Picasso: Flexible RF and Spectrum...
2
Picasso: Flexible RF and Spectrum Slicing Steven Hong Stanford University [email protected] Jeffrey Mehlman Stanford University [email protected] Sachin Katti Stanford University [email protected] Categories and Subject Descriptors: C2.1 [Computer Systems Organization]: Computer Communication Net- works Keywords: Radio Virtualization, Interference Cancellation 1. INTRODUCTION Many applications can benefit from the capability to simul- taneously and independently use arbitrarily sized but separate spectrum fragments with a single radio and antenna. By this capability we mean that the radio can simultaneously trans- mit, simultaneously receive, or simultaneously transmit and receive on arbitrary but separate spectrum fragments. For ex- ample, we can use it for spectrum aggregation in fragmented ISM bands as shown in as shown in Fig. 2(A). A WiFi AP can run independent OFDM PHY and CSMA MAC proto- cols on two WiFi channels to simultaneously serve two legacy WiFi clients assigned to different channels and achieve signif- icantly higher throughput than a legacy AP that is restricted to use only one channel at a time. Similarly, a WiFi client radio with such a capability can use it to simultaneously con- nect to multiple WiFi APs on different channels and obtain a much higher aggregate throughput than current radios that can transmit or receive on only one channel at a time. Such a capability could also be used for radio sharing and coexistence. Portable consumer devices such as smartphones must accommodate a growing list of separate ISM band pro- tocols, such as WiFi, WiFi-Direct, Zigbee, NFC, Bluetooth, etc. Current practice is to use a separate radio and antenna for each protocol, shown in Fig. 2(B), but as the number of ra- dios increase, it becomes increasingly difficult to find enough space to separately place all the antennas these radios would need (e.g. IPhone 4 “antenna-gate” was caused by antennas placed too closely [1]). Instead, a radio with a single antenna that allows simultaneous TX/RX on arbitrary spectrum frag- ments could be shared among all the above protocols, WiFi would use one fragment, Zigbee another and so on, saving valuable real estate on space constrained devices. They can operate their own independent PHY/MAC protocols on the shared radio without interfering with each other. Picasso provides a clean architecture for multiple such net- worked applications to co-exist by allowing one to decouple their operation from each other using slicing. Specifically, Pi- casso enables us to slice each device’s radio into three slices. The first slice would be assigned to the WiFi network, which Copyright is held by the author/owner(s). SIGCOMM’12, August 13–17, 2012, Helsinki, Finland. ACM 978-1-4503-1419-0/12/08. WiFi Network Group Video Streaming Group Game Controller Group Figure 1: Coexistence in Dense Environments with Picasso: Picasso enables devices operating in multiple networks to flexibly share fragmented spectrum and RF frontends, yet operate in a decoupled manner. is now free to operate its own optimized PHY and MAC pro- tocols to maximize throughput. The second slice would be for the gameplay video streaming application (Xbox, smart- phone, tablet and TV), and if it so chooses, would operate independent PHY and MAC protocols optimized to maxi- mize video quality. Finally, a third slice is for the gameplay controller application (Xbox, smartphone, tablet), with an independent PHY/MAC optimized for low latency. Thus Pi- casso decouples the operation of the networked applications from each other, and enables them to gracefully coexist yet independently design their protocols to optimize their respec- tive objectives. However, it is non-trivial to design such a radio. The key obstacle is that current radios cannot simultaneously transmit and receive on different arbitrary spectrum fragments with a single shared RF frontend and antenna. The reason for this is that the transmitted signal causes high powered self- interference, which ends up saturating the RX antenna and ADC, consequently nulling the received signal. While the standard solution is to utilize static analog RF filters to elim- inate the interference, such an option is infeasible because spectrum fragmentation is dynamic - available spectrum in the ISM band varies in space and time, depending on the presence of other wireless networks. Consequently, if a radio wants to leverage all the available spectrum and be able to simultaneously transmit and receive on different fragments, the shared analog frontend would need programmable ana- log filters that can be dynamically configured to let only the received signals through and filter out the self interference. Analog filters however are typically statically configured and programmable analog filters that can be changed dynamically are expensive and impractical to deploy in current radios [2]. 2. OUR APPROACH AND CONTRIBUTIONS: In this demonstration, we will present the design and imple- mentation of Picasso, a novel full duplex circuit design that 283
Transcript of Picasso: Flexible RF and Spectrum...
Picasso: Flexible RF and Spectrum Slicing
Steven HongStanford University
Jeffrey MehlmanStanford University
Sachin KattiStanford University
Categories and Subject Descriptors: C2.1 [ComputerSystems Organization]: Computer Communication Net-worksKeywords: Radio Virtualization, Interference Cancellation
1. INTRODUCTIONMany applications can benefit from the capability to simul-
taneously and independently use arbitrarily sized but separatespectrum fragments with a single radio and antenna. By thiscapability we mean that the radio can simultaneously trans-mit, simultaneously receive, or simultaneously transmit andreceive on arbitrary but separate spectrum fragments. For ex-ample, we can use it for spectrum aggregation in fragmentedISM bands as shown in as shown in Fig. 2(A). A WiFi APcan run independent OFDM PHY and CSMA MAC proto-cols on two WiFi channels to simultaneously serve two legacyWiFi clients assigned to different channels and achieve signif-icantly higher throughput than a legacy AP that is restrictedto use only one channel at a time. Similarly, a WiFi clientradio with such a capability can use it to simultaneously con-nect to multiple WiFi APs on different channels and obtaina much higher aggregate throughput than current radios thatcan transmit or receive on only one channel at a time.
Such a capability could also be used for radio sharing andcoexistence. Portable consumer devices such as smartphonesmust accommodate a growing list of separate ISM band pro-tocols, such as WiFi, WiFi-Direct, Zigbee, NFC, Bluetooth,etc. Current practice is to use a separate radio and antennafor each protocol, shown in Fig. 2(B), but as the number of ra-dios increase, it becomes increasingly difficult to find enoughspace to separately place all the antennas these radios wouldneed (e.g. IPhone 4 “antenna-gate” was caused by antennasplaced too closely [1]). Instead, a radio with a single antennathat allows simultaneous TX/RX on arbitrary spectrum frag-ments could be shared among all the above protocols, WiFiwould use one fragment, Zigbee another and so on, savingvaluable real estate on space constrained devices. They canoperate their own independent PHY/MAC protocols on theshared radio without interfering with each other.
Picasso provides a clean architecture for multiple such net-worked applications to co-exist by allowing one to decoupletheir operation from each other using slicing. Specifically, Pi-casso enables us to slice each device’s radio into three slices.The first slice would be assigned to the WiFi network, which
Copyright is held by the author/owner(s).SIGCOMM’12, August 13–17, 2012, Helsinki, Finland.ACM 978-1-4503-1419-0/12/08.
WiFi Network Group
Video Streaming Group
Game Controller Group
Figure 1: Coexistence in Dense Environments withPicasso: Picasso enables devices operating in multiplenetworks to flexibly share fragmented spectrum andRF frontends, yet operate in a decoupled manner.
is now free to operate its own optimized PHY and MAC pro-tocols to maximize throughput. The second slice would befor the gameplay video streaming application (Xbox, smart-phone, tablet and TV), and if it so chooses, would operateindependent PHY and MAC protocols optimized to maxi-mize video quality. Finally, a third slice is for the gameplaycontroller application (Xbox, smartphone, tablet), with anindependent PHY/MAC optimized for low latency. Thus Pi-casso decouples the operation of the networked applicationsfrom each other, and enables them to gracefully coexist yetindependently design their protocols to optimize their respec-tive objectives.
However, it is non-trivial to design such a radio. The keyobstacle is that current radios cannot simultaneously transmitand receive on different arbitrary spectrum fragments witha single shared RF frontend and antenna. The reason forthis is that the transmitted signal causes high powered self-interference, which ends up saturating the RX antenna andADC, consequently nulling the received signal. While thestandard solution is to utilize static analog RF filters to elim-inate the interference, such an option is infeasible becausespectrum fragmentation is dynamic - available spectrum inthe ISM band varies in space and time, depending on thepresence of other wireless networks. Consequently, if a radiowants to leverage all the available spectrum and be able tosimultaneously transmit and receive on different fragments,the shared analog frontend would need programmable ana-log filters that can be dynamically configured to let only thereceived signals through and filter out the self interference.Analog filters however are typically statically configured andprogrammable analog filters that can be changed dynamicallyare expensive and impractical to deploy in current radios [2].
2. OUR APPROACH AND CONTRIBUTIONS:In this demonstration, we will present the design and imple-
mentation of Picasso, a novel full duplex circuit design that
283
ZigBee Interference
ZigBee Interference
Legacy WiFi Client 1
Picasso AP
Legacy WiFi Client 2
Frequency
2.4
GHz
2.5
GHz
Ch15
(5MHz) WiFi 2WiFi 1Ch20
(5MHz)
Ch25
(5MHz)
ZigBee Interference
Bluetooth
Antenna
GPS
Antenna
WiFi Antenna
Primary
GSM
Antenna
(900M)
Extended GSM Antenna
4G GSM
Antenna
(DCS/PCS
Bands)
B)A)
Figure 2: A) Fragmented Spectrum is common inISM band:Picasso AP can aggregate multiple frag-ments and simultaneously service multiple clients ondifferent channels. B) Samsung Galaxy Printed An-tenna Layout: Instead of having an antenna per pro-tocol, Picasso enables multiple protocols to use thesame radio and antenna, saving space on the device.
sufficiently cancels (instead of filters) the self-interference inanalog and prevents antenna and ADC saturation, enablingthe radio to cleanly recover the received signal. Our key con-tribution here is a novel circuit design that leverages a circu-lator [3] to isolate TX and RX signals from a single antenna,and exploits the fact that self interference to the RX signaltravels through a fixed known circuit to design a staticallyconfigured self interference cancellation circuit. This allowsa radio to simultaneously transmit and receive on arbitraryspectrum fragments even while using a single RF-frontend andantenna.
Picasso leverages this full duplex capability to build a ra-dio slicing abstraction that allows one to flexibly slice a singleradio and antenna into separate independent slices operat-ing on different spectrum fragments. Each slice is associatedwith a specific spectrum fragment in the ISM band (whosewidth/position can be programmatically specified). The keyproperty is that the operation of each slice is decoupled fromthe other slices, i.e. the slice is free to run whatever nar-rowband PHY and MAC protocols it chooses, and the pro-tocol behavior is not impacted by any other slice that maybe present on the shared radio and antenna. Thus in theabove scenarios, the AP would have two slices correspond-ing to the two spectrum fragments and run two independentWiFi OFDM/CSMA protocols on the two slices in parallel.Similarly, a radio could be shared amongst multiple protocols(e.g. WiFi, Bluetooth, and NFC) by assigning independentslices to the appropriate spectrum fragment and running thecorresponding protocol.
Further, to ensure that each slice can use existing well-engineered narrowband PHY and MAC protocols on eachslice, Picasso designs a novel reconfigurable filter engine thattransparently takes signals spread over different spectrum frag-ments, and efficiently filters and resamples them so that thehigher layers just see a simple sample stream consisting of nar-rowband digital samples. The higher layers are then free toprocess these samples with any narrowband PHY techniquethey choose, and schedule access to the slice with a MACprotocol of their choice. The slice for all intents and purposesappears as their own piece of spectrum centered at zero op-erating on their own radio, Picasso thus completely abstractsout the complexity of spectrum fragmentation.
Spectrum Slicing (NI PXI-7695)
ADC/DAC (NI 5781)
Passive
Attenuation
+ Delay
Balun
Single Antenna
Circulator
Self Interference Cancellation
Figure 3: Picasso Radio Physical Setup: Built withoff the shelf components and software radios from Na-tional Instruments.
3. DEMONSTRATION SETUPWe design and implement a prototype of Picasso on Xilinx
Virtex-5 FPGA based software radios [4]. Our implementa-tion consists of both the radio design that provides the slicingabstraction, as well as a WiFi style contiguous OFDM PHYand CSMA MAC to operate on top of the slices. We show thatPicasso’s implementation of the slicing abstraction providesstrong decoupled operation, i.e. there’s no SNR loss becauseof either programmable filtering or simultaneous TX/RX. Inother words, a Picasso radio achieves the same throughput asone would have achieved by using several independent radiosstatically configured to operate on individual slices.
We have also implemented a WiFi style OFDM PHY anda CSMA MAC split across the FPGA and the realtime OS.The OFDM PHY can be configured to operate over differentbandwidths (from 20MHz to 5MHz), and runs in realtime onthe SDR platform. It supports all the WiFi constellations(from BPSK to 64-QAM) as well as channel coding rates(1/2, 2/3, 3/4 convolutional coding). We have not imple-mented an automatic rate adaptation algorithm, but in theexperiments we empirically pick the combination of constel-lation size and channel coding rate that maximizes the linkthroughput. Thus our prototype implements all the relevantfeatures from a WiFi PHY/MAC. Note that we can run mul-tiple instances of the WiFi PHY and MAC one for each sliceon the device.
We will be demonstrating this physical setup, first demon-strating that a typical node without Picasso cannot transmitand receive on different arbitrary bands. Without sharp ana-log filters (which are fixed in nature), the transmit signal over-whelms the received signal even when the two are on differentbands. We then test the performance of a Picasso node in thesame scenario - demonstrating that it can simultaneously op-erate independent connections over multiple spectrum slices.
4. REFERENCES[1] Gizmodo, ”IPhone 4 Antenna-Gate”,
http://gizmodo.com/5846638/giz-explains-whats-so-smart-about-the-iphone-4ss-antenna,2011.
[2] Matthaei, George L., Jones, E. L. and Young, Leo,”Microwave filters, impedance-matching networks, andcoupling structures”, Dedham, Mass: Artech House Books.ISBN 0-89006-099-1..
[3] D. Pozar, ”Microwave Engineering: 3rd Edition”, Wiley, 2004.[4] NI PXIe-8133 User Manual,
http://www.ni.com/pdf/manuals/372870b.pdf, 2011.
284
