Millimeter wave communications Robert Baldemair, Kumar Balachandran, Lars Sundstrém, and Dennis Hui Certain 3G METIS scenarios [1] such as Amazingly Fast, Best Experience Follows You, and Service in a Crowd create extreme requirements on data rate, traffic handling capability, and availability of high capacity transport respectively. These scenarios map 10 corresponding requirements that will entail support of over 10 Gbps, 10-100 times the number of connected devices, 1000 times the traffic, and 5 times lower end-to- end latency than possible through IMT-Advanced. The peak data rate requirements of these scenarios will entail acquisition of several hundreds of MHz of spectrum, These Tequirements do uot encompass 5G, but instead offer one avenue of stressing system capabilities along a limited set of dimensions, Several traffic forecasts [2][3) also predict a\tenfold increase in traffic volume from 2015 to 2020. The 5G requirements of interest to this chapter relate mainly to data rates and traffic volumes and can be met using techniques that are tried and tested in past generations of mobile networks. These are to (1) gain access to new spectrum, (2) improve spectral efficiency, and (3) densify the networks using small cells. In the case of 5G, these techniques are given new life using two means: the use of millimeter Wave (mmW) spectrum for the availability of large blocks of contiguous spectrum, and the subsequent adoption of beamforming as an enabler for high spectrum efficiency. The propagation of millimeter waves is naturally affected by physics to reduce coverage to shorter ranges. Ulira-Dense Network (UDN) deployments are therefore a consequence of the choice of frequency band, and will lead to a tremendous increase incapacity over the covered area The increase in spectral efficiency arises out of the drastic reduction of interference in relation to signal power due to the high gain beamforming Spectrum and regulations The primary motivation for using millimeter waves isthe promise of abundant spectrum above 30 GHz. While mmW spectrum spans the range from 30 GHz-300 GH, it is Widely believed that the reach of mass market semiconductor technology extends up to found 100 GHz and will inevitably surpass that limit with time. Microwave bands from ing extreme requirements for 5G, and much of 3GHlz-30 Giz are just as relevant to mes {2 Motte ans wiles Communications Technology. eA. Oncian, 1. F. Monsera, and F. Mach Dshed by Cambridge University Press. © Cambridge University138 Figure 6.1 Robert Baldemair, Kumar Balachandran, Lars Sundstrém, and Dennis Hui 10-30 GHz 30-100 GHz gaa 3GHz 30 GHz 300 GHz /_-— emw ——— >}. mmw—>} ‘The focus of SG deployment is on frequency bands up to 100 GHz, with the lower end ofthe range being favored; LTE expands reach into higher frequency bands around 6 GHz the discussion in this chapter is relevant to those parts of the ceutinieter Wave (emWj) band outside of the reach of existing systems as well, namely the region 10 GHz~30 GHz (see Figure 6.1). The technological abilities of the semiconductor industry do not however automatically translate to viable network architectures. In general, lower frequency bands are more appealing for implementation and pose fewer risks for systemization, while higher ranges of frequency are better for access to wide band- width, but worse for device and system complexity. Most frequency bands below 60 GHz are already assigned to various services, including mobile services; these assignments arc made through treaty arrangements within the three regions addressed by the ITU-R in periodic World Radio Conferences (WRC). Millimeter wave bands in current use are predominantly employed for radar, earth exploration, point-to-point services, and satellite communications, etc. Some of these bands are assigned in co-primary fashion to mobile services; there are of course no terrestris! mobile services in operation above 6 GHz. The 60 GHz ISM band does provide as much as 7 GHz of spectrum for unlicensed use in most parts of the world; the band has been used by IEEE 802.1 in the recent “ad” amendment to create a physical and Medium Access Control (MAC) layer capable of peak rates up to 7 Gbps [4] The specification is being utilized by WiGig for point-to-point mmW Links over 2.16 GHz channels for video and data transfer. In addition, the 802.1 Lay task group a IBEE 802.11 is examining channel bonding and MIMO as solutions for even hight throughput systems (over 30 Gbps) for video transfer, data center applications, and point-to-point communication [5]. There is considerable interest in industry in expanding mobile services further in'® microwave spectrum, including mmW spectrum. Inquiries by the Feder! Communications Commission (FCC) in the USA [6] and the UK regulator Ofcom [7] have also attempted to gauge industry seriousness about venturing into such spect” frontiers. These references and subsequent responses from industry partners provide a incomplete but evolving picture of regulatory issues. For a more in-depth discussion spectrum for SG systems see Chapter 12, At the time of writing Electromagnetic Field (EMF) exposure limits determined Y the FCC [8] and ICNIRP [9] independently create inconsistencies in allowable Po! limits above the transition frequencies of 6 GHz and 10 GHz, respectively {!0 The policy guidelines pertaining to EMF will likely need modification if om a mmW operation above the transition frequency is permitied for mobile serviees%») fe ro fer for ad se ns ter 10 62 fay Mitineer wave communications Channel propagation Ie ust be pointed out that eon frequency; the frequency. dependen 2 Ay mobile radio system at millinter en eS antenna arays of very high onder sence and attenuation ‘quate of operating frequency, lependent of frequency if the : Kept constant over the variation of min SOtopcaly refemed fee space loss on be tthe transmitter and receiver, ves will need beamforming us on. eases proportion al tothe s ling loss ean be kept © Kept ind stenificant Obstacles inthe signal path typically reflect nergy, andthe effet of oiage is rapid attenuation along the incident signal path and diffuse searing fom reflectors Diffaction attenuation increases as the wavelength gets shorter {11} Small-scale variations can be modeled using site-specific geometric models, sta tical models based on general characteristics of propagation, and hybrid approaches. Itis expected that narrow beamforming using high-gain antennas will reduce channel dis- persion. Ray tracing is a useful tool in modeling propagation, with statistical variations provided by the presence of diffuse scatterers in the envionment representing objects and non-smooth characteristics of surfaces, and by the modeling of comer effects due to diffraction, Building materials wil differ inthe absorption, reflectivity, and transmission characteristics and will be affected by incident angles to the surface. Unshielded windows can provide ingress to signals, while exterior walls are usually opaque, Wall losses within a building can be severe and outdoor to indoor connectivity will often need site planning through placement of antennas on all sides ofa building, especialy at the higher mmW frequencies. The effect of body loss and attenuation from mobile users or Mi detain onopgaton modeling sven Caner 13 of isbook For mmW propagation measurements, see e. [12][13] Hardware technologies for mmW systems Device technology Radio Frequency (RF) building block per! frequency. The power capability of powe yas ‘echnology roughly degrades by 15 4B per decade, s formance generally degrades with increasing i sircuit ar for a given integrated circui pee ywn in Figure 6.2. There is142 Figure 64 6.3.2 Robert Baldemair, Kumar Balachandran, Lars Sundstrém, and Dennis Hui FoM [tu/conv-step] 108 108 108 1010 fenyg (Hal ‘The Walden FoM representing the energy per conversion step is plotted against Nyquist sampling frequency for large number of published ADCs, The FoM envelope (dashed) represents the achievable lower limit using technology from around 2015. conversion beyond a few hundred MHz of sampling rate, roughly amounting to 10 times increase per decade. Although this FM envelope (dashed line in plot) is expected 1 be slowly pushed toward higher frequencies by continued development of integrated circuit technology, RF bandwidths in the GHz range are still expected to give poor power efficiency in the analog-to-digital conversion Larger signal bandwidths also impact the complexity and power consumption of he digital circuitry. While Moore's Law has enabled a virtually exponential complexity ‘growth for decades, in recent years concerns have gradually been raised on the longevity of this technology evolution. The problem lies in that pure scaling of geometric features 1s the engine for this progress is quickly approaching its limits, Solutions to this problem ‘considered include the introduction of INV materials, new device structure (FinFET, nanowire transistors, ete.), and 3D integration. However, none of these will serve ‘a vehicle for continued exponential improvements. Another problem is that the cost et digital transistor or function has been seen to flatten out or even increase as CMOS technology feature size goes below 28 nm. Nevertheless, afew more technology cycles «with a corresponding reduction of digital power consumption, are expected before 2 ‘There are many more building blocks and associated limitations beyond the ones mentioned above, but those treated are viewed as the most challenging, and are worth) of further study. Antennas “The small footprint of antenna elements at mmW frequencies is on the RF frontend eircuitry driving the antenna clement, This enables and sizeof ar with the si jes and post433 ey interconnects between chip oveve, sir fom permittivity and typically high doping of the Substrate. Talo z Peers atthe ‘ally Not possible as the sut e tity Sed whe tered cet tecnoy nae a dy orn of integrated circuit technology scaling on, WW efficiency due to high ng the substrate character- Larger antenna arrays help preserve the antenna Consequently, the directivity increases and the anten a phased array that can electrically steer the beam using a more general beamforming architecture Boamforming architecture A digital beamformer, see Figure 6.5(a), where each antenna element has its own eore- sponding baseband port offers the largest flexibility. However, ADCs and Digital-to- Analog Converters (DACs) operating at multi-GHz sampling rates are very power consuming; a full digital beamformer with several hundred antenna elements might be infeasible or at best feasible but very power hungry and complex. Therefore early mmW communication systems are expected to use analog or hybrid beamforming architectures, In analog beamforming, see Figure 6.5(b), one baseband por feeds an analog beam- forming network where the beamforming weights are applied either directly on the pad) spiiter [1 i ate ecoder 10. spliter i) - : ae a = base band port pac] ‘base band port (0) fa) fi) eamforming architectures () dtl (0) analog (hybrid.144 Robert Baldem: Kumar Balachandran, Lars Sundstrém, and Dennis Hui analog baseband components, at some intermediate frequency, orat RE. For RF beamforming network eee May consist of several phase shifter: element, and optionally also variable gain amplifiers. a ing network typically generates physical beams but Pattern. Especially ina multi-user environment this cat separation is not sufficient. Hybrid beamforming, see Figure 6.5(c), is a compromise between those two ah a digital beamformer operating an a few baseband ports j i beamforming network. This architecture enables a compromise with respect to bo complexity and flexibility between analog and full digital beamformer. A beamforming receiver provides spatial selectivity, ie. efficient reception of signals in desired directions while suppressing signals in other directions. Each individual antenna element, however, does not provide much spatial selectivity. For a digital beamforming receiver this means that each signal path, extending from respective antenna element all the way to the baseband port, will have to accommodate desired as well as undesired signals. Thus, to handle strong undesired signals, requirements on dynamic range will be high for all blocks in the signal path and that will have 1 corresponding impact on power consumption. In an analog beamformer, however, the beamforming may be carried out already at RF and thus all subsequent blocks will need less dynamic range compared to the digital beamforming receiver. While the digital beamformer requires a complete analog RF front-end including ADCs and DACs it does on the other hand not need to distribute RF signals over long ‘distances to a large number of antenna elements as is the case for analog RF beamform- ing. The power consumption saved by this does not compensate for the power footprint of a digital beamforming architecture The ability to create a physical beam in a given direction does not require high resolution and precision of the beamforming weights. An analog beamformer in many cases will be just as good as the digital counterpart with regards to antenna gain, The challenge rather lies in the degree of suppression of side lobes. Accu) of direoted nulls in the radiation pattem is even harder to realize, tn these mates analog beamforming is inferior to digital or hybrid beamforming, particularly higher frequencies. ‘The reader is referred to Chapter 8 for more information on beamforming. men ‘one pet antenng Inany case, an analog beamto, cannot generate a complex bean n lead to interference, if pure bean is followed by an analog Deployment scenarios ve Most outdoor deployments of 5G: mmW networks will intially ae ies “ 10 GHz, As the frequency of operation increases, they will be made a tee Justers of infrastructure nodes or UDNs in urban areas (cf. Chapter aca ae it UDNs). While the use of highly directive transmis a jignal-to-Noise Ratios (SNRs) and even reach a bei ration, area coverage will be limited by the rae me: : ‘hnological limitations of the more information about provide significant Si ‘a good channel realiz Jevels conducted into antenna ports due to the tec!Millimeter wave communications 145 Typical ble nents wile the ae spvonito migra eee the lower: iicrowave bands that LTE occupies woday, wit inter-site distances of 0g 200m, A deployment would primarily provide coverage using macro sites share rooftops, with fo caratden eng Deine Provided by sweet evel pico-base stations. The primary feaon tol Considet CmaW tectinology/sithe expecialien of improved Gelioe ay cae mance below 30 GHz : Spectrum above 30 GHz is useful for near line-of typically interior or exterior spaces that are com of electromagnetic felds. Coverage in such environments is provided by dense deploy ‘ment of infrastructure nodes within and around hot spots tha ten to concentrate traffic Millimeter wave bands ahove 60 GHz are well suited fur short range point-to-point {inks for backhaul. These bands will typically support higher bandwidth communication than access links, and can support high-reliability performance requirements for the data plane as well as provide excess bandwith for link management and radio system ‘monitoring, Such bands will also find use for short range applications that need very ‘wide bandwidth capability, such virtual reality applications. ‘One deployment scenario for systems in mmW bands is self-backhauling, which may be formally defined as use ofan integrated ar interface to provide multiple access as well as transport over one or more hops, Self:backhauling will nse the same hasic physieal layer as the access link and specialized MAC modes. Although not required, access and backhaul could share the same spectrum band. There are several scenarios that are enabled by self-backhauling in indoor and outdoor locations: ight environments; these are cted in a way that allows propagation as video transmission, virtual office or augmented and. + Macro-to-pico deployments, typically from above rooftop to below + Elevation coverage from ground level ) ‘ af + Outdoor to indoor coverage + Successive links along a roadway or an open space Ls i ny i ( ce (8) Outdoor self-backhaul for small-cell connectivity and (b) access and transport using self-backhaul along a road,Robert Baldemai extension of covers fiber-access, and th. $0 that base stations in the in ipported by the quick transfer 18 Over he © creation of gia Opportunities for connectiv connectivity, frastructute cans information, and mobility is su While self-backhauling is expected to share t access, the specific MAC protocol modes that Faguon Will differ from the MAC procedures used for mobile users. In mos ca, infrastructure nodes will be stationary, and self-backhaul links ean associate eich Hansmitter to multiple receiving nodes, with a scheduling algorithm to determina, Tate of active communication between any two nodes and the roles ofthe nls transmission o reception. The MAC protocol for selfbackhauling can change rus, and bandwidth assignments dynamically on the basis of changes in trafic distbuton and changes in the interference environment. Self-backhaul can therefore be pr: sioned for higher degrees of reliability than possible for the multiple access links, where interference statistics are more unpredictable. Several other deployment scenarios have been identified in the IEEE 802.15.3¢ and IEEE 802. lad amendments. A new project within IEEE known as 802, lay is examio- ing channel-bonding and MIMO for deployment scenarios such as backhaul. Millmsts wave radio is also being considered for inter-chip communication, communication between racks of'a data center [16], ete Architecture and mobility It is important to recognize that the 5G system will not be a complete aed of LTE with a new air interface, Future releases of LTE in lower Bese will provide a foundation for 5G that ensures wide coverage. It is eee new 5G air interfaces in higher spectrum bands are integrated into ee in a way where they can transform the ability of an operator to pi through the network, and to improve network utility and user exp or dat great flexibility, by enabling improvements in delay, reliabi pene lume, Some ofthese improvements willbe engineered by revolutions er. is he core network, e.g. the use of Software Defined Networ a aan jetwork resources, and employing Network Je. Air interes to configure and isolate ni ; p10 rtualization (NFV) to dyna rovision processing or storag ation (NFV) to dynamically pi PI . bands above 10 GHz will also need to be integrated with LTE '0 connectivity and mobility.652 6524 Millimeter wave communications 147 While Chapter 3 contains a P tains « comprehensive discuss jgcuses on three aspects of particular rel, jon of SG architecture, this chap cane ‘mobility management for phanton es levance for mi using phantom cel m cells appears in Dual connectivity When a new mmW air interfac Neo 2 terface is introduced, dual cone feature w prevent loss of coverage, Even for co-located denlommere Paves hen operation will provide eter coverage eto lawerdiatonoeninpened nay ssoch. The primary festure of dl connectivity iso ensue tt he terol aaa lower spectrum bands for control signaling, The Ber exotica iain ae igr at 1g. The primary purpose is fast fall-back to etter coverage if mmW band coni The requirements of 5G 10 enable dense deployment wih wiles af nltnting Fetes igh labor kate wl eat af ap a Giferent from LTE. As itis unitely iat he lower layer of te radio proaclsinck Sr ue yi Cla fis ee veteran Ge TER decess technology will ccuratan evolved muli:potcalconvergnclayer ator above Layer 3. The convergence layer would handle data plane and radio resource contro flows independently [17]. Dual connectivity is not meant to be a requirement, but a convenience. Mobility Phantom cell Ina heterogeneous deployment of a covering cell and small cells, itis convenient if mobility between small cells is transparently handled. Transparent tracking of terminal mobility ina small-cell layer covered by an overlay network is known asthe phantom cell concept [18][19]. The concept is already well known and proposed for LTE, and ean be applied in a similar form for the mmW air interface, although the exact timing between the overlay network and the 5G mmW air interface will not coincide as in a synchronized LTE deployment of small cells. q ‘The phantom cell concept, illustrated in Figure 6.1(b), is a special case of dual connectivity and involves separating the contol and user plane of the overlay cell and small cell so that the control and signaling pertaining 1 the small cell always fare orn the overlay cell, The opnscdvence of EE that radio level mobility is tracked only within the overlay i nding wi y erlay network, while the radio equipment correspo A the eee ae is of istinguishable from antenna vr layer ¢ small cells appears indi "Te differences in data rate and latency aS overlay network. Sc rl oe Specification of the phantom cell useful for 5G. d the 5G ai 1 concept, but the principles are without doubtFigure 6.7 Kumar Balachandran, Lars Sundstrém, and Dennis Hui i ae ) a) Ake (Ah. Lec EL ¥...._ Gusleannechiy re . “Clone (a) (o) 'y, while (a) co-located with LTE, (b) deployed ay or support phantom cell operation, with LTE assuring coverage, 5G mmW radios can support dual connectivity small cells for high capacity, ‘Terminal specific serving cluster At mmW frequencies, a terminal can quickly lose connection with its serving base station due to shadowing by other moving objects. It is important that a terminal hs 2 backup connection (either via another path to the same base station or to another base station) already prepared so that it can quickly switch if needed. In this way, the currently serving base station and those base stations with which the terminal has 4 prepared backup connection form a terminal-specific Serving Cluster (SC), which typically consists of base stations that are in the vicinity of the terminal. A releted concept [20] has also been proposed for communications in traditional cellular bands albeit for a different purpose A Principal Serving Access Node (P-SAN) in each SvC is responsible forthe connection between the SvC and its associated terminal, Most ofthe data flow between the terminal and the network passes directly through the P-SAN. Other base stations in the SvC are Assistant Serving Access Nodes (A-SAN) which act to provide diversity when the connection between the P-SAN and the terminal is lost (e.g. due to am obstacle), as illustrated in Figure 6.8. The P-SAN manages membership in the SvC and can proactively wake up sleeping base stations for inclusion in the SvC. A reliable backhaul connection with sufficiently low latency is required between the P SAN and each A-SAN in the SyC. 'A-SANs are typically lightly loaded nodes and must have spare radio resources and processing capacity to assist the SvC. It is additionally beneficial to provision a storage at the A-SAN to proactively buffer user data for immediate forwarding (0 terminal. sly be SvCs of different terminals may overlap. Thus, a base station ean simultaneous) an A-SAN and a P-SAN for two different SvCs. In the role of A-SAN, a base SY temporally assists a P-SAN to communicate with its terminal using ee ee ‘When a significant amount of data flows through an A-SAN for a prolonged Poget 6 at Mittimeter limeter wave communications 49 Sela SevingaN Assistant : eer IMustration ofa terminal-specific serving cluster with a serving beam from P-SAN blocked. time, the A-SAN will take on the role of P-SAN affer the completion of network provedutes for transfer of context from the original P-SAN to the new P-SAN, which may subsequently recruit new A-SANs or remove old A-SANs from the SvC. ‘The resulting migration of SvC provides a means for slowly tracking the long-term movement of a terminal across the network. It yields a desirable hysteresis effect in handling the potential rapid changes of connection poins as terminals move ina densely, and possibly irregularly, deployed network. Itis also well suited for distributed mobility management where no centralized mobility management entity is needed, which is particularly attractive for user-deployed, self-organizing networks. Beamforming Beamforming techniques As outlined in Section 63.3, beamforming functionality can be implemented in log beamformer where the desired beam several ways. Figure 6.5(b) shows an anal ee tirection is steered via the phase of the analog phase shifter. Created Beams #r ag s the system bandwidth, Data rate wideband, i.e, the beam direction is the same across tan be inerensed with two analog beamformers implementing spatial ‘multiplexing using polarization diversity In principle, ifthe propagation environment is rich enough and multiple strong paths exist between source and destination, each beam can convey upto two layers Eve ‘analog beamformer is required per layer. However ‘analog beamforming allows a Es simple beam shapes but does not enable advanced techniques such as creating nullsFmt 59 signal quality at each reso resource slot or beam identineg d Tepons back to the tification ranst This approach scaled inca bear sent te mont penne pes Oe | beams, this approach therefore log, NV bits ded for linear be mount of feedback nee ‘Tree scan A more efficient approach is ach isto divide the be linear beam scans and use an 8 Senne ces ino mile sge of Fae eae a iferent sang at etch age, illustrated in Figure 69 for i array of eight antenna elements. At te fist stage, these antenna clements may be next 10 each other, and thus forming wide beam pattems At subsequent stages, the spacing among adjacent antenna elements in each sub-array increases, and thus forming intentional grating lobes of reduced widths, At low SNR ‘where the receiver is power-limited, both linear scan and tree scan require the same ‘number of resource slots for the receiver to accumulate edequate amount of energy to identify the best beam direction. At high SNR, however, this approach only requires 4 (otal of 2log,N resource slots and is exponentially more efficent than linear beam scan. The amount of feedback needed for tee scan is roughly log bits about the same as that for the linear beam scan. a Stage 1 a Stage 2 Stage 3 3 10 3 0— ~ Bo oh 5-9 boli | 2% Te 000 100 tb) (c) Active: ) ene ‘used in (a) Stage 1, (b) Stage 2, and (c) Stage a Angular power distribution and the activated antenna el elements. te linear array of eight antenna ofteben oo ase6.6.2.3 67 674 67.2 Robert Baldemair, Kumar Balachandran, Lars Sundstrém, and Dennis Hui Random excitation ‘other efcient beam scanning method [24 st transmit a pilot signa) fandom direction in each resource stotby applying beamforming Weightcu, according toa pseudo-random sequence For example, the beamforming ng consist of pseudo-random phase shifts. Ifthe recei rr isn iver has the knowledge of the 5 : OF the sey Possible pseudo-random beamforming weights used over all resource soe es transmitter (e.g. through shared random seeds), re the reeeiver can determine wit ng scattering environment using compressive sensing techniques, From ten complex gains, the receiver can then derive the best beam in the beum codebook feeds back its index. The amount of feedback is again log, a N bits. Similar to tee scan, low SNR where the receiver is power-limited, both random excitation and linear tea Scan require the same number of resource slots to identify the best beam dieeios At high SNR, the number of resource slots needed for random excitatio Proportional to the number of significant scatterers that exist in the environ can be substantially smaller than the number of antennas at the transmit receive Probability the complex g: underly; mn is diteetly iment, Which ler. However, for this method require more complex processing than those for linear beam sean and tree scan, Physical layer techniques Duplex scheme Millimeter wave communication systems are expected to be deployed in small cells with infrastructure densities ranging from indoor deployments to densities corresponding to very dense macro cells of today. The number of users within the coverage area ofa cells. thus small leading to stiong variations in uplink and downlink traffic demands depend ing on the services currently used by the few covered users. This speaks in favor of dynamic resource partitioning, where resources can be assigned to the two transmission directions on demand. Millimeter wave communication systems are likely to operate over large bandwidths ranging from several hundreds of MHz up to one GHz and beyond. Such large amoun's of spectrum are likely to be unpaired for reasons of convenience and the unavailability of practical duplex filter technology in those bands. Flexible duplex assigns transmission resources for data dynamically to either trans- mission direction, allowing more efficient use of bandwidth for communication To enable power efficient terminal operation control resources will in most cases still follow a fixed structure, ‘Transmission schemes Mobile systems operating above 10 GHz will operate over vastly larger bands oe cellular systems of today do. Depending on terminal capabilities not all terminalsMitimete wave Communication sas nent support for the complete s quire the full bandwidth for its current ‘or power limitations [25}, transinssion, ei Based on these observ: » fraction of the system by ystem bandwidth, pepiolene me 'c Fourier Transform Spread OF a ady wed in LTE remain therefore good theses 2 nT) schemes for 5G are analyzed in Chapter 7. mae der to adopt Frequency Divison Malaga navi see dicing EDM(A) one needs icp in min tat many sin W sss il ea thin bemlonee ea te lnk tthe challenging propession condom preva ae nee Typically, only a single user will be in the covers : ither frequency select ‘beam directions or multiple wideband eum. The fst option requires digital beamforming while the latter can be implemented using. analog. hybrid eee analog or hybrid In [26], a mmW communication system is described for the 72 GHz band based ona single-carrier scheme that is based on Time Division Multiple Aecess (TDMA) and Time Division Multiplex (TDM), quite unrelated to DFTS-OFDM, The design is straightforward and utilizes very short transmission intervals for multiplexing users. A guard interval, also known as a null cyclic prefix allows easy frequency domain equalization. The Peak to Average Power Ratio (PAPR) of the scheme is ower than in OFDM and it has lower out-of-band emissions. Complexity is higher than in OFDM. (One attempt to harmonize these contradicting views can be based on operating frequency. The discussion in this book tries to cover the range from around 10 GH2~ 100 GHz while [26] discusses frequencies around 70 GHz. Systems operating at 70 Gitz are expected to cover fewer users and o use shorter transmission times ‘both facts that make pure TDM(A) more than systems operating closer to 10 GH2, attractive, The large operating range of a and the ability to support much higher bandwidths at 0 suggest changing numerology forthe signal vaveform. Irmight ‘well be desirable for ands lower than 1 GHz. Figure 6.10 illustrates three 5G systems to be introduced in bans Se ee mmW communication systems (10 GHz~100 GHz) Ine upper ranges of spectrum different choices of OFDM (or DFTS-OFDM) param 1 GH2-100 GHz [27]. Wider subcarrier Providing increased robustness t0 In the following, a short qualita! (incl, DFTS-OFDM) and OFDM is P Provided in [28] ‘bandwidth is specified for higher bands, thus Doppler and phase noise. ive comparison between single-carrier modulation rovided, and a more detailed comparison 15Figure 610 Robert Baldemair, 5 aa ae ee 10 100 Carrier frequency (GHz) ‘Three different OFDM numerologies ae proposed to cover the range from 1 GHz-100 GHz OFDM has a higher PAPR than single-carrier modulation and thus requires lax ower backoff.at higher output powers. This does not only limit peak transmit power but also biases the power amplifier into a less power efficient operating region. Futhermoe, dynamic range increases with PAPR, which requires higher-resolution ADCS forthe same quantization noise. In Section 6.3.1, itis shown that both the power capability of Power amplifiers and the resolution of ADCs ate important parameters for nmW hardware design, Link performance comparisons between OFDM and single-carrier modulation not considering hardware impairments often end up in favor of OFDM. OFDM typically outperforms single-carrier modulation over frequency selective channels while over flat fading channels the difference is much smaller, Frequency selective channels correspond to Non-Line-OF Sight (NLOS) conditions whereas in flat fading channels ‘one path — typically the LOS path — dominates. When adding impairments the comparison becomes less clear [28] Single-carrier modulation poses more restrictions on the system design than OFDM does since it restricts multiplexing of signals to the time domain. Single carrier modulation provides lower PAPR than OFDM, a key advantage, provided multiple signals (e.g. data and reference signals) are not transmitted simultaneously from the same power amplifier. No clear consensus on transmission schemes for 5G mmW communication systems exists, Given the large considered spectrum range from 10 GHz-100 GHz, this surptising and the answer can yery well depend on the operating frequency stemsMitimeter ‘er wavecommuneatons 45 5 both OFDM and singleve izband, sects seven te witless penny Sle-cartier modulation, Wireless Hg neil layers bi foe witeless transmission of high-definition wanes te ly 4 proprietary stand deaths y Standard aiming pbrysical layers based on OFDar 60 GHz band, defines multiple ; The perso 02.15.36, also operating in the 60 GHs henna 2 onl Networking standar Conclusions extent of emW frequencies, off e : However, itis clear that the unnge of millers woe ee unications, in fact also covering a large and access, in particular toward the upper rious challenges that have to be One of these is the propagation condition that milimeter waves experience, particular the effects of foliage, difraction, and body loss, rendering a large extent of beamforming necessary. These radio conditions imply tha for higher frequencies also control signals have to be beamformed, and beam finding and beam scanning techniques have to be applied Furthermore, there are various challenges on the hardware side that have to be overcome, namely the decreasing power efficiency and increasing phase noise toward higher carrier frequencies, and alto the decreasing A/D conversion power efficiency and increasing device complexity toward larger system bandwidths. These aspects require further innovation on the hardware side, and/or a 5G radio design that is able to compensate or alleviate their impact. On the positive sie, shorter signal ‘wavelengths enable a high antenna integration and hence the usage of on-chip oF in- package antennas. Cleary, the more hostile radio conditions for higher frequencies, such 25 the possibilty of sudden signal blockage, require tht mW communication sluions allow for a quick backup connection, for instance through another mm note O° Neve: weqtenoy ;adlo Invites 0 fo ennugae ea Ss Slapler hes pointed oat different dul sonmeee Ny QEnOME SEES and user plane spit between a lower and higher frequency io) called phantom cell concept. Ultimately, the chapter has compared diferent pis communication, where there is no consensus Yet PF based solutions are likely most relevant fora ae POT ON Tange. For the upper spectrum ranges single-carrier app the improved power efficiency. ifverent physical layer techniques for mW 3 er ead an indian that OFDM ‘of the considered frequency Jyes may be suitable due (©Miliveter wave communications 157 Conf Miami, December 2010, p, jog + “10a! Teleco Bertling, “Tight integration of new SG air fang ments,” in IEEE Vehicular’ Techno] = workshop un emerging technologies for te SG Rapes gies for LTE Advanced and Beyond-4G, formance Surve vey 1997-2915. idesurvey htm! Oi [Ontine} hutp:/iveb SR. Ryu, B. Bane "Hee, and R. Prakash, “Wireless data ynmunications n, J. Vikberg, and P. Schliva face and LTE to fulfil SG require ce, Glasgow, May 2015, fama, Tang Hai, Shen Xiaodong, ‘enhancements in LTE Advanced,” ne, vol. 51, no. 2, pp. 98-105, February 2013. 'e with local power negotiation for coordinated multi ‘Vehicular Technology Conference, Yokohama, May Prelei, and R, W. Heath Ir, “MIMO precoding and combining solutions for millimeter-wave systems.” IEEE Conmuntcaions Magazine, vol. 52, no. 12, pp. 122-131, December 2014, [23] L. Zhou and Y. Ohashi, “Efficient codebool-based MIMO beamforming for millimeter-wave WLANs,” in IEEE Intemational Symposium on Personal, Indoor and Mobile Radio Communication, Sydney, 2012, pp. 1885-1889. [24] W. U. Bajwa, J. Haupt, A. M, Sayeed, and R. Nowak, "Compressed channel sensing: A new approach to estimating sparse multipath channels,” Proceedings ofthe IEEE, Vol. 98, no. 6, Pp. 1088-1076, 2010 (25] R. Baldemair, T; Imich, K. Balachandran, E. Dahiman, G. Mildh, Y. Son, S. Parkvall, M. Meyer, and A. Osseiran, “Ultra-dense networks in millimeter-wave ffequencies,” IEEE Communications Magazine, vl, $3,n0. 1, p. 202-208, January 2015 26) A. Gi sa dak, R_ Ratasuk, P. Moon, F. W. Vook, Pe a TG NasCaraey J 8 Sim and 8 Ne, “Milinerwave T. S. Rappaport, G. R. MacCartney Jt, ‘ Tigh dat-ate approach for future wireless enhanced local area systems: A hig Ce Setworks:" IEEE Journal on Selected Areas in Communications, vol. 32, 00. 6, pp. 1152-1163, June 2014. [27] ICT-317669 METIS project, ‘A ‘ Deliverable D2.4, February 2015, hitpsi/ww deliverables! (28] TS. Rappaport, Wave Wireless Communications, proposed solutions for new ratio acess” " /w.metis2020.com/documents/ N. ‘k, Millimeter at sR . Dail ad J.N- Mure cu ee ‘New Jersey: Prentice Hall, 2014.