput of 0.916 can be reached at position C4 in Set 4 but only 0.672 in Set 6. Comparing Set 1 and Set 3 (Table 3), they should have the same power imbalance (8.2dB), but different power levels; the results re-inforce the indication that the drop in throughput is not because of low received power, but due to the imbalance. It can also be observed that the throughput performance does not fall any further than around 0.65 to 0.7, even when the received power imbalance problem becomes much more severe. For example, in Position A4, the power imbalance is as high as 30 dB in Set 6 and 19 dB in Set 2, but the normalized throughput for the two Sets are not very different, both being around 0.67.
RAU2’ and RAU1 is used in these experiments, and the distance between them is 9 m. The predicted received power from each RAU is also shown, obtained from the ITU path loss model. The receiver sensitivity for 64QAM is −63 dBm to −61 dBm for different coding rates in 802.11n . We can notice that the received powers from both RAUs are mostly above −63 dBm. For these positions, the drop in throughput must be due to the power imbalance, which leads to the receiver only processing one stream. For other positions, one of the received powers is lower than −63 dBm, see set 6 (Table 5) in particular. At position C4, for example, the received powers are −53.29 dBm from RAU1 and −67.32 dBm from RAU2’. However, the throughput at C4 remains at a similar level to other, larger power-imbalance positions, such as A4, where the received powers are −41.29 dBm from RAU1 and −71.10 dBm from RAU2’. There is no trend of reduced throughput as one of the received powers falls. Therefore, we deduce that both of these two situations are because of the large power imbalance, that the laptops cannot process the data with two streams. Besides, as the laptop can still receive one stream with good quality, it can work with only one branch. Hence, the commercial AP must adaptively change to a single-input single-output (SISO) system, according to the current channel state. To support the confirmation of this result, we obtained some simulation results using the software OPNET 17.5 to observe the throughput reduction that would result when changing from 2-stream transmission to 1-stream transmission for 802.11n. The results showed that one-stream transmission would result in 68% of the throughput of two-stream transmission.
It should be noted that any real environment will lead to received powers that are different to the results of a wireless propagation model. Also, there is no exact value that can be defined as a limitation for power imbalance for 2 × 2 MIMO system. Our results simply match the standard indicated limit range 12-15dB.
We also perform the same experiments for the Diversity-supported AP. The power imbalance problem does not affect the throughput performance, as seen in Table 5, as the Diversity method does not combine the signals from the two antennas. High throughput is obtained as long as there is at least one good link within its coverage.
4.2 Coverage discussion
Feeding MIMO signals to remote antennas via optical fibers enables large separation of these antennas and increases the coverage area. In Fig. 9, we simulate the throughput distribution in the room. The simulation set-up is the same as the Set 3 measurements. The dark blue indicate the peak bit-rates area, where the highest data rate modulation and coding scheme can be obtained, and the stars are the laptop positions drawn in Fig. 3. There are two restrictions for the peak bit-rates area (for IEEE 802.11n 64QAM): the first is that the received power for both antennas is above −63dBm, and the other is that the power imbalance should be below a certain level, with 13dB selected in this analysis. We can see that the simulated peak-rate area for Set 3 matches the experiment results of Table 3 very well.
From our results, we can consider more scenarios to make clearer the relationship between peak bit-rate area and the antenna separation. Different antenna separation arrangements are simulated in Fig. 10. The antennas are in the middle of the circle (red points), and transmit powers at both antennas are the same (we assume 10 dBm in this case). D is the distance between the RAUs. We assumes the bit-rate is mainly dependent on the received power, the noise layer is very low. The different circles represent different received power levels, 6 typical modelling and coding schemes  are chosen in our simulations, whose receiver sensitivities are −63dBm, −67dBm, −71dBm, −74dBm, −76dBm, −79dBm for each scheme. The dark blue area in the middle is calculated as the peak bit-rates area. It shows that D-MIMO can enlarge the wireless coverage at moderate bit-rates and improve the performance by decreasing the correlation between spatial streams (shown in [10–13]). However, the peak bit-rates area decreases as the antenna separation increases, as the large-received-power overlap area reduces and the power imbalance problem exists. Hence, there is a trade-off between the coverage and total throughput performance. As mentioned in Fig. 1, the proposed system can provide dynamic resource allocation, which means that any antenna port in the CU can connect to any RAU. Therefore, the radiation power and cell size (the distance between RAUs) could be adjustable depending on the bandwidth demands. In other words, we can install a number of RAUs in the building. At busy times (e.g. during daytime in an office area), the cell could be small using adjacent RAUs. At less busy times (e.g. during the night in an office area), the cell could be larger.
In this paper, we first compared the throughput performance of MIMO signals and spatial diversity signals in a RoF-based DAS with different length fiber links. The results show that MIMO signals can maintain high throughput performance when the fiber length difference between the two RAUs was under 100m but fell quickly when the fiber length difference was greater than this. For the spatial diversity signals, the high throughputs could be maintained even when the difference was 150m. Then, with regard to the effect of power imbalance of MIMO signals in a RoF-based DAS, results indicate that the throughput performance drops in specific positions when the received power imbalance is above 12-15dB. This leads to a trade-off between the wireless coverage area at moderate bit rates and the area over which peak bit-rates can be achieved. As we mentioned in the Introduction, the DAS architecture inherently provides dynamic resource allocation and effective management of network elements. However, if RAU B was being used with RAU A for a particular user, it might have a different transmit power than if it was being used with RAU C at a different time for a different user, because the path losses and resulting power imbalances would be different. A similar consideration can be applied for the fiber-length difference effect. Therefore, both the analysis of fiber-length difference effect and the received power imbalance problem can greatly assist the CU to make allocation decisions. The decisions can include which distributed antennas can cooperate and which cannot, and how much power each should transmit so that power imbalances can be reduced.
The authors are grateful to Anthony Nkansah for advice on the measurements, and to the China Scholarship Council for support for Y Fan. This work was in part of NSFC Program (61431003, 61302086), National 973 Program (2012CB315705), Specialized Research Fund for the Doctoral Program of Higher Education (2013005120007).
References and links
1. T. Alade, H.-L. Zhu, and H. Osman, “”The impact of antenna selection and location on the performance of DAS in a multi-storey building,” presented at IEEE Wireless Communications and Networking Conference (WCNC), Shanghai, China, April.2013, pp. 3213 - 3218.
2. M. J. Crisp, S. Li, A. Wonfor, R. V. Penty, and I. H. White, “Demonstration of radio over fibre distributed antenna network for combined in-building WLAN and 3G coverage,” in Proceedings of OFC/NFOEC 2007, Anaheim, California, United States, Mar.2007, paper JThA81. [CrossRef]
3. H.-L. Zhu, “Performance comparison between distributed antenna and microcellular systems,” IEEE J. Sel. Areas Comm. 29(6), 1151–1163 (2011). [CrossRef]
4. M. J. Koonen and E. Tangdiongga, “Photonic home area networks,” J. Lightwave Technol. 32(4), 591–604 (2014). [CrossRef]
5. J.-Z. Wang, H.-L. Zhu, and N. J. Gomes, “Distributed antenna systems for mobile communications in high speed trains,” IEEE J. Sel. Areas Comm. 30(4), 675–683 (2012). [CrossRef]
6. P. Monteiro, S. Pato, E. López, D. Wake, N. J. Gomes, and A. Gameiro, “Fiber optic networks for distributed radio architectures: FUTON concept and operation,” presented at IEEE Wireless Communications and Networking Conference Workshops (WCNC), Sydney, Australia, April.2010. [CrossRef]
7. C. Liu, J. Wang, L. Cheng, M. Zhu, and G.-K. Chang, “Key microwave-photonics technologies for next-generation cloud-based radio access networks,” J. Lightwave Technol. 32(20), 3452–3460 (2014). [CrossRef]
8. W. Qing, D. Debbarma, S. H. De Groot, I. Niemegeers, and A. Lo, “Cell switching mechanisms for access point sharing in WLAN over radio-over-fiber systems,” presented at IEEE 20th Symposium on Communications and Vehicular Technology(SCVT), Benelux, Namur, November.2013.
10. J. Q. Li., Y. T. Fan, H. Chen, K. Xu, Y.T. Dai, F. F. Yin, et al., “Radio-over-fiber-based distributed antenna systems supporting IEEE 802.11N/AC standards,” presented at 12th International Conference on in Optical Communications and Networks (ICOCN), Chengdu, China, July, 2013.
11. K. Zhu, M. Crisp, S. He, R. V. Penty, and I. H. White, “An experimental investigation of RoF-enabled MIMO DAS in a non-light-of-sight environment,” in Proceedings of the Conference on Lasers and Electro-Optics (CLEO), Baltimore, USA, May.2011, paper CFH3. [CrossRef]
12. K. Zhu, M. Crisp, S. He, R. Penty, and I. White, “MIMO system capacity improvements using radio-over-fibre distributed antenna system technology,” in Proceedings of OFC/NFOEC 2011, Los Angeles, USA, May.2011, paper OTuO2. [CrossRef]
13. T. Yamakami, T. Higashino, K. Tsukamoto, and S. Komaki, “An experimental investigation of applying MIMO to RoF ubiquitous antenna system,” in Proceedings of MWP 2008 Gold Coast, Australia, Oct, 2008. [CrossRef]
14. A. Das, M. Mjeku, A. Nkansah, and N. J. Gomes, “Effects on IEEE 802.11 MAC throughput in wireless LAN over fiber systems,” J. Lightwave Technol. 25(11), 3321–3328 (2007). [CrossRef]
15. “802.11n: The Next Generation of Wireless Performance,” Cisco Public document.
16. Y. T. Fan, A. E. Aighobahi, N. J. Gomes, and K. Xu, “Performance of commercial MIMO access point in distributed antenna system with different fiber lengths,” in Proceedings of MWP/APMP 2014, Sappro, Japan, Oct, 2014. [CrossRef]
17. H. Boelcskei, D. Gesbert, C. B. Papadias, and A. J. van der Veen, Space-time wireless systems: from array processing to MIMO communications (Cambridge University Press, 2006).
18. M. Sauer and A. Kobyakov, “Fiber-radio antenna feeding for MIMO systems,” in Proceedings of the Optical Fiber Communication & Optoelectronic Exposition & Conference, AOE 2008, Shanghai, China, Oct, 2008. [CrossRef]
19. A. Kobyakov, D. Thelen, A. Chamarti, M. Sauer, and J. Winters, “MIMO radio signals over fiber in picocells for increased WLAN coverage,” in Proceedings of OFC/NFOEC 2008, San Diego, USA, Feb, 2008, paper JWA113. [CrossRef]
20. G. S. D. Gordon, M. J. Crisp, R. V. Penty, and I. H. White, “Experimental evaluation of layout designs for 3×3 MIMO-enabled Radio-Over-Fiber distributed antenna systems,” IEEE Trans. Vehicular Technol. 63(2), 643–653 (2014). [CrossRef]
21. T. Yamakami, T. Higashino, K. Tsukamoto, and S. Komaki, “An experimental investigation of applying MIMO to RoF ubiquitous antenna system,” in Proceedings of MWP/APMP 2008, Gold Goast, Australia, Oct, 2008. [CrossRef]
22. E. M. Vitucci, L. Tarlazzi, F. Fuschini, P. Faccin, and V. Degli-Esposti, “Interleaved-MIMO DAS for indoor radio coverage: concept and performance assessment,” IEEE Trans. Antenn. Propag. 62(6), 3299–3309 (2014). [CrossRef]
23. P. Assimakopoulos, A. Nkansah, and N. J. Gomes, “Use of commercial access point employing spatial diversity in a distributed antenna network with different fiber lengths,” in Proceedings of MWP/APMP 2008, Gold Goast, Australia, Oct.2008. [CrossRef]
24. T. Chrysikos, G. Georgopoulos, and S. Kotsopoulos, “Site-specific validation of ITU indoor path loss model at 2.4 GHz,” IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks & Workshops, Kos, Greece, Jun, 2009. [CrossRef]
25. “Coverage or capacity – making the best use of 802.11 n,” Juniper Networks.