Master in Multimedia and Communications

Academic Course 2015-2016

Master of Science Thesis

Analysis of off-the-shelf Millimeter Wave Systems with Phased Antenna

Arrays ___________________

Guillermo Bielsa

Director/s

Joerg Widmer

Eva Rajo

Leganés, July 18th 2016

Keywords: 802.11ad; 60 GHz networks; practical performance; beamforming; interference; frame level protocol analysis

Summary: In this thesis, we present the first in-depth beamforming, interference, and frame level protocol analysis of off-the-shelf millimeter wave systems with phased antenna arrays. We focus on (a) the interference due to the lack of directionality of consumer-grade antennas, and (b) the degree of data aggregation of current devices. Regarding (a), our beam pattern measurements show strong side lobes that challenge the common conception of high spatial reuse in 60 GHz networks. We also show that reflections in realistic settings worsen this effect. Further, we measure weak directionality when beamforming towards the boundary of the transmission area of an antenna array. Regarding (b), we observe that devices only aggregate data if connections require high bandwidth, thus increasing medium usage time otherwise.

Preface

This Master Thesis presents the paper ‘Boon and Bane of 60 GHzNetworks: Practical Insights into Beamforming, Interference, and FrameLevel Operation’ written by Thomas Nitsche, Guillermo Bielsa, Irene Tejado,Adrian Loch and Joerg Widmer. This paper was published in the conferenceACM CoNEXT 2015, which took place at Heidelberg, Germany in December2015.

The student was heavily involved in all the measurement campaignspresented in this work, including processing and writing the documentationabout them. This makes possible for the student to present this work as hisMaster Thesis.

0

Boon and Bane of 60 GHz Networks: PracticalInsights into Beamforming, Interference,

and Frame Level Operation

Thomas Nitsche, Guillermo Bielsa,and Irene Tejado

IMDEA Networks Institute andUniversidad Carlos III, Madrid, Spain{firstname.lastname}@imdea.org

Adrian Loch and Joerg WidmerIMDEA Networks Institute

Madrid, Spain{firstname.lastname}@imdea.org

ABSTRACTThe performance of current consumer-grade devices for60 GHz wireless networks is limited. While such net-works promise both high data rates and uncomplicatedspatial reuse, we find that commercially available de-vices based on the WiHD and WiGig standards may suf-fer from their cost-effective design. Very similar mecha-nisms are used in upcoming devices based on the IEEE802.11ad standard. Hence, understanding them wellis crucial to improve the efficiency and performance ofnext generation millimeter wave networks. In this pa-per, we present the first in-depth beamforming, interfer-ence, and frame level protocol analysis of off-the-shelfmillimeter wave systems with phased antenna arrays.We focus on (a) the interference due to the lack of di-rectionality of consumer-grade antennas, and (b) thedegree of data aggregation of current devices. Regard-ing (a), our beam pattern measurements show strongside lobes that challenge the common conception of highspatial reuse in 60 GHz networks. We also show thatreflections in realistic settings worsen this effect. Fur-ther, we measure weak directionality when beamform-ing towards the boundary of the transmission area of anantenna array. Regarding (b), we observe that devicesonly aggregate data if connections require high band-width, thus increasing medium usage time otherwise.

CCS Concepts•Networks → Network experimentation; Physicallinks; Wireless local area networks;

Permission to make digital or hard copies of all or part of this work for personalor classroom use is granted without fee provided that copies are not made ordistributed for profit or commercial advantage and that copies bear this noticeand the full citation on the first page. Copyrights for components of this workowned by others than the author(s) must be honored. Abstracting with credit ispermitted. To copy otherwise, or republish, to post on servers or to redistribute tolists, requires prior specific permission and/or a fee. Request permissions frompermissions@acm.org.

CoNEXT ’15, December 01 - 04, 2015, Heidelberg, Germanyc© 2015 Copyright held by the owner/author(s). Publication rights licensed to

ACM. ISBN 978-1-4503-3412-9/15/12. . . $15.00

DOI: 10.1145/2716281.2836102

Keywords802.11ad; 60 GHz networks; practical performance; beam-forming; interference; frame level protocol analysis

1. INTRODUCTIONDeveloping robust consumer-grade devices for mil-

limeter wave wireless networks is challenging. In con-trast to 802.11n/ac networks operating in the tradi-tional 2.4 and 5 GHz ranges, devices based on the re-cent 802.11ad standard [1] for operation in the unli-censed 60 GHz range must overcome significant hurdles.This includes handling 2.16 GHz wide channels, usingdirectional beamforming antennas to overcome the in-creased attenuation at these frequencies, and dealingwith a highly dynamic radio environment [2, 3]. Ex-isting work studies the individual factors that such asystem must address regarding both the characteristicsof 60 GHz communication—fading [4], reflections [5],frequency selectivity [6], and multipath effects [7,8]—aswell as the design of hardware such as phased antennaarrays [9–12]. The challenge is that consumer-grade de-vices must be able to handle the resulting complexityof the overall system while still being cost-effective.

This challenge has naturally led to the use of cost ef-fective components in consumer-grade 60 GHz systems.For instance, current devices use electronic beam steer-ing with relatively low order antenna arrays, that is,with only a limited number of active antenna elements.As a result, while the transmitters use directional beampatterns, they do not fully achieve the 60 GHz vision ofextreme pencil-beam focusing and imperceptible inter-ference impact. While this is qualitatively well-known,it raises a crucial question: how large is the practical,quantitative impact of such limitations?

Understanding this impact is fundamental since theaforementioned limitations may undermine common 60GHz assumptions. This in turn is key to design pro-tocols that can reliably operate on millimeter wave fre-quencies. As a quantitative analysis of these limitationsis missing, the following issues remain unanswered:

10.1145/2716281.2836102

24 time units

6 time units6 time units

Agg

r. on

Agg

r. of

f

Aggr. off: 6 pkts / 24 units = 0.25 pkts/u,6 pkts / 12 units = 0.5 pkts/u,Aggr. on:

max. delay = 1 umax. delay = 9 u

d1 a1 d2 a2 d3 a3 d4 a4 d5 a5 d6 a6

d1-3a1-3

d4-6a4-6

di = Data ai = ACK

t

tMedium free Medium free

Figure 1: Aggregation primer. In general, it in-creases throughput and reduces medium usage,but worsens delay. It’s impact is particularlylarge in 60 GHz due to the very high data rates.

A. Directional communication. How directionalare consumer-grade phased antenna arrays? Howlarge is the actual impact of side lobes?

B. High spatial reuse. How close can devices op-erate without experiencing collisions? How strongis the impact of interfering reflections?

C. High data rates. Which coding and modulationschemes are feasible? What is the impact of dataaggregation at such data rates (c.f. Figure 1)?

Previous work using consumer-grade 60 GHz devicesis limited to studying the impact of human blockage [13]and transmission range [14]. In this paper, we provideanswers to the questions above. To this end, we use adown-converter to overhear and analyze the communi-cation of 60 GHz devices. This gives us unprecedentedinsights into link utilization1, beam patterns, and framelevel operation, since the existing devices themselves donot provide any of this information. This analysis allowsus to determine the key limitations of 60 GHz consumer-grade hardware, which opens the door to future workon mechanisms that address these limitations. In par-ticular, our contributions are as follows:

1. We provide a frame level analysis of WiGig, study-ing its frame length and showing that WiGig onlyuses data aggregation if a connection requires highthroughput. Otherwise, it does not aggregate, evenfor traffic that is not necessarily delay-constrained.

2. We measure the beam patterns that our devicesuse. We show that quasi omni-directional antennapatterns suffer significant imperfections. Further,directional beam patterns exhibit very strong sidelobes of up to −1 dB compared to the main lobe.

3. We evaluate the impact of interference due to sidelobes. We observe a perceptible impact for dis-tances of up to five meters. We show that thiseffect worsens at the boundary of an antenna ar-ray’s transmission area due to increased side lobes.

1We use link utilization/medium usage interchangeably.

4. We study the impact of reflections in realistic wire-less settings. We find that interfering reflectionsfrom neighboring, unaligned devices may reducethe achievable TCP throughput by more than 20%of the value that would be achieved otherwise.

The paper is structured as follows. Section 2 gives anoverview on millimeter wave communication. Section 3provides details on the tested devices and our measure-ment setup. In Section 4, we present the details of ourmeasurement campaign, and in Section 5 we discuss theresults. Finally, Section 6 concludes the paper.

2. BACKGROUNDIn this section, we describe the characteristics of mil-

limeter wave communications, and present related workin this area. Millimeter wave frequencies have been usedfor commercial wireless systems for several years. Thesefirst generation systems, however, were targeting mainlystatic or pseudo static application scenarios, like back-haul links (for example the HXI Gigalink 6451 systemused in [14]) or transmission of uncompressed high def-inition video data [13]. With formation of the WiGigAlliance [15], the latter use case broadened to dock-ing station applications and finally incorporated generalWiFi use cases when, in 2012, WiGig was merged intothe IEEE 802.11ad amendment. This amendment de-fines a unified millimeter wave communication standardfor a variety of use cases, that include dense deploymentscenarios with (limited) mobility [2].Transmission Characteristics. Communication inthe millimeter wave range has distinctly different com-munication characteristics than those of legacy ISM fre-quencies below 6 GHz. These differences mainly resultfrom the increased attenuation of free space propagationand signal blockage in case of obstacles [16]. The in-creased attenuation of around 20-40 dB is typically over-come by highly directional antennas. These may also beused to circumvent blockage, using a propagation pathvia a first order reflection that certain highly reflectivematerials provide [5]. Further, directional communica-tion and blockage significantly lower the amount on in-terference on millimeter wave frequencies and allow forhigh levels of spatial reuse.Beam Steering. With the central role of directionalcommunication in millimeter wave systems, a device’sability to steer its directional beams becomes essential.As the size of antennas scales with the wave length,millimeter wave systems can integrate antenna arrayswith a high number of elements even into small hand-held devices. Theses antenna arrays allow electronicconfiguration of the antennas’ beam direction and pro-vide very high directional gain. In contrast to lowerfrequency beamforming mechanisms, millimeter wavesystems usually rely on beam steering via codebooksof predefined beam patterns that implement differentdirections, which reduces the complexity of transceiversand of the beam training process. While such millime-

ter wave antenna arrays have been in use since firstgeneration devices, little is known about their impacton system level network performance.Work on Practical 60 GHz Networks. Insights intopractical 60 GHz networks are limited since 802.11adhardware is not available yet. However, related workuses WiGig and WiHD devices to study study the per-formance of 60 GHz links, similarly to this paper. Forinstance, Zheng et al. [14] uses off-the-shelf hardware tocharacterize 60 GHz links, dispelling a number of com-mon beliefs regarding millimeter wave communication.This includes showing that the range of such networks islarge enough for outdoor communication, and that elec-tronically steerable antenna arrays can deal with block-age as well as user motion. Still, in contrast to our work,they do not study and measure the antenna patterns ofoff-the-shelf hardware to investigate interference in 60GHz networks. Further, they focus on high-level metricsof 60 GHz links such as throughput, while we analyzethe frame level operation of WiGig to study the impactof data aggregation in 60 GHz networks.

Ramanathan et al. [17] follow a different approachthan [14] and, hence, us. Instead of measuring the per-formance of off-the-shelf devices, they use a custom-built software-defined radio that allows them to performsignal strength measurements at 60 GHz using horn an-tennas. Among other results, they show that differenttypes of link breaks should be treated differently. Forinstance, a transmitter should change beam direction toavoid human blockage, whereas it should widen its beamto deal with mobility. Most interestingly, they showhow the signal strength can be used for early detectionof each type of link break. In contrast, we study actualdata transmissions using hardware with electronicallysteerable phased antenna arrays. While [17] focuseson blockage, beam steering, and spatial reuse assuminghorn antennas, we investigate topics such as reflectionsfor range extension, data aggregation, and interferencedue to imperfections of commercial 60 GHz hardware.

3. MEASUREMENT SETUPNext, we present the evaluated 60 GHz systems and

the used measurement equipment. Further, we describethe setups for the frame level analysis, as well as thebeamforming, reflection and interference measurements.

3.1 Devices and Measurement EquipmentWe evaluate two different millimeter wave systems in

order to analyze their behavior and performance in real-world settings and investigate inter-system interference.As of today, no off-the-shelf 60 GHz system allows ac-cess to any significant MAC or PHY level informations.Hence, we use a 60 GHz down-converter together withan oscilloscope to overhear the communication.Devices Under Test. Our first device under test is theDell D5000 wireless docking station, which follows theWiGig standard. The docking station allows connec-

tions by Dell notebooks with a compatible WiGig cardand antenna module. We use Latitude E7440 notebooksas remote stations. The system can connect multipleUSB3 devices using the wireless bus extension (WBE)protocol, as well as multiple monitors. The servicedarea with best reception is in a cone of 120 degree widthin front of the docking station. In indoor environments,over short link distances, and with reflecting obstacles,we found it, however, to perform over a much widerangular range. The maximum achievable distance de-pends on the environment and fluctuates between 12and 18 meters. During disassembly, we found that bothdocking station and notebook module are manufacturedby Wilocity. Both sides consist of a baseband chip con-nected to an upconverter and a 2x8 element antennaarray. The docking station comes with an applicationthat provides limited configurability (e.g., channel se-lection) as well as PHY data rate readings.

Our second system is a WiHD-compatible DVDO Air-3c system for the transmission of HDMI data streams.The system has transmitter and receiver modules thatdo not allow for any configuration and do not providelink state information. When testing the transmissionrange and link stability, we found that it performs betterthan the D5000. Indoors, we could transmit video over20 meters, even with 90 degree misalignment and block-age on the direct path. Upon disassembly, we found onboth sides of the link a 24 element antenna array with ir-regular alignment in rectangular shape. The bandwidthof both the D5000 and the Air-3c is 1.7 GHz. They bothoperate on center frequencies 60.48 GHz and 62.64 GHz.Measurement Equipment. To collect data for framelevel analysis and received signal power measurements,we use a Vubiq 60 GHz development system in conjunc-tion with an Agilent MSO-X 3034A oscilloscope. Weuse this setup to obtain traces of the analog I/Q outputof the Vubiq receiver. In most experiments, we under-sample the signal at 108 samples per second. While thisprevents decoding, it allows us to extract the timing andamplitude of different frames by processing the tracesoffline in Matlab. The frontend supports downconver-sion of 1.8 GHz modulated bandwidth at the commonIEEE 802.11ad/aj frequencies [18]. Further, it has aWR-15 wave guide connector which allows us to con-nect horn antennas with different levels of directivity.For beam pattern measurements, we use a 25 dBi gainhorn antenna. To obtain a wide beam pattern for pro-tocol analysis, we use the open wave guide.

3.2 Measurement SetupIn the following, we explain the setup of the four mea-

surement studies that we present in this paper, namely,the analysis of the frame level protocol operation, beampatterns, interference, and reflections.Frame Level Protocol Analysis. In order to gaininsights into the protocol operation of the devices un-der test, we down-convert their signals to baseband andanalyze them in Matlab. To this end, we use the wide

reception pattern of the open wave-guide of the Vubiqsystem. This allows us to overhear the frames of boththe transmitter and the receiver.

For the Dell D5000 case, we use the laptop as thetransmitter and the docking station as the receiver. Toidentify which frames come from which device, we placethe Vubiq down-converter behind the docking stationand point it towards the notebook’s lid. As a result, thedown-converter receives the frames of the notebook viaa direct path, and the frames of the docking station viaa reflection from the notebook’s lid. Thus, the averageamplitude of the notebook frames is larger than the oneof the docking station frames, and we can easily separatethem. We use Iperf2 to generate TCP traffic on theWiGig link that connects the laptop with the Ethernetadapter at the docking station. We run the Iperf serveron the laptop and the Iperf client on a second systemconnected via Ethernet to the docking station.

For the wireless HDMI case, we place the Vubiq down-converter close to the transmitter. No reflector setup isneeded since the frames of the receiver inherently have alarger amplitude. Again, we post-process the collectedsignal traces in Matlab to analyze their structure. Wecarry out most of the trace analysis using automatedalgorithms. However, we also use manual inspection todraw some of our conclusions in Section 4.1.Beam Pattern Analysis. To analyze the directivityand side lobes of the antenna patterns, we use the Vu-biq system with a highly directional horn antenna asdescribed in Section 3.1. By aligning this setup to thedevice under test, the impact of the second device in anactive link is almost imperceptible, which allows for ac-curate beam pattern measurement. As beam patternschange once data transmission starts after link initial-ization, we also measure the patterns of trained links.

We measure the azimuthal plane of the beam patternson a large outdoor space to avoid unwanted reflections.In particular, we capture signal energy on 100 equallyspaced positions on a semicircle with radius 3.2 m. Tothis aim, we rotate the entire setup—i.e., the Vubiq andthe oscilloscope—along all measurement locations, andcollect signal traces at each position. Figure 2 shows oursetup. We place the device under test in the center ofthe semicircle, and ensure a clear line of sight betweentransmitter and receiver throughout the experiment toprevent unwanted beam training. Due to the high di-rectivity of the horn antenna that we use, we found thatthe most powerful data frames always belong to the de-vice under test. When processing the traces, we ensurethat we extract signal strength from data frames only.We discard periodic control frames, which are transmit-ted with higher power and wider antenna patterns. Thesignal strength in every location is then averaged overthe filtered frames recorded over the span of one minute.

Further, we analyze the device discovery behavior ofthe Dell D5000 docking station. When disconnected,

2https://iperf.fr/, version 2.0.5-3

TX

RX

Figure 2: Beam pattern analysis setup.

Time (ms)0 0.2 0.4 0.6 0.8 1

−1

0

1

Vol

ts (V

)

321 2 3 ... Sub−elements

Figure 3: Dell D5000 device discovery frame.

the system frequently emits a device discovery framethat is transmitted over multiple antenna patterns tocover an area as large as possible. An example for thisframe type is shown in Figure 3. It can be seen thatthe frame consists of 32 sub-elements, each with rela-tively constant amplitude. Each of these sub-elements istransmitted with a different antenna configuration. Asthe pattern sequence of the sub-elements is the same forall frames, we can measure them using the same averag-ing approach as for active data transmissions. We thensplit the frame into its sub-elements during post pro-cessing to find the beam pattern of each sub-element.Reflection Analysis. Next, we analyze the impact ofreflections in a realistic wireless setting. This addressesthe common assumption that 60 GHz reflections arevery limited compared to the 2.4/5 GHz case, and resultfrom quasi-optical propagation in the direction of trans-mission. To this end, we set up a single 60 GHz linkin an empty conference room, either using the D5000or the WiHD system. We then measure the energy re-ceived from all possible directions at six different loca-tions {A . . . F} in the room, as shown in Figure 4. If noreflections occur, we expect to receive energy only fromthe direction in which the transmitter is located. For in-stance, at location A in Figure 4, we should only observeenergy coming horizontally from the right. Additionallobes in the corresponding angular profile indicate re-flections. To analyze the impact of different materials,we perform the experiments in a room which has brick,glass, and wood walls. Figure 4 shows the material lay-out. To measure the angular profile at each location,we mount the Vubiq receiver on a programmable rota-tion device and place it at each of the six locations inFigure 4. Moreover, we attach a highly directional hornantenna to the receiver. At each location, we then mea-

https://iperf.fr/

9 m

3.25

m

1.85 m

1.3

m1.

6 m

1.85 m 1.85 m 1.85 m

Brick GlassWood

RX

TX

C

B A

D E

F

Figure 4: Reflection analysis setup.

LaptopDock

Reflecting wall

2 m0.5 m

1 m

Figure 5: Reflection throughput setup.

sure the incident signal strength in each direction andassemble the result to an angular profile.

While the above setup allows us to show that reflec-tions are significant, we also perform a case study toanalyze whether those reflections may help to extendthe coverage of a network if the line-of-sight is blocked.To this end, we set up a link parallel to a wall usinga D5000 docking station and its corresponding laptop.Additionally, we place an obstacle in between both. Wethen measure the angular energy profile at the receiverto verify that the line-of-sight path is actually blocked,and that all energy arrives via the reflection off the wall.Finally, we use Iperf to measure the achievable rate onsuch a reflection. Figure 5 depicts our setup.Interference Analysis. To analyze how interferenceaffects 60 GHz communication, we operate multiple 60GHz systems in parallel on the same channel. In partic-ular, we use two pairs of notebooks connected to D5000docking stations and the DVDO Air-3c WiHD system,as shown in Figure 6. The Dell D5000 systems do notinterfere with each other since they use CSMA/CA toshare the medium. However, we use two of them inparallel to increase wireless medium utilization, andthus raise the probability of observing interference withthe WiHD system. The WiHD system does not useCSMA/CA and blindly transmits data causing colli-

Laptop A Laptop B

Dock A Dock B

HDMI TX

HDMI RX

6 m

eter

s

6 m

eter

s

8 meters

HDMI TX

HDMI RX

We vary the distance of the docking station to the HDMI system from

0 to 3 meters

Inter-System Interference

1m

Figure 6: Interference analysis setup.

sions and retransmissions at the D5000 systems. Thatis, the inter-system interference in Figure 6 is due tothe impact of the “HDMI TX” on “Dock A” and “DockB”. We set the distance between the WiHD transmitterand receiver to eight meters to ensure that the trans-mitter transmits frames with sufficiently high power.A scenario comparable to this setup could be causedfor example by two close-by millimeter wave systemsconnected to different access points in a multi-AP net-work. Further, we vary the horizontal distance betweenthe D5000 and the WiHD system in the range from 0to 3 meters to analyze the impact of the interferenceincidence angle. As we found that the WiHD systemtransmits with a much wider antenna pattern than theD5000, this procedure creates interference whenever aside lobe of the D5000 system matches the interferesdirection. To measure the effect of interference we mea-sure link utilization, reported link rate and the time oftransmission of a file with a size of 1 GB. To obtainlink utilization measurements we collect seven minutesof channel traces and use a threshold based detectionapproach to calculate the ratio of idle channel time.While the formerly described reflection analysis allowsus to assess the existence and strength of reflections, weuse a second setup to determine the impact of those re-flections on data transmissions. In particular, we set uptwo geometrically non-interfering 60 GHz links close toa metal reflector, as shown in Figure 7. To eliminate theinfluence of side lobes on the measurement, we positionshielding elements close to the WiGig devices. Further,we make sure that we do not block the reflected signalresulting from the metallic surface behind the WiHDreceiver. We then analyze the coverage area of this re-flection using the Vubiq transceiver to ensure that thedocking station is located inside. Finally, we performa TCP throughput measurement, with frame flow fromthe laptop to the docking station. By powering on andoff the WiHD devices, we can evaluate the impact ofthe reflected signal on the WiGig TCP connection.

1.5 m1.9 m

WiHDTX

WiHDRX

0.7

m

0.2

m

Blo

ckag

e

Dock

LaptopReflector Blockage Blockage

elementsprevent

direct in-terferencefrom sidelobes of

the WiHDtransmitter

Figure 7: Reflection interference setup.

4. RESULTSIn this section we present the results of our measure-

ments using the setups described in Section 3. We firstpresent our findings on protocol operation and data ag-gregation. Second, we analyze the beamforming capa-bilities of the D5000. Third, we investigate the impactof reflections. Finally, we evaluate the interference be-tween WiGig and WiHD, including reflections.

4.1 Protocol AnalysisWe study the flow of frames between the devices un-

der test using the Vubiq receiver. In particular, we an-alyze the frame structure of both the D5000 and theWiHD system, as well as the frame length and the burstlength of the D5000. The latter allows us to get insightsinto the impact of data aggregation.Dell D5000. The Dell D5000 follows the WiGig stan-dard. This is particularly interesting since the laterversions of the standard are closely related to the IEEE802.11ad standard. Hence, the behavior of existingconsumer-grade WiGig devices reveals the issues thatfuture devices based on IEEE 802.11ad will face. TheWiGig protocol description is not freely available, butwe observe that it consists of three phases, namely, de-vice discovery, link setup, and data transmission. Inthe first stage, the docking station emits a characteristicdevice discovery frame that is transmitted over severalquasi omni-directional beam patterns shown in Figure3. The frequency of these beam sweep frames is given inTable 1. A detailed beam pattern analysis of this framefollows in Section 4.2. At the second stage, a complexassociation and beamforming process between dock andremote station takes place. Finally, when the link is setup completely, data transmission begins.

Our frame level analysis shows that the data trans-mission phase contains bursts, similarly to the IEEE802.11ad EDCA transmit opportunities (TXOP). Themaximum length of such bursts is 2 ms. Each burst be-

Frame type Repeat intervalD5000 Device Discovery Frame 102.4 msD5000 Beacon Frame 1.1 msWiHD Device Discovery Frame 20 msWiHD Beacon Frame 0.224 ms

Table 1: D5000 and WiHD frame periodicity.

Time (ms)0

−0.5

0

0.5

Vol

ts (V

) DataBeacon

Acknowledgement0.1 0.2 0.3 0.4 0.5 0.6

Figure 8: Dell D5000 frame flow.

gins with two control frames, which have a different am-plitude than the subsequent series of data and acknowl-edgment frames, as shown in Figure 8. These controlframes are most probably an RTS/CTS exchange, whichis crucial due to the deafness effects resulting from thedirectional transmission. Outside the bursts, the chan-nel is idle except for a regular beacon exchange betweenthe docking station and the notebook. The transmis-sion frequency of these beacons is given in Table 1.

To study data aggregation, we measure the length ofdata frames for different TCP throughput values. Wecontrol the TCP throughput by adjusting its windowsize in Iperf. Figure 9 depicts the CDF of the framelengths for each throughput value. The CDF revealsthat frames are either short (around 5 µs) or long (15 to20 µs). That is, we can divide them into two categories.The length of long frames varies more than the lengthof short frames, which may be due to different levels ofaggregation. The highest level we observed correspondsto a frame duration of 25 µs. Further, the amount oflong frames increases with throughput—the higher thetraffic load, the more data aggregation. This matchesFigure 10, which shows the fraction of long frames, i.e.,longer than ≈ 5 µs, for increasing throughput values.

Moreover, we investigate the level of medium usagefor increasing throughput values. Surprisingly, Figure 11shows that, beyond a relatively low throughput value3,all oscilloscope traces contained data frames. That is,the transmitter transmitted continuously. Hence, thethroughput increase is not due to a higher medium us-age but could be either due to a higher modulation andcoding scheme (MCS), or a higher level of data aggre-gation. Figure 12 answers this question—it depicts theraw physical layer data rate reported by the D5000 overa timespan of ten minutes. For short links, Figure 12

3We choose values in the order of kilobits per second toobserve low link utilization. We achieve these values bysetting a small TCP window size (≈ 1 KB) in Iperf.

0.8

0.6

0.4

0.2

0.0

1.0C

DF

0 5 10 15 20 25Frame length (μs)

9.7 kbps171 mbps372 mbps601 mbps806 mbps930 mbps

Figure 9: WiGig data frame length.

80604020

100

Long

fram

es [%

]

0

9.7 kb

ps

40 kb

ps

171 m

bps

183 m

bps

372 m

bps

601 m

bps

806 m

bps

831 m

bps

930 m

bps

934 m

bps

TCP Throughputc.f. Footnote 3

Figure 10: Percentage of long frames in WiGig.

shows that WiGig uses a very high MCS. This explainsthe low link usage for 9.7 and 40 kbps in Figure 11,which were also measured at a short distance. Further,it suggests that the MCS is not adapted to the link load,as expected. Hence, we conclude that high throughputvalues are achieved exclusively by means of data aggre-gation, as the MCS naturally only depends on the signalstrength. That is, for a constant MCS and medium us-age, WiGig can scale throughput from 171 mbps to 934mbps by aggregating only 25 µs of data, which is 320×less than what 802.11ac needs for just a 2× gain [19].

Figure 12 also shows the impact of the link lengthon the physical layer rate. As expected, the longer thelink, the lower and unstable the data rate. While Fig-ure 12 depicts the link rate, in Figure 13 we show theaverage Iperf throughput at increasing distances. Weobserve that link rate is approximately stable up to acertain distance d and then falls abruptly. This distanced varies significantly for different experiments, rangingfrom 10 to 17 meters. As a result, while the individualexperiments clearly exhibit the aforementioned abruptbehavior, the average falls gradually. From Figures 12and 13 we conclude that the MCS drops gradually withdistance to about 1 gbps but then falls abruptly. InFigure 13 we do not observe results beyond roughly900 mbps because the Gigabit Ethernet interface at thedocking station limits the achievable throughput.

100806040200

9.7 kb

ps

40 kb

ps

171 m

bps

183 m

bps

372 m

bps

601 m

bps

806 m

bps

831 m

bps

930 m

bps

934 m

bps

TCP Throughput

Med

ium

usa

ge [%

]

c.f. Footnote 3

Figure 11: WiGig medium usage.

0 2 4 6 8 101

2

3

4

PHY

Lin

k R

ate

(gbp

s)

Time (minutes)

2 m 8 m 14 m16-QAM, 5/8

QPSK, 3/4QPSK, 5/8

BPSK, 3/4QPSK, 1/2

Figure 12: MCS with low traffic.

Unless indicated otherwise, we carry out all of our ex-periments for links below two meters. Interestingly, therate reported by the WiGig driver matches the MCS lev-els defined in the standard for single-carrier mode [15],which suggests a direct relation. Figure 12 includes theMCS that corresponds to each of the rates we measure.While we reach 16-QAM with 5/8 coding—the secondhighest MCS in the standard—we never observed thehighest MCS. We measure the rates in Figure 12 in astable environment without mobility. Naturally, our re-sults are mostly constant for short link lengths. Still,Figure 12 only shows a relatively short time interval.Figure 14 depicts the rate and amplitude under simi-lar conditions for roughly one hour. In this case, weobserve that the link rate varies occasionally. More-over, this occurs precisely when the signal amplitudechanges4. Since the environment is constant, such am-plitude variations are most probably due to a beam pat-tern realignment. This suggests that rate adaptationand beam pattern selection are implemented as a jointprocess in the Dell D5000 system.

4In Figure 14, the bit rate decreases when the amplitudeincreases. This counterintuitive behavior is because wemeasure the bit rate at the D5000 and the amplitude atthe Vubiq receiver. Due to the directivity of the signal,a decrease at the Vubiq does not necessarily mean thatthe signal also has less amplitude at the D5000.

AverageExample for low rangeExample for high range

0

200

400

600

800

1000

0 5 10 15 20

Thro

ughp

ut [m

bps]

Distance

Figure 13: Throughput decrease with distance.

00

0.5

1

Ampl

itude

(V)

00

2

4

Link

Rat

e (G

bps)

Interface Bit Rate

Amplitude Laptop Frame

Amplitude D5000 Frame

Time (minutes)10 20 30 40 50 60 70 80

Figure 14: D5000 frame amplitudes and rate.

DVDO Air-3c. Next, we analyze the frame flow forthe DVDO Air-3c WiHD system. We observe the samecommunication stages as for the Dell D5000 system.The frequency of device discovery frames transmittedby the WiHD devices is given in Table 1. When analyz-ing the data transmission stage, we found the beaconframe frequency to be much higher than for the D5000system. Also the data transmission process significantlydiffers. Figure 15 shows an example frame flow for theWiHD system. In contrast to the D5000, there is noclear data/acknowledgement frame exchange. Instead,the transmitting device emits data frames of variablelength following periodic beacons of the receiver. When-ever no data is queued for transmission, we only observebeacon frames. The trace shows the transition from anactive video data transmission period to an idle period.The WiHD system does not seem to perform channelsensing, which has a significant impact regarding inter-ference (c.f. Sections 4.3 and 4.4).

Time (ms)0 1 2 3 4 5

−0.5

0

0.5

Volts

(V)

Change from data transmissionto idle periodBeacon

Data

Figure 15: DVDO Air-3c WiHD frame flow.

Our results show that data aggregation is key for 60GHz communication. In particular, we observe that ag-gregation improves throughput from 170 mbps to 930mbps, i.e., a 5.4× gain. Our measurements also suggestthat, for distances beyond 10 m, links become unstableand often break before the transmitter switches to ratesbelow 1 gbps. Finally, link rates may fluctuate even instatic scenarios due to beam pattern realignments.

4.2 Beam PatternsNext, we measure the antenna patterns of the D5000.

We first investigate the quasi omni-directional patternsused for device discovery. After that, we analyze thehighly directional patterns used for data transmission.Quasi Omni-directional Search. Implementation ofomni-directional antenna patterns is a major challengefor millimeter wave communication [9]. However, thiskind of pattern is needed for device discovery and beamtraining [2]. The Dell D5000 system sweeps 32 differ-ent quasi omni-directional patterns during its remotestation search before the link setup. Each of these pat-terns is used during a sub-element of the device discov-ery frame. The device discovery frame is continuouslyrepeated by the D5000 until a connection is established.

Figure 16 shows four out of the 32 different antennapatterns that the D5000 sweeps during link establish-ment. The irregularities of the beam patterns are a nat-ural result of our measurement setup—since we manu-ally move the Vubiq receiver along the 100 measure-ment positions (c.f. Section 3.2), small deviations areinevitable. Nevertheless, Figure 16 clearly depicts therough shape of the lobes. While the half power beamwidth (HPBW) can be as wide as 60 degrees, each pat-tern contains several deep gaps that may prevent com-munication with devices at this specific angle. Thesegaps are due to the limitations of consumer-grade phasedantenna arrays. The remaining 28 patterns are compa-rable in terms of directional focus and received signalpower. The device discovery frame of the WiHD systemalso sweeps several quasi omni-directional beam pat-terns. However, their order changes with every trans-mitted device discovery frame. Hence, measuring thebeam patterns with our setup is not practicable.

Device discovery in 60 GHz systems proves to be achallenging process. We find that the two systems thatwe test repeatedly sweep multiple quasi omni-directionalbeam patterns to reliable reach a pairing device. In par-ticular, one of the systems sweeps 32 different patterns.

−6

−4

−2

0

30

210

60

240

90

270

120

300

150

330

180 0

Received Signal Power [dB]

−6

−4

−2

0

30

210

60

240

90

270

120

300

150

330

180 0

Received Signal Power [dB]

−6

−4

−2

0

30

210

60

240

90

270

120

300

150

330

180 0

Received Signal Power [dB]

−6

−4

−2

0

30

210

60

240

90

270

120

300

150

330

180 0

Received Signal Power [dB]

Figure 16: Quasi omni-directional beam patterns swept by the D5000.

−8

−6

−4

−2

0

30

210

60

240

90

270

120

300

150

330

180 0

Received Signal Power [dB]

−8

−6

−4

−2

0

30

210

60

240

90

270

120

300

150

330

180 0

Received Signal Power [dB]

D5000 Rotated

D5000

Figure 17: Laptop (left) and D5000 (right) beampatterns. Main lobe points always to 0 degrees.

Directional Transmission and Side Lobes. Nextwe evaluate the beam pattern used by the D5000 systemduring the data transmission stage. There, in contrastto the device discovery frames during the link setup,highly directional antenna configurations are used. Fig-ure 17 shows the measured transmit patterns of a DellE7440 notebook and the D5000 docking station. Thepatterns are of highly directional nature with a HPBWbelow 20 degree. Despite the strong signal focus, signif-icant signal energy is measured from side lobes. Theseside lobes can have a transmit power in the range of−4 to −6 dB compared to the main lobe, and thus cancause major interference effects. The asymmetry of thepattern—especially for the notebook—results from theantenna being placed at the side of the notebook’s lid.

In a second measurement we introduce a misalign-ment of 70 degrees between docking station and note-book. Figure 17 shows the resulting docking stationbeam pattern as an overlay. While measuring this pat-tern, we had to increase the receiver gain by 10 dB.That is, beamforming towards the boundary transmis-sion area of the antenna array significantly reduces linkgain. Also, we observe a much higher number of sidelobes as strong as −1 dB with respect to the main lobe.

While antennas are already highly directional, strongside lobes exist. Further, the placement of the antennainside a device has a noticeable impact on the antennapattern. When beamforming close to the outer limit ofan array’s transmission area, directionality is reducedand the number of strong side lobes increases.

4.3 ReflectionsIn the following, we present our reflection results for

the setup shown in Figure 4. Figures 18 and 19 depictthe angular profiles that we measure for the D5000 andthe WiHD system, respectively. In both cases, mostangular patterns have at least two clearly identifiablelobes—one pointing to the transmitter and one pointingto the receiver. The reason for the latter is that the re-ceiver not only receives data frames but also transmitsthe corresponding acknowledgments. However, a sig-nificant number of angular patterns feature additionallobes that do not point to any of the devices in theroom. This is a clear indication of reflections off thewalls. In contrast to common assumptions regarding60 GHz communications, the lobes show a significantamount of incidence energy. For instance, the angularpattern at position F in Figure 18 has a lobe directlypointing to the lower wall. Geometrically following thereflection of the signal off the nearby window suggeststhat this lobe is due to the transmitter. Further, theangular pattern at position B features a lobe pointingtowards the wooden wall. However, this lobe can onlyarise from a second order reflection originating at thereceiver, and bouncing off both the glass wall as wellas the wooden wall. In comparison to the D5000, theWiHD system shows similar effects. However, the angu-lar patterns in Figure 19 feature more and larger lobesthan in Figure 18. This suggests that the WiHD systemis less directional than the D5000, and thus producessignificantly more reflections. As a result, the impacton spatial reuse is even higher. This matches earlier ob-servations throughout our measurement campaign. InSection 4.4, we analyze this effect in detail.

Reflections are strong and may thus undermine the al-leged spatial reuse of 60 Ghz systems. Moreover, secondorder reflections occur and may have a large impact.

Additionally, we study whether reflections may helpto extend the range of 60 GHz links beyond obstacles.To this end, we use the setup depicted in Figure 5.First, we ensure that the line-of-sight path is blocked,and thus the docking station and the laptop can onlycommunicate via reflections. Figure 20 validates this—the angular energy profile does not include any lobe

Dock

Laptop

C

B A

D E

F

0 dB-2 dB-4 dB-6 dB-8 dB

Legend

Figure 18: Reflections for Dell D5000.

HDMIRX

C

B A

D E

F

0 dB-2 dB-4 dB-6 dB-8 dB

Legend

HDMITX

Figure 19: Reflections for DVDO Air-3c WiHD.

on the line-of-sight. Second, we use Iperf to measurethe achievable throughput in this setup. We obtain 550Mbps (±18 Mbps with 95% confidence) which is morethan half of what we measure on line-of-sight links.

We show that commercial 60 GHz devices can achievesignificant data rates on links with evidenced no line-of-sight component. Related work [14] shows similar find-ings but lacks validation via an angular energy profile.

4.4 InterferenceIn this section, we analyze the cross platform inter-

ference between the WiGig compliant Dell D5000 dock-ing solution and the WiHD system. Both systems areforced to operate on the same channel, leading to imper-fect channel usage coordination and interference. Whilethe directionality of 60 GHz communication is supposedto severely limit interference we analyze two aspectsof electronically steered indoor systems that can breakthis assumption. First, interference caused by the se-vere side lobes identified in Section 4.2 is analyzed bya setup with WiGig and WiHD devices operating inparallel. Second, we investigate the effect that can becaused by strong reflections as described in Section 4.3.To this aim, a combined setup of WiGig and WiHDdevices is used, where the direct path between the twodevice classes is shielded. The measurement setups forboth experiments are described in Section 3.2. We be-

-8-6-4-20

30

210

60

240

90

270

120

300

150

330

180 0

Received Signal Power [dB]

Laptop

Line-of-sightblockage

All energy arrivesvia reflection off the wall

Figure 20: Angular energy profile at the dockingstation with link blockage (c.f. Figure 5).

gin this section with the description of frame level effectsobserved for the interfering links which are the same forboth setups. Then, details on the impact of link perfor-mance are given.Effects observed in frame level analysis. From aframe level analysis of the traffic flow between two inter-fering links, which follow different communication stan-dards, we distinguish the following four different cases.First, data transmission of one system may occur duringan idle period of the other system. In this case, there isa minor impact from the beacon frames transmitted bythe interfering system, but we do not observe frame re-transmissions or increased link utilization. Second, dueto highly directional beam patterns, data streams maycoexist without provoking retransmissions on the DellD5000 link. Third, when links are close by, the D5000shows enlarged data transmission gaps that are occu-pied by frames of the WiHD system. We assume this tobe caused by a carrier sensing feature integrated in theDell D5000. The same effect was found when two dock-ing station links coexist. Finally, for close by links andoverlapping transmissions, the D5000 shows missing ac-knowledgments and delays in the data frame flow. Thisindicates frame loss and retransmissions. In the lattertwo cases we also observe increased link utilization.

Figure 21 shows an example where both systems op-erate in parallel. Here, a typical D5000 data frame flowis interrupted by frames from the WiHD system. In par-ticular, we observe collisions between 0.25 and 0.35 ms(first enlarged interval in Figure 21). The elevated noisefloor indicates the presence of a weakly received WiHDframe, while at the same time D5000 data frames aretransmitted. As the acknowledgments for these framesare missing, it is very likely that the three frames areretransmissions due to corrupted packet reception. Fur-ther, between 0.76 and 0.95 ms a dense series of WiHDdata frames can be found (second enlarged interval inFigure 21). This series interrupts the usual D5000 dataflow and does not suffer interference from the D5000 fora substantial time period. We attribute this behaviorto the aforementioned carrier sensing of the D5000.

0 0.2 0.4 0.6 0.8 1

x 10−3

−0.5

0

0.5

Time (s)

Vo

lts (

V)

2.6 2.8 3 3.2 3.4 3.6 3.8 4

x 10−4

−0.5

0

0.5

Time (s)

Vo

lts (

V)

Interference and Retransmission (a)

7.5 8 8.5 9 9.5

x 10−4

−0.5

0

0.5

Time (s)

Vo

lts (

V)

Carrier Sensing by Dell D5000 (b)

a) b)

Figure 21: Inter system interference effects.

Side Lobe Interference Impact. We now discussthe performance impact observed from side lobes forlinks in a parallel setup (compare Section 3.2), sufferingfrom inter-system interference. The impact is evaluatedin terms of observed link utilization, time for transmis-sion of a file on the Dell 5000 system and the link ratereported by the D5000 WiGig driver application.

Figure 22 depicts the results for two setups: 1) analigned D5000 link interfering with the WiHD systemand 2) the same setup with the docking station mis-aligned by 70 degrees (denoted by ‘rotated’). In thesecond setup, the docking station is beam-forming to-wards the outer limit of its serviceable transmission areaand beam-forming performance is significantly worse asshown in Section 4.2. When comparing link utilizationpercentages, we find them to be significantly increasedin both interference scenarios. We measure the inter-ference free link utilization to be 38% and 42% in thealigned and rotated setup, respectively. Thus, we finda maximum link utilization increase of 62% and 58%for the two interference scenarios, which is significantlyhigher than the link utilization of the WiHD link alone,that we found to be 46%. The additional increase re-sults from the frame collisions and retransmissions asdescribed earlier in this section.

Even though we use two parallel Dell D5000 dock-ing station links, the link utilization never saturatesthe channel. Thus, the measured transmission timestayed approximately constant despite retransmissionsand carrier sensing induced delays. Therefore, we relyon the link utilization to asses the impact of interfer-ence. For higher network densities that saturate thechannel and/or higher rate requirements of the wirelessapplications, interference is expected to have a signifi-cant impact on the throughput rate.

The difference of link utilization over the measuredinterference distances reveals a high interference regimefor distances of up to two meters. When further increas-

0 0.5 1 1.5 2 2.5 340

60

80

100

Link

Util

izat

ion

(%)

Distance between Interfering Links (m)

0 0.5 1 1.5 2 2.5 3

2.6

2.8

3

3.2

Link

Rat

e (G

bps)

0 0.5 1 1.5 2 2.5 31.8

2

2.2

2.4

Link

Rat

e (G

bps)

Link UtilizationInterface RateLink Utilization (rotated)Interface Rate (rotated)

Figure 22: Side lobe interference impact.

ing the distance of the second link, utilization reducesbut only reaches interference free levels at distances be-yond 5 meters. In the high interference regime, linkutilization in the misaligned docking station setup ishigher by around 10% compared to the aligned setup.At some measurement locations it reaches values of upto 100% and also shows a strongly varying pattern. Forthe aligned link we find a strong utilization increase to97% at an interferer distance of 1.6 meters. We con-jecture that this behavior is due to the side lobes thatwe observe for the data transmission antenna patternsof the D5000 link (compare Section 4.2). The fluctuat-ing link utilization in the high interference regime forthe misaligned link correlates well with its measuredbeam pattern that shows many strong side lobes. Un-fortunately, we can not directly compare measured sidelobes to the interference impact as it is not possibleto measure receive antenna patterns. Also, since wecannot influence the selection of antenna patterns, it isnot possible to ensure that the docking station uses thesame pattern for both measurements.

From the reported link rates it can be clearly seenthat the misaligned link performs worse due to the re-duced beamforming capability at the limit of the an-tenna arrays transmission area, as explained in Section4.2. Further, in the high interference regime below twometers link distance, an inverse correlation between linkrate and link utilization is found. We assume the dock-ing station link to adjust the link rate according to SINRmeasurements and packet loss statistics, thus the ratedecreases under high link utilization that leads to anincrease in collisions.

We find significant impact of inter-system interfer-ence on link utilization, especially when interferers withwide antenna patterns are as close as two meters. How-ever, as current millimeter wave links are far from sat-urating the wireless link, there is little impact on theoverall throughput achieved by the end systems. As rate

requirements and number of communicating devices in-crease, we expect to see a significant impact on through-put. We further find increased interference for antennaarrays focusing their beam close to the transmission areaboundaries since patterns are less directional. The in-terference that we observe correlates well with the strongside lobes of the antenna patterns in Section 4.2.Reflection Interference Impact. In this section wediscuss TCP throughput measurements for a setup ofWiGig and WiHD devices suffering inter-system inter-ference via reflected paths. To this aim, we use a setupwith a strong reflector. We shield the line-of-sight pathbetween the links to protect from direct path and sidelobe interference. The setup is described in Section 3.2.

The Iperf server used a 250 KByte TCP window andwas configured to fully load the underlying Ethernetlink, which was tunneled over the 60 GHz wireless link.While the WiGig tunneling managed to provide full gi-gabit speed for the Ethernet link, we found that it al-most completely saturated the link. Most probably, thisis because the D5000 system tries to minimize the de-lay of the Ethernet traffic. That is, instead of aggregat-ing data to reduce the medium usage, the transmittersends a larger number of packets, as discussed in Sec-tion 4.1. Because of this high medium usage, we expectthe TCP link to be very sensitive to interference effects.Figure 23 shows the Iperf TCP throughput results forour reflection measurement. We observe that the TCPthroughput increases significantly after about 90 sec-onds, which is when we power off the WiHD link. Theperformance degradation due to the WiHD reflection isabout 200 mbps compared to the interference-free trans-mission. The strongest throughput loss occurs about20 seconds after the start of the measurement—at thatpoint, the throughput drops by almost 300 mbps. Fur-ther, we observe that the throughput fluctuates stronglyfor the case with interference. Most probably, thesevariations result from attempts of the D5000 to adaptto the high interference by switching among MCS levels.

When analyzing the effect of reflections on 60 GHzcommunication, we find that, apart from an increase ofthe coverage area, also interference increases. For anisolated experiment setup with direct path blockage be-tween links we find a severe impact on TCP throughputdue to inter-system interference. The average through-put reduction was about 20%, and reached up to 33%.

5. DISCUSSIONAnalyzing commercial 60 GHz devices with the Vu-

bIQ system allows us to get new fundamental insights.This methodology makes our results stand apart fromrelated work. In the following, we discuss these insightsand derive design principles for 60 GHz practitioners.

Antenna patterns. We observe that the antennapattern of consumer-grade 60 GHz devices not only ex-hibits significant side lobes due to its cost-effective de-sign but also significantly depends on the specific loca-

TCP

Thro

ughp

ut (m

bps)

600

700

800

900

1000 20 40 60 80 120Time (s)

WiHD on WiHD off

Figure 23: Reflection interference impact.

tion of the antenna array on the device. This hindersany assumptions regarding channel access at the MAClayer—in scenarios where devices with certain beampatterns do not interfere, others may cause collisions.As a design principle, we derive that 60 GHz networksshould implement multiple MAC behaviors and choosethe one which is most suitable for the beam patterns ofthe individual devices in the network.

Reflections. Our measurements show that 60 GHzsignals may cause interference even after reflecting twiceoff walls. As a result, MAC layer designs which exploitthe sparsity of 60 GHz signals to increase spatial reusemay incur unexpected collisions. Such designs are oftenbased on geometric principles, that is, the beam patternand range of a node define its area of interference. Asa design principle, we derive that such protocols shouldextend this geometric approach to include up to twosignal reflections off walls or obstacles if possible.

Aggregation. Our frame-level analysis reveals that802.11ad aggregation provides large throughput gainsat much smaller timescales than legacy systems. Forinstance, while 802.11ac achieves a 2× gain with a framelength of 8 ms (c.f. Table 1 in [19]), 802.11ad achievesa 5.4× gain by aggregating only up to 25 µs due to itsvery high data rates. Hence, the impact of aggregationon delay is much smaller for 802.11ad, while reducingdramatically channel usage time. As a design principle,we derive that the frame length should not only dependon the desired throughput and delay, but also on howmany nodes share the medium. If many nodes share itdue to, e.g., wide beam patterns, a higher aggregationlevel helps to provide channel time for all nodes.

Range. We show that, even in the same setup, therange of 60 GHz networks often varies significantly be-tween experiments (c.f. Figure 13). This may be due to,e.g., different atmospheric conditions on different days.As a design principle, we derive that devices may needto adjust their transmit power to control interferenceeven in quasi-static scenarios, such as homes.

6. CONCLUSIONWe present an in-depth analysis of consumer-grade

off-the-shelf 60 GHz systems. Our goal is to investigatethe impact of the cost-effective designs of such devices.While these effects are often qualitatively well-known,we quantify them in order to provide a practical intu-ition on how important they are. This contributes cru-cial insights for the design of 60 GHz networking pro-tocols. In particular, we investigate the impact of dataaggregation, coding and modulation schemes (MCSs),beam patterns, reflections, and interference. We findthat data aggregation yields throughput gains up to5.4×. Further, MCS fluctuations may occur even instatic environments due to beam realignments. Beam-forming itself often results in strong side lobes due to thelimitations of consumer-grade antennas. As a result, weobserve significant interference in scenarios with two ormore allegedly non-interfering 60 GHz links. We con-clude that, while some common 60 GHz assumptionshold (e.g., data aggregation), others become critical forconsumer-grade devices (e.g., impact of side lobes).

7. ACKNOWLEDGMENTSThis work is partially supported by the European Re-

search Council grant ERC CoG 617721, the Ramon yCajal grant from the Spanish Ministry of Economy andCompetitiveness RYC-2012-10788, and the Madrid Re-gional Government through the TIGRE5-CM program(S2013/ICE-2919).

8. REFERENCES[1] IEEE, “Wireless LAN Medium Access Control

(MAC) and Physical Layer (PHY) SpecificationsAmendment 3: Enhancements for Very HighThroughput in the 60 GHz Band,” IEEE Std802.11ad-2012, 2012.

[2] T. Nitsche, C. Cordeiro, A. Flores, E. Knightly,E. Perahia, and J. Widmer, “IEEE 802.11ad:directional 60 GHz communication formulti-Gigabit-per-second Wi-Fi [Invited Paper],”IEEE Communications Magazine, 2014.

[3] C. Cordeiro, D. Akhmetov, and M. Park, “Ieee802.11Ad: Introduction and PerformanceEvaluation of the First Multi-gbps WifiTechnology,” in Proc. ACM mmCom, 2010.

[4] N. Moraitis and P. Constantinou, “Indoor channelmeasurements and characterization at 60 GHz forwireless local area network applications,” IEEETransactions on Antennas and Propagation,vol. 52, no. 12, 2004.

[5] H. Xu, V. Kukshya, and T. S. Rappaport,“Spatial and temporal characteristics of 60-GHzindoor channels,” IEEE Journal on Selected Areasin Communications, vol. 20, no. 3, 2002.

[6] H. Yang, P. Smulders, and M. Herben, “FrequencySelectivity of 60-GHz LOS and NLOS IndoorRadio Channels,” in Proc. 63rd IEEE VTC, 2006.

[7] T. Zwick, T. Beukema, and H. Nam, “Widebandchannel sounder with measurements and modelfor the 60 GHz indoor radio channel,” IEEETransactions on Vehicular Technology, vol. 54,no. 4, 2005.

[8] T. Manabe, Y. Miura, and T. Ihara, “Effects ofantenna directivity and polarization on indoormultipath propagation characteristics at 60 GHz,”IEEE Journal on Selected Areas inCommunications, vol. 14, no. 3, 1996.

[9] K. Hosoya, N. Prasad, K. Ramachandran,N. Orihashi, S. Kishimoto, S. Rangarajan, andK. Maruhashi, “Multiple Sector ID Capture(MIDC): A Novel Beamforming Technique for60-GHz Band Multi-Gbps WLAN/PAN Systems,”IEEE Transactions on Antennas and Propagation,vol. 63, no. 1, 2015.

[10] A. Lamminen, J. Saily, and A. Vimpari, “60-GHzPatch Antennas and Arrays on LTCC WithEmbedded-Cavity Substrates,” IEEETransactions on Antennas and Propagation, 2008.

[11] X.-P. Chen, K. Wu, L. Han, and F. He, “Low-CostHigh Gain Planar Antenna Array for 60-GHzBand Applications,” IEEE Transactions onAntennas and Propagation, vol. 58, no. 6, 2010.

[12] S. B. Yeap, Z. N. Chen, and X. Qing,“Gain-Enhanced 60-GHz LTCC Antenna ArrayWith Open Air Cavities,” IEEE Transactions onAntennas and Propagation, vol. 59, no. 9, 2011.

[13] X. Tie, K. Ramachandran, and R. Mahindra, “On60 ghz wireless link performance in indoorenvironments,” in Passive and ActiveMeasurement, ser. Lecture Notes in ComputerScience, 2012, vol. 7192.

[14] Y. Zhu, Z. Zhang, Z. Marzi, C. Nelson,U. Madhow, B. Y. Zhao, and H. Zheng,“Demystifying 60GHz Outdoor Picocells,” in Proc.ACM Mobicom, 2014.

[15] C. Hansen, “WiGiG: Multi-gigabit wirelesscommunications in the 60 GHz band,” IEEEWireless Communications, vol. 18, no. 6, 2011.

[16] P. Schmulders, “Exploiting the 60 GHz Band forLocal Wireless Multimedia Access: Prospects andFuture Directions,” IEEE CommunicationsMagazine, vol. 40, no. 1, 2002.

[17] S. Sur, V. Venkateswaran, X. Zhang, andP. Ramanathan, “60 ghz indoor networkingthrough flexible beams: A link-level profiling,” inProc. ACM SIGMETRICS, 2015.

[18] Vubiq. V60WGD03 60 GHz WaveguideDevelopment System. [Online]. Available:http://www.pasternack.com/60-ghz-development-systems-category.aspx

[19] S. Byeon, K. Yoon, O. Lee, S. Choi, W. Cho, andS. Oh, “Mofa: Mobility-aware frame aggregationin wi-fi,” in Proc. ACM CoNEXT, 2014.

http://www.pasternack.com/60-ghz-development-systems-category.aspxhttp://www.pasternack.com/60-ghz-development-systems-category.aspx

IntroductionBackgroundMeasurement SetupDevices and Measurement EquipmentMeasurement Setup

ResultsProtocol AnalysisBeam PatternsReflectionsInterference

DiscussionConclusionAcknowledgmentsReferences