Figures of Merit for Active Antenna Enabled 5G Communication Networks

Transcription

1 Figures of Merit for Active Antenna Enabled 5G Communication Networks Jan M. McKinnis Ball Aerospace Westminster, CO, USA Dr. Ian Gresham Anokiwave, Inc Billerica, MA, USA Randy Becker Keysight Technologies Santa Rosa, CA, USA Abstract Active antennas enable novel spatial techniques and beam-forming technology crucial to overcoming millimeter wave propagation challenges for fifth generation (5G) communication systems. At millimeter wave frequencies, a large effective aperture can be accommodated in a physically small area to overcome the high channel loss in these spectrum bands. The ability to dynamically steer and shape active antenna beam(s) to track users, overcome changing channel conditions, and focus the radiated energy into the desired direction provides additional degrees of flexibility and enables better performance for 5G radio system designs. To provide an over-the-air interface for previous generations of radio access networks, traditional architectures have relied upon separate, passive antennas connected by radio frequency cables to active radio transceivers. Active antenna systems, also known as phased array antenna systems, are an advancement from these previous radio access architectures. To implement an active antenna, a array of active radiating elements is utilized to combine passive antenna functions with active amplification and signal conditioning capabilities. Active antennas are an enabling technology for millimeter wave 5G communication systems that create a fundamental architecture shift requiring new Figures of Merit (FoMs). The 5G active antenna FoMs defined in this paper provide methods for antenna performance comparisons and wireless system evaluation. Keywords Active antennas, phased arrays, planar electronically steered arrays (ESA), 5G, New Radio (NR), Figures of Merit, spatial techniques, beam steering, beam forming, error vector magnitude (EVM), millimeter wave, radio networks, mobile communications I. INTRODUCTION To support a wide variety of new applications and market opportunities, fifth generation (5G) wireless systems must provide far higher data rates than previous communication networks. Unlocking more bandwidth and improving efficiency is required to deliver these higher data rates. Bandwidth is available in the millimeter wave spectrum, but propagation challenges at these frequencies will compel the wireless industry to implement innovative technologies. For the next-generation, millimeter wave, multi-gigabit networks, it is essential for radio access technologies to use new spatial techniques. Dynamic beam management and beamforming active phased array antenna systems are revolutionary changes that are rapidly coming to the wireless industry. Active antennas are software-enabled, adaptive systems that combine antenna capabilities with a number of transceiver functions. These active antenna systems will fundamentally alter the relationship between antenna and radio, requiring novel over-the-air (OTA) test methods and performance metrics. New 5G active antenna figures of merit (FoMs) are required to allow comparison of various solutions and predict system performance. II. 5G MARKET DEMANDS The communication industry and the 3rd Generation Partnership Project (3GPP) are actively planning for a new, wirelessly connected 5G world [1] [2] [3] [4]. Previous wireless applications have primarily focused on connecting people and allowing users mobile access to the internet. The 5G vision of the future dramatically expands on these past efforts to develop systems that will also rapidly and reliably connect an extremely large number of things to other things (also referred to as machine-to-machine communication). Novel wireless applications and verticals are being developed to implement concepts such as connected cars, massive mobile broadband, virtual and augmented reality, Internet of Things (IoT), remote manufacturing, and remote health care. These new markets will demand reducing latency and increasing data rates by orders of magnitude. The requirement for significantly more data capacity is driving 5G radio access technologies into the higher frequency millimeter wave spectrum to access the large amount of bandwidth available. Operating in the millimeter wave spectrum presents technical challenges. Higher transmission losses due to blockage, non-line of sight conditions, and increased atmospheric, material (e.g., foliage, buildings), and weather-related propagation losses will require beam-forming antennas to achieve robust and flexible coverage. The wireless industry has announced near-term plans for commercial 5G deployments utilizing millimeter wave spectrum. The industry is aggressively evaluating innovative solutions, and millimeter wave active antennas are rapidly becoming an essential, enabling technology for next-generation base stations, small cells, customer premise equipment (CPE), and supporting channel sounding test equipment. XXX-X-XXXX-XXXX-X/XX/$XX.00 20XX IEEE /18/$ IEEE

2 III. ACTIVE ANTENNA TECHNOLOGY AND BENEFITS Active phased array antennas have been widely used for a variety of purposes ranging from medical imaging to military applications. Active antennas are composed of an array of active radiating elements. By controlling the signal phase at each element, a combined, highly directive beam can be created. The resulting beam is electronically steered, so it can be rapidly pointed from one desired direction to the next without mechanical motion. To support 5G applications, active antennas will integrate phase and amplitude control stages for both transmit and receive paths. An active antenna s radiation performance is determined by a combination of the hardware implementation choices and its commanding software. Active phased array antenna technology provides numerous advantages [5] [6] [7]. Electronic beam steering supports fast, agile, and accurate beam pointing to meet the low latency requirements of 5G. Array technology is scalable to meet various connectivity demands. Increasing the number of radiating elements will improve receive sensitivity and increase transmitted radiated power. By properly applying amplitude and phase tapers across the radiating elements, active phased array antennas can be software controlled to generate narrower beam widths for individual data channels and wider beam widths for control channels. Focused narrow beams will control interference and maximize a user s signal-to-noise ratio to improve data throughput and thus optimize network performance. Wider beams allow flexibility for the control channels, which are required to maintain radio links, to be broadcast to the many users in a serving cell s coverage area. Array tapers can also be applied to control sidelobe levels to reduce radiation outside the desired main beam. Sidelobe control can be used to decrease network-level interference, resulting in improved link quality, increased wireless channel availability, and improved system capacity. At millimeter wave frequencies, active antennas are compact, lightweight, and easy to install. Their small physical volume provides an easier path to satisfying site installation restrictions and the various local planning laws and regulatory board requirements. Higher reliability is made possible since active antennas have a distributed amplifier architecture. Active antennas support 5G beam management and network densification requirements to maximize the reuse of spectrum, provide spatial reuse, and enable higher performance communication systems utilizing greater bandwidths. IV. 5G ACTIVE ANTENNA FIGURES OF MERIT A set of 5G active antenna FoMs is defined below to provide for objective performance comparisons across various active antenna technologies and facilitate performance evaluation of wireless networks employing these antennas. A. Effective Isotropic Radiated Power (EIRP) For an active antenna, P1dB EIRP can be defined as power transmitted with the main beam scanned to boresight and the antenna s power level specified at the output referenced 1dB compression point (OP1dB). Saturated output power (P SAT ) is another power term used in wireless link budgets as a measure of available radiated power from the antenna aperture. P SAT influences the upper limit for available signal power from the transmitter and therefore the feasible maximum for the EIRP. However, the saturated output power level is strongly technology and design dependent, whereas the output 1dB compression point is a standard, unambiguous reference point that facilitates comparison of various systems that will likely operate at different transmit back-off values from OP1dB. Care should be taken in drawing comparative values since either P SAT or OP1dB could be used in the definition of available EIRP. P SAT is often considered more relevant for transmitter systems where some form of pre-distortion algorithm is used to reduce the required level of output power back-off (OBO). EIRP = Pt + Gt + 10*log 10 (N) L (1) Pt: Conducted power per element (OP1dB in dbm) determined with a continuous wave (CW) stimulus Gt: Total antenna array directivity due to spatial power combining of all radiating elements (dbi) N: Total number of radiating elements L: Total losses from element amplifier output to radiating surface (includes path loss after signal amplification, through radome losses) Larger EIRP values enable higher data throughputs and better link quality at the edge of cell coverage. This can translate directly into both cell size and data throughput to the operator and therefore has a direct link to remuneration. Table 1 provides measured P1dB EIRP values for two product classes of 28GHz active antenna developed jointly by Ball Aerospace and Anokiwave. TABLE I. MEASURED TRANSMIT PERFORMANCE FOR 28GHZ ACTIVE ANTENNAS 28GHz Active Antenna P1dB EIRP TXeff (dbm) (1/cm^3) 64 Element Element A second, equally important measure is Linear EIRP, which is defined by the amount of OBO from OP1dB required to meet an error vector magnitude (EVM) threshold. The Orthogonal Frequency Division Multiplexing (OFDM) waveforms specified by 5G New Radio (NR) have large crest factors (8 to 15 db) because they are composed of many orthogonal sub-carriers. Since most active systems with amplifiers will start to compress approximately 5 to 10 db below OP1dB, applying OFDM signals above these OBO levels will result in distortion and EVM degradation. Determining Linear EVM is reflective of operational usage and requires the precise definition of the signal waveform and modulation; channel bandwidth; and relative constellation error contributions of other signal chain sources of error. These concepts are shown in Fig. 1. B. Transmit Efficiency (TX eff ) A second active antenna transmit FoM that is as important as the level of radiated power (P1dB EIRP and Linear EIRP),

3 is the efficiency with which the radiated power is generated relative to the required DC supply power and aperture area. is a Base Band Unit (BBU) that generates formatted baseband 5G NR channels. The data sink is the physical air interface for radiated transmission of 5G NR millimeter wave signals. Optimization of the overall RRHU system efficiency considers the contribution of both signal path and non-signal path devices (including devices such as control circuitry and power supply and distribution). Non-signal path devices may contribute substantially to the mass (volume) of the unit, especially with increased size required for Power Supply Units (PSU) associated with inefficient device supply signal conditioning. The transmit efficiency of the RRHU active antenna system as a whole captures how efficiently the system generates transmitted millimeter wave signals relative to the required DC supply power and total antenna volume. TX eff = EIRP / (P diss * Vol Ant ) (2) Fig. 1. Reference Power Levels for Linear EIRP definition Active antenna systems incorporated into a Remote Radio Head Unit (RRHU) are thermally limited systems and are required to operate for extended periods (>10 years) in a reliable manner without active thermal management. This means excess energy generated within the RRHU is expected to be dissipated through convection and radiation methods only, without forced air cooling systems requiring moving mechanical parts. This requirement is also reflected in the objective for 5G systems to form the basis of sustainable green networking through the increased deployment of small cells and enhanced energy efficiency gain [4] [8]. Deployment regulatory requirements and the nature of small-cell network implementations will set an upper limit on the allowable volume of the RRHU. The efficiency of heat transfer to the environment through surface heat flux as a function of heat transfer modes has been long established, and a given unit volume implies a maximum allowable power dissipation to meet reliability Mean Time Before Failure (MTBF) or Failure in Time (FIT) requirements [9]. One of the distinguishing capabilities of an active antenna is that EIRP is generated using many radiated elements to create array aperture gain along with distributed signal conditioning and electronic amplification in each of the element signal paths. This approach significantly diverges from the traditional transmit signal chain architecture where the efficiency of signal generation is dominated by a single power amplifier and a FoM such as Power Added Efficiency (PAE) could be utilized to characterize performance. To correctly characterize the distributed architectures of active antenna systems, more comprehensive transmitter efficiency FoMs must be considered. A useful vehicle for doing so is the notion of Consumption Factor theory whereby the energy efficiency of a system is considered at a more holistic level [10]. Fig. 2 illustrates how the notion of Consumption Factor may be applied to (a) the Remote Radio Head Unit (RRHU) and (b) an Active Antenna subsystem. The RRHU data source EIRP: Effective Isotropic Radiated Power (Watts) at output reference P1dB. Value defined (1) converted from dbm to Watts P diss : DC supply power required by active antenna system in transmit mode to create a beam with EIRP (Watts) Vol Ant : Total active antenna system volume including aperture, electronics and thermal management (cm^3) A higher value of TX eff indicates: lower DC power supply requirements, higher reliability due to less required thermal dissipation, and smaller physical size of the active antenna. These factors reduce operating costs for the radio access network and facilitate simple RRHU active antenna system installations meeting regulatory deployment constraints. Table 1 includes TX eff values for two rapid-prototype 28GHz active array designs that have not yet been optimized to reduce volume. If we wish to inspect the transmit efficiency performance of the millimeter wave circuitry more directly, then we can consider the active antenna subarray network shown in Fig. 2(b). A hybrid beam-forming approach implemented using Fig. 2. Figurative representation of (a) the RRHU [10] and (b) an Active Antenna sub array

4 multiple subarrays may be selected for the millimeter wave RRHU active antenna systems to generate multiple independent beams radiating from the overall combined system. Each subarray in the system will consist of N-elements (where N typically ranges from 16 to 256) and will require separate frequency conversion and baseband circuitry to generate the independent data streams. The overall power dissipation of the RRHU active antenna system includes contributions from each subarray. To create a subarray component metric for the efficiency of RF signal conditioning network at the subarray level, TX SAeff can therefore be determined by: TX SAeff = EIRP / (P SAdiss * N) (3) P SAdiss : DC supply power required by all components in the active antenna RF subarray signal path (not including frequency conversion) N: Number of elements in sub-array EIRP follows a 20log 10 (N) relationship for active array antennas where N is the number of elements in the array. This 20log 10 (N) relationship includes two terms: 10log 10 (N) for array aperture gain and 10log 10 (N) due to the inclusion of conducted millimeter wave power amplifiers for every radiating element. The efficiency of the subarray is therefore normalized by the number of elements in the subarray. C. Transmit and Receive Error Vector Magnitude (EVM TX, EVM RX ) over Beam Scan Volume Error Vector Magnitude (EVM) is a comprehensive signal quality metric used to quantify the performance of transmitters (EVM TX ) and receivers (EVM RX ) used in digital wireless communication networks. EVM measurements include the signal degradation effects of component impairments (distortion, phase noise, carrier leakage, etc.) throughout the radio network and are used to determine link quality and expected data handling capabilities of wireless systems. Specific 5G NR reference signals and test models will need to be defined along with unambiguous test methods to generate EVM measurements for active antenna enabled networks. To quantify performance for active antennas, measurements of EVM TX and EVM RX will need to be conducted as beams are steered over scan volume. Fig. 3 displays measured EVM data for a 28GHz, 256- element active antenna developed by Ball Aerospace and Anokiwave. Performance features of this 5G active antenna include: operating with a single 256-element beam or four independent 64-element beams; TX/RX half duplex function; full two-dimensional electronic beam steering and programmable beam widths. EVM data was measured using Keysight Technologies 5G Testbed for Design Validation (including the M9383A PXIe microwave vector signal generator and M9393A PXIe microwave vector signal analyzer). A 3GPP 5G NR waveform with 256 QAM and a single 100 MHz component carrier was used as the test signal for these EVM measurements. To conduct these tests, the active antenna was mechanically rotated to maintain direct pointing link to test system s standard gain horn as the active Fig. 3. Measured 28GHz Active Antenna EVM performance over scan volume (a) Transmit EVM for 64-element beam (b) Receive EVM for 64- element beam beam was scanned. These over-the-air EVM measurements were conducted in an office environment to provide representative data for realistic channel environments. For the high-order, digitally modulated signals being defined for 5G NR, EVM is a comprehensive measurement of how an active antenna will impact the signal quality of the signals it receives and transmits for the wireless network. D. Azimuth and Elevation Beam Scan (BS AZ, BS EL ) Active antennas electronically steer their main beam to point in different directions without mechanically moving the antenna. Two useful FoMs to define an active antenna s scan volume are Beam Scan over Azimuth (BS AZ ) and Beam Scan over Elevation (BS EL ). These FoMs measure the angular extent off boresight that a beam can be scanned in the Azimuth and Elevation planes without grating lobes occurring in the field of regard. For most 5G NR radio access networks, active antennas typically need to be able to scan their beams across the horizon

5 (Azimuth plane) to a greater range than they are required to scan above and below the horizon (Elevation plane). Typical required values for BS AZ are +60 degrees to -60 degrees to allow beam coverage over a 120-degree sector. Requirements for BS EL often only range from +30 degrees to -30 degrees or less. One exception would be for active antenna RRHU serving urban areas with tall buildings. To provide service for the higher stories of buildings in urban areas, full two-dimensional beam steering may be required where both BS AZ and BS EL must provide larger angular ranges (typically +60 degrees to -60 degrees). The 5G NR 28GHz active antenna described for Fig. 3 provides full two-dimension beam steering over a wide scan volume (BS AZ and BS EL both +/- 60 degrees). E. Beam Scan Loss (BSL TX ) An active antenna s main beam peak power will decrease as the beam is scanned from boresight to the outer edge of the antenna scan volume. This is known as scan loss and is defined here for antenna transmit performance. BSL TX = EIRP Peak EIRP Scan (4) EIRP Peak : EIRP of main beam scanned to boresight in dbm EIRP Scan : EIRP of main beam scanned to edge of array scan volume in dbm BSL TX (scan loss) defines how much coverage is available from the active antenna. Higher values of scan loss result in lower available transmit power (EIRP) as the beam is steered off boresight. Lower EIRP reduces link robustness and data throughput. Scan loss can similarly be defined for active antenna in receive mode. Fig. 4 provides measured beam scan performance for the 28GHz rapid-prototype, 64-element active antennas jointly developed by Ball and Anokiwave. Radiation patterns are shown for seven beam positions generated by programming the antenna to point the main beam in seven Fig. 4. Measured 28GHz, 64-element Active Antenna Beam Scanning directions from boresight to 60 degrees off of boresight. F. Beam Steer Execution Time (BSET) Beam Steer Execution Time (BSET) defines how rapidly the active antenna can transition its beam to point in a new direction to complete the previously issued beam steer command. BSET is the duration of time that the beam is not available as it is actively transitioning from one beam state to the next. It is critical for wireless networks that active antennas minimize the time spent in an active beam transition state since the channel served by the beam will be unavailable during this time. To best support 5G beam acquisition and refinement requirements, an active antenna s BSET should be less than the cyclic prefix duration. At the beginning of a symbol period (or earlier), the 5G NR network can send a command to transition the active antenna beam to a new direction during the following symbol s cyclic prefix duration. Once the cyclic prefix duration is completed, an active antenna s beam should be pointed in the correct direction and available for use during the remainder of the symbol period. For the 5G NR supported numerologies, a sub-carrier spacing of 240 khz results in a cyclic prefix duration of only 290 nanoseconds. 5G applications will require precise control over when a beam changes state. For example, to support hybrid beam forming, beam commanding must accurately align beam steer changes of multiple subarrays beams. Active antennas must provide rapid and precise beam steer execution to allow the radio access network to optimize data throughput. G. Number and Polarization of Beams (N b, Pol b ) More independent data channels in the wireless network allow for higher system capacity, increased data throughput, and more reliable radio links. Multiple beams can be produced by 5G NR active antennas, but these beams must provide sufficiently low correlation to support simultaneous operation of multiple, independent data channels. Techniques to reduce beam correlation include spatial diversity, orthogonal polarizations, separation of beam pointing direction, and adaptive radiation patterns. Envelope Correlation Coefficient (ECC) is an antenna radiation pattern correlation metric that can be used to characterize how independent two active antenna beams are. To facilitate determining the number of independent data channels that may be supported, joint FoMs specifying Number of Beams (N b ) and Number of Orthogonal Beam Polarizations (Pol b ) are useful. An active antenna supplying a larger number of individual beams with orthogonal polarization will be better able to support a larger number of independent data channels. Currently, various 5G NR field trials are evaluating systems with multiple beams supporting orthogonal slant +/- 45-degree polarization (Pol b =2). H. Radiated Pattern Management Active antennas must be able to rapidly optimize beam widths by software commanding to meet channel conditions. In addition to controlling the scan angle (pointing direction) of an active antenna system s main radiated beam, other radiation pattern characteristics can be software commanded by

6 adjusting the phases and amplitudes in each of the system s radiating element paths. Two key radiation pattern characteristics for 5G NR networks are antenna Beam Width Control (BWC) and Side Lobe Suppression (SLS). BWC is the ability of an active antenna to create narrow-to-wide beam widths to support both data and control channels. Radiated SLS is the ratio of the peak of the far field radiation pattern main beam to that of the maximum sidelobe level. This quantity is a measure of an active antenna s ability to control sidelobe levels to reduce radiation outside the desired main beam. SLS = Β p / SL p (5) B p : Peak of main beam scanned to boresight in dbi SL p : Peak of sidelobes with main beam scanned to boresight in dbi Lower sidelobes will decrease network-level interference, resulting in improved link quality, increased wireless channel availability, and improved system capacity. Fig. 5 illustrates some examples of these characteristics in the measured performance of a Ball Anokiwave 28GHz, 256- element array. Measurements of radiated antenna patterns are shown as a function of the Azimuthal plane angle for the active antenna programmed into four different radiated pattern configurations. The active antenna performance is software enabled with rapid commanding of the programmable vector modulators in the radiating element paths. The highest gain response pattern, labeled Beam 1, shows the antenna beam steered to boresight and stimulated with a Uniform Illumination (UI) array taper. In this condition, each of the signal path vector modulators is programmed to their maximum amplitude setting, and the resulting pattern displays the expected nominal -13dBc first sidelobe level (SLL). UI provides the maximum gain available from the radiating antenna aperture, and it is commonly used in many transmitter applications to maximize the EIRP. The relatively high sidelobe level in this setting may violate the stringent demands of telecommunications regulatory requirements, although spatial radiation energy restrictions are still in a state of definition. A reduction of the SLL of the radiated antenna pattern can be achieved by imposing a taper across the array elements where the relative magnitudes of the signals in the radiating element paths are weighted against each other to effect a reduction in the energy summation in the far field. Beam 2 in Fig. 5 illustrates the effect of imposing a 25 db (nominal) taper on the first sidelobe of the pattern, and the overall energy contribution of the sidelobes can be seen to be significantly reduced across the entire azimuth plane. The associated cost of this taper is a reduction in the peak gain and a slight broadening of the half-power beam width of the main lobe, thereby also reducing the available signal power from the antenna aperture in the desired pointing direction. Beam 3 and Beam 4 in Fig. 5 illustrate active antenna BWC. Beam spreading amplitude and phase tapers were applied to create wider beam width radiation patterns compared to the narrower beam shown for Beam 1. I. Additional Active Antenna Design Considerations There are additional active antenna design considerations that are vital to evaluate for optimizing 5G NR network performance. Over the next few years as millimeter wave 5G networks are implemented and performance is assessed, active antenna system requirements will mature and be further defined. For previous wireless access networks constructed with passive antennas, the far-field radiation patterns could not be software controlled to point beams and change beam widths. Since passive antennas have static radiation patterns, 3GPP used receiver metrics such as Block Error Rate (BLER) and throughput to characterize system receive sensitivity and performance without requirements to additionally define associated beam states for the test models. To establish comprehensive and unambiguous receive performance characterizations for active antenna enabled 5G NR networks, test models and methods will need to be further refined to specify beam management and active spatial components. Other key parameters for active antennas that must be defined over the antenna s scan volume include beam pointing accuracy, cross polarization isolation, adjacent channel leakage ratio (ACLR), intermodulation performance, and spatial interference rejection. For example, to support two-layer Multiple-Input Multiple-Output (MIMO) with orthogonal data streams on each polarization of an orthogonally polarized active antenna, acceptable cross polarization leakage as the beam is scanned must be defined. Another area of vigorous discussion has focused on whether there is a need for active antenna calibration requirements. Calibration activities, if needed, will likely impact many areas, including time to market, production yield, performance stability, and active antenna cost. Fig. 5. Measured radiation patterns of a 256-element array operating at 28GHz under different beam-configuration settings V. CONCLUSION A comprehensive set of 5G active antenna figures of merit is required to allow comparison of various active antenna solutions and predict network system performance. To overcome the challenges of operating in millimeter wave

7 channels, active antennas are an enabling technology that will assist with unlocking the full potential and supporting the diversity of new 5G applications. The FoMs defined in this paper facilitate objective assessments for active antennas enabling 5G wireless communication networks, and support the wireless industry goals to accelerate the deployment of 5G and grow the connectivity markets. REFERENCES [1] T. S. Rappaport et al. Millimeter wave mobile communications for 5G cellular: It will work! IEEE Access vol. 1 pp [2] 3GPP TR v14.2.0, Study on new radio access technology, physical layer aspects. [3] 3GPP TR v14.3.0, Study on scenarios and requirements for next generation access technologies. [4] P. Nikolich et al. Standards for 5G and beyond: their use cases and applications, IEEE 5G Tech Focus: Volume 1, Number 2, June [5] R. McMorrow, D. Corman, and A. Crofts, All Silicon mmw planar active antennas: the convergence of technology, applications, and architecture, 2017 IEEE International Conference on Microwaves, Antennas, Communications and Electronic Systems (COMCAS) [6] S. Kutty and D. Sen, Beamforming for millimeter wave communications: an inclusive survey, IEEE Commun. Surveys & Tutorials vol. 18 no. 2 pp , [7] E. Onggosanusi et al. "Modular and high-resolution channel state information and beam management for 5G new radio," IEEE Commun. Mag. vol. 56 no. 3, March [8] M. Yao, M. M. Sohul, X. Ma, V. Marojevic, and J. H. Reed, Sustainable green networking: exploiting degrees of freedom towards energy-efficient 5G systems, Wireless Networks Journal, Springer US, Nov [9] A. D. Kraus and A. Bar-Cohen, Thermal Analysis and Control of Electronic Equipment, McGraw-Hill, January [10] J. N. Murdock and T. S. Rappaport, Consumption factor and powerefficiency factor: a theory for evaluating the energy efficiency of cascaded communication systems, IEEE Journal on Selected Areas in Communications, February 2014.