Whitepaper UE-Side Virtual MIMO Using mm-wave for 5G · FTTH further. As mmWave is strongly blocked by walls, a BS small cell transmitting in the mmWave band would be required in

UE-side Virtual MIMO Using mmWave for 5G © by Fraunhofer IIS, August 2014. No part of this document may be reproduced, published, distributed or displayed in any form.

1

Shared UE-side Distributed Antenna System (SUDAS) for 2 UEs connected to 2 mobile networks, supported by 3 so-called SUDACs to increase the data rate

UE-SIDE VIRTUAL MIMO USING MM-WAVE FOR 5G

Whitepaper FRAUNHOFER-INSTITUTE FOR INTEGRATED CIRCUITS ( I IS ) , ERLANGEN/GERMANY

FRIEDRICH-ALEXANDER UNIVERSITY ERLANGEN-NÜRNBERG, GERMANY

Fraunhofer-Institute for Integr. Circuits (IIS) F.-A.-University Erlangen-Nürnberg Communication Systems Division Institute for Digital Communications

Prof. Dr. Robert Schober Am Wolfsmantel 33 Cauerstr. 7 91058 Erlangen (Germany) 91058 Erlangen (Germany)

www.iis.fraunhofer.de www.idc.lnt.de

Contact: Thomas Heyn, +49 9131 776 6311, [email protected]


2

Fig. 1: Simulated achievable data rates in downlink for two proposed schemes compared to state-of-the-art (LTE-like) link. Channel models: Winner+ outdoor-to-indoor and IEEE 802.11ad “living room”. Distance to base station: 100m. System parameters: see Table 1.

Executive Summary

The ambitious data rate target of 10 Gbit/s for the 5G standard is elusive despite the use of advanced air interface technologies such as massive MIMO, two-tier networks with small cells, and employing mmWave links. In order to approach this high target data rate for user equipments (UEs) inside a building, we propose in this whitepaper for 5G to use an infrastructure of many low-price relay nodes installed in fixed indoor locations. These nodes relay signals from the base station and a mobile band at ultra-high frequency (UHF) via mmWave links to the UE, and vice versa. We refer to this scattered infrastructure as a Shared UE-side Distributed Antenna System (SUDAS). A SUDAS translates massive spatial multiplexing in a licensed mobile band into massive frequency multiplexing in an unlicensed mmWave band. This new concept is derived from Virtual MIMO schemes known from literature, such that the mmWave band is used for relaying and the infrastructure can be shared between multiple UEs and multiple mobile network operators (MNOs). In an internal research project, we have considered three different options for the out-of-band relay nodes in this innovative system: amplify-and-forward (AF), or compress-and-forward (CF) either based on IEEE 802.11ad or on a 5G-specific air interface dedicated to mmWave relaying.

All three options have been analyzed with respect to various aspects for their use in a practical 5G system. Among others, the following points have been examined:

Mechanisms for resource allocation considering the SUDAS-sharing between multiple UEs and MNOs.

Realization of multi-user MIMO in such a setting.

Potential usage scenarios. Considerations of the cost of the

hardware implementations. Channel estimation and

synchronization of the relayed signals.

Fig. 1 shows the results from system simulations using realistic models for the UHF and mmWave channels. As can be seen both SUDAS with AF and CF relaying realize an extraordinary high gain in data rate compared to a system without SUDAS, and bringing 10 Gbit/s to the homes and offices appears to become realistic. The compelling results obtained from the analyses and simulations imply that the novel SUDAS-like infrastructure could be highly beneficial for 5G and should be seriously considered as part of it. Within 5G, an air interface profile dedicated to mmWave relay links should be specified. As of today, only the surface has been scratched of many practical aspects of such a system. Therefore, it is strongly recommended that future 5G research projects address the open issues that are named in this whitepaper.

2 4 6 8 10 12 14 160

2.5

5

7.5

10

12.5

15

Total basestation EIRP (dBW)

Ave

rage

sys

tem

dat

a ra

te (G

bit/s

)

CF SUDASAF SUDASState-of-the-art without SUDAS

x 4.4Performance gain: x 5.4

Advantages of SUDAS:

Significantly higher data rates than for massive MIMO

Higher data rates and/or lower cost than for small cells

High degree of diversity and high robustness

Transfer of 5G deployment cost from MNO to end user

End user can install infrastructure parts according to his data rate needs

No regulatory issues about using the mmWave band for 5G


3

Current vision of 5G

Besides reduced latency, an increased number of devices, better energy-efficiency, and reduced cost, another primary goal of the 5th generation (5G) mobile communication standard is to strongly increase data rates at up to 10 Gbit/s per device. The data rate originally envisaged for 4G mobile communication standards was already 1 Gbit/s, and Release 10 of the Long Term Evolution Advanced (LTE-A) standard extends this to even 3 Gbit/s, but the ambitious goal of 5G will enable the download of movies within seconds and offer very high data rates to many simultaneous users inside a cell. This tenfold increase in data rate from 4G to 5G is for instance reflected in the key performance indicator objectives issued by the European Commission for 5G research within the Horizon2020 programme. However, with today’s technology, it appears very difficult to reach such high data rates even in favorable conditions. It is unlikely that a single mobile network operator (MNO) will possess more than a total of about 200 MHz of licensed spectrum (see also [1]), such that spectral efficiencies as high as 50 bit/s/Hz need to be reached. Three technologies are currently considered as top candidates for 5G: massive MIMO (Fig. 2), using a two-tier architecture with macro-cells and small cells, and use of the unlicensed mmWave bands. We will show in the sequel that it is highly questionable whether these technologies alone are able to reach the elusive goal of 10 Gbit/s.

Massive MIMO and small cells without mmWave cannot offer 10 Gbit/s

State-of-the-art smartphones are typically equipped with 2 antennas. Other user-equipments (UEs) might possess even 4 internal antennas, but assuming more than 4 antennas may not be realistic. Given the above number of a total of 50 bit/s/Hz required spectral efficiency, this means that each of the at most 4 spatial streams must carry 12.5 bit/s/Hz, which would theoretically require a signal-to-noise ratio of at least 38 dB and (taking into account FEC redundancy) the use of at least 16384-QAM in each stream. This statement is valid, whenever we are limited to the assumed 200 MHz total bandwidth. In other words, this problem concerns base stations (BS) using massive MIMO and small cells schemes, as long as they operate in the licensed mobile bands. Only an extension of the available bandwidth by using (unlicensed) mmWave spectrum could solve this problem.

Backhauling of small cells at 10 Gbit/s is extremely challenging

Thanks to the very large available bandwidth in mmWave, e.g., in the 60 GHz bands, it appears possible to deliver 10 Gbit/s from a small cell BS to a UE. However, the bottleneck now becomes the backhauling of this data to the small cell BS. The last mile from the backbone network to the homes cannot support such high data rates, except if optical fibers can be used (fiber-to-the-home – FTTH). However, the cost to equip all homes with FTTH

UE fmm

Small cell 2

Room 1

Small cell 1

Small cell 3

Small cell 4

Room 2

Room 3

Room 4

Optical fiber

FTTHRouter

To backbone network

Fig. 2: House equipped with small cells in every room

UEfUHF (Macro- or

Small Cell)Base stationfUHF

Fig. 3: Massive MIMO to a UE inside a building


4

is very high; for instance for Germany alone, the cost to equip every building (not home!) with FTTH is estimated to be around 67 billion € [2].

Moreover, most buildings do not possess an internal optical fiber infrastructure or a wired LAN (Ethernet) to distribute the data received from FTTH further. As mmWave is strongly blocked by walls, a small cell BS transmitting in the mmWave band would be required in all rooms of a building, see Fig. 3. Hence, the cost of connecting each room of each building must be added on top of the 67 billion € for connecting the buildings. Another difficulty is to find a suitable location for the small cell BS that is in line-of-sight (LOS) to all potential UE positions in the room. Without such a LOS connection, using mmWave links is hardly possible. Observe that if a small cell BS belongs to a single MNO, the envisaged high data rate is only available to UEs in the small cell, which are associated with this MNO. In order to allow mobile access at high data rates also to guests from other MNOs, the small cell BS must be neutral as regards the MNO, and its owner should therefore be the owner of the building or room.

Combinations of massive MIMO and mmWave

Besides the aforementioned technologies, further approaches are currently being considered:

Backhauling of small cells by massive MIMO over mmWave. The problems of this scheme are the high penetration loss of building walls from outdoor to indoor, the high propagation losses over larger distances suffered at mmWave frequencies, and that a small cell is needed in every room.

Massive MIMO to UEs based on mmWave. Similarly, the problems of this are the high penetration loss of the walls, which renders this approach only suitable for outdoor use, and that many BSs are needed. The latter is due to the fact that a LOS connection is required from any UE location to at least one base station at any time.

Virtual MIMO could be part of the 5G solution

Most of the data rate increase in last decade’s mobile communication systems is owed to the use of MIMO with an increasing number of antennas. However, we saw above that the small number of ultra-high frequency (UHF) antennas that can be integrated into a UE limits the MIMO gain for improving the spectral efficiency. On the other hand, the mmWave bands can offer the desired data rates but suffer from very difficult propagation conditions. Therefore, we propose a scheme, which combines the advantages of MIMO over a mobile UHF band channel with its favorable propagation conditions and mmWave with its large bandwidths. Our novel scheme is derived from the idea of Virtual MIMO (VMIMO) [3], [4]. It extends this concept to the use of mmWave for relaying the MIMO signals transmitted in the UHF band, and to sharing an infrastructure between multiple UEs and multiple MNOs. VMIMO is also known as Cooperative MIMO, Distributed MIMO or Virtual Antenna Arrays [5], [6]. The basic idea is that surrounding communication devices act as virtual antennas for a UE by relaying received MIMO signals

Fig. 4: Two ideas for the realization of SUDACs for electrical outlets: (a) integrated into the socket, (b) adapter to be plugged into the outlet

State-of-the-art massive MIMO, small cells and mmWave are not sufficient to reach the 5G data rate target.


5

(prior to MIMO decoding, which takes place in the destination UE). A variant of this technology is already used in LTE-A on BS-side under the name of Coordinated Multi-Point (CoMP), where the received signals are relayed or rather forwarded to one entity (e.g. a BS) over the backhauling network. In our proposed concept, VMIMO is used also on the UE-side by relaying each signal transmitted from the BS in the mobile bands to non-overlapping channels in the mmWave band, see Fig. 5(a). Our scheme limits relaying to only two hops, namely between BS and relay node, and between relay node and UE. This simplification allows the end-to-end latency to remain low, and it avoids the significant overhead for exchanging routing information. The relay nodes are dedicated very low-cost devices and scattered in a room. Such relays could be integrated into many devices with continuous power supply such as electrical outlets (see Fig. 4), lamps, and light outlets or other devices that will carry Machine-Type-Communication (MTC) circuits in the future like TV sets and fridges. Fig. 6 explains this scheme in a diagram. As shown, all links in the mobile bands from BS “A” use the same frequency resource fUHF,A in the UHF range, while

every mmWave link occupies its individual frequency band fmm,i. The relays thus form a scattered infrastructure, which is in contrast to the centralized infrastructure realized by a small cell BS per room. As all UEs in a room can share these relay devices, we denote the relaying system as a Shared UE-side Distributed Antenna System – SUDAS. In particular, each relay is called a Shared UE-side Distributed Antenna Component – SUDAC. As Fig. 6 shows, the SUDAS is not only shared among all UEs in a room, but possibly shared also between BSs of different MNOs. Thanks to the high attenuation of stone walls, multiple SUDAS contained in different rooms do hardly interfere with each other, when mmWave is adopted. Note that LTE-A already enables both in-band and out-of-band relaying. However, both take place in licensed bands. Such relays are (high-cost) regenerative (i.e. decode-and-forward) relays and represent a centralized infrastructure in contrast to the scattered structure of a SUDAS.

A

s/c payload

B

s/c payload

C

s/c payload

solitarys/c

f

C A B

f

A

payload

B

s/c payload

fAB

f

mmWave band mobile bands (UHF)

(b)

(c)

1 2(a)

f f

3

BS transmits/receives a single MIMO signal

Signals relayed to/from the UE by relay nodes #1,2 and 3

Fig. 5: Spectrum occupation in mmWave band and mobile bands (UHF): (a) a single (V)MIMO carrier in the mobile bands is relayed to multiple non-overlapping signals in the mmWave band; (b) relaying between three channels A, B, C in the mmWave band and the mobile (UHF) bands: channelization in mmWave band for amplify-and-forward (AF) SUDACs: status/control (s/c) and payload channels form pairs, additional solitary s/c channels exist, e.g. as rendezvous channels during the discovery procedure; (c) aggregation of multiple mobile carriers in mmWave

SUDAS is an infrastructure of low-cost relay devices scattered over a room or building.


6

SUDAS may become an important brick in the 5G framework

Besides the installation into indoor devices, SUDACs could be mounted in buses, trains and cars. As most mobile traffic is consumed indoors (especially at home in the evening, see e.g. [7]) and while commuting, such SUDAS installations would cover the most important mobile use cases. We find the SUDAS concept particularly promising, because it translates spatial multiplexing in the mobile bands (we call this the backend) into frequency multiplexing in the mmWave bands (dubbed frontend); while spatial multiplexing works very well in the mobile bands, frequency multiplexing is highly suitable for mmWave. The end-to-end connection between BS and UE is hence split into two segments, and the SUDAS acts as a translator between them. Of course, this does not preclude the existence of direct links between the BS and the UE, which further improves the VMIMO gain. Advantageously for the SUDACs, they do not need to carry out the (V)MIMO decoding. They simply forward the received signal, which will be decoded

at the UE (downlink) or the BS (uplink). Similarly, MIMO precoding takes place only in the BS and UE, respectively. The SUDACs operate hence in either Amplify-and-Forward (AF) or Compress-and-Forward (CF) mode. Since the signal reception and transmission at each SUDAC have a very large frequency separation (one in the UHF band and the other one in the mmWave band), both can take place simultaneously, i.e., relaying can be instantaneous. Different from traditional relaying systems, a SUDAS utilizes both a licensed and an unlicensed frequency band in parallel. As the spectrum in the licensed mobile bands is scarce and costly, the spectral efficiency should be maximized here. On the contrary, in the very wide unlicensed mmWave band, such a high spectral efficiency is not needed as the unlicensed band is virtually free of cost. SUDAS exploits these properties very well. Thanks to its scattered nature, a SUDAS realizes a markedly higher degree of space diversity than a small cell BS. This is particularly pronounced in the mmWave band. While a human body may completely shadow the mmWave links between a small cell BS and a UE, it is rather unlikely that the links to all SUDACs in a room will be shadowed at the same time.

SUDAC

SUDAC

SUDAC

UE 1

UE 2

fUHF,A

fUHF,A

fUHF,A

fUHF,A

fUHF,B

fUHF,B

fmm,1

fmm,2

fmm,3

fmm,4

SUDAC

fUHF,Bfmm,5

MNO „B“

MNO „A“

Room in a buildingUHF links in mobile bands (backend)

mmWave links (frontend)

Fig. 6: Basic principle of a SUDAS for the downlink: multiple UEs of different MNOs share the scattered infrastructure represented by the SUDACs. Direct links from the BS exist to the UEs. Only 2 channels are needed in UHF, while 5 are needed in the mmWave band.

SUDAS translates massive spatial multiplexing in the licensed mobile UHF bands into massive frequency multiplexing in the unlicensed mmWave bands.


7

Usage scenarios

The main scenario for the use of a SUDAS is inside buildings, such as a living room or a large open-plan office. We expect that SUDACs could be integrated into electrical outlets and many electrical devices, such that 8 SUDACs or more per room appears a realistic scenario. Besides providing very high data rates for one UE, a SUDAS helps to improve multi-user (MU)-MIMO performance for supporting high-speed communication for simultaneous users.

Moreover, a SUDAS could enable the data rates and spectral efficiencies that future broadcast systems such as eMBMS strive to achieve in order to deliver UHDTV content to fixed TV sets

(which act as a receive-only UE) with indoor antennas (Fig. 7). To achieve these goals, the use of MIMO is indispensable, and the MIMO gain brought to the TV set could be increased by a SUDAS. Fig. 8 shows another application of a SUDAS: a private Home-eNodeB (HeNB) that is coupled to an optical fiber in one room of the building can transmit high data rates to UEs in other rooms in the building by first transmitting to the SUDACs in the first room by mmWave. These SUDACs then relay the signal to a sub-6 GHz or UHF band like the 2.4 GHz ISM band, where they all transmit in the same (VMIMO) channel to a SUDAS in a second room using spatial multiplexing. The second SUDAS finally relays these MIMO signals to the destination UE once again over mmWave. The

Fig. 7: SUDAS used for broadcast TV reception with a MIMO TV transmitter and shared by a UE for mobile reception. The TV set uses internal or external indoor antennas.

Fig. 8: SUDAS in two rooms used for connecting a UE to a Home-eNB in another room.

The same SUDAS installation can serve multiple usage scenarios, even at the same time.


8

mmWave transmissions in the first and second room do hardly interfere with each other thanks to the high attenuation of the walls; the same mmWave band can be re-used. The two innovations in this scenario are that the HeNB is assisted by several virtual antennas, which can be used in addition to its (probably not more than 4) built-in antennas, and that the UE extends its 2 internal antennas by the surrounding SUDACs in the same way. The advantage of SUDAS over the existing HeNB concept is hence that the rather small devices (HeNB and UE) can employ more antennas and achieve higher VMIMO gains. With such an approach, connecting an optical fiber to a single HeNB could suffice to cover the complete building with high-rate services. It should be noted that a single SUDAS infrastructure can cover two usage scenarios: that of a UE connected to a macro-cell BS, and that of a UE connected to a HeNB. Another important usage scenario is transport systems. Besides the obvious installations of SUDAS in trains and buses, cars could be installed with SUDAS, too. Manufacturers are equipping cars with multiple antennas to achieve receive diversity for the on-board units (OBU) today. However, these antennas do not provide gains to further UEs carried by the passengers. As an alternative, the installed antennas could be replaced by SUDACs and the communication between the antennas/SUDACs and the OBU could thereby be replaced by a standardized wireless air interface operating in mmWave. Then, the UEs of all passengers in the car could benefit from the improved car infrastructure.

Possibly, the passengers could employ not only their own car’s SUDAS, but even the SUDAS infrastructure of near-by cars could be used when needed, as long as such a resource sharing is fair and reasonable. Using an approach similar to the one described above for HeNBs, a device-to-device (D2D) or car-to-car communication between the OBUs of two cars could be established over such a VMIMO link between the two SUDAS, such that higher data rates and/or increased diversity can be reached. Outdoor usage is a third scenario. The SUDAS concept might not be as beneficial as in the previous scenarios. One could imagine that multiple SUDACs belonging to one MNO or shared between several MNOs are mounted to each street lamp, e.g., along the pole separated by tens of cm.

CAPEX and ownership

The European Commission shows in [8] that in 2012, telecom revenues have dropped throughout Europe while the investments have mostly increased, possibly because of the LTE roll-out. As a consequence the Average Revenue Per User (ARPU) is relatively low in Europe. End users request ever better QoS and higher data rates – this is why 5G is needed – but they do not want to pay higher prices. Therefore, it would be highly beneficial for the MNOs, if part of the capital expenditure (CAPEX) for 5G deployment could be borne by the users. In a SUDAS, a user can “install” the SUDACs himself. “Install” means here: buy a SUDAC-enabled device and plug it into the mains. This is similar to the ownership concept of a Wi-Fi system, but in a SUDAS, the user still uses the mobile network of his MNO. The deployment cost is thus partially transferred to the user. Every user can decide himself, what service quality he needs and how much he wants to spend on his infrastructure. To make the installation of a SUDAS attractive for a user and the integration of a

Fig. 9: SUDAS installed in a car

SUDAS allows the MNOs to lower the CAPEX for the 5G roll-out.


9

SUDAC attractive for a device manufacturer, the price of a SUDAC must be very low. We reckon that the end user price should not exceed 5 € per SUDAC, including all RF circuitry, antennas, fitting and integration. This solution has a second advantage: if SUDAS are installed by the end users, it is legitimate to use the unlicensed shared mmWave spectrum. Current regulations do not need to be changed, even if the SUDAS specification is part of a 5G mobile communication standard.

Realization options of a SUDAC

As stated before, a SUDAC can perform Amplify-and-Forward (AF) or Compress-and-Forward (CF) relaying. We will now take a closer look at the characteristics of these realizations.

Amplify-and-Forward (AF)

The main advantage of an AF SUDAC is its potentially low price. Fig. 10 shows the block diagram of such a device. In the displayed simple realization, the payload signal (red lines) is only

processed in the analog domain. In an alternative realization, digital processing can be used, which requires down-conversion of the payload signal to the baseband. The processing (analog or digital) in the SUDAC comprises bandpass filtering in order to avoid the amplification and relaying of noise or interference surrounding the useful signal. When the signals can have a flexible bandwidth as in LTE, then the passband of the bandpass filter has to be adjusted accordingly. A SUDAC needs to be configured to know the carrier frequency and bandwidth of the relayed signal. This information can be provided by the UE over a dedicated status/control (s/c) channel in the mmWave band that is frequency-multiplexed with the payload signal. Fig. 5(b) shows such a channelization. Of course, the s/c channel needs to be filtered out prior to forwarding the payload channel in the mobile (i.e. UHF) bands to the basestation. The s/c channels should carry pilots that enable channel estimation in the mmWave links. Thanks to the high path loss in the mmWave band, the impulse response of an indoor channel is rather short (10 to 20ns [9]). This leads to little frequency selectivity over typical bandwidths of up to 20 MHz

PA

LNA

TDDFDD

PA

TDDFDD

LNA

PLLXTAL

s/c channel Rx/Tx+ SUDAC controller

Ctrlsignals

Powerctrl

BWctrl

BWctrl

Powerctrl

fcar,mm

ctrl

TDDctrlTDD

ctrl

s/c channels

mmWave UHF

PLL fcar,UHF

ctrl

fixed IF

baseband

Fig. 10: Block diagram of an Amplify-and-Forward (AF) SUDAC, where the payload is processed only in the analog domain. Band selection filters in UHF and mmWave have been omitted for simplification; potentially IQ-modulation is used for SSB operation.

SUDAS may use the unlicensed mmWave spectrum without changing current regulations.


10

and simple channel estimation. In FDD mode, a UE needs to transmit and receive in the mmWave band at the same time. In order to avoid self-interference due to leakage, the transmit and receive frequency bands have to be separated sufficiently. For applications such as TV reception, a unidirectional transmission is used. However, also in this case there needs to be an s/c channel in the return direction to allow handshaking between the UE (i.e. TV set) and the SUDAC for control information exchange. This solitary s/c channel without accompanying payload channel of course must be sufficiently separated in frequency from the signal in the forward direction to avoid self-interference. Such a solitary s/c channel might also be needed to realize a return link for exchanging control information continuously in the mmWave band, when a TDD mode is used by the BS. Additionally, special rendezvous channels may be needed for the discovery procedure. Besides the cost, further arguments in favor of an AF SUDAC could be reduced power consumption and a shorter end-to-end latency. However, the price to pay is a lower flexibility than for the solutions presented hereafter. For instance, AF SUDACs are limited to use FDM(A) for the mmWave

links, while CF SUDACs can employ further multiplexing and multiple access schemes. Moreover, for purely analog AF SUDACs (Fig. 10) it is difficult to realize fully flexible bandwidths (e.g. from 1.4 to 20 MHz as in LTE) and to equalize the mmWave channel, e.g. compensate the Doppler shift and the slight frequency selectivity – see further below.

Compress-and-Forward (CF) using IEEE 802.11ad (WiGig)

The mmWave IEEE standard 802.11ad could be a good option for realizing a compress-and-forward (CF) relay. Fig. 11 shows the block diagram of such a solution. In the downlink, the signal from the BS received in the mobile (UHF) band is sampled and the binary representation of the IQ samples is transmitted over an 802.11ad link to the UE. In the uplink, the payload of the 802.11ad signal received by the SUDAC from the UE contains the binary representation of the IQ samples that are simply transmitted over the mobile (UHF) band to the BS. This scheme is straightforward and it is flexible. Its advantage is that future UEs of a 5G standard might support 802.11ad anyway and that large numbers of 802.11ad chips will be produced, which lowers the price.

mmWave

mmWave Rx/Txincl. digital base-band processing

e.g. 802.11ad

PA

TDDFDD

LNA

Ctrlsignals

Powerctrl

TDDctrl

UHF

PLL fcar,UHF

ctrl

DAC

ADC

analogbaseband

XTAL

digitalbaseband

Fig. 11: Block diagram of a Compress-and-Forward (CF) SUDAC (either using IEEE 802.11ad or a dedicated air interface). Band selection filters in UHF have been omitted for simplification.


11

Moreover, we can expect efficient inter-operability and co-existence with other users of the mmWave band. However, 802.11ad was not designed with the SUDAS application in mind. A single High Rate PHY channel defined by this standard occupies a bandwidth of over 2 GHz, such that Analog-to-Digital converters with very high clock frequency must be employed. This might lead to higher cost and/or higher power consumption than for alternative SUDAC realizations. The link budget of 802.11ad envisages the use of beamforming. As this requires a mmWave antenna array at the SUDAC, it could be difficult to realize cheap SUDACs. Moreover, the 802.11ad channelization, access protocol, and scheduling might involve higher overhead, less flexibility and higher latency (even though the theoretical latency of 802.11ad is only in the order of ten µs). Latency could be a critical aspect for use in 5G, as a significantly reduced latency is necessary for mission-critical machine communication like car-to-X.

Compress-and-Forward (CF) using a dedicated 5G air interface for mmWave relaying

This option can avoid the shortcomings for the application of 802.11ad in SUDAS. The definition of an air interface for the relayed mmWave links could be an integral part of the 5G specification. Similar to using 802.11ad, a SUDAC would adopt compress-and-forward in the downlink and apply FEC encoding prior to transmission over the mmWave link. In the uplink, it would receive a FEC encoded mmWave signal, whose binary payload represents the IQ samples to be forwarded over the mobile (UHF) band. However, this air interface could be simpler and lead to a cheaper chip: an LDPC decoder designed for more than 6 Gbit/s (as used in 802.11ad) is more costly than a simpler scheme for lower data rates, an ARQ scheme and its

associated buffers are possibly not needed for the relay links, and omitting beamforming (at least in the SUDACs) would further reduce the cost. Future UEs could include this air interface by default and share their mmWave antenna with an 802.11ad chip. Then the same UEs could make use of indoor SUDAS relaying macro-cells or Home-eNBs, and of SUDAS in trains or cars. As a consequence, lower cost, lower power consumption, and lower latency than for 802.11ad could be realized by a specific profile of the 5G air interface, which is dedicated to VMIMO relaying in mmWave.

Aspects and challenges of SUDAS

When the above three options are considered in detail, many questions and challenges appear that need to be addressed individually. In this section, we will discuss some basic considerations for a SUDAS. More details are presented in the Annexes to this whitepaper.

Keyhole effect and MIMO gain

The SUDAC performance stands and falls with the assumption that it is possible to have a VMIMO system from an outdoor BS to an indoor SUDAS. However, it is known that the so-called keyhole effect can lead to rank reduction of the MIMO matrix because of three phenomena described in [10], [11]:

spatial keyholes diffraction-induced keyholes modal keyholes

[12] shows that there is only a negligible keyhole effect if an opening of at least 30x30 cm exists in a metal wall. Assuming that the SUDAS is used in an environment that meets this condition (e.g. rooms with windows) we can safely assume that spatial keyholes will not affect a SUDAS.

5G should specify an air interface profile dedicated to VMIMO relaying in mmWave.


12

[11] relates diffraction induced keyholes to outdoor scenarios and also proposes a solution for the placement of the BS antennas to counter this effect. Modal keyholes will occur when the waves propagate along hallway-like structures. This could be the case in an indoor scenario or a street canyon. [10] proposes that a relaying system for MIMO could mitigate such effects. It appears therefore that a SUDAS will probably not suffer from the three described keyhole effects. However, no channel measurements have been found in literature that fit exactly to the SUDAS scenario. This is hence a topic of future research.

Carrier aggregation

As carrier aggregation will probably be used in 5G, a SUDAS must allow relaying several carriers simultaneously. A simple solution is of course to assign one SUDAC to one mobile carrier frequency only and to assign further carriers to other SUDACs. Alternatively, several independent RF chains could be bundled in a single SUDAC device. Similarly, several mobile downlink channels could be aggregated inside a SUDAC. For instance, a “double” AF SUDAC could bundle two carriers that are separated in frequency in the mobile bands to adjacent channels in the mmWave band, see Fig. 5(c).

Resource allocation for one or multiple MNOs

As stated before, a SUDAS should allow any UE to connect to the network of its associated MNO. One SUDAS can host different UEs from several MNOs at the same time. This must be taken into account by the resource allocation. The resources to allocate in a SUDAS are not only the time, frequency, code, and space resources occupied in mmWave band, but also the relaying (i.e. hardware) resources represented by the pool of SUDACs in the room. A fair resource

allocation must ensure that the available data rates are assigned to each UE according to the QoS requirements of its services, to the number of carriers that can be relayed, and to the UE’s and SUDAC’s capabilities to support certain carrier frequencies, carrier spacings, bandwidths, channel switching times, transmit powers etc. Moreover policies defined by the MNO concerning the data rates granted to certain users or services can be taken into account. Of course, the resource allocation must ensure that interference between multiple mmWave links or other users of the mmWave bands (such as 802.11ad) is avoided. The resource allocation can be centralized (e.g. an entity in the core network cooperating with the other MNOs) or distributed, cf. e.g. [13]. It can restrict itself to the mmWave band only or also affect the Radio Resource Management (RRM) in the mobile bands. According to our investigations, the best solution appears to be a distributed resource allocation carried out in the UEs but supported by the BSs. The motivation for a distributed resource allocation is among others lower latency and lower control plane overhead compared to a centralized approach. In the distributed approach, the UEs can interact with the RRM carried out in the BSs in order to assist, e.g., in the carrier selection for the mobile bands. For the case of 802.11ad, the resource allocation is part of the standard already. Hence, any interaction with the BS/MNO-side is not possible, such as taking the link quality between BS and each SUDAC (or the policies from one or multiple MNOs for the traffic prioritization) into account in the resource allocation for the mmWave links. Furthermore, 802.11ad does not carry out any assignments about which SUDAC should relay which UE. Moreover, the 802.11ad standard offers only 4 High Rate PHY or 12 Low Rate PHY channels operated in TDMA mode. In scenarios with a very large number of UEs and SUDACs using the mmWave links as in large open-plan

IEEE 802.11ad might not offer the resource allocation that is required for SUDAS.


13

offices, air ports, concert halls, or cars stuck in a traffic jam, these 802.11ad channels might be exhausted and result in excessive latencies, which are not acceptable anymore because of the stringent 5G latency requirements.

Mobility

Fig. 12 shows a typical scenario, where a UE moves from one room equipped with two SUDACs to another room with one SUDAC. One might wonder whether this system contains two individual infrastructures (one SUDAS per room) or a single larger SUDAS with three SUDACs. It becomes clear for the shown trajectory of the UE that both rooms cannot simply be treated as two separate SUDAS. In position 2, all three SUDACs have a LOS connection to the UE, and the resource allocation algorithm has to cover such cases. Hence, a SUDAS can become a quite extended network of SUDACs, and any UE sees only its local network environment at any one time. As the UEs move, the SUDACs must constantly scan passively for approaching UEs and contact them. Then, the resources must be re-allocated locally in the immediate neighborhood of the UE and potentially a hand-over to another SUDAC has to be initiated. Passive scanning should be chosen for the SUDACs in this discovery procedure in order to reduce the electromagnetic pollution and to improve the energy-efficiency of the system. With regards to the channel characteristics, the user mobility causes relatively high Doppler shift due to the very high carrier frequency in the mmWave band. Moving the UE at, e.g., 10 km/h when turning around, causes an instantaneous Doppler shift of 550 Hz.

Synchronization and channel estimation

The aforementioned momentary Doppler shift is high but it is still within the range supported by LTE, where

maximum Doppler frequencies up to 840 Hz are covered. Therefore, for an AF SUDAC the UE or the BS are in principle able to compensate the channel variations over the complete end-to-end link. However, the time-derivative of the Doppler shift can be very high in mmWave, which might exceed the capabilities of the BS or UE. Moreover, indoor SUDACs are fortunately not mobile but installed in a more or less fixed location. Therefore, the channel between the BS and the SUDAC in the mobile bands is frequency selective but quite static, while the channel between the SUDAC and the UE in the mmWave band is time-variant but affected only by relatively flat fading (over up to 20 MHz bandwidth). Sophisticated AF SUDACs could exploit these facts and compensate the channel variations over the mmWave link, e.g. remove the Doppler shift if required. The situation changes when it comes to the usage of SUDACs in trains, buses, and cars. Now, also the channel between the BS and the SUDAC is time-variant with possibly high Doppler shift. This would add to the Doppler shift originating from the mmWave link, e.g. when the user moves in a train. Hence, the compensation of the Doppler shift by the SUDAC as described above becomes indispensable in the case of AF relaying. On the other hand, for fixed indoor SUDACs installed in an industrial environment like a factory building, the delay spread may be much larger than the aforementioned 10 to 20 ns [9]. The longer delays are due to the larger

2

3Position 1

Position 2 Position 3

Trajectory of UE WallUE SUDAC

LOS to 1, 2

LOS to 1, 2, 3 LOS to 1, 3

1

Fig. 12: UE discovering new SUDACs during its move


14

spaces lacking absorbing walls and including metallic reflectors. For such scenarios, the mmWave channel can be quite frequency selective, and an AF SUDAC is unable to compensate these effects. Instead, equalization must be performed at the BS or the UE. Additionally, the phase noise mask of a mmWave carrier is significantly worse than that typically met in mobile bands in UHF. An AF SUDAC could compensate part of this phase noise by utilizing the pilots embedded in the s/c channel. Without such compensation, it is questionable whether a BS or a UE would be able to track and compensate the phase noise. This would require the inclusion of features dedicated to this purpose in the 5G air interface like additional pilots that assist in the phase noise compensation. Finally, the potentially very large initial frequency offset of the SUDACs needs to be tackled. The SUDAC discovery procedure could reduce such an offset by synchronizing the SUDACs to the UEs, which in turn are synchronized to the mobile network anyway.

Pilot overhead

As stated before, the channel between the BS and an indoor SUDAC is frequency selective but quite static thanks to the fixed location of the SUDACs. Therefore, pilots can be interpolated over long time spans, and the pilot density in time can be chosen rather low. Consequently, the pilot overhead seems to be acceptable even for spatial multiplexing with, e.g., 16 transmit antennas.

Implementation issues

As explained above, an AF SUDAC chip could be cheaper and consume less power than a CF SUDAC chip incorporating an 802.11ad component or a CF SUDAC chip implementing a dedicated 5G relaying air interface. On the other hand, the SUDAC device cost will (at least in the short and medium term) not be governed by the chip cost

but by the cost for integration, packaging, and fitting of the chip for mmWave input and output signals. In the long term, this might change, however. With regards to power consumption, according to [14], the total power consumption of a transceiver chip (that processes similar data rates in receive and transmit directions) appears to be governed largely by the power needed for the power amplifier. The power required by the digital baseband processing (e.g. FEC decoding) is probably a minor part compared to this, and once again, the apparent advantage of an AF SUDAC solution might not materialize in practice. The filtering of relayed signals with flexible bandwidth required in AF SUDACs is furthermore still difficult to realize in the analog domain today, and it is a topic of current research.

First conclusions

Based on the above and further considerations (see Annexes), it appears that all three candidates, i.e., AF SUDACs, CF SUDACs using 802.11ad, and dedicated 5G CF SUDACs are viable solutions. However, the advantages of AF SUDACs are probably not decisive in the near future, such that the other candidates might be the better overall choice. However, these first statements are only the outcome of an initial theoretical analysis. To obtain reliable conclusions, the advantages and drawbacks of each solution need further analysis and possibly testing in demonstrators/prototypes.

Previous work

Numerous papers have analyzed various theoretical properties of Virtual MIMO systems and cooperative diversity, e.g. [3], [4], [5], [6], [10], [13]. On the other hand, the literature lacks to some extent the consideration of practical aspects such as those outlined in the previous sections.

Thanks to the stationarity of the SUDACs, pilot overhead remains acceptable even for massive spatial multiplexing.


15

State of the art

Existing work on Virtual MIMO covers among others the aspects of channel modelling, analysis of channel capacity and energy efficiency, resource allocation, and routing algorithms. The focus of VMIMO studies has mostly been on wireless sensor networks with accordingly low data rates. Most analyzed systems use multi-hop relaying. In some system models, all network nodes have the same functionalities, i.e., any node can be source, destination, or relay, while in other models, there are dedicated source and destination nodes (also dubbed “data-gathering nodes”) as well as sheer relaying nodes. Moreover, some models are hierarchical and allocate the nodes to groups or layers/stages according to their location or distance to the source node, while simpler models represent a plain uniform network. The considered systems are partly using in-band relaying, e.g. by using time-division-multiplexing (TDM) with a first time slot reserved for the hop from a node X to a node Y and a second slot for the hop from Y to Z, or they use short range links for connecting nearby nodes within one of the aforementioned groups or layers. The model in [5] uses a combination of both, i.e., in-band relaying between the groups and short-range links within a group. Our new proposal uses a simpler system model compared to most of the aforementioned approaches: a SUDAS is a relaying system with only two hops in order to ensure the low latencies required by 5G. As a consequence, the routing is not an issue here (it is clear that a SUDAC shall forward the signals to the UE or BS, respectively), and the routing algorithm and associated signaling overhead can be omitted. An important property of SUDAS is its use mmWave for the short range links to connect each individual “virtual antenna” to a UE. For many devices in a room, many such links are needed and the required spectrum must be wide. The mmWave band is literally

perfectly suited for this purpose, as the links are almost completely insulated from those in the neighboring rooms. It should be noted that 5G research already considers D2D links and (multi-hop) relaying by UEs for other UEs at the cell edge. In contrast to our novel scheme, 5G does not yet envisage a scattered infrastructure realized by SUDACs that serve not only one mobile network but are shared between multiple MNOs. When relaying is realized by other UEs (as suggested for 5G) then this depends strongly on the goodwill of their owners, which might not be willing to spend their battery energy for relaying the signals of other users. If instead, SUDACs are fixed and installed in places with electric power supply (no batteries), these problems do not exist. The infrastructure simply belongs to the owner of the apartment who can make it available to his guests. Interestingly, [6] shows several parallels to our SUDAS proposal. At that time, 10 years ago, the topic was the evolution from 3G to 4G. The data rate requirements for 4G were two orders of magnitude higher than for the contemporary 3G standard. Either a significant densification of the network infrastructure had to take place (compare this to the two-tier model with a large number of small cells as proposed for 5G), or – as proposed in [6] – the massive use of relaying by a so-called Wireless Media System (WMS). Even the use of unlicensed spectrum from relay to UE was proposed. A significant difference to our SUDAS proposal is that we abstain from using multi-hop relaying as WMS for the sake of shorter latency. Our second hop is only very short (within the same room), such that mmWave can be used, and we explicitly target the use of Virtual MIMO for higher data rates.

SUDAS differs significantly from the multi-hop relaying and the mesh and ad-hoc networks currently considered in 5G.


16

Our work so far

In a joint internal research project, the Institute for Digital Communications of the Friedrich-Alexander-University Erlangen-Nürnberg and the Fraunhofer Institute for Integrated Circuits (IIS) in Erlangen/Germany have considered the implications of employing SUDAS as part of a practical 5G system. These aspects include the case that multiple UEs in a room are connected to multiple different MNOs, accounting for RF issues, consideration of what can be implemented in practice in hardware at acceptable cost, and the consequences of higher Doppler shift, phase noise and frequency offset in the mmWave band. Moreover, the device discovery procedure, resource allocation, the realization of MU-MIMO, and the performance for realistic channel models were topics of these investigations. Part of the results are exhibited in this whitepaper and in its Annexes. For realistic channel models, simulations have been carried out. The corresponding system and channel parameters are listed in Table 1 and the results yielding an extraordinarily increased data rate are visualized in Fig. 1. For this simulation, it was assumed that the downlink signal occupies 200 MHz of UHF spectrum, which is motivated by [1]. For simplicity, we assume that all carrier frequencies of the aggregated 200 MHz are located around 800 MHz, i.e. between 700 and 900 MHz. In practical carrier aggregation, the carrier frequencies would instead be separated significantly. The simulation goal was to analyze the potential capacity of a SUDAS system for 5G independently of the actual air interface to be chosen later. Therefore perfect channel estimation and no implementation losses were assumed. The result is of course an upper bound on the performance, but the interesting point is the comparison with a non-SUDAS system. The assumption of 14 SUDACs with LOS link to the UE is also very

Table 1

Parameters used for the simulation

FDD-downlink 1 BS and 1 UE in the macro-cell no interference from adjacent cells BS has a Distributed Antenna System

(DAS) with 16 antennas 14 SUDACs that are at line-of-sight to

the UE SUDACs have one UHF and one

mmWave antenna UE has two internal UHF antennas

and one (non-beamforming) mmWave antenna

downlink channels from BS are mutually statistically independent for any pair of BS antennas

UHF and mmWave channels of any pair of SUDACs are mutually statistically independent

distance BS <--> SUDAC: 96 m distance BS <--> UE: 100 m distance SUDAC <-->UE: 4 m channel model for UHF links to/from

BS: Winner+ outdoor-to-indoor channel model for mmWave links:

IEEE 802.11ad "living room" BS carrier frequency: 800 MHz useful signal bandwidth: 200 MHz

(actually occupied bandwidth excluding any guard bands)

OFDM with LTE-like parameters mmWave links carrier frequency: ~ 60

GHz BS antenna height: 25 m total EIRP of the DAS in the direction

towards the SUDAS: shown on x axis of the graph

SUDAC’s and UE’s height: 1.5 m SUDAC's and UE's UHF antenna gain:

0dBi SUDAC's and UE's mmWave antenna

gain: 3dBi (semi-isotropic) SUDAC’s and UE’s noise figure: 8 dB SUDAC's power in mmWave: 13 dBm synchronization and channel

estimation: perfect overhead for guard intervals, pilots

and signalization: not taken into account

MIMO pre-processing in BS: equal power allocation (i.e. no MIMO precoding)

MIMO decoding in UE: perfect


17

optimistic. However, very favorable conditions are always employed to reach the claimed maximum data rate in 3G, 4G, and 5G systems. Fig. 1 shows data rate curves for a system using AF SUDACs, and for CF SUDACs with slightly improved performance. For comparison, an LTE-like system is shown with the same 16 BS-antennas but without the use of SUDACs, such that the UE can exploit only its 2 internal antennas. As expected, the (V)MIMO gain is very significant. For instance, at 8 dBW, the non-SUDAS system achieves 1.9 Gbit/s, while the CF SUDAS-System reaches 10.5 Gbit/s. Even if this 450% (sic) gain does not materialize fully in a real system, the expectable gain is so large that using SUDAS should be seriously considered in 5G. Annex 1 shows further results, among others the data rate achieved for a varying number of SUDACs connected to the UE. The objective of our internal research project was to provide initial results. In the next step, the promising SUDAS concept should be investigated in depth in 5G research projects in the coming years. Table 2 shows a non-exhaustive list of topics that could be considered in these projects.

Conclusions

The investigations carried out by the Institute for Digital Communications at the University of Erlangen and the Fraunhofer Institute for Integrated Circuits (IIS) in Erlangen/Germany revealed that the proposed innovative Virtual MIMO scheme using the mmWave band for the relayed links has great potential to achieve the ambitious data rate targets of 5G, especially in indoor and car environments. Our simulation results imply that the achievable data rates inside buildings can be multiplied when using a SUDAS infrastructure compared to the conventional case without SUDAS. From the current point of view, the most promising way forward is the introduction of a mmWave air interface into the 5G standard that allows compress-and-forward relaying of signals in both up- and downlink. Furthermore, the air interface should define the mechanisms for the allocation of the (time and frequency) resources and the assignment of the individual relays to the UEs. The resource allocation could be carried out by a distributed algorithm executed by the UEs to allow the sharing of the scattered relay infra-structure by UEs from multiple MNOs. The algorithm could and should carry out this resource allocation in tight inter-action with the base stations. In order to verify the suitability of the proposed scheme for a 5G standard, further research is necessary including channel measurements, analysis and simulation, and finally implementation of prototypes and demonstrators.

Table 2: Acronyms

BS Base Station EIRP Equivalent Isotropically

Radiated Power LOS Line-Of-Sight RRM Radio Resource

management SUDAC Shared UE-side

Distributed Antenna Component (= device)

SUDAS Shared UE-side Distributed Antenna System

VMIMO Virtual MIMO


18

References

[1] http://nsn.com/news-events/press-room/press-releases/lte-throughput-leader-nokia-sets-world-record-with-sk-telecom-of-close-to-4-gbps-using-tdd

[2] http://www.ftthcouncil.eu/ documents/Reports/2013/Cost_Model_CAR_Germany_August2013.pdf

[3] Jing Jiang; Dianati, M.; Imran, M.A.; Yan Chen, "Energy Efficiency and Optimal Power Allocation in Virtual-MIMO Systems," Vehicular Technology Conference (VTC Fall), 2012 IEEE , vol., no., pp.1,6, 3-6 Sept. 2012

[4] S. K. Jayaweera, “Energy analysis of MIMO techniques in wireless sensor networks,” in 38th Annual Conf. on Inform. Sci. and Syst. (CISS 04), Princeton, NJ, Mar. 2004.

[5] Mischa Dohler: "Virtual Antenna Arrays", PhD thesis, King’s College London, University of London, Nov. 2013

[6] Pabst, Ralf; Walke, Bernhard H.; Schultz, D.C.; Herhold, P.; Yanikomeroglu, H.; Mukherjee, S.; Viswanathan, H.; Lott, M.; Zirwas, W.; Dohler, M.; Aghvami, H.; Falconer, D.D.; Fettweis, G.P., "Relay-based deployment concepts for wireless and mobile broadband radio," Communications Magazine, IEEE , vol.42, no.9, pp.80,89, Sept. 2004

[7] Ericsson Mobility Report, November 2013

[8] EU Commission staff working document, „Implementation of the EU regulatory framework for electronic communications – 2014“, July 2014

[9] Rappaport, T.S.; Murdock, J.N.; Gutierrez, F., "State of the Art in 60-GHz Integrated Circuits and Systems for Wireless Communications," Proceedings of the IEEE , vol.99, no.8, pp.1390,1436, Aug. 2011

[10] Souihli, O.; Ohtsuki, T., "Cooperative Diversity Can Mitigate Keyhole Effects in Wireless MIMO Systems," Global Telecommunications

Conference, 2009. GLOBECOM 2009. IEEE , vol., no., pp.1,6, Nov. 30 2009-Dec. 4 2009

[11] D. Chizhik, G. J. Foschini, M. J. Gans and R. A. Valenzuela, “Keyholes, correlations, and capacities of multi-element transmit and receive antennas,” IEEE Trans. Wireless Commun., vol. 1, no. 2, pp. 361 - 368, Apr. 2002.

[12] Almers, P.; Tufvesson, F.; Molisch, AF., "Keyhole Effect in MIMO Wireless Channels: Measurements and Theory," Wireless Communications, IEEE Transactions on , vol.5, no.12, pp.3596,3604, December 2006

[13] Ramirez, Rodrigo A.Vaca; Altman, Eitan; Thompson, John S.; Ramos, R.Victor M., "A stable marriage framework for distributed virtual MIMO coalition formation," Personal Indoor and Mobile Radio Communications (PIMRC), 2013 IEEE 24th International Symposium on , vol., no., pp.2707,2712, 8-11 Sept. 2013

[14] Jensen, AR.; Lauridsen, M.; Mogensen, P.; Sørensen, T.B.; Jensen, P., "LTE UE Power Consumption Model: For System Level Energy and Performance Optimization," Vehicular Technology Conference (VTC Fall), 2012 IEEE , vol., no., pp.1,5, 3-6 Sept. 2012


19

About the Institute for Digital Communications at the Friedrich-Alexander-University (FAU) Erlangen-Nürnberg

The Institute for Digital Communications (IDC) at FAU was founded in 2012 through the Alexander von Humboldt Professorship program of the Alexander von Humboldt Foundation (AvH) and the German Ministry for Education and Research. IDC conducts research in the broad areas of communications, signal processing, and information theory. Besides the Head, Prof. Robert Schober, IDC has currently two further professors, three postdoctoral fellows, and 12 PhD students on its staff. Before joining FAU in 2012, Prof. Schober was a Full Professor, Canada Research Chair, and Associate Department Head Research and International Outreach at the University of British Columbia (UBC), Vancouver, Canada. He is a Fellow of the IEEE, the Canadian Academy of Engineering (CAE), and the Engineering Institute of Canada (EIC). He serves currently as the Editor-in-Chief of the IEEE Transactions on Communications. He received numerous best paper awards from international conferences and journals as well as career awards for his research. His career awards include the 2002 Heinz Maier-Leibnitz Award of the German Science Foundation, the 2006 Killam Research Award from UBC, the 2007 Wilhem Friedrich Bessel Research Award from AvH, the 2008 Charles McDowell Award for Excellence in Research from UBC, and a 2012 NSERC E.W.R Steacie Fellowship from the Canadian Science Foundation. IDC’s current research focus areas include 5G, massive MIMO, small cells, cooperative diversity, and resource allocation.

About the Fraunhofer Institute for Integrated Circuits (IIS) in Erlangen

With 66 institutes and independent research units and liaison offices in Europe, the USA and Asia, the Fraunhofer society is Europe’s largest application-oriented research organization. 22,000 staff members work with an annual research budget of over 1.9 billion euros. Founded in 1985, Fraunhofer Institute for Integrated Circuits (IIS) in Erlangen, Germany, ranks first among the Fraunhofer Institutes concerning headcount and revenues. As the main inventor of mp3 and universally credited with the co-development of the AAC audio coding standard, Fraunhofer IIS has reached worldwide recognition. Software implemented at Fraunhofer IIS is currently installed in over 7 billion devices. The communication systems division within Fraunhofer IIS comprises 90 employees and is active in the fields of mobile broadband and broadcast, terrestrial and satellite backhauling, mobile satellite services, defense and telemetry. Among others, part of the air interface and receiver ASIC chipset of the SiriusXM radio system was designed by the communication systems division, and it contributed significantly to several DVB standards. Furthermore, the communication systems division has a long track record in development and implementation of prototypes, demonstrators and test equipment. By the end of 2014, the Fraunhofer IIS will be equipped with an LTE-Advanced (Rel. 10) over-the-air testbed with two macro-cells, and with an anechoic chamber realizing a virtual electromagnetic reality for multi-antenna devices.


20

Authors: M. Breiling, D. W. K. Ng, C. Rohde, F. Burkhardt, R. Schober, T. Heyn

Table 3

Selected topics for future 5G research projects

Analysis, simulation and design

Channel measurements for a SUDAS-type VMIMO setup; measuring the average power and correlations within the MIMO channel matrix and clarifying whether there is a keyhole effect in typical rooms

Resource allocation algorithm design for SUDAS infrastructure connected to multiple mobile network operators, including: QoS requirements and operator policies for user and service prioritization Choice of the used carrier frequency in the mobile bands and the mmWave band

based on the channel qualities in the mobile bands and the mmWave carriers Definition of the control plane between UEs and base stations (and possibly SUDACs)

to allow the aforementioned resource allocation Optimal MU-MIMO precoding for the Virtual MIMO system, especially in view of

SUDACs that are shared among several UEs Specification of a dedicated 5G air interface for Compress-and-Forward over the

mmWave link (as an alternative to using IEEE 802.11ad WiGig) Realization of mostly analog SUDACs for Amplify-and-Forward: Feasibility of flexible frequency-conversion between a configurable UHF frequency

and a configurable mmWave frequency Feasibility of flexible bandpass filtering to relay only the useful payload signal and

suppress surrounding noise and pilot/beacon/control signals

Implementation in prototypes and demonstration

Implementation of SUDAC prototypes and modified base stations and UEs Testing and comparing the achieved performance of Amplify-and-Forward,

Compress-and-Forward relaying using IEEE 802.11ad, and Compress-and-Forward using a dedicated 5G relaying air interface for the mmWave band

Determining the requirements: What kind of SUDAC infrastructure is needed in a typical home to satisfy the high data rate requirements of 5G: How many SUDACs are needed? In which locations and in which devices should SUDACs be integrated?

Verifying robustness and co-existence with other systems (like IEEE 802.11ad) in the same mmWave band

Measurement of the actual energy-efficiency of the overall system (including base stations, SUDACs, and UEs)

Documents

Whitepaper UE-Side Virtual MIMO Using mm-wave for 5G · FTTH further. As mmWave is strongly blocked by walls, a BS small cell transmitting in the mmWave band would be required in