"OUT OF PRINT" F4i11 HEWLETT ~~PACKARD. HIPERLAN An Air Interface Designed for Multi-Media

Transcription

1 "OUT OF PRINT" F4i11 HEWLETT ~~PACKARD HIPERLAN An Air Interface Designed for Multi-Media Timothy A. Wilkinson Networks and Communications Laboratory HP Laboratories Bristol HPL May, 1995 HIPERLAN, radio LANS, wireless LANS The target applications of HIPERLAN are discussed and the target performance of HIPERLAN is specified. The HIPERLAN PHY and MAC designs, to deliver this performance, are described. Areas for further research are suggested. Internal Accession Date Only Copyright Hewlett-Packard Company 1995

2 HIPERLAN An Air Interface Designed for Multi-Media T A Wilkinson Hewlett Packard Laboratories, FUton Road, Stoke Gifford, Bristol, BS12 6QZ, UK Tel: Fax: taw@hplb.hpl.hp.com Abstract The target applications of HIPERLAN are discussed and the target performance of HIPERLAN is specified. The HIPERLAN PHY and MAC designs, to deliver this performance, are described. Areas for further research are suggested.!.introduction Currently, there are two standards for radio LANs (Local Area Networks) being formulated; these are IEEE and HIPERLAN. IEEE is a standard for systems that will operate in the ISM (Industrial Scientific and Medical) bands, these being MHz (in the USA only) and GHz (available worldwide). These bands are not dedicated spectrum and hence these systems are required to use spread spectrum techniques to reduce the effect of considerable mutual interference between these and other ISM band systems. The transmission rate of these systems is around 1Mbit/s. HIPERLAN (High PErformance Radio LAN) is a European standard for operation in the bands GHz and GHz. These bands are dedicated spectrum allocated to the HIPERLAN CAl (Common Air Interface) standard. The provision of considerable dedicated spectrum for HIPERLAN should enable it to provide higher performance than IEEE in terms of reliability and throughput. The initial aim was that HIPERLAN should deliver wired LAN like performance which implies a transmission rate of 10-20Mbit/s. This paper presents an introduction to the design of the HIPERLAN standard in relation to its ability to support multi-media communications. The target applications of HIPERLAN are discussed in section II. The target performance of HIPERLAN is specified in section III. The HIPERLAN PHY (physical layer) and MAC (Medium Access Control layer) designs are described in sections IV and V respectively. Areas for further research are suggested in section VI. II.T.rget applications In 1992 the CEPT (Conference European des administrations des Postes et des Telecommunications) allocated the HIPERLAN bands GHz and GHz with EIRPEP (Equivalent Isotropic Radiated Peak Envelope Power) limits of OdBW and -10dBw respectively (1). ETSI (European Telecommunication Standards Institute) started work on the HIPERLAN standard in early 1992 and finished the draft standard for the 5GHz band in early Figure 1 Wired LAN replacement 1

3 At conception two quite different applications were perceived for HIPERLAN, the wired LAN replacement and the ad-hoc network. These are illustrated in figures 1 and 2 respectively. The wired LAN replacement application provides the user with the convenience of mobility and the elimination of cable management problems around his work area. However, initial installation could be more expensive than a wired LAN if the system is not carefully designed. The ad-hoc network allows the user to create a network anywhere without the need for a wired infrastructure. A wired infra-structure is unlikely in the environment shown in figure 2. However, more typical environments are at work or in the home or anywhere where people may wish to work together or play together using networked computers. lit has a wired infra-structure with access points and the ad-hoc network does not. Many of the other attributes follow from this. If there is no wired Infra-structure then a mesh communication topology is appropriate and a distributed medium access control is compatible with this. In the wired LAN replacement the access points could be placed in elevated positions and hence a line-of-sight radio propagation path is likely, whereas in the ad-hoc networkthe antennas will be buried in the room clutter and a line-of-sight will be unlikely. The existence of a line of sight reduces the degree of fading and time dispersion (illustrated in figure 3). Fading causes the link to be unreliable and can cause the hidden node problem where connection is difficult between stations that are in close proximity with one another. With aline of sight the fading statistics are Rayleigh, without they are Ricean. Time dispersion causes lsi (Inter-Symbol-Interference) which ultimately limits the transmission rate in the channel or forces the use of multipath counter-measures such as equalisation. The delay spread in indoor environment is usually less than 1oons. Typically, this limits the transmission rate to around 1Mbit/s without multipath counter-measures. Figure 2 Ad Hoc network _ LAIl M-HOC _ Il!1'\ACEMeHT INFM STIlUCTUIlE YES NO TO_Y ITAII ME'" COIlMUHl ME_ACC;III CINTMUIID ~ CONTIlOl. (OIl DIITNMJTIDI UNI-Of_ LIKELY UNI.lKILY 'AIlING IlIC1AH MYLI1GH MEO OELAYPMAO LOW - COULD..PLUelID '''~_1ITT1Il CAPACITY IlIQUlll!MfllT HIGH LOW T_1e AIY~ -- _ Table 1 Comparison of applications The attributes of these two applications are compared in table 1. The most important difference between the two applications is that the wired LAN replacement -~ DllTANCI011TWI nme DISPERSION - _..~- DllTANCI0Ill.- Figure 3 Fading and time dispersion In both cases the medium re-use should be self managing so that these networks are easy to deploy, increasing their usability. The support of multi-media applications requires the support of all types of traffic, asynchronous and isochronous. 2

4 The communication of asynchronous traffic such as conventional LAN traffic is typically connectionless, as the traffic is bursty. With asynchronous traffic, minimising the medium access delay and maximising the transmission rate are important. Currently, wired LAN standards such as Ethernet have transmission rates of around 10Mbit/s. The communication of isochronous traffic such as audio or video is typically connection oriented as the traffic is constant rate. With isochronous traffic minimising the transfer delay is important. Typical requirements are; for audio 32kbit/s with a maximum transfer delay of 10ms, and for video 2Mbit/s with a maximum transfer delay of 100ms. Typically the capacity requirement of the wired LAN replacement will be higher than that of the ad-hoc network. III. Performance targets The aim of ETSI was to create a standard that could support both of the two extreme applications discussed previously. Based on the above requirements the following performance targets were set for HIPERLAN. Range - 50m Collocation tolerance - 5m Mobility tolerance - 10m/s, Is Delay spread tolerance - 100ns requirements. IV.HIPERLAN PHY Adaptive equalisation was chosen as the HIPERLAN multipath counter-measure to facilitate high transmission rates. It is envisaged that the system will use a DFE (Decision Feedback Equaliser) although this is not specified in the standard. HIPERLAN transmission can be at one of two data rates. The high data rate is 23.5Mbit!s. The modulation is pre-coded non-differential GMSK (Gaussian Minimum Shift Keying) with a BT (bandwidth bit-period product) of 0.3. This is transmitted in a channel of 23.5MHz bandwidth. There are five channels in the 5GHz band. The low data rate is 1.5Mbit/s, 1/16 of the high data rate. This will be transmitted using the same modulation and direct sequence spread spectrum. This can be detected with a matched filter or correlator. The HIPERLAN data packet consists of three parts a low data rate header, a synchronisation and training sequence, and the data. This is shown in figure 5. The low data rate header contains a truncated version of the destination 48bit MAC address which is transmitted at 1.5Mbit!s. This can be demodulated without equalisation. This is to facilitate power saving strategies where the receiver is only activated when the MAC address is detected in the low data rate header. Packet failure ratio - < 10-3 MAC - distributed PACKET Asynchronous support - asynchronous MAC ldw l MrE_ Asynchronous data rate Mbit!s Isochronous support - priority in MAC Isochronous data rate kbit!s Target format - PCMCIA Power consumption - a few hundered mw The following sections describe the HIPERLAN MAC and PHY design to meet these Figure 5 HIPERLAN packet The synchronisation and training sequence is 450 bits long. 3

5 The data consists of m data blocks of 496 bits where m is from 1 to 47. (the maximum packet length of 47 blocks is calculated for a relative velocity of 1.4m/s.) The packet is designed in this way to make it compatible with Ethernet. A block of 416 bits is divided into 16 segments and FEC (Forward Error Correction) encoded with a BCH (31,26) code to give a block of 496 bits. The HIPERLAN acknowledgement packet is also transmitted at the low data rate so that it can be demodulated without equalisation. There are three classes of transmitter 30dBm, 20dBm and 1OdBm EIRPEP. Antenna diversity is an option and directional antennas can be used but the transmit power must be scaled accordingly. If antenna diversity is used, the antenna used for CCA (Clear Channel Assessment) must also be used for transmitting. The MAC protocol CCA uses an adaptive threshold based on the signal strength measured in the environment. To enable the MAC protocol to function efficiently, the TX to RX turn around time must be less than 5ps. Concerns have been expressed about the complexity of the HIPERLAN PHY and the associated power consumption. These concerns can only be investigated with HIPERLAN hardware demonstrators. V. HIPERLAN MAC The only workable way to implement a distributed MAC protocol was to use a variant of CSMA (Carrier Sense Multiple Access) or Iisten-before-talk that is used to great effect in Ethernet. The difficulty with implementing this on a radio link is that the large dynamic range between transmit and receive signal strengths makes CO (Collision Detection) or IIsten-whlletalk very difficult. Hence, most radio versions of CSMA use CA (Collision Avoidance), which is a randomisation of the collision probability rather than actual avoidance. The HIPERLAN MAC protocol uses a form of Iisten-before-talk with a random listen-talk sequence prior to data packet transmission. The protocol also has priority capability designed to ensure delivery times of isochronous traffic. The MAC protocol mechanism is shown in figure 6 and described below. There are three phases in packet transmission; these are, the prioritisation phase, the contention phase and the transmission phase, all of which are of variable duration. If no activity has been sensed on the channel for 1700 bit periods, transmission can take place immediately. Otherwise, it is assumed that all stations wishing to transmit will synchronise to the end of the last transmission and execute the three phases. StatIon 1 Station 2 Station3 Station4 priorltlsatlon contention transmission stops ~OJ:. TIme ~~ datanacket Figure 6 HIPERLAN MAC protocol mechanism The prioritisation phase exists to prioritise transmissions. The prioritisation phase consists of a maximum of 5 slots of 10ps (256bits) duration. Priority is asserted by transmitting in a slot, 1 to 5, 1 for highest priority, 5 for lowest priority. Stations listen until they transmit. The priority phase ends when one or more stations assert their priority and listening stations with lower priority hear this and defer. Packet priority is related to the residual lifetime of the packet, number of hops to final destination and application assigned priority. So asynchronous traffic such as file transfer will have a long lifetime and low priority, whereas isochronous traffic such as audio or video will have a short lifetime and high priority. The priority for a packet increases until the lifetime expires and then the packet is discarded. When this happens the MAC informs the higher layers. This has interesting implications for multimedia applications. The contention phase exists to resolve contention between stations with the same priority. The contention phase consists of an elimination phase and a yield phase. The elimination phase consists of a maximum of 12 slots of 10ps (256bits). Stations entering the 4

6 elimination phase after the prioritisation phase will continue to transmit in successive slots with probability 0.5 then listen for a single slot. If they hear nothing, they then enter the yield phase. The duration of this phase is 1 to 12 slots. The yield phase consists a maximum of 14 slots of 2.5ps (64bits). Stations entering yield phase from elimination phase will continue to listen in successive slots with probability 0.9. If they hear nothing they then enter the transmission phase. The transmission phase simply consists of packet transmission. Packet reception is acknowledged by transmission of an acknowledgement. Figure 6 shows four stations contending for the channel. Station four has a lower priority packet than the other stations and is eliminated in the prioritisation phase. Stations 3 is eliminated in the elimination part of the contention phase. Station 2 is eliminated in the yield part of contention phase and station four gets the channel. The big question is whether this MAC specification satisfies the original requirements specification. On paper it looks promising, but the answer to this question will be known only when HIPERLAN hardware is available. However, early simulation results from ETSI studies have shown that the MAC protocol can adequately support asynchronous and isochronous traffic. For example it was shown that 25 audio links at 32kbit/s (10ms delivery time), 25 audio links at 13kbit/s (20ms delivery time), 1 video link at 2Mbit/s (1ooms delivery time) and asynchronous file transfer at 13.4Mbit/s can simultaneously be supported by the HIPERLAN MAC. The HIPERLAN standard has other elegant features. HIPERLAN has multi-hop relaying capability facilitated by neighbour tables. These are updated periodically when hello packets are transmitted by neighbour stations. There is also a power saving strategy involving wake sleep cycles specified. VI. Conclusions and areas for further research multi-media traffic. In addition to an outline of the standard, one of the most important things required by academia and industry at this time is a guide to what additional work is required before the standard can be realised as a viable product. The list of research areas below is by no means exhaustive. Diversity antennas for PCMCIA realisations This paper has described the HIPERLAN standard, which should be one of the most [2] powerful and flexible air interfaces yet designed and one of the first radio CAls designed specifically for LAN system architectures and Diversity strategies Low profile RF components for PCMCIA realisations Local oscillator frequency offset measurement and compensation Equaliser technology (high speed and low power) Equalisers tolerant to local oscillator frequency offset Equaliser training algorithms CCA optimisation LAN trafficmodels for MAC optimisation Simulation studies for MAC optimisation General power saving strategies With attention given to these areas we should see HIPERLAN become as successful as other ETSI standards such as GSM (Global Standard for Mobile) and DECT (Digital European Cordless Telecommunications). VII. References [1] CEPT, "Relating to the Harmonised Radio Frequency Bands for High PErformance Radio Local Area Networkn CEPT Recommendation T/R 22-06, ETSI, "Radio Equipment and Systems (RES) High PErformance Radio Local Area Network (HIPERLAN) Functional Specification", RES10TTG 95/07. 5