The FUTON Prototype: Broadband Communication through Coordinated Multi-Point using a Novel integrated Optical/Wireless Architecture

Fabian Diehm , J org Holfeld , Gerhard Fettweis , Nathan J. Gomes , David Wake , Anthony Nkansah , Esther Lopez Casariego
Vodafone Chair Mobile Communications Systems, Technische Universit at Dresden, Germany {fabian.diehm, joerg.holfeld, fettweis}@ifn.et.tu-dresden.de Broadband and Wireless Communications Group, University of Kent, United Kingdom ACORDE Technologies, Santander, Spain
AbstractThe growing relevance of wireless communications has been driving research and development to enable costefcient support of very high data rates to be delivered to a large number of users. The high bandwidth targets for future networks are well reected in the ITUs call for IMT-Advanced. As interference poses the main limitation in todays networks, cooperative signal processing (often referred to as coordinated multi-point, CoMP) is seen as a key enabler to achieve these targets. However, cooperation among base stations in traditional cellular infrastructures is problematic because interconnections (referred to as backhaul) are often very limited in capacity and investigations on more exible and cost-efcient future architectures are in progress. A promising concept was proposed in the European research project FUTON. It is based on the use of radio-over-ber (RoF) technology to create a distributed antenna system (DAS) that is capable of supporting the high requirements of future wireless networks. In the scope of the project, a prototype system was built as a proof of concept. In this contribution, we present measurement results from the prototype system that demonstrate the potential of the FUTON architecture. Index TermsFUTON, optical/wireless integration, RoF, distributed antenna systems

I. I NTRODUCTION The rapid growth of mobile communications has led to a highly increasing demand for wideband high data rate services. Many services enabled by wired networks are now pushing into the wireless domain and are generally characterized by high quality of service requirements like high bandwidth and low latency. Examples of such sophisticated services are video streaming and video conferencing. The provision of these services to a large number of users is currently the main driver for research and development of advanced network technologies. While the industry is trying to meet increasing user demands, it is simultaneously facing diminishing revenues, as users become more and more accustomed to the cheap availability
This work has been performed in the framework of the ICT project FP7 ICT-2007-2015533 FUTON, which is partly funded by the European Union. The authors would like to acknowledge the contributions of their colleagues in FUTON, although the views expressed in this paper are those of the authors and do not necessarily represent the project.

of wireless services. Thus, from an economic perspective, next generation systems are required to meet high user demands, while keeping the total cost of ownership (TCO) of networks as low as possible. The objective of true mobile broadband wireless access has also been addressed by the ITU in their call for IMTAdvanced (IMT-A). In particular, IMT-A species data rates of up to 1Gbps for low mobility users (e.g. pedestrians) and up to 100Mbps for high mobility users (e.g. traveling by car or train) [1]. The technologies considered for the succession of the currently deployed third generation wireless cellular networks (LTE and WiMAX) will not be able to meet the IMTA requirements yet. Steps towards a radio access technology that is able to realize these targets are now being undertaken with the advent of LTE-Advanced. Since spectrum is a scarce resource, the IMT-A requirements can only be met if spectral efciency is signicantly increased. As high spectral efciency requires high degrees of spectrum reuse in cellular systems, interference strongly increases and puts a tight cap on the achievable system capacity. To overcome this limitation, cooperative signal processing is seen as a key technology for LTE-Advanced. In cooperative networks, signals transmitted or received by spatially separated antenna sites are jointly processed, enabling so called coordinated multi-point (CoMP) techniques. These techniques allow for mitigating interference or even exploiting it for gains in spectral efciency. In traditional cellular systems, cooperation is difcult to realize as backhaul capacity is often scarce. Thus, new architectures are of interest to support the requirements of future wireless technologies. In the wired domain, optical bers play a vital role in delivering high bandwidth data services. While the core and metro networks already rely on optical links completely, traditional copper-lines to the subscribers premises are gradually being replaced by deployments of Fiber-To-The-Curb (FTTC), Fiber-To-The-Building (FTTB) or even Fiber-To-The-Home (FTTH). The triumph of optical bers in many segments of telecommunication originates from its ability to support very high data rates in a cost efcient manner. This is mainly

RAU fiber MT

CU

Fig. 1.

System concept

due to the mediums inherent broadband and low-attenuation characteristics, which allow transportation of data over large distances. The European research project FUTON (Fiber Optic Networks for Distributed, Extendible, Heterogeneous Radio Architectures and Service Provisioning) has proposed a novel integrated wireless/optical system architecture that enables wireless services to meet the IMT-A requirements in a costefcient manner through the usage of centralized CoMP techniques. The FUTON consortium is comprised of 16 partners from industry and academia from various European countries, bringing together strong expertise in wireless and optical technology [2]. To demonstrate the capabilities and the feasibility of the FUTON architecture, a small prototype has been realized as a proof of concept. In Section II, we describe the FUTON architecture in more detail, before we introduce the prototype system and its components in Section III. Finally, Section IV discusses measurement results that were obtained with the prototype system, before the paper is concluded in Section V. Notation: We denote matrices with boldface uppercase letters. The operator (.)H , denotes transpose conjugate of the operand. If H represents a matrix, H[a:b,c:d] selects rows a to b and columns c to d of it. Furthermore, we denote an identity matrix of size N N as IN . II. T HE FUTON A RCHITECTURE In the FUTON architecture, low complexity remote antenna units (RAUs) are connected via optical links to a central unit (CU), where joint signal processing and resource management is performed. Figure 1 depicts the system concept. As the RAUs form a distributed antenna system (DAS), this infrastructure allows deployment of various CoMP techniques such as joint decoding in the uplink to greatly enhance throughput and coverage in the service area. Furthermore, the architecture is not specic to a particular radio access technology (RAT) so that multiple wireless standards can make use of the infrastructure. This not only enables efcient cross-layer algorithms to be implemented at the CU, but also cross-system algorithms and mechanisms for vertical handovers between the systems. The support of legacy systems also makes the concept very attractive from a TCO point of view. The broadband (up to 100 MHz) RAT to utilize the FUTON architecture for

high spectral efcieny is referred to as Distributed Broadband Wireless System (DBWS) and is the focus of the prototype. Outside the FUTON project, the concept of remote radio heads (RRHs) has already garnered much attention. As it allows for exible deployments and greater CAPEX and OPEX savings due to low complexity hardware and low site acquisition costs, it has found widespread adoption. In this context, two specications have emerged recently, namely, Common Public Radio Interface (CPRI) [3] and Open Base Station Architecture Initiative (OBSAI) [4]. They specify digitized serial interfaces between the RRHs and a digital base station (DBS) containing the baseband processing functions of traditional base stations. Currently, the specications cover UMTS, WiMAX and LTE. The links between the DBSs and the RRHs use digital optical transmission. Still, current standards do not make use of the increased degrees of freedom offered by distributed antennas. Unlike the above concepts, the RAUs in the FUTON architecture are connected to the CU via Radio-over-Fiber (RoF) technology, where the light in the bers is modulated by analog radio signals. The advantages of analog ber transmission include support of higher bandwidths and more signicantly, higher cost-efciency [5]. Signals from the CU to different antennas of a site and to different RAUs can be multiplexed onto the same optical ber by employing Wavelength Division Multiplexing (WDM) and Subcarrier-Multiplexing (SCM). Another advantage of RoF technology is that it shifts complexity from the RAUs to the CU, as digital-to-analog conversion (DAC) and IQ-modulation components are not required at the RAUs [5]. On the downside, analog optical transmission limits the achievable dynamic range performance in wideband systems and non-linear distortions degrade the signal quality. However, dynamic range requirements can be reduced through uplink power control and automatic gain control [5] and the nonlinear distortions are often negligable (cf. Section IV) and can be compensated in the digital baseband as e.g. described in [6]. For more details on the FUTON architecture and the FUTON project, please refer to [2], [7], [8]. III. T HE FUTON P ROTOTYPE The DBWS that was specied in the FUTON project comprises a collection of MAC-, PHY- and cross-layer algorithms that utilize the FUTON architecture to deliver the high data rates envisaged for future wireless systems. One of the key performance enablers in this respect is the usage of CoMP techniques to exploit spatial degrees of freedom. Besides demonstrating the benets that can be achieved by such techniques in combination with the distributed radio/ber architecture, the FUTON prototype was built to show the general feasibility of the concept for practical implementation. To demonstrate the architectures ability to exploit spatial degrees of freedom to cancel or avoid interference for gains in spectral efciency, the following scenario was selected for demonstration. The prototype features downlink data

Central Unit

Uplink RAU1 MT1 MT IF Sorbas Digital BB Unit

JPUe Sorbas Digital BB Unit

RAUe

2 receiver noise vector with covariance matrix nn = n INR . This model is applicable for each of the OFDM subcarriers. The MMSE precoder, which employs the Wiener lter approach is computed as

E/O

O/E

W = HH
Sorbas Digital BB Unit RAU2 JPUe RAUe MT IF E/O O/E MT2

2 H HH + n I

(2)

with =
2 ETx /x H H ( HH (HH + 2 I)2 ) n

(3)

Uplink

Fig. 2.

Demonstration setup

transmission from two single-antenna RAUs to two singleantenna mobile terminals (MTs). The users are jointly served by both RAUs, employing spatial division multiple access (SDMA), i.e. both users are served on the same time and frequency resources. In the compound system, there are thus NT = NR = 2 transmit and receive antennas, respectively. In this setup, each user can be supplied with one data stream. To separate the users spatially, an MMSE precoding lter is applied at the central unit. In the downlink, pilots and data are transmitted to the MTs, whereas the uplink is used to feed channel state information from the MTs back to the CU to adapt the precoding lter. Figure 2 depicts all components of the demonstration setup. In the following, we describe the main functional blocks of the setup in more detail. JPUes, RAUes, MT-IFs and antennas have been custom designed and manufactured for the prototype. A. Baseband processing The digital baseband processing for CU and MTs has been implemented on the Signalion Sorbas digital baseband platform that has basic compliance with LTE Release 8. The basic physical layer parameters are summarized in Table I.
Bandwidth FFT size Subcarriers for trans. Sampling rate Subcarrier spacing Transmit Time Interval (TTI) OFDM symbols/TTI Subcarriers per resource block Total available resource blocks Modulation Schemes 20 MHz 2048 1200 30.72 MHz 15 kHz 1ms 14 12 100 QPSK,16-/64-QAM

where the scalar limits the transmit power according to the sum transmit power constraint ETx . The value of needs to be estimated at the receiver to equalize the data stream. For this purpose, the 3GPP-LTE pilot structure was modied to include precoded pilots. As in 3GPP-LTE the 1st, 8th and 5th, 12th OFDM symbols within a 1ms transmit time interval (TTI) (with a total of 14 OFDM symbols) are used for pilots to estimate the channel to the two transmit antennas. The pilots have a spacing of 6 subcarriers, which leads to a total of 8 pilots in a physical resource block (PRB) that is dened as 14 OFDM symbols in time and 12 subcarriers in frequency domain. However, unlike the LTE standard, the other subcarriers in the above mentioned OFDM symbols are not used for data, but contain precoded pilots [9]. These are = HW. Equalization used to estimate the effective channel H is then performed as x =R x+n , H (4)

where R = diag (r1,1 r2,2 ) with /( h j,j rj,j = h j,j

2 2 + n ) 1 . j

(5)

TABLE I PHY-L AYER T RANSMISSION PARAMETERS

For downlink transmission, the following frequency at system model is considered to explain the baseband signal processing steps: x =R H W x+n , (1)

where x, x C21 are the transmitted and estimated user data vectors, respectively, H CNR NT is the channel matrix, W CNT NR is the precoding matrix and n CNR 1 is the

All signal processing steps that are performed at the MTs are depicted in Figure 3. Before channel estimation and equalization, the data burst is detected by coarse timing estimation from the incoming baseband samples. In the next step, the coarse CFO compensation reduces phase drifts and carrier frequency offsets (CFOs) which occur due to unsynchronized local oscillators at the MTs. Finally, the ne timing estimation chooses the proper sample window, before the conversion into frequency domain is carried out. Through estimation the mobile terminal j = 1, 2 acquires the local channel state information (CSI) H[1:2,j ] of a set of allocated PRBs. These channel estimates are sent back to the central unit in the uplink direction. At the CU, the local channel knowledge of both MTs is combined to obtain the compound channel matrix H. In our implementation, each estimated channel coefcient is quantized with 8 bits per complex dimension at the MTs. The entire feedback message (containing all channel estimates) is then sent back to the CU using the uplink direction, where each MT allocates distinct physical resources. Note that a separate uplink which uses a direct link to the CU at 2.5 GHz is used for channel feedback in the FUTON prototype. By changing the contents of the feedback messages, the precoding behaviour of the system can be changed. For the conducted measurements, two different modes were used: The

Coarse timing estimation

Coarse CFO compensation

Fine timing estimation

DFT

Channel estimation H DeMux Channel ~ estimation H PDCCH decoding

Feedback with CSI Channel equalization R

Uplink

sufcient for the IFs used. The radio over ber links have been separately tested with other wireless signals, such as emulated 802.11 signals, to ensure that they meet the required PHY layer performance. Confer to Section IV for measurement results of the complete baseband to baseband link. D. RAUe The RAUe is the electrical part of the RAU. At the RAUe, the detected signal from the photodiode receiver is demultiplexed from the IF subcarrier multiplex, and then each demultiplexed signal is frequency translated to the 4 individual RF signals at 3.5 GHz to be fed to the antennas. As stated above, only one 3.5 GHz downlink signal is required here. The RAUe thus contains locked oscillators (locked to a multiple of the frequency reference signal generated at the JPUe, which is also transmitted over the optical link), frequency converters (mixers), lters and ampliers. Monitoring functions are also provided. The RAUe also contains uplink RF electronics to convert received 3.5 GHz signals to the uplink IFs used, and the necessary amplication and ltering, but the uplink has not been used in this demonstration. E. MT-Interface The main function of the MT-Interface is to frequency translate the 3.5 GHz signals to baseband. Synthesizer, frequency conversion, amplication and ltering functions are used. F. Antennas Dual polarized patch antenna arrays [10] operating in the 3.5 GHz band are used at the RAUs. For this paper, each RAU uses only one polarization. The arrays have been designed to provide a beam suitable for illuminating a street corridor. Small, planar Inverted-F Antennas (PIFAs) [11] with a broadband input match designed to offer the capability of a frontend Band Pass Filter (BPF) in transceivers were used at the MTs to receive the 3.5 GHz signals. The range of operation was designed to be wide because it was felt that a mobile terminal might need to be capable of connecting to different radio access technologies. The MT antennas are also largely omnidirectional. Both RAU and MT antennas have been designed for low cost and ease of fabrication. IV. M EASUREMENT S ETUP & R ESULTS A. Radio-over-Fiber Links The full RoF downlink, comprising JPUe, laser module, 10 dB optical attenuator, photodiode module and RAUe, was tested using a vector signal generator (VSG) and vector signal analyser (VSA). Error vector magnitude (EVM) was measured as a function of input power using a 64-QAM IEEE802.11n signal with a bandwidth of 40MHz and 128 OFDM subcarriers (MCS7). This signal was the closest in complexity to the signals from the Sorbas baseband platform using the VSG and VSA test equipment available. The results are shown in Figure 6, where the measured EVM performance is compared to a baseline case where the VSG and VSA are linked using a short coax cable. For the RoF link, at low levels of input power

x ^

Decoding

Modulation and coding scheme

Fig. 3.

MT signal processing steps

full precoding mode, where all channel coefcients are taken into account for the calculation of W. Henceforth, we refer to this mode as cooperative mode, as both user data streams are transmitted from both RAUs. In the second mode, MT1 and MT2 only feed back the channel coefcients of RAU1 and RAU2, respectively (the other coefcient is set to zero in the feedback message). We refer to this mode as non-cooperative mode, as in effect, the user streams are transmitted separately from the two RAUs. This mode thus mimics the behaviour of conventional cellular systems with frequency reuse 1, where each base station only transmits data to its own assigned user. Finally, to obtain the transmit data at the MTs, the equalized received symbols are decoded with a Turbo-Decoder. Information about the modulation and coding scheme used, is obtained through the evaluation of the Physical Downlink Control Channel (PDCCH), that uses a specic part of the downlink spectrum. B. JPUe At the JPUe (electronic part of the Joint Processing Unit), up to 4 baseband channels can be multiplexed to different intermediate frequencies (IFs) for transport over the radio over ber link. The JPUe thus contains frequency synthesis, mixer, lter and amplier functions. The amplication is set to drive the optical link in a linear operating regime. A frequency reference signal is also transmitted, and so is a control and management information channel (on a separate IF subcarrier). In the uplink direction, the JPUe will translate up to 4 different received IF signals back to baseband, feeding each resultant signal to a different uplink output. It also provides for the uplink control and management channel. The 4 channels are there to provide for feeding separate antennas (e.g. sectorization) and/or different antenna polarizations. In this paper, only downlink transmission has been used, and only one antenna (and one polarization) is employed. C. Optical Links The optical links consist of an electrical to optical converter (E/O) in the form of a laser diode, a 100m length of single mode ber and an optical to electrical (O/E) which is a PIN photodiode. The laser diode is directly analog-modulated at the IF generated at the JPU. As stated above, the link is used in a generally linear regime - the drive signal to the laser is not high enough to cause serious distortion. The modulation bandwidths of the laser and photodiode receiver are easily

the EVM increases due to noise (reducing carrier to noise ratio) and at high levels of input power the EVM increases due to distortion. The distortion limit at around -15 dBm input power is in this case due to the power ampliers used in the RAU and higher input power could be accommodated with more expensive ampliers. Any other viable alternatives to the RoF link, such as antennas directly connected to the base stations, or digitized links similar to the CPRI or OBSAI specications would show inferior performance to the baseline case due to power amplier nonlinearity and carrier to noise ratio reduction through signal loss or quantization noise. The EVM penalty of about 1% is expected and not a serious impairment; for example Figure 6 shows that there is a 20 dB range of input power over which the EVM remains below 5%. This is also within the design limits for the RoF link [12], since the downlink transmit power can remain relatively xed and the uplink power can be controlled within the radio parts of the DBWS. In [12], it was shown that the RoF link has minimal impact on wireless range as a consequence of uplink power control. In the complete prototype setup, the input power to the optical transmission chain is -22 dBm and -24 dBm for JPUe 1 and 2, respectively. Thus, the links can be seen as being virtually transparent. B. Complete Setup For the measurments conducted in this paper, the FUTON prototype system was setup in a lab environment. The equipment as described above can be seen in Figure 4. Measurement data was collected for two different positions of the RAUs in the lab. These can be seen in Figure 5. For each of these setups, the RAU antennas were directed towards the middle of the lab. The mobile terminals were then moved to various locations within the room to collect measurement data. In total, 240 measurements were obtained for each MT. For each measurement, incoming baseband samples for 100 TTIs were collected once for the cooperative precoding mode and once for the non-cooperative precoding mode with a time delay of 2 seconds between the modes. In each TTI, 10 consecutive PRBs (120 subcarriers 14 OFDM symbols) were allocated to both MTs for data transmission and the transmit power at the output ports of the RAU-antennas was measured to be 8 dBm. After collecting measurement data, the baseband samples were evaluated by applying the same processing as implemented for the MTs in software. Using the channel estimates from the measurements and the estimated noise power (measured on unused subcarriers) we can depict the exerienced user link qualities in terms of the link SNR (Figure 7), where the SNR values on the abscissa represent the measured SNR from MT1 and MT2 to RAU1 and RAU2, respectively and the values on the ordinate represent the measured SNRs of each MT to the corresponding residual RAU. As we can see, similar links were observed for both measurement setups. Thus, we do not distiguish the setups in the following. The observed channels can be classied into land asymmetric channels 2 groups: symmetric channels A

Fig. 4. Equipment (left: all components, lower left corner: MT-IF and MTantenna, right: RAU and RAU-antenna)

CU RAU1 1/2 6m

RAU2

MT2 MT1 RAU2 8m 2

Fig. 5.

Demonstration setups 1 & 2

l , where the link to the MTs respective RAU is stronger B than the link to the interfering RAU. To estimate the maximum throughput that could be achieved for each MT at a certain measurement position, the CU was programmed to send to each UE with different modulation and coding schemes (MCS) every TTI. Twenty different MCS where selected and cyclicly transmitted (see Table II)1 . In this way, each MCS was observed 5 times for every measurement. The achieved maximum throughput for a measurement (100 TTIs) was then obtained by multiplying the data rates associated with each individual MCS with the decoding success rate (number decoding successes/5) and selecting the maximum. As the measurements were obtained in a lab environment, measured channels were very static during each measurement.
Modulation QPSK 16 QAM 64 QAM Code rates 0.27, 0.41, 0.53, 0.68, 0.76, 0.86 0.50, 0.60, 0.67, 0.77, 0.84, 0.89 0.47, 0.53, 0.62, 0.66, 0.71, 0.74, 0.81, 0.84

TABLE II U SED MODULATION AND CODING SCHEMES

The statistics of the throughput measurements (both users) for both precoding modes and channel types are depicted in Figure 8 in the form of a cumulative distribution function (CDF). From the gure, we can see that the MMSE precoding (cooperative mode) provides a large overall throughput gain compared to non-cooperative precoding. The MMSE precoder is able to mitigate the mutual interference of the RAUs, so that both MTs can receive data at high data rates using the
1 Note that coding is applied over all allocated subcarriers (120) and data OFDM symbols in a TTI (9). This yields a code block size of 1080 log2 (M ), where M is the size of the modulation alphabet.

15

10 EVM (%)

P(user throughput <= X)

RoF link baseline

1 0.8 0.6 0.4 0.2 0 B cooperative A noncooperative

0 45

40

35

30 25 Input power (dBm)

20

15

10

X (Mbps)

Fig. 6. EVM as a function of input power for the full RoF link and a short length of coax cable (baseline).

Fig. 8. User throughput CDF for both users sharing a frequency allocation of 1.8 MHz (10 consecutive PRBs)

50
link SNR (interfering RAU) [dB]

45 40 35

Setup 1 Setup 2

answered. For example, the high downlink performance that was demonstrated in this paper comes at the cost of signicant overhead in the uplink. Practical systems will thus have to nd good tradeoffs between CSI quality and uplink overhead. Furthermore, outdoor measurements must conrm the feasibility of the concept in typical cellular propagation scenarios.
A

ACKNOWLEDGEMENTS
B

30 25 25

30

35 40 45 link SNR (reference RAU) [dB]

Observed user channels

50

The authors would like to thank Ainoa Navarro-Caldevilla for her signicant support with the software tools employed and Sven-Einar Breuer for his great help in setting up the equipment and collecting measurement data. R EFERENCES
[1] [2] [3] [4] [5] Recommendation ITU-R M.1645, http://www.itu.int/ITU-R FUTON website, http://www.ict-futon.eu CPRI Specication v4.0, http://www.cpri.info OBSAI Specication v2.0, http://www.obsai.com D. Wake, S. Pato, J. Pedro, E. L opez, N. Gomes, P. Monteiro, A Comparison of Remote Radio Head Optical Transmission Technologies for Next Generation Wireless Systems, LEOS 2009 Annual Meeting, Belek-Antalya, Turkey, 4-8 October, 2009 A. Hekkala, M. Lasanen, Performance of Adaptive Algorithms for Compensation of Radio over Fiber Links, Wireless Telecommunication Symposium (WTS), Prague, Czech Republic, 22-24 April, 2009 S. Pato, P. Monteiro, N. Gomes, A. Gameiro, T. Kawanishi, NextGeneration Distributed and Heterogeneous Radio Architectures: the FUTON Project, Asia-Pacic Microwave Photonics Conference, Beijing, China, 22-24 April, 2009 G. Heliotis, I.P. Chochliouros, G. Agapiou, Fibre optic networks: the case of the FUTON programme, The Journal of The Institute of Telecommunications Professionals, volume 2, part 3, pp.113-118, 2008 J. Holfeld, V. Kotzsch and and G. Fettweis, Order-recursive precoding for cooperative multi-point transmission, in International ITG/IEEE Workshop on Smart Antennas (WSA2010), Bremen, Germany, 23.-24. February 2010 I.G. Garcia Zuazola, J.C. Batchelor, A. Nkansah, D. Wake, N.J. Gomes, Microcell MIMO RoF System using Low Cost Array Antennas and Improved DFB Matching Lines, Loughborough Antennas & Propagation Conference, LAPC 2009, Loughborough, pp. 373-376, Nov 2009 I.J. Garcia Zuasola, J.C. Batchelor, J.M.H. Elmirghani, N.J. Gomes, UWB PIFA for Simplied Transceivers, Electronics Letters Volume 46, Issue 2, pp.116-118, Jan 2010 D. Wake, A. Nkansah, N. J. Gomes, Radio Over Fiber Link Design for Next Generation Wireless Systems, J. Lightwave Tech., vol. 28, pp. 2456-2464, 2010.

Fig. 7.

same time and frequency resources. As the non-cooperative mode reects the signal processing that is done in conventional cellular systems with frequency reuse 1, the measurements conrm the potential performance gains offered by the centralized FUTON concept. By comparing the results for the two different channel types, we can make a second observation: As expected, cell-edge users (i.e. users between two RAUs) lbenet more from the joint processing than cell-center A l users B (i.e. users close to one RAU). Note, that the system throughput calculates as the sum of the user throughputs depicted here. As throughput results will generally scale close to linear with the allocated bandwidth (up to 100 MHz in the FUTON concept), the measured results render the envisaged high data rates of the FUTON concept feasible. Adding more RAUs and users will further increase spectral efciency. V. C ONCLUSION By building a small prototype system of the FUTON concept, we were able to validate its general feasibility. Our measurement results show that centralized joint processing enabled by the proposed RoF infrastructure can provide signicant gains when compared to conventional cellular systems. However, before we can come to a nal conclusion on the concept, there are still many questions that need to be

[6] [7]

[8] [9]

[10]

[11] [12]