TELE4652 Mobile and Satellite Communication Systems

Transcription

1 TELE4652 Mobile and Satellite Communication Systems Lecture 10 IS-95 CDMA A second generation cellular standard, based on CDMA technology, was proposed by Qualcomm in the early 1990s. It was standardised as Interim Standard 95 (IS-95) in 1993, and first deployed in under the commercial name cdmaone in IS-95 was designed to coexist with AMPS, and as such shared the MHz Uplink and MHz downlink spectral allocation. Within the AMPS spectrum, frequency bands of 1.25 MHz groups could be assigned to cdmaone operators. As a CDMA based systems, all user shared the same spectral region, begin distinguished at the receiver by orthogonal spreading codes. The chip rate was Mcps, resulting in spreading factors of 128 for both forward and reverse channels. One cdmaone channel thus corresponds to 41 AMPS channels. Considering guard bands, the spectral allocation might look like the diagrams shown below.

2 CDMA has since been adopted as the standard multiple access technique over the air interface. Some of the attractive properties of a CDMA based cellular system are: inherent frequency diversity, since the signal bandwidth is now much larger than the channel coherence bandwidth; multipath diversity, through the use of RAKE receivers; privacy, since a user s signal appears below the noise floor; the ability to implement soft handoffs and macro diversity to improve signal quality; and the idea of the soft capacity limit. Network Structure A diagram illustrating the components of an IS-95 CDMA network is shown below. Apart from some slight differences in nomenclature, the components and their functionality are identical to that of GSM, so we do not need to repeat the description here. One of the biggest attractions of CDMA is a re-use factor of unity every frequency channel can be re-used in every cell throughout the network. This produces a much higher overall system capacity than a TDMA based cellular system, with capacity improvements of the order of times claimed. There are no co-channel interference issues in CDMA networks. The issue is to ensure there are a sufficient number of orthogonal spreading codes available across the network. IS-95 CDMA features two orthogonal spreading codes, one set to distinguish users within a cell (Walsh codes for the downlink and PN sequences for the uplink), and another set to distinguish between base stations (long PN sequences).

3 Network Functionality The basic functionality of the cdmaone network is very similar to GSM in the sense of call establishment and authentication procedures. The diagrams below illustrate the procedures of call establishment from a mobile and call directed towards a mobile. The pertinent features are still readily apparent: there is a fair degree of signalling overhead required prior to the assignment of a voice frequency channel for data communication (and in this, a traffic channel is associated to a Walsh code). In addition, as in GSM, there is the need to define certain traffic and logic channels to perform specific network tasks. When an IS-95 mobile is first turned on it sets its receiving code sequence to the all zero sequence and searches for the strongest pilot channel. This pilot channel gives the mobile a phase and timing reference. Then it sets its receiving code for synchronisation channel, to obtain cell specific information. Then the MS switches itself to the idle state and listens to the paging channel, from which it can receive

4 packets identifying system and access channel parameters the requisite information necessary for the MS to make a call and use network resources. Finally the mobile will place registration request message on an access channel, following which an authentication procedure will follow, as in GSM. As was discussed in an earlier lecture, power control is of critical importance in CDMA systems. The signals of other users appear as noise in the detection of a given user s signal. This leads to the danger of the near-far effect, whereby the high power signal received from a nearby base station can completely swamp and drown out the weaker signals received from far away mobiles. IS-95 CDMA uses both fast open loop power control, with a slower, closed loop power control for fine-tuning. Fast open loop power control involves the mobile altering its transmitted signal strength based on the power level it receives from the base station. Fast open loop power control is able to quickly compensate for rapid shadowing events, since there is no need to wait for instructions to be received from the base station. To facilitate closed loop power control, the base station measures the received signal strength from each mobile unit relative to the interference and sends a high-low indication to each mobile unit on the forward channel. Mobile units adjust their transmitted power level in discrete steps for each power control bit. These power control bits are sent at a rate of 800 Hz, and are inserted into the transmitted bit stream by stealing encoded data bits, akin to puncturing used in the 2.5G GSM adaptations.

5 Soft handoffs (SHO), whereby a mobile station can maintain connection to both base stations during a handoff, and can combine the independently received signals to obtain an overall enhanced combined signal. Essentially, the signals from independent base stations will merely appear to the mobile station as independent multipath components, and can be resolved by the fingers of the RAKE receiver. The diagram below gives a sampling of the signalling involved in the decision to instigate a handoff. To build further on this soft handoff idea, CDMA systems lend themselves to the implementation of macro-diversity. Since all transmissions occur at the same frequency, a sector base station can receive the signals from a mobile station in an adjacent sector by using the PN offset of that sector. A diversity gain could then be achieved by combining the signals received from several base stations. The implementation details of these macro diversity schemes are at the discretion of the network operators, but the point is that CDMA systems lend themselves readily to this possibility. A mobile station can also exploit macro diversity techniques. Depending on signal strength, sectors are continually added or removed to the list, called the active set, of servicing base stations. Members of the active set are managed and negotiated with the network over a signalling channel so that both ends of the radio link have a consistent view. All members of the active set implement macro diversity in both directions.

6 Channel Structure In IS-95 CDMA, the physical channel is the spreading code. For the forward link this is the user assigned Walsh code, while for the reverse link it is the user-specific long code sequence. Common reverse channels use a cell-specific long code sequence. As in GSM, certain logical formats are defined for the insertion of data onto these physical channels, called logical channels. These logical channels are characterised by the physical layer functionality that each provides, broadly classified as either a control channel or a logical channel.

7 The pilot channel is always assigned Walsh code 0 on the downlink. It is used as a phase reference for coherent demodulation on the mobile units and also for signal strength measurements (such as to decide on whether to attempt a handoff). Its message consists of the all zero sequence. The synchronisation channel is assigned Walsh code 32. This Walsh code sequence is half 0 s and then half 1 s. It transmits a 1200 bps message, that allows the mobile station to obtain system identification information, like system time, long code state, and network specific features. Walsh codes 1 through 7 are assigned as Paging channels, to carry user specific data packets. In the reverse direction the control channels are predominantly access channels, used to carry packets to the base stations. The content of these packets could contain location update requests, call initiation, paging response, etc. These channels will be discussed in more details when the specifics of the physical layer are described below. Protocol Stack As in all communication systems, the implementation of the sub-components of the IS-95 CDMA network is done in the form of a protocol stack. Each layer of the protocol stack performs a requiste operation, from the lowest physical layer, that handles transmission between the two points in the link (for instance, the air interface between the mobile station and the base station antenna), to the highest application layer, that ultimately formats data in a way that can be correctly perceived by the end users. The protocol stack of IS-95 CDMA follows the general OSI model, as illustrated in the diagram below. The specific implementation details of the protocol layers will not concern us here. The diagram below shows the protocol stack for communication between the Base Station Subsystem (BS) and the Mobile Switching Centre (MSC), as an example. In general the protocol used in IS-95 is very similar to that in GSM. Our main concern in this course is on the lowest physical layer, specifically the form of the physical channels and the logical mapping of data onto these physical channels (the logical channels).

8 Physical Layer A feature of IS-95 CDMA is that the spreading process is quite different in the forward and reverse channels. The physical origin of these differences comes from the ability to synchronise code sequences in the two transmission directions. In the downlink, the base station controls the transmission to all mobiles within the cell, and as such it can ensure the synchronisation off all user codes. However, the reverse link features many mobile stations transmitting at random times, and their signals arriving at the base station antenna after random delays. Synchronisation of spreading codes cannot be maintained, as so stronger, longer spreading sequences with low crosscorrelation must be used.

9 The downlink spreading sequences is illustrated in the diagram below. The data is first passed through a rate-½ convolutional encoder and interleaved, and then modulated onto one of the 64 orthogonal Walsh sequences. Next, a synchronised spreading sequence unique to each cell is superimposed on top of the Walsh sequence to reduce interference between cells. The base station specific code requires synchronisation between base stations, which is handled by the network. The spreading procedure for the reverse link is shown below. The data signal is firstly passed through a rate-1/3 convolutional encoder, providing stronger error protection than the forward link, to account for a drop in performance with spreading sequence mismatch and a generally weaker RF link. This is followed by interleaving, modulation onto the 64-bit Walsh code, and finally spreading with a user and base station specific PN sequences. Unlike the Walsh codes these PN sequences are not perfectly orthogonal, but almost orthogonal. However, unlike the Walsh codes this almost orthogonality property is preserved on time shifts of the two sequences. The Walsh codes are used a little differently on the reverse channel. Rather than representing a spreading code sequence as in Direct Sequence CDMA, the Walsh functions here are used to perform 64-ary orthogonal modulation. Each group of 6 data bits are mapped to a corresponding 64-bit Walsh code, and this Walsh code sequence gives the next transmitted 64 bits. The motivation here was to provide stronger noise performance on the reverse link through the employment of this orthogonal modulation scheme. The receiver recovers the 64 bits of the transmitted

10 Walsh code, decides on the Walsh sequence that matches it most closely, and then recovers the original 6 bits. The use of orthogonal Walsh codes in modulation thus facilitates the non-coherent detection on the uplink. There are three basic codes in CDMA. The channelisation codes for the downlink are the 64-ary Walsh codes, formed as rows of the familiar Hadamard matrices. 1 1 H N H N H 2 = and H 2 N = 1 1 H N H N In addition there is 42 bit long code, which is the same sequence for all users except an individual user uses a unique 42 bit sequence to mask this sequence. The 42-bit mask is formed from the electronic serial number(esn), which is 32 bits long and passed to the network on registration, with ten additional bits appended. This long code is often termed the scrambling code. Finally, there is short code, a 15 length maximal sequence PN code, of period bits. The role of the short code is to distinguish between cells. Interestingly, in IS-95 CDMA the same short code sequence is used in every cell, with each cell being distinguished by a unique 512 bit offset. This is exploiting the low autocorrelation properties of PN sequences. A detailed description of the processing for the forward link is shown in the diagram below. The CELP speech codec outputs speech at a data rate of 8.6 kbps, or 172 bits per 20 msec speech frame. To each speech frame 12 cyclic redundancy check (CRC) check bits are added, to detect errors that may occur in transmission, along with 8 tail bits to flush the (2,1,8) convolutional encoder after use. The polynomial to form the 12 bit CRC is, known as the frame quality indicator is: g ( x) = x + x + x + x + x + x + x + 1 The structure of the IS-95 CDMA (2,1,8) convolutional encoder is shown in the diagram below.

11 The output of the convolutional encoder is at a bit rate of 19.2 kbps. Interleaving is used to reduce the impact of burst errors. Block interleaving of depth 32 bits is specified to randomise bit errors resulting from burst errors, common in the mobile channel due to fading events. The convolutional code employed is extremely efficient at correcting random bit errors, but not burst errors. The binary sequence out of the interleaver is then modulo-2 added to the user-defined long code sequence, in effect providing ciphering and security of user data. Then the data sequence is spread by the user assigned Walsh code, by a factor of 128 to the transmitted chip rate of Mcps. Finally, prior to the RF stage, the data sequence is XOR-ed with the cell specific short code sequence. Quadrature Phase Shift Keying (QPSK) is used as the modulation scheme. In general, coherent detection can be employed on the mobile station, thanks to access to the common pilot channel transmitted by the base station. However, on the reverse link the base station must in general implement non-coherent detection, and this is achieved by the use of Walsh codes to form a 64-ary orthogonal modulation scheme. On the reverse link the Q-phase is offset by half a symbol period, so the reverse modulation scheme is really Offset Quadrature Phase Shift Keying (OQPSK).

12 Due to the impossibility of synchronising the transmissions of independent mobile units, the user-specific long code is used to provide channelisation on the up-link rather than the Walsh code, since Walsh codes have poor cross-correlation properties. The output of the speech codec, plus CRC and parity check bits, is passed through a (3,1,8) convolutional encoder, thus providing stronger error robustness than in the forward link. The output data of the convolutional encoder, at 28.8 kbps, is passed through the 64-ary orthogonal modulation scheme, to result in a bit rate of kbps. As previously discussed, the modulator here maps 6 bit patterns to Walsh code sequences, to improve error robustness. The user-defined scrambling sequence is then used to spread the signal to the air data rate of Mcps, and the short code is employed to distinguish between adjacent cells. As also indicated in the diagram below, differential encoding must be employed to facilitate non-coherent detection at the base station.

13 Let s now turn our attention to the characteristics of the control channels. The only control channel in the reverse direction is the access channel, which performs much the same role as the RACH in GSM. These access channels are common (shared by all mobile stations), and as such have common codes. The messages on the access channel are each 88 bits long, and the spreading process is then identical to that used for reverse traffic channels. The long code mask for the access channel is common, however. The 42 bit mask is made up of 16 bits of base station ID, the 9 bit pilot PN sequence offset (Pilot_PN), a 3 bit paging channel number, a five bit access channel number, and an additional 9 bit prescribed sequence. A modified ALOHA access scheme is used on the Access Channel. The mobile makes an access attempt by randomly selecting amongst the available access channels. An access attempt consists of a sequence of access probes, each with increasing power and delayed by a random time from the previous probe. The access attempt finishes upon the receipt of an acknowledgement on the corresponding paging channel. Every access message contains the 34 bit Mobile Information Number (MIN) and the 32 bit ESN. There are a variety of access channel messages, ranging from response to a paging message, location update, and call initialisation.

14 For the forward direction, the pilot channel carries the all zero message, spread with Walsh code 0 (which is itself the all zero sequence). It is thus nothing more than an unmodulated carrier wave spread with the base station cell-specific sequence. Sync channel data is transmitted at a rate of 1.2 kbps and assigned to Walsh code 32. It is spread in the same fashion as forward traffic data. The sync channel carries a single message, consisting of the system time, the 9-bit Pilot Offset Number (Pilot_PN), system and network ids, and the data rate for the paging channels (4.8kbps or 9.6 kbps). Paging channels are used to transmit a variety of system control messages and user specific messages. Some possible messages on the paging channel are: a page to a mobile unit on the event of an incoming call; a response to a access attempt by a mobile station; a channel assignment message, to allocate a dedicated traffic channel to a user; a system parameters message, that contains the base station id, network and system ids, Pilot_PN, and the number of paging channels available; an access parameters message, containing power information for the mobile, the number of access channels available for this paging channel, Pilot_PN, information about the access procedure (number of probes and timeout procedure), access message format; a Neighbour list message; a CDMA channel list message; and others. Some of these messages are sent periodically, some only when needed. Paging channel messages can be classified as either broadcast messages, call management messages, authentication messages, and operations and maintenance messages.

15 Let s now turn out attention to the format of these various messages in cdmaone. On the reverse traffic channel each transmitted frame is 172 bits in length. The first four bits of the frame indicate whether this particular frame carries only user traffic, in which case the first bit is 0 and the following 171 bits are data, (the output of the speech codec), or some combination of data traffic and signalling traffic. Thus, IS-95 CDMA multiplexes control data with traffic data during a call. A mixed frame is indicated with the first bit 1, and the following three bits then indicate the distribution of data traffic and signalling data in the remaining 168 bits. Signalling messages can be anything from 16 to 2016 bits in length, and as such can be distributed over multiple frames. To each signalling message an 8-bit message length header is attached, and a 16 bit CRC added at the end of the data. The diagram below illustrates a 296 bit length signalling message, that is multiplexed with traffic data in a series of mixed mode frames. The mixed mode indicated here (1000) corresponds to there being 80 user data bits per frame, and 88 signalling bits per frame. Examples of reverse traffic signalling messages are: Pilot Strength Measurement message, a 6 bit indication of the received pilot power; a Power Measurement Report,

16 a set of power measurements of each pilot channel in the active set, and as such used to decide on handoffs; an Origination Continuation message, containing any dialled digits not included in the original message sent over the access channel; a Status message, and a variety of Order Messages (to release, etc.). Messages on the access channel, such as an origination message, paging response, or registration message, can vary between 2 and 842 bits in length. Each message is appended with the mobile id (MIN) and ESN, and fitted into the access channel message format shown in the diagram below. The access messages are all padded to be an integer multiple of 88 bits in length, so that it can be broken into 88 bit frames for transmission. The first 8 bits of the access message carry the message length, and 30 bits CRC is added at the end to detect bit errors.

17 Each access channel transmission is preceded by a preamble, an assigned number of frames containing all zeros, to allows the base station to synchronise to the access channel burst. The diagram below shows a possible access channel structure, with three frames of preamble followed by four access frames, as an example. Forward traffic channel messages are also transmitted in 172 bit length frames, and like the reverse link, can carry a combination of user data and signalling information. Examples of messages on the forward traffic channel include varieties of Order messages, Authentication challenge response message, a handoff direction message, a neighbour list update message, and a power control parameters message. The sync channel message, 162 bits in length, is continuously transmitted over the sync channel. The message contains, among other things, a 15 bit system id, a 16 bit network id, a 9 bit Pilot PN sequence offset unique to each base station, a 36 bit system time parameter, and an indication of the paging channel bit rate. The diagram below then illustrates the formatting of the sync channel message, ultimately transmitted over three 80 ms frames, with 96 bits on each frame. This gives the sync channel the corresponding bit rate of 1.2 kbps. Finally we have paging channel formats. Paging channel messages are transmitted as 384 bit or 768 bit groups every 80 ms, depending on whether the data rate is 4.8 kbps or 9.6 kbps. The structure of a paging channel message is illustrated below.

18 Signal Processing A three-finger RAKE receiver is specified to provide diversity and to compensate for multi-path induced ISI. Given the transmitted chip rate of Mcps, the cdmaone system should be able to resolve multipath components spaced by approximately 160ns. In terms of path length, this is a distance of the order of 50 metres. It is thus very easy for the cdmaone system to resolve these independent arriving multipath components. The spreading sequences used have very sharp autocorrelation properties, meaning that a multipath signal arriving as little as one chip period late will appear as noise at the output of the receiver. These individual multipath components can be resolved by shift the sequence offset chip by chip. The performance of the RAKE receiver depends on the number of fingers, the combining technique employed, and the accuracy of estimating the delay and path loss of the channel components arriving at the receiver.

19 The IS-95 standard specifies a three finger RAKE receiver be employed, though combining and estimating techniques are at the discretion of the equipment manufacturer. Students will be familiar from earlier lectures that Maximum Ratio Combining (MRC) is the theoretically optimum combining technique. As discussed in an earlier lecture, IS-95 CDMA employs the very efficient Code Excited Linear Predictive Coding (CELP) for its speech codec. As mentioned above IS-95 CDMA uses a variable rate codec, but here we ll only describe the full rate operation. As in the GSM speech codec, the CELP encoder in IS-95 takes 20 ms speech frames and breaks them into 5 ms subframes for analysis. The long term pitch prediction filter is updated every subframe, and is quantised with 7 bits, while the long term gain takes up 3 bits per subframe. Thus, the long term predictor requires 40 bits per 20 ms speech frame. The LPC filter is updated once per frame, and quantised with a total of 40 bits. IS-95 uses a single stochastic excitation codebook, with 128 entires. The optimal excitation is found using the minimal mean square error criterion eight times per frame. The excitation is represented with 7 bits, and the gain with 3 bits. Thus, the IS- 95 CDMA codec produces 160 bits per 20 ms speech frame. At this layer, 12 additional parity bits are appended to bring the output to 172 bits per speech frame, as quoted earlier.

20 A particularly attractive feature of CDMA systems is the employment of voice activity detection, VAD. VAD systems are able to detect the presence of human speech, and can reduce the output data rate when speech is not present. When data rates are reduced the output data blocks are repeated. For example, an eight rate speech block is repeated eight times. Since it will be received eight times independently at the receiver, the overall bit energy is increased by a factor of eight. This allows the lower data rate speech to be transmitted at a reduced power level, reducing interference and thereby increasing network capacity. In particular, the 2.4 kbps traffic channel can be used to carry background noise information (the equivalent of the silence descriptor in GSM). The 4.8 kbps is used to mix digitised speech and signalling data. The evolution of the IS-95 network to 3G saw multiple spreading codes assigned to a single user. A max of 8 could be assigned, achieving a data rate of kbps, though in practise only about 64 kbps was achieved. This standard became known as IS-95b.