Principles of Digital Mobile Communication Systems - The GSM System

Transcription

1 Principles of Digital Mobile Communication Systems - The GSM System Contents: by Petri Jarske Principles of Cellular Mobile Communications Systems 2 The mobile radio environment 8 The GSM System 20 Basic Architecture 20 Architecture Evolution 24 Transmission inside GSM 31 The Radio Interface 38 Principles of Signalling 61 Radio Resource Management 70 Mobility & Security Management 82 Communication Management 89 Network Management 93 Evolution of the GSM System 101 CDMA Systems Intro 107

2 Principles of Cellular Communications Systems A typical mobile communication environment can be described with the following assumptions: The communication network operators and service providers want to provide mobile communications services to a large number of customers. The customers are distributed over a (possibly) large geographical area. The customers want to be able to access the services while moving around in the service area (the degree of mobility may vary, depending on the system, from 0 to 250 km/h). The operators can use certain limited band of radio frequencies for the wireless part of the communication. Now, the problem in the exponentially growing markets of mobile communications is: How can the capacity of the communications system be increased (and increased, and increased, and )? In this context capacity = number of customers receiving services with satisfactory quality of service 2

3 Straightforward solution: The same frequencies used over the whole geographical area. For increasing capacity, use more efficient source coding and modulation (compression, efficient modulation, TDMA, CDMA, etc.) No need to worry about the location of the customer. Cellular solution: Divide the geographical area into small subareas (cells), and assign each cell enough frequency resource to serve the customers in this area (see figure below, real cells are not that regular in shape and size). The same frequency resource may be used in many cells provided that they are separated by enough distance. Capacity can be increased simply by making the cells smaller and smaller... Are there other advantages, in addition to the higher capacity? The cellular solution also introduces new problems. What could these be? Think before turning the page. 3

4 Other advantages: Well, at least lower transmitter powers > less interference to other users of the same system, less interference to other systems, also longer battery life Flexible coverage: small cells for densely populated areas, large cells for rural areas. Problems to be considered: In order to provide service anytime, the network has to know the location of each customer to some accuracy (location management) at least when a call is coming to the mobile unit. To provide continuous service even when the customer is moving, handover procedures are needed. That is, the service connection is passed to a new base station every time the quality of the existing connection gets too low. There are two extreme alternatives to handle the previous problems. (1) The location of the customer is not known prior to the call but a paging message is sent to the whole network when a call arrives, or (2) the location of the customer is kept in a central database with the precision of one cell. Many of the existing solutions, such as GSM, are something between these extremes because the signalling load can be minimized that way. Also, the network becomes more complicated and expensive when smaller cells are introduced. The cost issues are not much emphasised in this text but they are very important for the operator, and also to the customer. 4

5 Some further tricks to improve system efficiency in general (not only capacity): transmitter power control: When the transmitted power level of each transmitter is kept to the minimum required for satisfactory quality, the interference caused to other cells sharing the same resource is also minimized. This way, the cells sharing the same radio resource can be built closer to each other, and capacity increases (compared to the noncontrolled case). frequency hopping: The interference degrading the transmission quality is not equally distributed to all radio channels. By changing the channel frequency periodically for each user, the quality can be made approximately equal for everybody. This way many users can be served with satisfactory quality, rather than serving a few customers with good quality, and leaving some without service. discontinuous transmission: In speech communication, the active speech covers only about 25 40% of time. During the silent periods, it is sufficient to transmit only very little information, for example, one or two frames per second. This again reduces the average transmitted power, and reduces the interference caused to other users. The cost, however, is increased complexity because voice activity detection (VAD) has to be implemented. mobile assisted handover: This reduces the network complexity by giving the responsibility of monitoring the signals from neighboring cells to the mobile terminal. These measurements are needed for the handover decision, and would, otherwise, require constant message exchange between neighboring cells. 5

6 Overview of mobile services Service provision of a particular user depends on: contents of the subscription held by the user capabilities of the serving network capabilities of the user equipment Examples Services in GSM: Speech probably the most important also in the future Circuit switched data currently up to 38,4 kbits/s commercially available, or more Packet data available Short messages point-to-point & broadcast Supplementary services call forwarding, barring, etc. Multimedia messages Services in IEEE wireless LANs: Packet data, up to 11 Mbits/s in 2.4GHz ISM frequencies, 54 Mbits/s available Packet data, up to 54 Mbits/s in 5-6 GHz frequencies available IP based core network, operator not necessary No special speech channel, voice over IP (VoIP) 6

7 Presumed evolution of Wireless Cellular Communication: ETSI: GSM evolution to WCDMA (UMTS) ANSI: US-CDMA evolution to cdma2000 3GPP 3G ARIB: selection of 3G technology for Japan etc. ITU-T: IMT GPP: 3 rd Generation Partnership Project is a co-operation project between the standardisation bodies mentioned above. Global 3G did not happen. There is also 3GPP2 for U.S. and some other areas. 7

8 The mobile radio environment General description Radio propagation mechanisms are strongly affected by the wavelengths used, and the environment (natural or human-made). Buildings are wave scatterers. The sizes of buildings are typically many wavelengths of the used frequency, creating reflected waves at that frequency. Typically, the antenna height of a mobile unit is much lower than the average height of houses. Given the conditions above, and the propagation frequency clearly above 30 MHz, the environment forms a multipath propagation medium. The base-to-mobile link is usually less than 25 km, so the radio horizon need not be considered. Actually, earth's curvature reduces interference from distant sources. For large cell designs (radius km) the height of the base station antenna is usually m. The height of a mobile unit antenna is about 2 3 m. The base station antenna is usually clear of its surroundings, whereas the mobile-unit antenna is embedded in them. Base station antenna 8

9 From this description of the environment, we might imagine that the mobile site will receive many reflected waves and (possibly) one direct wave. We can assume that the reflected waves received at the mobile site come from different angles equally distributed throughout 360. If the direct signal is strong compared to reflected signals, the received signal level can be described with Rician statistical model. If the direct signal is weak (or non-existent), the received signal level can be described with Rayleigh statistical model. Path loss and fading In free space, signal attenuates 6 db / octave (of distance). That is, if the distance from the transmitter is doubled, the free space path loss will be 6 db more. The signal strength r(x) or r(t) can be, for modelling purposes, separated into two parts called long-term fading m(t) or m(x), and short-term fading r 0 (t) or r 0 (x) as or r(t) = m(t) r 0 (t) r(x) = m(x) r 0 (x) The long term fading is the envelope of the fading signal, or local mean. x+ L x+ L 1 1 mˆ ( x) = r( ξ ) dξ = m( ξ ) r0 ( ξ ) dξ 2L 2L x L x L 9

10 When the length L is properly chosen, this becomes x+ L 1 mˆ ( x) = m( x) r0 (ξ ) dξ 2L x L The long-term signal fading m(x) is mainly caused by terrain configuration and the built environment between the base station and the mobile unit. Terrain configurations can be classified, for example, as Open area Flat terrain Hilly terrain Mountain area and the human made environment as Rural area Suburban area Urban area Short-term fading is mainly caused by multipath reflections of a transmitted wave by local scatterers such as buildings or natural obstacles. 10

11 Classification of channels In a dispersive medium, there are two kinds of spread: Doppler spread F and multipath spread δ. Doppler spread F is spreading in frequency, and multipath spread δ is spreading in time. In a strict sense, all media are dispersive. We can classify a medium's characteristics based on the signal duration T and the signal bandwidth W. Nondispersive channels A nondispersive but fading channel is created if F 1 1 << and δ << T W In many practical systems, the values of T and W can be chosen so that the channel can be considered nondispersive. Time-dispersive channels 1 >> T and δ >> W δ but F << 1 T 11

12 Frequency-dispersive channels but F >> W δ << 1 W and F >> 1 T Guess what is doubly-dispersive channel. 12

13 Delay spread The mean delay time T d of a channel can be calculated as T d = 0 t e( t) dt and the delay spread as 2 = 0 t 2 e 2 ( t) dt Td where e(t) is the impulse response of the channel. Typical values for the delay spread are: Type of environment Delay spread In-building < 0.1 µs Open area < 0.2 µs Suburban area 0.5 µs Urban area 3 µs 13

14 Prediction of propagation loss As we saw earlier, the local mean (long-term fading) of the received signal level can be obtained by averaging a suitable spatial length over a piece of raw data. The choice of suitable L is essential for obtaining a good estimate of the local mean. In practice, L in the range 20λ 40λ is acceptable samples in an interval of 40 wavelengths is adequate for obtaining the local means. The measurements are usually recorded while the mobile units are travelling along a road (street). The recorded signals from the mobile paths have to be converted to radio path. 14

15 Models for path loss Note: path loss model is only for path loss prediction and not for multipath fading. Assume that the characteristics of a rough earth surface are random in nature and that the radius of curvature of the surface irregularities is large compared to the wavelength of the incident wave. Then the received signal can be represented by a scattered field E s which can be approximated by combining the direct wave and the reflected wave. E s = (1 + a v e j ψ )E The reflection coefficient is a v and ψ is the phase difference between the direct and reflected wave. The phase difference can be expressed as 2π ψ = β d = d λ where β is the wave number and d is the difference between the two radio path lengths. E is the direct wave received at the mobile antenna. 15

16 According to the free-space propagation path loss, the received power from a direct wave is P 0r = λ 4 Pt π d 2 In the mobile radio environment, the incident angle is usually small, and therefore, the reflection coefficient is approximately a v = 1 and ψ<<1. The received power of the scattered field becomes P r λ = P 4π t d 2 1 cos ψ For d >> h 1 +h 2 we can approximate j sin ψ 2 λ P 4π t d 2 ( ψ ) 2 This gives P 4π h1h λd ψ 2 r P t h h d This is an imperfect formula since it does not involve wavelength. It indicates two correct facts the equation shows a path loss of 40 db/dec which has been verified from the experimental data to be roughly true the equation shows a 6 db/oct rule for an antenna height gain at the base station, i.e. doubling the antenna height at the base gains 6 db which also seems to be roughly true within certain limits but there are two weak points, too the wavelength term is missing but the measured data show that the path loss is a function of frequency the equation shows a 6 db/oct rule also for an antenna height gain at the mobile unit which is not true in practise 16

17 An area-to-area path loss prediction model An area-to-area prediction is sometimes used to predict path loss over a general flat terrain without knowing the particular terrain configuration. The area-to-area path loss prediction requires two parameters: (1) the power at the reference (1-mile) point of interception P r0 and (2) a path loss slope γ. The field strength of the received signal P r can be expressed as or in db P r P r = P r γ n r0 α 0 r 0 f 0 f r f = P 0 γ log log r n + α 0 r0 f0 where r is in miles or kilometers and r 0 equals 1 mile or 1.6 km. γ is expressed as γth power in the linear formula, and γ db/dec in the db formula. α 0 is an adjustment factor. This is a general formula that can be used for different frequency ranges above 30 MHz. The assumed default conditions are: frequency f 0 = 900 MHz base station antenna heigth = m (100 ft) base station power at the antenna = 10 watts base station antenna gain = 6 db above dipole gain mobile unit antenna height = 3 m (10 ft) mobile unit antenna gain = 0 db above dipole gain 17

18 The adjustment factor is used for different conditions as follows: α 0 = α 1 α 2 α 3 α 4 α 5 (or α 0 = α 1 +α 2 +α 3 +α 4 +α 5 in db) where α 1 new base station antenna height (m) = 30.48m new mobile station antenna height (m) α 2 = 3m α = new transmitter power 3 10W new base station antenna gain with respect to λ/2 dipole α 4 = 4 α 5 = antenna gain correction factor at the mobile unit The parameters γ and P r0 are found from empirical data: Terrain P r0 (mw) P r0 (dbm) γ γ (db/dec) free space open area suburban Philadelphia Newark Tokyo The values of n and v is also found from empirical data. In suburban or open area with frequencies < 450 MHz n=20 db/dec. In urban areas with >450 MHz frequencies n=30 db/dec is recommended. 2 for new mobile unit antenna height > 3m ν = 1for new mobile unit antenna height < 3m 2 ν 18

19 The model of Okumura et al. (From M. Hata, "Empirical Formula for Propagation Loss in Land Mobile Radio Services", IEEE Trans. Vehicular Tech., VT-29, No. 3, August 1980.) The standard formula for propagation loss is L p (db) = log f c log h b a(h m ) + ( log h b ) log R where f c is the used frequency MHz, h b is the base station antenna height m, R is distance 1 20 km, h m is the mobile antenna height, and a(h m ) is a correction factor for h m given by a( h m (1.1log fc 0.7) hm (1.56log fc 2 ) = 8.29(log1.54hm ) (log11.75hm ) ) for medium -small city for large city, for large city, f f c c 400MHz 400MHz In suburban areas the loss is L ps = L p {urban area} 2 (log (f c /28)) and in open areas L po = L p {urban area} 4.78 (log f c ) log f c Street orientation channel effect The signal strength received from a street in line with the base station is about 10 db higher than the signal from a street perpendicular to the base. This phenomenon diminishes at about 8 km distance. Note, that the previous description gave only examples of how radio path loss is modelled in mobile communication systems. It is not a complete list, and the exact numbers are not relevant for this course. 19

20 The GSM System In the following text, we will concentrate mainly on the GSM system. Basic Architecture The GSM system, as originally specified in 1991, has a hierarcical architecture, typical for 2 nd generation cellular systems: BTS BTS BSC TRAU HLR AC EIR OSS MSC VLR PSTN ISDN BTS BSC TRAU SMSC VMS BTS BTS = base transceiver station BSC = base station controller TRAU = transcoder & rate adapter unit MSC = mobile (services) switching centre VLR = visitor location register AC = authentication centre HLR = home location register OSS = operation sub-system including network management (NMS) SMSC = short message service center VMS = voice message system EIR = equipment identity register 20

21 GSM Network Elements: The Mobile Station (MS) MS = ME + SIM Mobile Equipment (ME): generic radio and processing functions to access the network, human interface and/or interface to other terminal equipment. Subscriber Identity Module (SIM): a smart card containing all the subscriber related information, confidentiality related information. Something to think about: What advantages follow from making the ME and SIM separate entities? The Base Station Subsystem (BSS) BSS = BSC + BTS + TRAU Base Station Controller (BSC) is in charge of the radio interface management, allocation and release of radio channels, handover management (up to some tens of BTS s). Base Transceiver Station (BTS): radio transmission and reception from antennas to the radio interface specific signal processing, handling 1 10 radio carriers at a time. Transcoder & Rate Adapter Unit (TRAU): GSM-specific speech encoding and decoding, bit rate adaptation. 21

22 The Network & Switching Subsystem (NSS) NSS = MSC + VLR + HLR + AC + EIR Mobile services Switching Center (MSC): performs the basic switching function, coordinates the set-up of calls to and from GSM users, manages communications between GSM and other telecommunications networks. Visitor Location Registers (VLR): database storing temporarily subscription data for those subscribers currently located in the service area of the corresponding MSC, holds data of their current location area. Home Location Register (HLR): database holding subscriber information relevant to the provision of telecommunications services, some information related to the current location of the subscriber (mainly under which MSC/VLR the user can be found). Authentication Centre (AC): database maintaining security related information of the subscriptions. Equipment Identity Register (EIR): database maintaining security related information of the mobile equipment (separate from subscriptions). 22

23 The Operation Sub-System (OSS) Operation Sub-System (OSS): (1) network operation enabling the operator to observe system load, blocking rates, handovers, etc. and providing means to modify network configuration, (2) equipment maintenance aiming at detecting, locating and correcting faults, (3) subscription management for registering new subscriptions, modifying and removing subscriptions, as well as billing information. Tasks (1) & (2) are major part of the Network Management System (NMS). Task (3) is more service management, not directly related to network status. Value Add Services The services offered in the basic GSM network are similar to those available in a sophisticated PSTN network. Mobility is the main feature differentiating the basic GSM system from fixed telephony systems. On top of this, the first services adding value to the GSM network, have been Short Message Services (SMS) and Voice Messaging System (VMS). Especially, the success of SMS has been surprisingly good. More value add services have been, and will be built as the capabilities of the GSM network improve over time. Intelligent Network (IN) features are added to the GSM networks, in order to enable tailored services to different customer groups, or individual subscribers. 23

24 Architecture Evolution The GSM specification is evolving constantly. Some major development lines are reviewed here, from architecture point of view. The functionality of these new features will be discussed more, later on. Step 1: Higher data rates The basic GSM offers circuit switched data transfer services with rates up to 9.6 kbits/s which is not sufficient for many services. Higher data rates are possible by changing channel coding, and using several physical channels (time slots) for a high rate connection. High Speed Circuit Switched Data (HSCSD) is an implementation of this concept in GSM. HSCSD will be discussed in more detail later. For the GSM architecture, as presented on page 20, HSCSD does not introduce any visible changes in the block diagram (so let's not redraw it here). It does, however, require HW and SW changes in most of the network elements shown in the architecture block diagram.... BTS BSC/TRAU MSC... HSCSD principle I W F PSTN ISDN PDN 24

25 Step 2: Packet data Wired data networks have typically used packet data transfer. In order to connect smoothly to these networks, and to use the radio resource more efficiently, General Packet Radio Service (GPRS) has been specified to GSM. From architecture point of view, in addition to HW & SW changes in the existing network elements, GPRS also introduces new network elements called Serving GPRS Support Node (SGSN) and Gateway GPRS Support Node (GGSN). BTS BTS BSC TRAU HLR AC EIR OSS MSC VLR PSTN ISDN BTS BSC TRAU SMSC VMS BTS SGSN IP GGSN IP IP networks 25

26 Step 3: Higher data rates (again) When we want to go beyond HSCSD in data rates with minimal changes in the frame structure and protocols, it is necessary to change the modulation used in the physical layer to represent the transmitted bit stream. In the GSM case, this is done by packing 3 bits per symbol on the physical layer, instead of 1 bit per symbol of basic GSM. This is called Enhanced Data rates in GSM Environment (EDGE). For the architecture presented on previous page, EDGE does not introduce any visible changes in the block diagram (so let's not redraw it here). It does, however, require HW and SW changes in most of the basic network elements shown in the architecture block diagram. Step 4: Completely new radio interface For the 3 rd generation (3G) cellular networks, the core of the network, at least in Europe and Japan, will be based on GSM. The air interface, however, will be based on CDMA technology which is completely different from basic and enhanced GSM. For controlling the CDMA radio network, similar network elements are needed, as in GSM but different terminology is used in order to draw distinction between 2G and 3G systems. Instead of BTS, we have Base Station (BS), and instead of BSC, we have Radio Network Controller (RNC), in the 3G network. Of course, the detailed functionality of these elements is also different from GSM. 26

27 BTS BTS BSC TRAU HLR AC EIR MSC VLR PSTN ISDN BTS BSC TRAU OSS SMSC VMS BTS SGSN IP GGSN IP IP networks BS RNC IWU BS The 2G and 3G radio interfaces will co-exist for a long period of time. Also other radio interfaces, such as DECT or wireless LANs, may utilise the same core network. 27

28 Step 5: All-IP BTS BTS BSC TRAU HLR AC EIR BS RNC IP NETWORK IWU MSC VLR SMSC VMS PSTN ISDN BS SGSN GGSN IP networks It is expected that eventually GSM, and 3G networks will evolve into all-ip architecture. Majority of the traffic will use packet transfer. IP will support mobility management, and quality of service (QoS) features. 28

29 In the following, we will concentrate on the basic GSM functionality, and will revisit HSCSD, GPRS, and EDGE later on. Architecture & GSM Functional Planes Functionally, the GSM system can be divided into five planes: Transmission: provides the means to carry user information (speech or data) on all segments along the communication path, and to carry signalling messages between entities. Radio Resource Management (RR): establishes and releases stable connections between mobile stations and an MSC, and maintain them despite user movements. The RR functions are mainly performed by the MS and the BSC. 29

30 Mobility Management (MM): functions are handled by the MS (or SIM actually), the HLR/AuC, and the MSC/VLR. These include also management of security functions. Communication Management (CM): is setting up calls between users, maintaining and releasing them. In addition to call control, it includes supplementary services management, and short message management. Operation, Administration & Maintenance (OAM): enables the operator to monitor and control the system. The following figure tries to illustrate the relationship between the network elements and functional domains: GMSC = Gateway MSC, a switching centre which is able to find the corresponding HLR based on the called number. GMSC and MSC/HLR may be physically one unit. 30

31 Transmission inside GSM On the network side, the GSM system is designed to be compatible with ISDN where the transmission rates are multiples of 64 kbit/s. On the air interface, however, the net bit rate per channel is less than 16 kbit/s. For adapting the different rates, the Transcoder / Rate Adaptor Unit (TRAU) has been introduced. For speech, TRAU includes the speech codecs. TRAU belongs functionally to the BTS but its actual location is not strictly specified. Transmission of speech and data is next briefly described in the radio interface BTS - TRAU interface interface between TRAU and point of interconnect with other networks (IWF) 31

32 Speech on the radio interface Speech processing for transmission over the air interface includes the following functions: Speech coding Error protection (codec specific) Error detection (CRC) Bad Frame Handling (substitution) Voice Activity Detection / Discontinuous Transmission (VAD/DTX) Manufacturer specific audio features - noise cancelling - spectrum equalization - echo cancellation For spectrum efficiency, as low bit rate as possible on the radio path (but with acceptable quality, of course), is required. Speech coding takes care of this. In the first phase of GSM spec, a full rate speech channel was defined, with provision of half rate in the second phase. Why? Today, the GSM standard includes the following codecs: Full rate (FR), 13 kbit/s RPE-LTP Half rate (HR), 5.65 kbit/s VSELP Enhanced full rate (EFR), 12.2 kbit/s ACELP Adaptive Multi Rate (AMR), ACELP 12.2, 10.2, 7.95, 7.4, 6.7, 5.9, 5.15, 4.75 kbit/s AMR wideband codec 32

33 As an example, we can take a brief look at the original full rate codec. The full rate speech encoder, compressing 64 kbit/s > 13 kbit/s, is a so called RPE-LTP (regular pulse excitation - long term prediction) encoder. Speech is encoded in blocks of 20 ms, that is, 160 samples having 8 bits each (in A-law representation) are encoded into 260 bits as illustrated in the figure below. Since this is not a speech processing course, we will not go into details of the speech codec. Decoder basically the previous stuff in reverse order: Up-sampling LTP filter LPC filter de-emphasis filter 33

34 The transmitted parameters, after speech encoding, are NOT equal in importance. Therefore, they are divided into 3 classes of importance, each protected against transmission errors in a different manner. This will be described later. Discontinuous transmission When the user is speaking, speech is encoded at the normal rate 13 kbit/s (260 bits / 20 ms). Otherwise a bit rate around 500 bit/s (260 bits / 480 ms) is used which is sufficient to encode the background noise. The background noise is regenerated to the listener. Why? Discontinuous transmission (DTX) requires voice activity detection (VAD). What are the advantages of discontinuous transmission? 34

35 Speech on the BTS-TRAU interface If TRAU is physically distant from BTS, the 13 kbit/s stream is carried to the TRAU over standard digital links making use of 16 kbit/s circuits. The 20 ms frame synchronization cannot be derived from the 13 kbit/s flow. Therefore, some auxiliary information is added. This also includes information for speech/data, full/half rate and bad frame indication. Total 316 bits / 20 ms. Speech on the TRAU-IWF interface On a 64 kbit/s link, the standard G.711 speech transmission is used with A-law coding. 35

36 Data in the basic GSM system Several connection types are provided. Why? Basic division is into T and NT modes. In T (or transparent) mode, the error correction is entirely done by a forward error correction (FEC) mechanism. In NT (or non-transparent) mode, an additional scheme is used where information is repeated when it has not been correctly received by the other end. The T mode connection types of the basic GSM are summarized in the following table: User rate Intermediate rate Channel type Residual error rate 9600 bit/s 12 kbit/s full rate 0.3 % 4800 bit/s 6 kbit/s full rate half rate 0.01 % 0.3 % 2400 bit/s or less 3.6 kbit/s full rate half rate % 0.01 % (residual error rates for typical urban conditions with frequency hopping) For NT mode, the Radio Link Protocol (RLP) is added which is basically a link protocol of repetition-when-needed type. 36

37 The following table summarises all basic GSM data connection types: Type QoS two-way delay TCH/F9.6 T low 330 ms TCH/F9.6 NT high > 330 ms TCH/F4.8 T medium 330 ms TCH/F2.4 T medium 200 ms TCH/H4.8 T low 600 ms TCH/H4.8 NT high > 600 ms TCH/H2.4 T medium 600 ms (don t take the quality estimation too literally - note differences) The T mode of transmission is derived from the ISDN specifications (but we will not discuss these much in this course). In the NT approach, the transmission is considered as a packet data flow (although the offered service, end-to-end, is a circuit service). 37

38 The Radio Interface The radio interface, in addition to the fact that the users move, is the source of many difficult problems that need to be solved in the GSM system (just as in any mobile communication system). The radio interface needs to be specified in very detail, in order to achieve full compatibility mobile stations and networks of different manufacturers. Spectral efficiency of a cellular system is one of the key economic factors. The multiple access scheme used in GSM is a combination of TDMA and FDMA. FDMA is mainly used to share spectrum between neigboring cells. The basic time division is into 8 time slots but the actual time division scheme is more complicated, as we will soon see. Logical channels The basic division between logical channels is: Traffic channels / Control channels. 38

39 The main task of the communication system is to transport user information. For the speech and different types of data communications, the radio interface accommodates bidirectional connections. For these purposes, Traffic CHannels (TCH) are assigned to the user. Full rate traffic channels may be denoted by TCH/F, and half rate channels by TCH/H. All the other logical channel types can be regarded as control or signalling channels. One exception to the previous statements is the transfer of point-to-point short messages, which is implemented in a similar way as signalling. When a mobile station is connected to the network (whether or not there is a user communication in progress), signalling messages are exchanged between the mobile station and other network elements. For signalling in connection with a call, two possibilities are offered: Each assigned traffic channel comes with an associated low rate signalling channel called Slow Associated Control CHannel (SACCH). This bi-directional channel is capable of carrying about 2 messages / second, with a transmission delay of about 0,5 second. 39

40 The other alternative is (surprise!) Fast Associated Control CHannel (FACCH) which is actually not a separate logical channel but uses the traffic channel (TCH) by replacing a user data frame with a signalling frame when necessary. A signalling frame is marked with one bit called stealing flag. Signalling connection is often necessary also when there is no call in progress (supplementary services management, short messages, location updating, etc.) For this purpose, a Standalone Dedicated Control CHannel (SDCCH) is set up. Sometimes, this is also referred to as TCH/8 since its characteristics are very close to the traffic channel but it uses only 1/8 of the capacity of a full rate traffic channel. TCH/8 also has an SACCH associated with it. For spectrum efficiency, traffic channels are allocated to users only when needed (in PSTN you always have the connection to the network). Therefore, we can distinguish between dedicated mode and idle mode of the mobile system. A mobile station is in dedicated mode when it has a TCH assigned to it. In idle mode (but power on), the mobile station is far from idle. It must continuously listen to one base station, and also monitor up to 6 other base stations. 40

41 Before a mobile station can communicate with a base station, it must become and stay synchronised with it. For this purpose, two logical channels are broadcast from each base station: the Frequency Correction CHannel (FCCH), and the Synchronisation CHannel (SCH). General information concerning each cell (identity, which network it belongs to, which frequencies are used, etc.) is broadcast regularly on the Broadcast Control CHannel (BCCH). After the mobile station has synchronised itself with the base, it can access the network through the Random Access CHannel (RACH). Paging messages are sent on the Paging CHannel (PCH) and messages indicating the allocated channel on Access Grant CHannel (AGCH). Because these are similar and never used simultaneously, they can be treated together as PAGCH. Cell broadcast short messages are broadcast on the Cell Broadcast CHannel (CBCH). This requires about 80 bytes every 2 seconds. The common channels FCCH, SCH, BCCH, PAGCH as well as the CBCH are downlink (from base to mobile) only. The RACH is uplink (from mobile to base) only. The other channels, called dedicated channels (TCHs ans SACCHs), are bi-directional. 41

42 The multiple access scheme The radio interface of GSM uses a combination of Frequency Division Multiple Access (FDMA) and Time Division Multiple Access (TDMA) with slow frequency hopping. The basic unit of transmission on the radio path is a sequence of about 156 modulated bits called burst. They are sent in time and frequency windows called slots. The center frequencies of the slots are placed 200 khz apart within the frequency band reserved for GSM, and the duration of one slot is 15/26 ms ms. All slots in a cell are aligned in time. This is illustrated in the following figure. 42

43 The time axis is divided into 8 distinct slots, numbered 0 7. The information of certain logical channels is mapped to certain time slot number. For example, if the shadowed burst in the previous figure belongs to certain logical channel, the next time we can find information belonging to the same logical channel is at least 8 15/26 ms later. The frame structure is as follows: 8 consecutive time slots form a TDMA Frame. 26 or 51 TDMA Frames form a Multiframe. 51 or 26 Multiframes form a Superframe Superframes form a Hyperframe. The length of a hyperframe is 3 hours 28 minutes 53,76 sec. At the base station, the transmitted and received bursts are synchronized such that the received burst arrives 3 15/26 ms after the burst with the same time slot number is transmitted. Transmission Reception So, this is the base station viewpoint. The figure looks similar for a mobile station very close to the base station. The purpose of this arrangement is to avoid simultaneous transmission and reception in the mobile station. For a mobile station several kilometers away from the base station, the propagation delays have to be considered (30 km distance => 200 µs round trip delay). This is compensated in the mobile station by transmitting the bursts earlier. The timing is adjusted with the timing advance parameter. 43

44 This timing arrangement also has an impact on the future development of the GSM system. Think about, for example, increasing the user data rates without changing the air interface totally. In the following description, each rectangle denotes one slot with certain time slot number. The slots with other time slot numbers are not shown. So, adjacent slots in the figures are separated by 8 15/26 ms. The following figures try to show how different logical channels are grouped on respective time sequences: TCH/F + SACCH T T T T T T T T T T T T S T T T T T T T T T T T T TCH/H + SACCH T T T T T T T T T T T T S T T T T T T T T T T T T S Here the slots denoted with bold and italic characters belong to two different logical channels. 44

45 TCH/8 + SACCH (8 channels grouped) T 1 T 1 T 1 T 1 T 2 T 2 T 2 T 2 T 3 T 3 T 3 T 3 T 4 T 4 T 4 T 4 T 5 T 5 T 5 T 5 T 6 T 6 T 6 T 6 T 7 T T 7 T 7 T 8 T 8 T 8 T 8 S 1 S 1 S 1 S 1 S 2 S 2 S 2 S 2 S 3 S 3 S 3 S 3 S 4 S 4 S 4 S T 1 T 1 T 1 T 1 T 2 T 2 T 2 T 2 T 3 T 3 T 3 T 3 T 4 T 4 T 4 T 4 T 5 T 5 T 5 T 5 T 6 T 6 T 6 T 6 T T 7 T 7 T 7 T 8 T 8 T 8 T 8 S 5 S 5 S 5 S 5 S 6 S 6 S 6 S 6 S 7 S 7 S 7 S 7 S 8 S 8 S 8 S 8 T TCH/8 + SACCH (4 channels grouped, with common ch.) T 1 T 1 T 1 T T 2 T 2 T 2 T 2 T 3 T 3 T 3 T 3 T 4 T 4 T 4 T 4 S 1 S 1 S 1 S 1 S 2 S 2 S 2 S T 1 T 1 T T 1 T 2 T 2 T 2 T 2 T 3 T 3 T 3 T 3 T 4 T 4 T 4 T 4 S 3 S 3 S 3 S 3 S 4 S 4 S 4 S

46 The empty slots in the previous figure are used for common channels. FCCH + SCH F S F S F S F S F S BCCH + PAGCH/3 B B B B P P P P P P P P P P P P OK, maybe it is not necessary to show all possible channel combinations. 46

47 Some examples of possible cell configurations follow (here TN = timeslot number). A small capacity cell with a single transmitter/receiver: TN0: FCCH, SCH, BCCH, PAGCH/3, RACH/H; 4x(TCH/8+SACCH) TN1 7: one TCH/F+SACCH each. A medium capacity cell with 4 transmitters/receivers: one TN0: FCCH, SCH, BCCH, PAGCH/F, RACH/F; 2x8x(TCH/8+SACCH) 29x(TCH/F+SACCH) A large capacity cell with 12 transmitters/receivers: one TN0: FCCH, SCH, BCCH, PAGCH/F, RACH/F one TN2, TN4, and TN6: BCCH, PAGCH/F, RACH/F 5x8x(TCH/8+SACCH) 87x(TCH/F+SACCH) 47

48 The frequency band The so called primary band of GSM includes two 25 Mhz subbands. Other bands: Extension to 33 MHz with MHz and MHz GSM1800 bands MHz and MHz In the US GSM1900 The carriers spacing is 200 khz The border frequencies are usually not used, which limits the number of frequencies to 122 in the 25 MHz band. There may be additional national limitations. 48

49 Frequency hopping The radio interface of GSM uses slow frequency hopping. Each burst is transmitted with one frequency, in GSM. This provides at least two advantages: Frequency diversity: Mobile radio transmission is subject to severe multipath fading, but different frequencies fade independently. For example, when a mobile is standing still or moving very slowly, the signal may fade for several burst periods, and the connection may be lost. If different frequency is used for each burst, consecutive frames are probably not lost, and the connection quality may be acceptable. Interferer diversity: Cells using same frequencies interfere each other less if their hopping sequences are independent. Less interference means better re-use of the radio resource (cells sharing the same resource may be closer to each other), and thus, better spectrum efficiency. 49

50 Frequency hopping is not used on common channels (FCCH, SCH, BCCH, PAGCH, RACH and CBCH). The downlink common channels all use the same frequency. Also, signal on the frequency of the common channels is transmitted continuously, even if no information is to be transmitted, because mobile stations in neighboring cells continuously measure the signal level from the base stations. When there is not information to be transmitted, dummy frames are used. Hopping Sequences With or without frequency hopping, always the uplink frequency = downlink frequency + 45 MHz. For a set of n frequencies 64 x n different hopping sequences can be built, in GSM. They are described by two parameters, the Mobile Allocation Index Offset (MAIO, n different values), and Hopping Sequence Number (HSN, 64 different values). Two channels having the same HSN but different MAIO never use the same frequency at the same time. On the other hand, two channels having the same MAIO but different HSN interfere with the probability 1/n. The sequences are pseudo-random, except for the one with HSN = 0 which uses the frequencies in increasing order. Channels in one cell usually have the same HSN but different MAIO. Adjacent cells use different set of frequencies. Distant cells using the same frequency sets should use different HSN to minimise interference. 50

51 From source data to radio waves As an example, let us look at speech. Speech Digitizing and source coding Channel coding Interleaving Ciphering Burst formatting Modulation Speech Source decoding Channel decoding De-interleaving Deciphering Burst decoding Demodulation Note that in the source (speech) codec, encoding is more complicated than decoding. On the other hand, in the channel codec, decoding is much more complicated than encoding. Also, demodulation (including equalisation, synchronisation, etc.) is computationally intensive. 51

52 The following blocks are common to all transmission modes. channel coding introduces redundancy into the data flow by adding information calculated from the actual data, in order to allow correction, or at least, detection of transmission errors. interleaving mixes the bits of several code words such that consecutive bits are spread over several bursts. This is done because transmission errors often occur in bursts such that many consecutive bits (sometimes hundreds) are lost, and on the other hand, channel codecs perform better on uncorrelated errors. ciphering modifies the contents of the burst by performing an x-or -operation between a pseudo-random bit sequence and 114 bits of a normal burst. De-ciphering is done exactly the same way. The pseudo-random sequence is derived from the burst number, and a session key with a simple but confidential algorithm. burst formatting adds some tail bits at the ends, and a training sequence in the middle of the burst, in order to help synchronisation and equalisation of the received signal. modulation transforms the binary signal into an analog waveform which is mixed into the selected frequency and the selected timeslot. The receiver end is more or less logically the reverse operations in reverse order. 52

53 Channel coding provides protection against bit errors in the transmission channel. Error correction is mainly done with the convolutional codes, and block (parity) codes are for detecting remaining errors. In common channels, a so called Fire code is used which is capable of correcting errors occurring in groups. The following figure (next page) summarises basic GSM channel coding schemes. Some explanation to the figure: In each box, the last line indicates the chapter of GSM spec defining the function. In the case of RACH, P0 = 8 and P1 = 18; in the case of SCH, CSCH, CTSBCH-SB and CTSARCH, P0 = 25 and P1 = 39. In the case of data TCHs, N0, N1 and n depend on the type of data TCH. Interfaces: 1) information bits (d); 2) information + parity + tail bits (u); 3) coded bits (c); 4) interleaved bits (e). 53

54 interface 0 interface 1 interface 2 interface 3 TCH/HS (half rate speech TCH) speech frame 112 bits 3.2 cyclic code + tail in: 112 bits out: 121 bits speech frame 244 bits 3.1 cyclic code + repetition in: 244 bits out: 260 bits reordering and partitioning +stealing flag in: 228 bits out: 4 blocks interface 4 block diagonal interleaving in: 4 blocks out: pairs of blocks TCH/EFS (Enhanced full rate speech TCH) convolutional code k=7, 2 classes in: 121 bits out: 228 bits SACCH, FACCH, TCH/FS BCCH, CBCH, PCH (full rate AGCH, SDCCH speech TCH) CTSAGCH, CTSPCH data TCHs speech frame 260 bits 3.1 cyclic code + tail in: 260 bits out: 267 bits convolutional code k=5, 2 classes in: 267 bits out: 456 bits TCH/FS, TCH/EFS TCH/F2.4, FACCH block diagonal interleaving in: 8 blocks out: pairs of blocks 3.1.3, encryption unit message 184 bits Fire code +tail in: 184 bits out: 228 bits convolutional code k=5, rate 1/2 in: 228 bits out: 456 bits reordering and partitioning +stealing flag in: 456 bits out: 8 blocks 3.1.3, 4.1.4, TCH/F2.4 others block rectangular interleaving in: 8 blocks out: pairs of blocks data frame N0 bits 3.n.1 +tail in: N0 bits out: N1 bits 3.n.2 convolutional code k=5, rate r in: N1 bits out: 456 bits 3.n.3 others message P0 bits 4.6, 4.7, cyclic code + tail in: P0 bits out: P1 bits 4.6, 4.7, convolutional code k=5, rate r in: P1 out: P2 bits 4.6, 4.7, diagonal interleaving + stealing flags in: 456 bits out: 4 blocks diagonally interleaved to depth 19, starting on consecutive bursts 3.n.4 PRACH PTCCH/U RACH, SCH CTSBCH-SB, CTSARCH PDTCH(1-4), PBCCH, PAGCH, PPCH, PNCH, PTCCH/D RLC block Q0 bits 5.1.n.1 CS-1 others cyclic code + tail in: Q0 bits out: Q1 bits 5.1.n.2 others CS-4 convolutional code k=5, rate r in: Q1 bits out: 456 bits 5.1.n.3 reordering and partitioning +code identifier in: 456 bits out: 8 blocks

55 Bad frame substitution In speech channels, an important matter affecting speech quality is the bad frame substitution. After error correction, even in good conditions, several % of the speech frames may still be errorneous. In the method originally proposed in the GSM specification, a lost frame is substituted by the previous frame. If several speech frames are lost, they are substituted by attenuated versions of the previous good frame. Each good frame is reproduced with full amplitude regardless of the condition of neighboring frames. This kind of approach causes strange sound effects in poor channel conditions. One can imagine a situation where the system manages to get a correct speech frame through only occasionally. The correct frame is reproduced with full amplitude, and the missing frames after it are replaced by attenuated versions of the previous correct frame. This can be heard as some kind of ringing. Later, this strategy has been improved, and audio signal processing is developed to improve the sound quality in poor channel conditions. 55

56 Bursts Normal burst Tail Information 3 58 Training sequence 26 Information 58 Tail 3 Access burst Tail Training sequence 3 26 Information 36 Tail 3 Synchronisation burst Tail Information Training sequence Information Tail Frequency correction burst All zeros 148 Some notes: When modulated, the frequency correction burst produces almost pure sine wave signal. Training sequences are pseudo-random sequences with narrow autocorrelation function. Adjacent base stations use different training sequences. The mobile station has to switch off its transmitter between bursts. Is this a problem? 56

57 Modulation The modulation chosen in GSM is Gaussian Minimum Shift Keying. This is a quadrature phase modulation scheme where the phase φ (t) of the signal E( t) = cos( ω 0t + φ( t)) is changed according to the input data. One can think that the function describing the phase change is a ramp filtered by a low-pass filter whose impulse response is a gaussian pulse. The filtering spreads the phase change over 3 bit periods. Later when higher data rates (up to almost 400 kbits/s) are introduced to GSM, 8- PSK modulation is adopted. This will be described later. In GMSK of GSM, the modulating symbol rate is 1/T = 1 625/6 ksymb/s (i.e. approximately ksymb/s), which corresponds to 1 625/6 kbit/s (i.e kbit/s). Before the first bit of the bursts enters the modulator, the modulator has an internal state as if a modulating bit stream consisting of consecutive ones (d i = 1) had entered the differential encoder. Also after the last bit of the time slot, the modulator has an internal state as if a modulating bit stream consisting of consecutive ones (d i = 1) had continued to enter the differential encoder. Each data value d i is differentially encoded. The output of the differential encoder is: dˆ d d ( d {0,1}) i = i i 1 i where denotes modulo 2 addition. 57

58 The modulating data value α i input to the modulator is: α = 1 2d ( α { 1, + 1}) i i i The modulating data values α i excite a linear filter with impulse response defined by: t g() t = h()* t rect T where the function rect(x) is defined by: t 1 rect = for t < T T T 2 t rect = 0 T otherwise and * means convolution. h(t) is defined by: h( t) = 2 t exp 2 2 2δ T (2π ) δt where ln( 2) δ = and BT = 03. 2πBT where B is the 3 db bandwidth of the filter with impulse response h(t), and T is the duration of one input data bit. 58

59 The phase of the modulated signal is: t' it = i i ϕ(') t α πh g( u) du where the modulating index h is 1/2 (maximum phase change in radians is π/2 per data interval). The time reference t' = 0 is the start of the active part of the burst. This is also the start of the bit period of bit number 0 (the first tail bit). The modulated RF carrier, except for start and stop of the TDMA burst may therefore be expressed as: 2Ec x( t' ) = cos(2π f0t' + ϕ( t' ) + ϕ0) T where E c is the energy per modulating bit, f 0 is the centre frequency and ϕ 0 is a random phase and is constant during one burst. 59