Field of the art

The present invention generally relates to a method and a system for lossless compression and decompression of baseband digital signals in distributed LTE- Advanced radio access networks, and more particularly to a method and a system for compression of the bandwidth required for fronthaul links in order to reduce the cost and complexity associated to cloud-RAN architectures.

Prior State of the Art

Long-Term Evolution (LTE) is the next step in cellular 3G systems, which represents basically an evolution of present mobile communications standards, such as UMTS and GSM [1 ]. It is a Third Generation Partnership Project (3GPP) standard that provides throughputs up to 50 Mbps in uplink and up to 100 Mbps in downlink. It uses scalable bandwidth from 1.4 to 20 MHz in order to suit the needs of network operators that have different bandwidth allocations. LTE is also expected to improve spectral efficiency in networks, allowing carriers to provide more data and voice services over a given bandwidth.

LTE-Advanced (LTE-A), an evolution of LTE, is being standardized in LTE

Release 10 and beyond. It is aimed at fulfilling IMT-Advanced requirements, whose capabilities go beyond those of IMT-2000 and include enhanced peak data rates to support advanced services and applications (100 Mbps for high mobility, and 1 Gbps for low mobility).

Some of the most advanced functionalities devised for LTE-Advanced require tight cooperation between multiple cells, as in Cooperative Multi-Point (CoMP), load balancing, enhanced Inter-Cell Interference Coordination (elCIC) and other Centralized Radio Resource Management (C-RRM) techniques. C-RRM techniques comprise any radio resource management procedure involving cooperation between several nodes at the radio access level. This kind of cooperation in general involves tight synchronization, as well as the deployment of high-capacity, low-latency links between them.

C-RRM techniques usually require heavy information exchange between the nodes involved, or between them and other aggregation nodes in charge of performing some centralized processing. The rate of information exchange with these schemes is usually very high and the latency should be very low, as lots of channel and state information should be simultaneously known by the relevant nodes and real-time operation is a must.

In order to alleviate the stringent requirements imposed by C-RRM techniques, new distributed architectures have been proposed in which a centralized entity (sometimes called Baseband Hostelling Unit, or BBU) performs all the baseband processing corresponding to a collection of cells, and is connected to multiple remote radio heads (RRHs) through optic fiber links [6]. This kind of architecture is sometimes known as "cloud radio access network", or cloud RAN [8]. The remote radio heads convert the baseband digital signals into useful radio frequency (RF) signals by passing to the analog domain and translation to the desired frequency of operation. Each of the signals being radiated by the different antenna elements present in a base station should be carried in an appropriate way by these links.

In order to promote interoperability and push widespread use of these architectures, an industry cooperation called Common Public Radio Interface (CPRI) has been introduced to specify a precise format for the baseband digital signals between the RRHs and the central baseband unit [2]. The CPRI interface covers the necessary items for transport, connectivity and control of user plane and control plane data, as well as synchronization, with a clear focus on layer 1 and layer 2.

Figure 1 shows a simplified diagram showing the CPRI connectivity between the Radio Equipment Control (REC) and the Radio Equipment (RE), where REC corresponds to the BBU mentioned above and RE represents the remote radio element(s). This architecture gives rise to what is sometimes called "Fiber-to-the- Antenna" (FTTA) systems [7]. Distributed Antenna Systems (DAS) are also another typical solution where signal data from a main node is distributed to multiple remote antennas via coaxial cables or fiber optic links.

The CPRI interface was designed with an emphasis put on hardware dependent layers, so that any high level interaction between transmit and receive peers may rely upon the basic defined procedures. In order to make it as simple as possible, l/Q data representing the physical signals are obtained through sampling and quantization of the complex baseband signals prior to up-conversion. The sampling rate of the CPRI signal should therefore correspond to the physical sampling rate of the signal to be carried (although in practice a higher rate is required for the transport of other control information), and the quantization depth can be chosen from a list of defined possibilities for downlink and uplink [2]. Discarding any control plane data, the minimum bandwidth required for the CPRI links should be at least equal to the bandwidth required for transmission of user plane data. In the case of the downlink of LTE, this bandwidth can be calculated in an easy way by considering a 20 MHz signal with a sampling rate of 30.72 Msamples/s and a sample width of N _b its bits per real l/Q component:

Minimum CPRI bit rate for single-antenna 20 MHz LTE downlink signal = 30.72 x 2 x N _bits (Mbps)

The factor 2 in the above calculation comes from the fact that each complex baseband sample should be represented by a pair of two components (real and imaginary), each of which is in turn represented by a digital sample of N _bits bits. If a sample width of 16 bits is considered, this results in a total of 983.04 Mbps per single- antenna sector, both for downlink and uplink. In a typical site comprising three sectors and two antennas per sector, the resulting bandwidth will thus be 5.89 Gbps, only for the downlink.

In addition to that, it is apparent that the above bit rate is needed irrespective of the number of connected users, because no statistical multiplexing gain is available and a constant bit rate is needed. The same physical signal is to be sampled and l/Q converted even if no users are present, because control channels are always transmitted therefore resulting in the links being always fully utilized even in the absence of cell load.

The absence of statistical multiplexing precludes any possibility of dimensioning the links according to the average cell bit rate (as is usually done in the aggregation part of the transport network for the backhaul), therefore requiring always the maximum bandwidth per radio site. As shown in figure 2, a typical transport network comprises three parts:

· The "last mile", covering individual base stations;

• the aggregation part, comprising multiple links from base stations which are aggregated in a lower number of high-capacity links; and

• the core part, finally connecting multiple aggregation links towards the relevant core network nodes.

The last mile is usually dimensioned according to the cell's "quiet time", which corresponds to situations in which few users in good radio conditions may obtain full resources from the cell. In this case the backhaul links should have a capacity at least equal to the available peak cell rate, corresponding to 100% use of resources and the highest spectral efficiency. In HSPA cells this corresponds to 14.4 Mbps per sector in the downlink, and in LTE cells with two antennas this roughly corresponds to 150 Mbps per sector in the downlink. Similar figures can be obtained for the uplink.

The aggregation part in the transport network can however benefit from the fact that it is highly unlikely that multiple sites are simultaneously serving traffic at the maximum bit rate. Given a sufficiently high number of sites, the corresponding aggregation capacity can therefore be dimensioned according to the cell's "busy time", in which the average spectral efficiency can be included in the calculations thus yielding much lower data rates. The core part also benefits from this statistical multiplexing gain, and the result is a transport network with much lower bandwidth requirements than it would have if peak rates were always taken into account in the calculations.

In the case of CPRI links in a cloud-RAN environment this multiplexing is not available. Due to the huge bandwidth requirements when aggregating multiple cells, specific solutions are thus required for compression of the baseband signals between the central processing nodes and the remote radio heads. These links are sometimes denoted as "fronthaul links" in cloud-RAN scenarios, as opposed to the backhaul links in traditional radio access architectures connecting the base stations to the core network nodes.

Other solutions exist relying upon compression of the digital information based on entropy coding, for instance patent application US 2009/0290632, "Compression of Signals in Base Transceiver Systems". Although, this patent application doesn't take into account any considerations on the physical properties of the signals. Other solution, for instance patent application US 2012/0207206, "Method and Apparatus for Signal Compression and Decompression", rely upon reduction of the sampling rate followed by block scaling and quantization, taking into account only the spectral characteristics of the physical signal. Other solutions are even less specific, for instance patent application US 2010/0177690 "Wireless Communication Unit", omitting parts of the characters representing l/Q samples according to some redundancy detection algorithm, also not considering the physical properties of the baseband signals. These solutions, although applicable for many different radio technologies, have the drawback of not exploiting specific signal properties that could improve the attainable compression ratios, even if lossless compression is pursued.

Yet another solution proposes header compression techniques for optimization of the required bandwidth WO 2010/147528 "Backhaul Header Compression", that only exploit redundancy in protocol headers and do not optimize the transmission for specific radio technologies. More efficient solutions for compression of the bandwidth required for fronthaul links are therefore needed in order to reduce the cost and complexity associated to cloud-RAN architectures. Summary of the Invention

The proposed invention deals with the problem of reducing the data rate required for transmitting full LTE/LTE-A baseband signals through the fronthaul links connecting a central processing node and one or multiple remote radio heads in Cloud-RAN scenarios. The aim is to reduce the required bandwidth and to benefit from the statistical multiplexing gains resulting from the aggregation of multiple cells, in order to avoid full utilization of the links even in low load situations. When compared to plain CPRI baseband signals, the proposed scheme can achieve high compression ratios especially in situations where the average spectral efficiency is far from its peak value (characteristic of busy time), with no loss of information.

In the present invention a compression scheme is proposed for the fronthaul links connecting baseband processing units (BBUs) to remote radio heads (RRHs) in an LTE/LTE-A Cloud-RAN scenario. The compression is achieved by introduction of the proposed Baseband Fronthaul Termination nodes at both ends of the fronthaul links.

The Baseband Fronthaul Termination nodes shall perform the proposed processing over the LTE/LTE-A physical signals, in order to reduce the amount of information and exploit statistical multiplexing when aggregating multiple cells. Different compression ratios are achieved for the downlink and uplink signals, due to their fundamentally different properties that can be exploited.

To that end, the present invention relates, in a first aspect, to a method for lossless compression and decompression of baseband digital signals in distributed LTE- Advanced (LTE-A) radio access networks. The method, as commonly known in the field, comprises at least one Remote Radio Head with a plurality of transmit and receive antennas connected to at least one Baseband Unit through a wired connection denoted as fronthaul link.

On contrary to the known proposals, the method of the first aspect in order to perform said lossless compression and decompression of a plurality of LTE-A baseband digital signals comprises in a characteristic manner:

using two processing nodes hosted at both ends of said fronthaul link in both an uplink direction and a downlink direction; transforming for said both uplink and downlink directions said plurality of LTE-A baseband digital signals into frequency domain; and

sending through said fronthaul link the information actually contained in the set of resource blocks occupied with said plurality of LTE-A baseband digital signals.

In an embodiment, in said downlink direction a set of precoding matrices are specified over a group of subcarriers and OFDM symbols sharing similar channel conditions, thus exploiting precoding redundancy across time and frequency and, prior to performing any modulation and/or precoding operations, representing constellation symbols by their corresponding bits according to any suitable constellation mapping.

The corresponding plurality of mapped bits representing said constellation symbols are 2, 4 or 6 bits for QPSK modulation, 16QAM modulation and/or 64QAM modulation, respectively.

In another embodiment, the resource blocks containing semi-static control channels are excluded. These control channels can be reconstructed at the peer side with the aid of suitable control information.

Other control information, such as cell reference signals (CRS), channel state information signals (CSI-RS), UE-specific reference signals, positioning reference signals, physical broadcast channel (PBCH) and primary/secondary synchronization signals (PSS/SSS), can be generated at the peer decompressing node of said at least one Baseband Unit adjacent to said at least one Remote Radio Head.

The operations in the downlink direction shall be repeated on a subframe basis of at least 1 ms.

In yet another embodiment, in said uplink direction a suitable Fourier transform is applied to said plurality of LTE-A baseband digital signals to convert them to said frequency domain. Then, the real and imaginary parts of the resulting samples of said Fourier transform in the OFDM symbols are quantized.

Finally, in the uplink direction other control information such as demodulation reference signals (DM-RS), Sounding Reference Signals (SRS) and/or random access information (PRACH) shall also be quantized in the frequency domain for the receiver to perform the necessary detection procedures.

A second aspect of the present invention, relates to a system for lossless compression and decompression of baseband digital signals in distributed LTE- Advanced (LTE-A) radio access networks, comprising at least one Remote Radio Head with a plurality of transmit and receive antennas and at least one Baseband Unit, said at least one Remote Radio Head being connected to said at least one Baseband Unit through a wired connection denoted as fronthaul link.

On contrary to the known proposals, the system of the second aspect to perform said lossless compression and decompression of a plurality of LTE-A baseband digital signals comprises having two processing nodes at both ends of said fronthaul link in both an uplink direction and a downlink direction, said plurality of LTE-A baseband digital signals being transformed into frequency domain, and sending through said fronthaul link the information actually contained in the set of resource blocks occupied with said plurality of LTE-A baseband digital signals.

In an embodiment, the two processing nodes are connected to said at least one

Baseband Unit and to said at least one Remote Radio Head by means of a logical connection.

In another embodiment, the two processing nodes can be physically implemented as part of said at least one Baseband Unit and said at least one Remote Radio Head.

The system of the second aspect is adapted to implement the method of the first aspect.

Brief Description of the Drawings

The previous and other advantages and features will be more fully understood from the following detailed description of embodiments, with reference to the attached, which must be considered in an illustrative and non-limiting manner, in which:

Figure 1 shows a simplified diagram showing the CPRI connectivity between the Radio Equipment Control (REC) and the Radio Equipment (RE), where REC corresponds to the BBU and RE represents the remote radio element(s).

Figure 2 shows an example of a typical transport network.

Figure 3 is a diagram with the logical interconnection of the elements of the present invention.

Figure 4 is an illustration of the main processing steps for the proposed compression scheme in the downlink, to be applied at the BFT node located at the BBU side.

Figure 5 is an illustration of the similar compression scheme for the uplink, which will be applied at the BFT node adjacent to the RRH.

Figure 6 is an illustration of the proposed processing steps to be performed by the BFT node located at the BBU, for the downlink case. Figure 7 is an illustration of the processing steps required at the BFT node located at the RRH side for reconstruction of the LTE baseband signal in the downlink.

Figure 8 is an example set of physical resource blocks scheduled for a number of users, in the case of an FDD frame with normal cyclic prefix.

Figure 9 is an example of LTE constellation encoding for the case of 16QAM modulation, taken from actual modulation mapping stage as specified in [3].

Figure 10 is the proposed transmission format for the downlink.

Figure 1 1 is the simplified format for the cases of PMCH, PDCCH and PHICH when the channels do not involve any precoding.

Figure 12 is the mapping procedure to be applied for the information contained in the PDSCH data layers and the PMCH.

Figure 13 shows the detailed structure for the proposed PREC field.

Figure 14 is an illustration of the proposed processing steps to be performed by the BFT node at the RRH, for the uplink case.

Figure 15 is an illustration of the processing steps that will take place in the peer

BFT node located at the BBU side in order to undo the compression procedure, for the uplink case.

Figure 16 and 17 are examples of the proposed transmission format for the uplink. Figure 16 corresponds to the case of performing user detection at the BFT node located at the RRH, and figure 17 corresponds to not having such user detection.

Figure 18 is an exemplary embodiment for the proposed invention at both BFT nodes located at the RRH side and the BBU side, in an LTE/LTE-A Cloud-RAN scenario.

Detailed Description of Several Embodiments

The present invention introduces new entities at both the central processing unit side and the remote radio heads side, aimed at alleviating the data rate requirements at the cost of increasing the computational load. In figure 3, the proposed Baseband Fronthaul Termination (BFT) nodes perform the tasks of converting the signals to the formats proposed in the present invention, both for uplink and downlink, for efficient transmission through the fronthaul. The BFT node adjacent to the BBU shall compress downlink signals for transmission by the RRH, and de-compress uplink signals for reception by the BBU. Conversely, the BFT node adjacent to the RRH shall compress uplink signals and de-compress downlink signals.

The compression is achieved by introduction of the proposed Baseband Fronthaul Termination nodes at both ends of the fronthaul links. These nodes will be logically connected to the BBU and the RRHs, but in practice they can be part of the corresponding nodes for ease of implementation.

The Baseband Fronthaul Termination nodes shall perform the proposed processing over the LTE/LTE-A physical signals, in order to reduce the amount of information and exploit statistical multiplexing when aggregating multiple cells. Different compression ratios are achieved for the downlink and uplink signals, due to their fundamentally different properties that can be exploited as is explained below.

The depicted diagram in figure 3 represents a logical interconnection of elements, but the BFT nodes can be physically implemented as part of the corresponding BBU/RRH nodes for ease of implementation. The BFT nodes shall perform the processing operations described below for the physical LTE signals.

For compression of the downlink signals by the BFT node located at the BBU side:

The frequency contents in the corresponding resource blocks for all the data channels can be compacted by specifying the information to be mapped on the set of resource blocks (RBs) dedicated for each user. The amount of information required is thus directly proportional to the cell resources in use, as only occupied RBs are sent through the fronthaul. Compression shall be applied at the BFT nodes adjacent to the BBU for transmission through the fronthaul.

2. Precoding information can be compressed by exploiting time and frequency coherence of the appropriate precoding matrices, in both cases of codebook and non-codebook based precoding [4].

3. Constellation symbols corresponding to each spatial layer of the physical downlink shared channel (PDSCH) can be represented by their corresponding mapping bits (2, 4 or 6 bits for QPSK, 16QAM and 64QAM respectively), neglecting any precoding information. The same also applies for the physical downlink control channel (PDCCH), physical HARQ indicator channel (PHICH) and physical multicast channel (PMCH, if applicable).

4. Other control information (such as cell reference signals, channel state information signals, physical broadcast channel, etc.) will be locally generated at the peer BFT nodes adjacent to the RRH, by appropriate generation of the frequency contents and mapping on the resource elements as specified in [3].

For compression of the uplink signals by the BFT node located at the RRH side: 1 . Uplink signals can be compressed by application of a suitable Fourier transform that converts the signals into the frequency domain, and subsequent quantization of the real and imaginary parts of the resulting samples. As in the downlink, the amount of required information will be directly proportional to the resources being used for the uplink, if user detection is performed. However in this case no time/frequency coherence can be exploited as the signals will be corrupted by noise, interference and other impairments to be taken into account at the BBU.

2. Control information (Demodulation Reference Signals -DM-RS-, Sounding Reference Signals -SRS- and/or random access information -PRACH-) shall also be quantized in the frequency domain for the receiver to perform the necessary detection procedures.

Figure 4 depicts schematically the main processing steps for the proposed compression scheme in the downlink, to be applied at the BFT node located at the BBU side. It is possible to extract data and control information from the frequency contents of the OFDM symbols, and in turn separate the information aimed at different users. For each of the users, both precoding information and constellation symbols are separately processed to exploit time and frequency redundancy, while information on other channels can also be efficiently compressed. Finally, suitable compressed messages for both control and data information can be constructed and sent through the fronthaul to the RRH.

It is apparent that these transformations will have to be undone at the BFT node located at the RRH side for reconstruction of the physical downlink signals.

Figure 5 depicts the similar compression scheme for the uplink, which will be applied at the BFT node adjacent to the RRH. The scheme in this case is somewhat simpler as no precoding information is applied. I/Q components of both data and control symbols are in this case quantized with a fixed number of bits, as no redundancy is available in the received signals if lossless compression is pursued. Operations shall be undone at the peer BFT node located at the BBU side.

It is apparent that the reduction in fronthaul bandwidth comes mainly from three parts:

o The conversion from a pair of l/Q N-bit samples to only 2, 4 or 6 bits for the constellation symbols in the downlink (neglecting any precoding); o the time and frequency redundancy in precoding information, which can be efficiently specified by the precoding matrix index (PMI) in case of codebook- based precoding, or the matrix elements in any suitable l/Q (or amplitude/phase) format; and

o the saved resources obtained by avoiding any unused subcarriers in the frequency domain.

The attainable compression ratio in downlink depends on the transmission mode in use for each user (as well as the resources reserved for each of them). In uplink the compression ratio is only dependent on the number of users and the cell resources reserved for each of them, if user detection is performed as will be explained later.

The highest compression ratio is obtained when the RBs do not experience any precoding other than transmit diversity. However it will be apparent that even in the case of spatial multiplexing significant compression ratios can be achievable. In the uplink significant compression ratios can also be achieved by ignoring any unused subcarriers in the frequency domain.

Fronthaul compression in the downlink: the main processing steps to be performed for downlink by the BFT node located at the BBU are:

1 . Control information corresponding to cell reference signals (CRS), channel state information signals (CSI-RS), UE-specific reference signals, positioning reference signals, physical broadcast channel (PBCH) and primary/secondary synchronization signals (PSS/SSS) will be locally generated at the BFT node adjacent to the RRH. Only the minimum information necessary for reconstruction of control signals will be delivered by the BFT node located at the central processing unit.

Generation of control information only involves appropriate mapping of the frequency contents on resource blocks, as specified in [3]. Some of these signals are completely static (CRS, CSI-RS, and PSS/SSS), while others are semi-static thus requiring limited interaction with the BBU (as happens with the PBCH and the UE-specific reference signals). This interaction can be accomplished by any means and will not be covered by the present invention.

2. Instead of applying the usual inverse DFT for generation of the baseband LTE signals through the fronthaul, frequency contents in the resource blocks can be compacted on a per-user basis by specifying the information to be mapped on the set of scheduled resource blocks (RBs). The frequency locations of the resource blocks can be specified by the start RB and the number of occupied RBs. Bandwidth is therefore saved by avoiding non-modulated and guard subcarriers, and the amount of information required is directly proportional to the cell resources in use.

If virtual resource blocks of distributed type are used, this shall be appropriately signalled to the peer BFT node so that it can apply the appropriate mapping.

It is possible to exploit redundancy in precoding information across both time and frequency domains. Due to the limited granularity of the channel state information as provided by the UEs, the BBU usually applies the same precoding matrix over a group of subcarriers that approximately share similar channel conditions. Thus only a set of precoding matrices shall be signalled along with the number of applicable subcarriers for each of them, and symbols corresponding to multiple subcarriers can benefit from the same basic precoding matrix. Moreover, all the OFDM symbols contained in a subframe will experience the same precoding operations.

The number of subcarriers for applicability of the precoding matrices can be specified in any suitable way. One possibility would be to express it as a number of resource blocks, similar to the parameter Precoding Resource Block Group (PRG) specified in 3GPP Release 10 [4], by which the eNodeB specifies the number of applicable resource blocks for each precoding matrix in the downlink. Any other possibility is not precluded depending on actual implementation needs.

If codebook-based precoding is applied, then for each precoding matrix it is sufficient to indicate the precoding matrix index (PMI) as well as the rank indication, and the BFT at the RRH side will pick the corresponding matrix as specified in [3] . If non codebook-based precoding is applied, then it is necessary to specify the amplitudes and phases of the matrix elements in any suitable format.

Users in transmission mode 2 (transmit diversity) need not specify any precoding matrix and only the symbols in one spatial layer should be sent, as the other layers can be easily obtained by application of Space-Frequency Block Coding (SFBC) (and eventually Frequency Switched Transmit Diversity -FSTD- for more than two transmit antennas). The same happens with the PDCCH.

Constellation symbols can be represented by their corresponding bits (2, 4, or 6 bits for QPSK, 16QAM and 64QAM modulation, respectively) as specified by the LTE modulation stage [3] [3], for the spatial layers in use by the physical downlink shared channel (PDSCH). The mapping shall be done neglecting any precoding information. Data on the physical downlink control channel (PDCCH) can also be encoded in a similar way taking into account that it is always QPSK-modulated and does not experience any precoding

(other than transmit diversity). The same applies also for the physical HARQ indicator channel (PHICH) and physical multicast channel (PMCH). The physical control format indicator channel (PCFICH) contains only an indication of the number of OFDM symbols devoted to the PDCCH (1 to 4 depending on the system bandwidth), so it can be conveniently sent in any suitable digital format (e.g. with 2 bits).

Figure 6 depicts the proposed processing steps to be performed by the BFT node located at the BBU, for the downlink case. After neglecting semi-static control information (CRS, CSI-RS, positioning signals, etc.), useful information aimed at each connected user is extracted from the frequency contents of the OFDM symbols, as well as dynamic control information (PDCCH, PHICH) and PMCH information (if present). Relevant precoding parameters are obtained for each of the users through inspection of precoding redundancy, and constellation symbols are also encoded. A compressed transmission format is constructed by adding the users' data information, PMCH and dynamic control information, and is finally sent through the fronthaul link.

The depicted operations shall be repeated on a subframe basis (1 ms). The compression ratio obtained in downlink depends on the number of connected users, the resources reserved for each of them and the amount of precoding information included, which in turn depends on the transmission mode for each user.

The BFT node located at the RRH side shall undo the described operations in order to reconstruct the baseband LTE/LTE-A signals that will subsequently be up- converted to the desired radio frequency. Figure 7 depicts the processing steps required at the peer BFT node for reconstruction of the LTE baseband signal. With the aid of the received precoding information as well as the constellation symbols it is possible to reconstruct the frequency contents corresponding to each user. Similarly it is possible to reconstruct dynamic control information and PMCH information (if present). Semi-static control signals and channels can be locally generated, and after adding the frequency contents of all channels and signals a suitable Inverse Discrete Fourier Transform (IDFT) converts it to the time domain. Finally, addition of the cyclic prefix at the beginning of each OFDM symbol completes the LTE baseband signals to be transmitted by each antenna.

Figure 8 illustrates an example set of physical resource blocks scheduled for a number of users, in the case of an FDD frame with normal cyclic prefix. For each group of resource blocks only appropriate information about the initial RB and number of RBs should be included for each user. Inside each resource block, information can be efficiently compressed by making use of precoding redundancy and constellation symbol mapping. In addition to the achieved compression ratio, it is possible to exploit the statistical multiplexing of multiple users and cells by dimensioning the aggregation part of the fronthaul links with the average capacity per site, instead of the peak capacity as is required for the CPRI.

Figure 9 illustrates an example of LTE constellation encoding for the case of 16QAM modulation, taken from actual modulation mapping stage as specified in [3] [3]. It is apparent that each symbol can be uniquely determined by its associated set of bits, given that the precoding information accounts for any additional amplitude/phase correction to be applied over the spatial layers. Mapping for QPSK and 64QAM would also be similar.

Proposed transmission format for the downlink: Figure 10 depicts a proposed format for the downlink information to be included on a per-subframe basis. Each subframe (1 ms) comprises several parts corresponding to the different channels to be compressed.

Information on PDCCH and PHICH channels is first included, followed by PMCH channel (if applicable, in dashed lines) and as many fields containing user data as users are scheduled on a given subframe. The proposed format for user data on PDSCH is also depicted. The length for the included fields considers the case of a 20 MHz system bandwidth. This format does not preclude any other possibility given that the basic ideas proposed in this invention are followed

The rationale for the proposed user data fields is as follows:

1. VRB. This field (1 bit) indicates if the virtual resource blocks are of localized type or distributed type. In the latter case the mapping to physical resource blocks shall follow the rules specified in [3] .

2. RBstart. This field (7 bits) indicates the starting resource block as scheduled by the BBU. NRB. This field (7 bits) indicates the number of resource blocks, irrespective of whether VRB is of localized or distributed type.

MOD. This field (2 bits) indicates the modulation according to the following possibilities: QPSK, 16QAM and 64QAM.

Rl . This field (3 bits) indicates the channel rank, which in turns corresponds to the number of spatial layers to be used. The actual length of this field will depend on the number of available transmit antennas (up to 8), as the rank will always be at most equal to the number of transmit antennas. If transmit diversity is applied (LTE transmission mode 2) this field will have the lowest value indicative of only one layer, although in 3GPP terminology it would be equal to the number of transmit antennas. This avoids sending more than one layer to the peer entity, as the other layers can be easily obtained at the remote side through SFBC/FSTD coding [3] .

TM. This field (4 bits) indicates the transmission mode in use, from TM1 to TM9.

PREC. This field (with variable bit length denoted as N _P ) specifies a common precoding matrix to be applied over a number of resource blocks. The precoding matrix can be specified as a precoding matrix index (if codebook- based precoding is applied) or as a specific complex matrix (in the case of non-codebook based precoding), along with the number of resource blocks where the precoding matrix applies. The proposed structure exploits the redundancy of the precoding matrix over a subset of the frequency resources and all the OFDM symbols reserved for the user. There will be as many precoding matrices as necessary so that the whole precoding operation is specified (n, precoding matrices for user / ^' ), each having a number of bits N _P dependant on the type of precoding. If no precoding is needed then this field will have no bits.

DATA LAYER #1 ... #L. These fields (with variable bit length denoted as N _L ) convey the constellation data corresponding to the spatial layers in use. Information can be encoded with a much lower number of bits than would otherwise be required for a plain l/Q quantized sample, as only 2, 4 or 6 bits are needed for a QPSK, 16QAM or 64QAM modulated symbol, respectively. From this information, as well as the precoding field, the BFT node at the RRH side will be able to apply appropriate precoding operations to obtain the signals that feed the antennas. Data will be sent from the lowest to the highest resource elements and from the lowest to the highest OFDM symbol in the subframe. The number of layers will be given by the rank indication, which in turn is always lower than (or equal to) the number of transmit antennas.

It will be clear for those skilled in the art that the length of the fields RBstart and

NRB can be conveniently reduced in case of lower system bandwidths, according to the maximum number of RBs in each case. Similarly, other modifications for the proposed format will also be possible depending on implementation needs, provided that the same basic ideas are followed.

The above described format applies for the physical downlink shared channel

(PDSCH) corresponding to each of the users. The case of PMCH, PDCCH and PHICH is much simpler as these channels do not involve any precoding (other than transmit diversity for PDCCH and PHICH). Therefore the corresponding format can be simplified as depicted in figure 1 1.

The rationale for the proposed fields is explained below.

1. RBstart. This field (7 bits) indicates the beginning of the scheduled resources (only applicable to PMCH).

2. NRB. This field (7 bits) indicates the number of resource blocks (only applicable to PMCH).

3. MOD. This field (2 bits) indicates the modulation employed (only applicable to

PMCH).

4. DATA. This field (variable length ) contains the constellation data in 2, 4 or 6 bits per symbol for QPSK, 16QAM and 64QAM, respectively.

The first three fields are only applicable for the PMCH channel, as indicated by the dashed lines. PDCCH and PHICH channels are characterized by a fixed modulation (QPSK) and a fixed resource mapping procedure as specified in [3], so no extra information is needed.

In the latter case of PDCCH and PHICH, an alternative for data mapping could be to specify the contents of the whole system bandwidth for the first OFDM symbols in the subframe containing PDCCH, PHICH and PCFICH. Because of the complexity in the resource mapping procedure for these channels [3], it is possible to encode all the QPSK REs with 3 bits per sample according to the convention shown in Table 1 . This would significantly decrease the complexity for these channels, and would also cover PCFICH.

Table 1. Constellation mapping for PDCCH, PHICH and PCFICH

The possibility of no transmission ("100" value) must be considered taking into account that PDCCH resource mapping involves spreading information along the system bandwidth for increased diversity, with the possibility of having unoccupied resource elements along the band.

Figure 12 depicts the mapping procedure to be applied for the information contained in the PDSCH data layers and the PMCH. Information is collected from the lowest to the highest index frequency and from the lowest to the highest OFDM symbol, excluding the symbol(s) devoted to PDCCH, PCFICH and PHICH, and skipping any other control channel or signal.

By separating precoding information and constellation symbols it is possible to account for any standard or non-standard-based beamforming operation, including open/closed loop spatial multiplexing, UE-specific beamforming, and more complex processing such as those dealt with in cooperative multi-point (CoMP) techniques.

It is important to note that eventual UE-specific reference signals shall not be included as part of the user information but will certainly be taken into account in the precoding process, because such reference signals are only received by the intended UE (in transmission modes 7, 8 or 9) and will therefore experience the same precoding process as the rest of the data signals.

In case of having Carrier Aggregation in LTE-A networks each proposed format should be replicated for each of the component carriers in use by the system.

Precoding information: Precoding information will be significantly different depending on the transmission mode and MIMO configuration in use.

In the common case of codebook-based open-loop spatial multiplexing (as in LTE transmission mode 3), precoding is not dependent on any feedback from the UEs but is completely specified by the standard [3]. No additional precoding information is thus needed in this case, and the PREC field will have no contents.

In case of codebook-based closed-loop spatial multiplexing (as in LTE transmission modes 4 and 6), precoding information can be conveniently represented by a set of precoding matrix indicators (PMI) for the user, as defined in [5]. Each precoding matrix will be applicable over a number of resource blocks which may eventually span the whole bandwidth reserved for the user. High compression ratios can therefore be achieved with this scheme.

In case of non-codebook-based spatial multiplexing (LTE transmission modes 7, 8 and 9), precoding information must be given in the form of complex (N _T x L) matrices where N _T is the number of transmit antennas and L the number of spatial layers, with complex matrix elements a _l} :

Any suitable representation format can be used for the matrix elements as e.g. I/Q components or amplitude/phase, with a given number of bits. Any other representation format will be equally valid without departure from the proposed ideas.

The number of resource blocks where each precoding matrix applies shall also be indicated in any suitable format. One possibility is to express it as a number of RBs, or to define a set of pre-defined possibilities according to actual implementations. Any other possibility is not precluded for the purposes of the present invention. The detailed structure for the proposed PREC field is shown in figure 13.

The meaning of the fields is as follows:

1 . NRBP. This field (7 bits) specifies the number of resource blocks where each precoding matrix applies. A lower size can also be envisaged for this format, depending on the actual number of possible values in real implementations. 2. PMAT #1 ... PMAT #n,. These fields are of variable length and specify the appropriate precoding matrices (up to n, in the figure). The format will depend on whether codebook-based or non-codebook-based precoding is considered.

For codebook-based precoding each PMAT field will contain an index according to the appropriate codebook as specified in [3]. For non-codebook-based precoding the matrix elements should be specified in any suitable format, such as l/Q or amplitude/phase with any given number of bits.

It will be assumed that the number of precoding matrices will be equal to

NRB NRBP where NRB is the number of resource blocks scheduled for the user, NRBP is the number of resource blocks for applicability of the precoding matrix, and represents the rounding towards infinity operator.

In case of applying transmit diversity (transmission mode 2), no precoding information is needed and only one layer shall be considered.

Fronthaul compression in the uplink: The following processing steps are proposed for the uplink:

1 . As uplink signals contain by definition the effects introduced by the mobile radio channel and the receive antenna, they can only be compressed by application of a suitable Fourier transform (DFT or FFT) that converts the signals to the frequency domain, and subsequent quantization of the real and imaginary parts of the resulting samples in the OFDM symbols. If user detection is performed (as explained below), the amount of required information will be directly proportional to the resources in use for the uplink (as in the downlink case), for the physical uplink shared channel (PUSCH) and physical uplink control channel (PUCCH). This also avoids spending resources on non-modulated subcarriers. In case that no user detection is performed a saving in resources is still obtained by neglecting guard bands in the frequency domain.

2. As opposed to the downlink, other control information (demodulation reference signals -DM-RS-, sounding reference signals -SRS- and/or random access information -PRACH-) shall also be quantized in the frequency domain for the receiver to perform the necessary detection procedures. Additionally, user detection of these transmissions would allow for statistical multiplexing in the aggregation network.

Figure 14 depicts the proposed processing steps to be performed by the BFT node at the RRH, for the uplink case. After the RRH down-converts the received radio frequency signals to baseband signals, appropriate analogue-to-digital converters (ADC) transform them to the digital domain for subsequent processing. Time synchronization is first performed to detect the beginning of the OFDM symbols, and after removal of the cyclic prefix a suitable Discrete Fourier Transform (DFT) can extract the frequency contents of the signal. Detection of the users' information is depicted with a dashed box in the figure as it may be considered optional for the purpose of the present invention. If present, it allows for neglecting any non-modulated subcarriers (containing only noise and interference from other cells) by locating active transmissions for the PUSCH, PUCCH, PRACH or SRS in the frequency domain. This could be accomplished through any suitable mechanism such as correlation with any relevant control signal (demodulation reference signals, sounding reference signals -if available-, and so on), but not precluding others. The difficulty in this box comes from detecting activity in low signal-to-noise conditions. If present, it would allow for statistical multiplexing also in the uplink. If not present, the whole bandwidth should be considered in the frequency domain (excluding guard bands) and no possibility of statistical multiplexing would be available in the aggregation part of the fronthaul links.

Finally, quantization of l/Q (or amplitude/phase) samples in the frequency domain with a specified number of bits results in a compressed digital signal to be efficiently sent through the fronthaul. Quantization involves both data and control samples corresponding to each of the active users, or the whole bandwidth if no user detection is accomplished (excluding non-modulated subcarriers at the guard bands). In the first case highest compression ratios will be obtained especially for low cell loads, as any unused subcarrier will be discarded.

Figure 15 schematically depicts the processing steps that will take place in the peer BFT node located at the BBU side in order to undo the above described compression procedure. Reconstruction of users' information (in dashed lines) is only applicable when user detection is performed at the peer BFT node, thus recovering the original frequency contents as received by the remote antennas. A subsequent Inverse Discrete Fourier Transform (IDFT) translates the frequency contents to the time domain, thus obtaining the OFDM time symbols. After insertion of the cyclic prefix the LTE baseband signal is ready to be sent back to the BBU.

It would also be possible to avoid the IDFT and CP insertion blocks because, depending on the implementation, the BBU might only need the frequency contents of the signal for subsequent SC-FDMA processing.

Proposed transmission format for the uplink: For the case of uplink there are two possible formats depicted in figures 16 and 17. Figure 16 corresponds to the case of performing user detection at the BFT node located at the RRH, and figure 17 corresponds to not having such user detection. Considering figure 16, and given a number M of receive antennas, the information can be simplified with respect to the downlink case by specifying only appropriate indicators for the resource blocks scheduled for each terminal. The fields containing the data received by each antenna are quantized in the frequency domain with a specified width (in number of bits), giving rise to l/Q samples that include the physical uplink shared channel (PUSCH), physical uplink control channel (PUCCH), demodulation reference signals (DM-RS), sounding reference signals (SRS) and physical random access channel information (PRACH), if applicable.

In figure 17, where no user detection is performed, the format consists only on the quantized samples of the full set of REs (excluding guard bands) for each of the receive antennas, as the applicable RBs will be given by the system bandwidth.

In all cases the quantized samples in the frequency domain will be collected from the lowest to the highest frequency index and from the lowest to the highest OFDM symbol, as in the downlink case (figure 12). There is no need to differentiate between data and control information as both shall be quantized in the same way.

The rationale for the depicted fields in figures 16 and 17 is as follows.

1. RBstart. This field (7 bits) indicates the starting resource block as detected by the user detection block.

2. NRB. This field (7 bits) indicates the number of resource blocks for the applicable transmission. Note that in LTE uplink only localized type is implicitly assumed, as the single carrier nature of SC-FDMA transmission precludes the use of distributed resource mapping.

3. DATA RX Antenna #1 ... #M. These fields carry the received l/Q samples contained in the resource blocks in use by the user, for each of the M received antennas. As the received signals will be affected by multiple impairments (noise, interference, effects of the radio channel, Doppler spread, antenna pattern, etc.) it is not possible to exploit any redundancy without incurring in some loss of information. Received data will include data and control signals, with a variable length (K bits).

In case of not having user detection the whole bandwidth is quantized over the receive antennas, excluding the guard bands. The number of bits K' in this case is fixed and given by the following expression:

i ^ KB sc ^{J V} sym ^{J V} bits ' where stands for the number of available resource blocks in the whole system bandwidth for the uplink, N ^™ is the number of subcarriers per resource block, is the number of OFDM symbols contained in a subframe for the uplink, and N _BITS is the number of bits for quantization of the samples. The factor 2 comes from the presence of a pair of l/Q components in each RE. The above formula allows for calculating the compression ratio with respect to a plain CPRI format, defined as the ratio between the CPRI bit rate and the bit rate resulting from the present invention:

^samples ^ ^ bits N s.amples

compression ratio _T UL

RB ^{J V} sc ^{J V} sym ^ ^{J v} I RB ^l sc ^l sym

where N, is the number of samples contained in an LTE subframe (1 ms). For the case of a 20 MHz system bandwidth (1 10 RBs) with normal cyclic prefix and a subcarrier spacing of 15 kHz, the above equation yields:

= no

N = 12

compression = 1.66 : 1

N: 14

N s.amples

This compression ratio is independent on the number of active users, and therefore in this case there is no statistical multiplexing gain in the uplink. However it allows for much simpler implementations in the uplink, leaving more efficient compression ratios for the downlink.

Attainable compression ratios: In order to assess the advantages of the proposed invention, simple calculations were made and are detailed below aimed at obtaining the different bit rates that result from applying the proposed compression technique on several cell configurations and Ml MO modes. Tables 2 and 3 summarize the assumptions considered in the calculations for downlink and uplink, respectively, for the case of a single sector. In Table 3 user detection was considered because otherwise the attainable compression ratio in uplink was fixed to the value 1.66:1.

Parameter Setting

Number of users Variable from 1 to 10

Modulation QPSK, 16QAM, 64QAM

Bandwidth 20 MHz (110 RBs)

Number of TX 1 , 2, 4

antennas

Transmission 1 , 2, 4, 8/9

modes

Number of spatial Equal to the number of TX antennas (worst case) layers in use

Number of OFDM 1 (worst case)

symbols for

PDCCH

Constellation 3 bits per RE over the whole system bandwidth (worst mapping for case)

PDCCH, PHICH

and PCFICH

Number of 1 1

scheduled resource

blocks per user

Number of resource 2

blocks per

precoding matrix

Number of bits for 16 bits for each real and imaginary component representation of

matrix elements in

non codebook- based precoding

Table 2. Parameters for calculation of the compression ratio in downlink

Parameter Setting

Number of users Variable from 1 to 10

User detection Active

Modulation QPSK, 16QAM, 64QAM

Bandwidth 20 MHz (1 10 RBs)

Number of RX 1 , 2, 4

antennas

Number of 1 1

scheduled resource

blocks per user

Number of bits for 16 bits for each real and imaginary component representation of

l/Q samples

Table 3. Parameters for calculation of the compression ratio in uplink with user detection The number of users was varied from 1 to 10, while the resources reserved for each of them was fixed to 1 1 RBs. Only one OFDM symbol was reserved for the PDCCH, and up to 4 transmit and receive antennas were considered.

Four transmission modes were studied in the downlink: TM1 (SISO), TM2 (transmit diversity), TM4 (codebook-based closed-loop spatial multiplexing) and TM8/9 (non-codebook-based spatial multiplexing, up to 2/8 layers respectively). TM1 was only applied when having a single TX/RX antenna at the RRHs. TM2, TM4 and TM8/9 were applied when having two and four TX/RX antennas at the RRHs, with different user distributions in the cell (TM8/9 only applicable for LTE-A). Quantization of the samples in uplink and the matrix elements in non-codebook precoding had a depth of 16 bits for each l/Q component. The results are detailed in what follows.

Compression ratios for 1 TX, 1 RX antenna: In this case only TM1 is applicable in the cell. Four cases were studied, the first three considering a single modulation and the fourth one comprising a mixture of modulations in the cell:

1. 100% users in QPSK;

2. 100% users in 16QAM;

3. 100% users in 64QAM;

4. Mixture of users comprising 80% in QPSK, 15% in 16QAM and 5% in 64QAM.

The number of users was varied between 1 and 10 thus giving rise to cell load values from 10% to 100%. From the results obtained, it was apparent that in the uplink the compression ratio was independent on the modulation. In the downlink however the compression ratio was slightly dependent on the modulation, with the best results obtained for QPSK. The global compression ratios presented little dependency on the modulation due to the influence of the uplink, which dominates the global tendency. At full cell load the global compression ratio was roughly 3:1 , while at very low loads it reached values of 26:1 - 29:1 , which is quite high. At 50% cell load the compression ratio was between 4.6:1 and 6.2:1 depending on the modulation.

Compression ratios for 2 TX, 2 RX antennas: In this case three different transmission modes were considered: TM2, TM4 and TM8. The cases considered were:

1. 80% users in TM2, 15% users in TM4, 5% users in TM8, all of them using QPSK.

2. The same distribution with all the users in 16QAM.

3. The same distribution with all the users in 64QAM.

The cell load was varied between 10% and 100% as in the single antenna case. In the downlink higher compression ratios were achieved by the use of two TX antennas compared to the single-antenna case. In the uplink the compression ratios were however almost independent on the number of receive antennas. Maximum global compression was around 30:1 for low cell loads, and around 6:1 for 50% cell loads.

Compression ratios for 4 TX, 4 RX antennas: This case is analogous to the two- antenna case, only changing transmission mode 8 to 9 for the support of four layers. Very large compression ratios were obtained in the downlink, but the global compression was again dominated by the uplink, with values around 31 :1 for low load and around 6:1 for 50% cell load.

The values obtained depended on the cell load, with little differences observed by varying the modulation and the transmission mode.

Figure 18 represents an exemplary embodiment for the proposed invention at both BFT nodes located at the RRH side and the BBU side, in an LTE/LTE-A Cloud-RAN scenario.

Block 31 1 represents a remote radio head (RRH) and block 319 a baseband unit (BBU), both of which are connected by suitable fronthaul links (block 315). Any actual network topology can be considered, as for the purpose of the present invention a single logical fronthaul link is seen by the peer entities. Block 312 represents the proposed Baseband Fronthaul Termination (BFT) Node adjacent to the RRH, in charge of compressing uplink signals towards the BBU (block 314), and de-compressing downlink signals towards the RRH (block 313). Block 316 represents the BFT node adjacent to the BBU, in charge of compressing downlink signals towards the RRH (block 317), and decompressing uplink signals towards the BBU (block 318).

Outside the BFT nodes the LTE baseband signals should be the same as if no compression/decompression scheme was present, with the only difference of a lower noise power in uplink compressed signals corresponding to the unused subcarriers actually removed.

The proposed embodiment can be implemented as a collection of software elements, hardware elements, firmware elements or any combination of them.

One of the main advantages of the proposed invention is that it deals with the problem of connecting LTE/LTE-A baseband processing units to a collection of remote radio heads in cloud-RAN scenarios. These connections, commonly referred to as fronthaul links, suffer the drawback of requiring very large bit rates after sampling and quantization of the baseband signals in downlink and uplink. In this invention a way of reducing the fronthaul bit rate is introduced, achieving significant compression ratios when compared to plain sampling and quantization as is usually done in other solutions.

The most important advantage of the proposed invention is, in addition to the bandwidth reduction in the fronthaul, the possibility to exploit statistical multiplexing when dimensioning the aggregation network between the remote radio heads and the central processing units. This allows for a dimensioning based on average (or busy time) data rates rather than peak data rates, with a clear impact on cost and complexity. Even in the case of full cell resources, the advantage of analyzing the frequency contents comes from the absence of any non-modulated subcarriers (corresponding to guard bands) in the resulting signals, which represent a significant percentage of resources.

In the downlink the present invention allows for a significant bandwidth reduction with no loss of information, as well as statistical multiplexing of users when aggregating fronthaul links from multiple cells. In the uplink it becomes more difficult to identify active transmissions corresponding to the own cell, but even without user detection it can account for a minimum bandwidth reduction of 1 .66:1 by excluding the guard bands in the frequency domain.

Lower bit rates for the fronthaul directly results in more efficient cloud-RAN architectures that could resemble usual backhaul architectures in traditional RAN deployments, which leverage on statistical multiplexing for connection of multiple cells to the core network without incurring in an excessive complexity over the fiber links.

ACRONYMS

3G Third Generation

3GPP Third Generation Partnership Project

ADC Analog-to-Digital Converter

BBU Baseband Unit

BFT Baseband Fronthaul Termination Node

C-RAN Cloud Radio Access Network

C-RRM Centralized Radio Resource Management

CoMP Cooperative Multi-Point

CP Cyclic Prefix

CPRI Common Public Radio Interface

CRS Cell Reference Signal

CSI-RS Channel State Information Reference Signal

DAS Distributed Antenna System

DFT Discrete Fourier Transform

DL Downlink

DM-RS Demodulation Reference Signal

elCIC Enhanced Inter Cell Interference Coordination eNB eNodeB

FDD Frequency Division Duplex

FFT Fast Fourier Transform

FSTD Frequency Switched Transmit Diversity

FTTA Fiber-to-the-Antenna

GSM Global System for Mobile Communications HARQ Hybrid Automatic Repeat Request

HSPA High Speed Packet Access

IDFT Inverse Discrete Fourier Transform

IMT International Mobile Telecommunications

LTE Long Term Evolution

LTE-A Long Term Evolution - Advanced

MI MO Multiple Input Multiple Output

OFDM Orthogonal Frequency Division Multiplexing

PBCH Physical Broadcast Channel

PCFICH Physical Control Format Indicator Channel PDCCH Physical Downlink Control Channel PDSCH Physical Downlink Shared Channel

PHICH Physical HARQ Indicator Channel

PMCH Physical Multicast Channel

PMI Precoding Matrix Index

PRACH Physical Random Access Channel

PRG Precoding Resource Block Group

PSS Primary Synchronization Signal

PUCCH Physical Uplink Control Channel

PUSCH Physical Uplink Shared Channel

RAN Radio Access Network

RB Resource Block

RE Radio Equipment

RE Resource Element

REC Radio Equipment Control

RF Radio Frequency

Rl Rank Indicator

RRH Remote Radio Head

RRM Radio Resource Management

RX Receive

SC-FDMA Single Carrier Frequency Division Multiple Access

SFBC Space Frequency Block Coding

SRS Sounding Reference Signal

SSS Secondary Synchronization Signal

TM Transmission Mode

TX Transmit

UE User Equipment

UL Uplink

UMTS Universal Mobile Telecommunication System REFERENCES

[1 ] 3GPP TS 36.300, Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Network (E-UTRAN); Overall Description, Stage 2 (Release 8)

[2] Common Public Radio Interface (CPRI); Interface Specification, V4.2

[3] 3GPP TS 36.21 1 , Evolved Universal Terrestrial Radio Access (E-UTRA); Physical Layer and Modulation (Release 10)

[4] S. Sesia, I. Toufik, M. Baker (editors), "LTE, the UMTS Long Term Evolution: From Theory to Practice - second edition", John Wiley & Sons, 201 1

[5] 3GPP TS 36.213, Evolved Universal Terrestrial Radio Access (E-UTRA); Physical Layer Procedures (Release 10)

[6] T. Pfeiffer and F. Schaich (Alcatel-Lucent Bell Labs Stuttgart), Optical Architectures for Mobile Back- and Fronthauling", OFC/NFOEC wireless backhauling workshop, Los Angeles, 5/3/2012

[7] "BrightJack Continuous Fiber Monitoring for Fiber-to-the-Antenna (FTTA)", JDSU Application Note

[8] "Discovery of Cloud-RAN", Nokia Siemens Networks, Cloud-RAN workshop, 23/04/2010