Presentation on theme: "Improved CCA for 80 and 160 MHz BSSs"— Presentation transcript:
1Improved CCA for 80 and 160 MHz BSSs September 2006doc.: IEEE /1458r0July 2010Improved CCA for 80 and 160 MHz BSSsDate:Authors:Hart et al (Cisco)Joonsuk Kim, Broadcom Corp.
2Outline Problem Statement Coexistence mechanisms September 2006doc.: IEEE /1458r0July 2010OutlineProblem StatementCoexistence mechanismsLet’s choose a CCA better suited to wider bandwidthsLet’s explore a new coexistence technique: Receiver CCASummaryHart et al (Cisco)Joonsuk Kim, Broadcom Corp.
3September 2006doc.: IEEE /1458r0July 2010Problem: WiFi can experience overlapping BSSs (OBSSs) anywhere, any time, on any channelThree sources of difficultyWiFi is increasingly used for QoS applications80 and 160 MHz BSSs affect, and are affected by, a larger proportion of the available spectrum; e.g. adjacent home/office legacy APsPersonal/mobile usage of WiFi is increasing, driven by smartphone uptake and 3G backhaul, using ad hoc mode, 3G/WiFi bridges (e.g. “MiFi”) and soon WiFi-Direct.It is harder to locally harmonize:the Primary channel of mobile devicesthe Primary channel of legacy 11a/11n APs with 80/160 MHz APsBasically, WLANs can experience overlapping BSSs (OBSSs) anywhere, any time, on any channelYet customers want reliable very-high throughputrather than unreliable ultra-high throughputHart et al (Cisco)Joonsuk Kim, Broadcom Corp.
411n coexistence was optimized for a mostly-free secondary channel September 2006doc.: IEEE /1458r0July 201011n coexistence was optimized for a mostly-free secondary channelUnequal CCA protection by channel-82/-79 dBm CCA sensitivity for valid 20/40 MHz packets that include the PrimaryOnly -62 dBm CCA protection on secondaryMore hidden nodes on the secondaryNo virtual carrier sense on the secondaryA mostly free secondary and tertiary and quaternary etc is much less probable with a mixture of 80/160 MHz, mobile and legacy BSSsWe want to do better in 11acHart et al (Cisco)Joonsuk Kim, Broadcom Corp.
5Unequal CCA thresholds introduce unfairness September 2006doc.: IEEE /1458r0July 2010Unequal CCA thresholds introduce unfairnessThe default extension of 11n to 11ac is:-82/-79/-76/-73 dBm CCA sensitivity for valid 20/40/80/160 MHz packets that include the primaryBut only -62 dBm per 20 MHz for all other channelsThis leads to CCA unfairness for devices on non-Primary channelsA and B are close enough for error-free communications even when colliding with C or D transmissions, yet far enough that the -62 dBm secondary CCA isn’t triggered by C or D.C and D see A and B and each other on their Primary so defer to everyone using -82 dBm CCAFull details of 20/40 MHz CCA unfairness in 07/3000r2The 20/40 MHz problem is very similar to the 40/80 MHz problem (and 40/160 and 80/160)Hart et al (Cisco)Joonsuk Kim, Broadcom Corp.
6Does CSMA/CA on the Primary channel only approach ALOHA? September 2006doc.: IEEE /1458r0July 2010Does CSMA/CA on the Primary channel only approach ALOHA?Imagine:Lots of OBSSs, a mixture of legacy and 160 MHz 11ac, on uncoordinated Primary channels with STAs at moderate distances (<-65 dBm)RTS/CTS and virtual carrier sense respected only on 20 MHz out of 160 MHzSensitive CCA only for signals that include the Primary, else -62 dBm CCA(and -62 dBm is not triggered in this example)On the non-primary channels, there is not much CS, nor much CA.Do we approach ALOHA-like efficiency and/or instability?Hart et al (Cisco)Joonsuk Kim, Broadcom Corp.
7Overview of Selected Coexistence Mechanisms September 2006doc.: IEEE /1458r0July 2010Overview of Selected Coexistence MechanismsPHY (physical carrier sense aka CCA)Ensures transmitter doesn’t collide with nearby transmittersVariants that exclude or include PLCP decoding per channelMay say little about the responder’s environment; no virtual carrier senseReceiver CCAAttempt to learn something about responder’s CCA environment via a frame exchangeNo virtual carrier senseMAC (virtual carrier sense aka NAV)All variants require one PPDU decoder per channelLoses NAV during transmissions, although mitigation techniques existIn this presentation we only address techniques that do not require multiple PLCP or PPDU decodersTechniques that involve more than one PLCP or PPDU decoder require careful analysis to determine if the benefit justifies the complexityHart et al (Cisco)Joonsuk Kim, Broadcom Corp.
8PHY Coexistence Mechanisms September 2006doc.: IEEE /1458r0July 2010PHY Coexistence MechanismsNon-PLCP decodingEnergy Detection, on each channelParallel filters and energy detection – low complexityMore false alarms if threshold is below -72 dBmPreamble Detection, on each channelParallel filters and short symbol detectors - low complexityBlinded while transmitting on a subset of channelsMid-packet Detection, on each channelParallel filters and cyclic extension detectors - low complexityResynchronizes quickly even after transmitting on a subset of channels(PLCP decoding, on each channel)High complexityMOSTLYHart et al (Cisco)Joonsuk Kim, Broadcom Corp.
9Example Scheme for Mid-Packet CCA for OFDM September 2006doc.: IEEE /1458r0July 2010Example Scheme for Mid-Packet CCA for OFDMMOSTLYOFDM looks like Gaussian noise yet can be identified by its regular cyclic extensionObscured by carrier frequency offsets and delay spreadNo complicated processing hereMany improvements and/or simplifications are possibleHart et al (Cisco)Joonsuk Kim, Broadcom Corp.
10Two independent sim studies show effectiveness September 2006doc.: IEEE /1458r0July 2010Two independent sim studies show effectivenessMOSTLYCh C/E, 1x1Ch D, 1x1See 07/3001r2 (Hart) and 10/0012r0 (Kim) for full results10/0012r0 shows that Pmiss is similar to or better than HT-SIG error rate of primary channelSummary: sensitive non-primary, non-PLCP CCA is feasibleHart et al (Cisco)Joonsuk Kim, Broadcom Corp.
11Problem with duplicated RTS/CTS September 2006doc.: IEEE /1458r0July 2010Problem with duplicated RTS/CTSInitiator sends duplicated RTS to 40/80/160 MHz responder40/80/160 MHz responder detects RTS on primary, and sends duplicated CTSWithout regard to conditions on the non-primary channels – e.g. if one or more are busyThere is actually no collision detection/avoidance on non-primary channels when the responder only considers the primaryThe purpose of RTS/CTS is collision detection with hidden nodesThe responder could try to detect the RTS on the non-primary, but has higher implementation complexity, still may not detect collisions, and was not required by 11nHart et al (Cisco)Joonsuk Kim, Broadcom Corp.
12Receiver CCA Coexistence Mechanism September 2006doc.: IEEE /1458r0July 2010Receiver CCA Coexistence MechanismHow it worksTXOP begins with an initFrame/initResponseResponder does non-PLCP CCA (ED/preamble/mid-pkt) on all non-primary channelsinitResponse indicates which channels were clear during the PIFS leading up to the initFrame (e.g. via 3 bits for ch2, ch3-4, ch5-8, each indicating all-clear or any-busy/incapable)initResponse could be:A CTS or Ack that includes 3 bits of multichannel CCA busy/free state in Service field or PHY paddingA new control frame, etcinitFrame/initResponse are duplicated packetsResponder reports “ch1 & 2 clear; 3 & 4 busy”Responder reports “ch1, 2, 3 & 4 clear”ch1Dup-initRespInitiator data to responderBAch2Initiator data to responder…Responder is performing non-PLCP CCA on non-Primary channelsBADup-initFrameDup-initFrameDup-initResponsech3…OBSSdataBAch4…Hart et al (Cisco)Joonsuk Kim, Broadcom Corp.
13Benefits of Receiver CCA Mechanism September 2006doc.: IEEE /1458r0July 2010Benefits of Receiver CCA MechanismLow PHY complexityOnly requires ED or mid-packet CCALow MAC complexityIntegrates well with existing RTS/CTS or initial Data/Ack exchangesBetter than just CCA at the transmitterEnables RTS/CTS or initial Data/Ack to detect collisions on non-Primary channelsThis reduces collisions with OBSS(s)Especially valuable as the bandwidth of the signal gets wider, since then:there is more potential for OBSS(s) andthe NAV state of the Primary channel is a much less complete picture.Hart et al (Cisco)Joonsuk Kim, Broadcom Corp.
14Summary July 2010 We should do better than 11n September 2006doc.: IEEE /1458r0July 2010SummaryWe should do better than 11nIt is feasible to do better than 11nPLCP decoding on primaryPlus multichannel ED or mid-packet CCAimplementer’s choicespec just defines a threshold such as TBD dBm for each 20 MHz channelIn order to avoid overlapped transmissions in neighborhood of transmitterNo known, reasonable virtual carrier sense solution existsParallel PPDU decoders is a tough requirementThe proposed Receiver CCA mechanism enables detection and avoidance of non-Primary-channel collisions at receiverComplexity is modestEnables duplicated RTS/CTS or initial Data/Ack exchanges to perform non-Primary collision detectionRecommended to the group for further studyHart et al (Cisco)Joonsuk Kim, Broadcom Corp.
15? Questions? July 2010 September 2006 doc.: IEEE 802.11-06/1458r0 Hart et al (Cisco)Joonsuk Kim, Broadcom Corp.
16Strawpoll Add to the Spec Framework Document a new requirement: September 2006doc.: IEEE /1458r0July 2010StrawpollAdd to the Spec Framework Document a new requirement:An 11ac device shall provide a CCA per 20 MHz channel, for all 20 MHz channels that the device is presently capable of transmitting over.The CCA sensitivity shall be:TBD (<-62) dBm for valid signals-62 dBm for any signal.Y15N0A21Hart et al (Cisco)Joonsuk Kim, Broadcom Corp.