GSM By Alcatel

TDMA Methods, page 1

Rohde & Schwarz Trainingszentrum, V 2.3 TDMA_system-e.doc

Access Methods in GSM

1. Fundamentals of Multiple Access

Frequency division multiple access FDMA

Time division multiple access TDMA

Code division multiple access CDMA

2. TDMA in GSM

RF data

TDMA structure in GSM, frames and multiframes

TDMA timers

3. Burst Structures

Information bits

Training sequence

Bit synchronization

Compensation of multipath reception

Guard time

Delay correction

Burst types

Frequency-correction burst

Synchronization burst

Dummy burst

Access burst

4. The Concept of "Channel" in GSM

Physical/logical channel

Physical channels and their definition

Main logical channels and their functions



1. Fundamentals of Multiple Access

The 200 kHz channel bandwidth of GSM systems seems fairly wide in comparisonwith that of conventional systems. This bandwidth is "divided up" using timeslotswhich allow one channel to be used by several subscribers (multiple access). Themultiple access methods available and their characteristic features are described inthe following.

Frequency division multiple access FDMA

For analog radio systems, the trend has always been towards a more efficientutilization of the available frequency spectrum by reducing the channel spacing. Thenumber of radio channels obtained at a channel spacing of 12.5 kHz is, of course,twice that obtained with 25 kHz. However, improvements usually have to be tradedoff against some drawbacks: the narrower the channel spacing, the higher therequired frequency accuracy. Consequently, the maximum deviation has to bereduced, which leads to poorer transmission quality due to the lower S/N ratio.Moreover, the gaps between the channels, which must be several kilohertz wide toprovide a "safety margin", also reduce the available transmission bandwidth.

Fig. 1: Channels in wideband and narrowband systems (fdma.dsf)

Dividing the available frequency spectrum into a number of frequency channelsenables several users to simultaneously access the various frequencies. This form ofmultiple access is called frequency division multiple access (FDMA). Consequently,all radio systems using a spectrum divided into channels are FDMA systems. Today,the technically useful limit is reached with a channel spacing of 10 to 12.5 kHz. If timeis considered as a third dimension, the following diagram, frequently used in GSMenvironments, is obtained:



A m p l i t u d e

F r e q u e n c y

C h a n n e l sT i m e N - 1 N N + 1 N + 2

Fig. 2: Frequency division multiple access FDMA

Advantages of FDMA:

- Simultaneous use of a given system bandwidth by many subscribers- More channels are available thanks to reduced channel spacing

Disadvantages of FDMA:

- Higher frequency accuracy required- Transmission quality decreases as the channel bandwidth is reduced- Better selectivity filters required- One transmitter and also one receiver is required per channel

Time division multiple access TDMA

With TDMA (time division multiple access) systems, the available bandwidth isdivided into considerably fewer and so wider channels than in FDMA systems. Itappears that each channel is simultaneously available to several subscribers but infact each subscriber can use the whole channel only for the period of a timeslot. Forthe rest of the time, he has no access. This serial access by several subscribers isrepeated over time.



Amplitude

n-1n

n+1 n+2

Frequency

Time

Fig. 3: Time division multiple access TDMA

Advantages of TDMA:

- Simultaneous use of a given system bandwidth by many subscribers- Depending on the number of available timeslots, several subscribers can be served by one transmitter/receiver unit- Transmitter and receiver are not permanently on (saves battery power)- The instrument can perform other tasks in the intervals between transmission and reception of call (e.g. monitoring the field strength of neighbouring channels)- Reduced susceptibility to frequency-selective fading with large channel bandwidths

Disadvantages of TDMA:

- Accurate time (and frequency) synchronization of intruments required- Higher processing capacity required- Broadband modulators required

Code division multiple access CDMA

The advent of powerful, cost-effective signal processors meant that a lessconventional multiple access method could be employed for mass communicationsystems. With code division multiple access CDMA, the whole system bandwidth isavailable to all subscribers all the time, i.e. all subscribers transmit and receivesimultaneously but each subscriber uses a different code. Logic 1 is represented by acertain bit sequence, logic 0 is the inverse of this sequence. The different signals aredistinguished in the receiver by cross-correlating the received sum signal, whichcontains the different codes of all active subscribers, with the bit sequence of thesubscriber whose transmission the receiver wants to detect. UMTS (Universal MobileTelecommunications System), the 3rd generation mobile telephone system uses thisaccess method.



Amplitude

Frequency

Time

Fig. 4: Code division multiple access CDMA

Advantages of CDMA:

- Simultaneous use of a given system bandwidth by many subscribers- Several signals can be received simultaneously by a single RF section- Reduced susceptibility to frequency-selective fading in the case of large channel bandwidths- More subscribers can be served- Reduced costs for radio network planning

Disadvantages of CDMA:

- Accurate time synchronization of subscribers required- Fast transmitter power control over a wide dynamic range required- No mass-market experience available

2. TDMA in GSM

RF data

In spite of the competition from other mobile telephone systems, it was possible todefine common frequency bands for GSM on a worldwide basis. All nations whosigned the GSM-MoU (memorandum of understanding) committed themselves to usethe standardized frequency ranges when they installed their GSM system. Thecompetition for frequencies mainly affects countries using NMT 900 (Nordic MobileTelephone), whose frequency range coincides with the GSM P band. The TACS(Total Access Communication System) band too has some overlap with the GSM Pband, and the G1 band (extended GSM 900) is completely within the TACS range.Cordless telephones to the CT1 standard also use the upper end of the GSM P band.CT1+ telephones which had been assigned a frequency range below the P bandyears ago to protect them against GSM are now being ousted by the G1 band.



GSM 900 GSM 1800

Frequency range P band G1 band Uplink (MS transmitting)

890 to 915 MHz 880 to 890 MHz 1710 to 1785 MHz

Downlink (BTS transmitting)

935 to 960 MHz 925 to 935 MHz 1805 to 1880 MHz

Duplex spacing 45 MHz 95 MHzSpectrum 2 x 35 MHz 2 x 75 MHzNumber of channels 124 49 374Channel No.s 1 to 124 975 to 1023 512 to 885Channel spacing 200 kHzModulation GMSK with B x T = 0.3Data transmission rate 270.833 kbit/sBit duration 3.69 µs

Fig. 5: RF data for GSM 900 and GSM 1800

Depending on the resources of the network operator and the technical facilities of themobile phone, up to 124 channels of the GSM 900 network are available in theP band (two frequencies per channel at a spacing of 45 MHz for uplink and down-link), and perhaps another 49 channels in the G1 band (also GSM 900 frequencies),and probably 374 channels in the E network (GSM 1800, duplex spacing 95 MHz).Normally, the network operators can use the frequencies (channels) assigned tothem for their base stations as they choose. Each base station requires at least onechannel (C0, also referred to as BCCH carrier) on which it continually sendssynchronization information at full power and - depending on the expected trafficvolume - additional frequency channels (traffic channels which are only used foractual calls).

Why is GSM referred to as a TDMA system when it uses different frequencychannels?The key point is that each frequency channel is divided into 8 timeslots. In the firstchannel (C0) of every base station, synchronization information is sent in timeslot 0while the remaining 7 timeslots are used for calls (traffic) or dummy bursts so thatpower is always being transmitted. This is also necessary for synchronization andensures that a telephone which is switched on in an area for the first time can find itsGSM base station.

To keep the hardware of the telephone as simple as possible and to ensure optimumutilization, it was decided that transmission and reception should not be simultaneous(intermittent operation of transmit and receive section). Nevertheless, this seems likeduplex mode to the subscriber because the large amount of speech data transmittedin compressed form in the timeslots fills all the eight timeslots when expanded.Operation seen from the mobile:



RX

TX

RX

TX3 time slots

* Duplex spacing

* GSM900: 45 MHz GSM1800: 95 MHz

RX

Fig. 6: Transmission and reception in "duplex" mode

The receive frequency and transmit frequency are generated by a single synthesizerin the mobile. The synthesizer lets the receiver "listen" to the base station in atimeslot and then three timeslots later lets the transmitter transmit (now the basestation should listen). This sequence is repeated after eight timeslots (a frameconsists of eight timeslots). The free slots in between are used, say, for field-strengthmeasurements on the C0 frequencies of neighbouring base stations. The measuredfield strength is the criterion for deciding whether an ongoing call is handed over toanother base station or not.

To simplify timeslot counting, the timeslots of the base station and of the mobile arecounted in the same way. For instance, if the downlink (base station sends) isassigned to timeslot 0, the mobile station must receive at the same time. The uplink(mobile station sends) for the phone is also assigned to timeslot 0, but because of theoffset of 3 timeslots described above, the uplink timeslot is delayed by 1.73 ms (fortiming see next section).

TDMA structure in GSM, frames and multiframes

GSM timing is based on 48 periods of a 13 MHz signal (approx. 3.69 ms, whichcorresponds to the transmission time of one bit). A certain number of these bits iscombined to form a burst and is transmitted in one timeslot. Eight timeslots form aframe. A certain number of these frames is combined to give a multiframe. Sincethere are several types of multiframe, certain numbers are grouped together to formstandard hyperframes and superframes.

Detailed description:



Fig. 7: Timeslot in TDMA frame

Two 57-bit information blocks, i.e. 114 bits, are transmitted in every timeslot. A TDMAframe contains eight of these timeslots and since a call can only use one timeslot perframe, the raw data transmission rate (coded speech or data signal plus errorcorrection code) is about

114 bits/4.62 ms = 24.7 kbit/s.

A known bit sequence, referred to as the training sequence, is transmitted betweenthe information blocks. It is used for synchronization to the bit stream and forassessing the current transmission characteristics of the radio channel. The trainingsequence makes it possible to set channel equalizers in the receiver to considerablyimprove decoding. Since the transmission conditions in the radio channel maychange rapidly, the training sequence is sent between the information blocks andtransmitted with each burst.

The guard periode has been inserted to prevent consecutive bursts time overlappingif signals are not fully time compensated (see further down). It is also required forramping up resp. ramping down the transmitter (power ramping). It is certainlyunusual to specify a guard time as a fraction of a bit transmission period (8.25), butinterpreting this information in terms of time (8.25 bit periods x 3.69 µs) has proved tobe useful.

Time slot(0,577 msec)

Information InformationFT TS F T Guard3 157 26 1 57 8,253

T=Tailbit, F=Flag, TS=Training Sequence, Guard=guard priod

number of bits:

TDMA frame(4,62 msec)

6 72 3 4 50 1



This "odd" number is obviously due to the maximum signal delays resulting from thecell size planned in the GSM definition phase.

The flag bit indicates whether the transmitted bits are normal information bits orwhether some of the transmitted bits are used for signalling, e.g. when there is anurgent need to perform a handover.

Frame numbers are also used on the control channel but the frames of controlchannels and traffic channels are numbered separately. This is necessary so thatcertain measurements can be performed, e.g. while a call is in progress. 26 TDMAframes are combined to form a 26-frame multiframe for all timeslots containing atraffic channel (i.e. voice and data signals and a small amount of signalling data tokeep the link up). All time slots that are exclusively for signalling are counted using51-frame multiframes.

This method of counting conceals the fact that physical channels (specified in termsof frequency channels and timeslots) and logical channels (e.g. traffic channels,TCHs) are handled in a different way. Thinking in terms of logical GSM channels isan approach that has turned out to be useful. Logical channels have an almostparallel existence and must, of course, be mapped serially onto the physical channelsby the hardware. The procedure is a bit confusing for a user who is not familiar withGSM but the approach has proved to be very useful for network operation becausesome logical channels are only transmitted when required and can be moved fromone physical channel to another according to the traffic volume. Only very few logicalchannels have always to be associated with the same physical channels to allow themobile to synchronize to an unknown GSM base station. If the first synchronizationattempt succeeds, the information in a few fixed signalling bursts will be sufficient todecode the whole data stream.

The TDMA structure provides for hyperframes and superframes above the 26-frameand 51-frame multiframes. The hyperframes and superframes can be used for bothtypes. These hyperframes and superframes are used, for instance, for encryptionalgorithms. The following structure is obtained:

26-frame multiframe= 26 frames for timeslots containing traffic channels

"in parallel" with51-frame multiframe

= 51 frames for timeslots containing control channels

are "combined" insuper frames:

= 51 x 26 frames (least common multiple, makes a "combination" possible)

= 1326 frames= 6.12 s

andhyper frames: = 2048 super frames

= 2,715,648 frames= 3 hours 29 min. 3.5 s



TDMA timers

The frames within the hyperframe are continuously counted so that the TDMA clockrestarts after approx. 3.5 hours. The frame number, therefore, represents the basictime unit for the GSM system. As it would be too cumbersome to use just seconds tomeasure time, larger units like minutes and hours are defined; the same thinghappens with frame numbers which are expressed in larger units – the various typesof timer. These timers are defined as follows:

T1: = FN div (26 x 51) value range: 0 to 2047

T2: = FN mod 26 value range: 0 to 25

T3: = FN mod 51 value range: 0 to 50

FN (Frame Number) value range: 0 to 2715647

FNmax = 51 x 26 x 2048 - 1

Fig. 9: TDMA timers

The original frame number can be calculated from the three counter readings. Forcertain signalling tasks, not all the counters are needed. In other words, this meansthat some signalling procedures can be performed without all the counter readingsbeing known.The division without remainder function "div" is used to calculate T1; div gives thewhole number obtained when FN is divided by 1326 = 26 x 51. T2 and T3 are equalto the FN mod 26 and FN mod 51 respectively, the timers are repeatedly countingfrom 0 to 25 and 0 to 50.

3. Burst Structures

Information is exchanged between the base station and mobile station in thetimeslots. In each slot a certain amount of information, i.e. a burst, can betransmitted. Depending on the task to be performed, different types of burst can beused, although the most frequently used type is the "normal burst" shown in Fig. 10.It is used for signalling as well as for speech and data transmission.



Fig. 10: Normal burst

Each part of the burst serves a specific purpose which will be described below:

Information bits

The normal burst is able to transmit a total of 114 bits so that a maximum data rate ofapprox. 24.7 kbit/s is obtained by 2nd generation GSM. The transmission rate withinthe system can only be increased when more than one timeslot is used (GeneralPacket Radio Service GPRS, High-Speed Circuit-Switched Data HSCSD) or anothermodulation method (8PSK modulation with EDGE, Enhanced Data Rate for GSMEvolution).The bit rate in the control channels is much lower, i.e. the above transmission rate isonly attained by the mobile if a traffic channel has been established. In this case themobile and the base station use a control channel in addition to the speech and datachannel, which uses up capacity and carries information on reception quality andpower ramping. Fig. 11 shows the bit rates for the various channels:

Useful data: Error protection: Total:

Traffic channel: 22.8 kbit/s - Voice (full-rate): 13.0 kbit/s 9.8 kbit/s - Data: 2.4 kbit/s 20.4 kbit/s

4.8 kbit/s 18.0 kbit/s 9.6 kbit/s 13.2 kbit/s14.4 kbit 8.4 kbit/s

Control channel: 0.95 kbit/s Idle frame: 0.95 kbit/s

Total: 24.7 kbit/s

Fig. 11: Transmission bit rates

1 time slot(0,577 msec)

Information InformationFT TS F T Guard3 157 26 1 57 8,253

T=Tailbit, F=Flag, TS=Training Sequence, Guard=guard period

number of bits:



Training sequence

In the middle of the normal burst, a 26-bit training sequence is sent, the bit sequencebeing known to the receiver. There are eight different sequences which are referredto as the training sequence code (TSC). The eight sequences must be stored in allreceivers and at the beginning of a transmission the base station decides on the TSCto be used. The training sequence performs two main tasks: bit synchronization andestimating the channel impulse response (instantaneous response of the radiochannel). Using this estimate, the channel equalizers in the receivers can be set foroptimum data stream decoding.

Bit synchronization

Data are transmitted via the air interface in asynchronous mode. The receiver mustbe able to regenerate the bit clock from the data stream and needs features in thedata stream to enable it to identify information units (block synchronization).Conventional data radio therefore uses data telegrams that start with a ...10101010...sequence so that the receiver can regenerate the bit clock. A predefined bit word tellsthe receiver when the actual information (block synchronization) starts. A receiversynchronized in this way is able to decode the data stream online.The GSM training sequence is used for fine bit synchronization and for blocksynchronization. Since the training sequence is not sent at the beginning of a burst,the received data stream must be buffered in the receiver and decoded later on.Synchronization itself makes use of cross-correlation, i.e. the stored data stream iscompared bit-by-bit with the expected training sequence. When the position of thetraining sequence is known, the timing of the information bits is also known and finetuning of the bit-clock is performed. A burst that does not contain the expectedtraining sequence cannot be synchronized and decoded.

Compensation of multipath reception

The signal from the transmitter (in the Fig. below the base station is transmitting tothe mobile, but the same explanation still applies, if the mobile is transmitting) arrivesat the receiver not only along the direct path but also on various other paths as aresult of reflections and diffraction caused by obstacles in the signal path. Thepropagation conditions on these additional paths are different to those on the directpath, for instance:

- longer travel time because of longer path- various receive levels (depending on reflections)- different Doppler shifts (possibly due to different relative velocities)



BTS

MS

Fig. 12: Multipath reception due to reflections and diffraction

Because of the different travel times, the signals have different phases at thereceiving antenna. Depending on their phase, components may compensate - i.e.the total level goes to zero - or reinforce so that a strong signal is received for a shortperiod of time. RF level variations are random, fading may be up to 40 dB.

In addition to RF level fading, there is another annoying effect which, withoutcompensation, would make correct signal decoding rather difficult.Because of the extra length of the indirect paths, the signals at the receiving antennanot only have different phases but the modulated information arrives at differenttimes. The sum of all the channel responses to a single pulse is called the CIR(Channel Impulse Response). If the indirect path is just one kilometer longer, theGSM echo bit reaches the receiver later than the directly received bit and sointerferes with the next bit to be received. This intersymbol interference may affectseveral consecutive bits. With delays greater than 15 µs, identification of the receivedsignal components and echoes becomes more and more difficult. This problem canalso be solved with training sequences. The echoes on the indirect paths alsocontain the echoes of the training sequence. The cross-correlation method used tofind the original training sequence may also be used to find the training sequenceechoes as well as their delay and attenuation. With the aid of this information, thereceived signal can be corrected by a channel equalizer.



t

Sent pulse

Received signal

T = bit durationT = propagation time BitD

Fig. 13: Channel impulse response

Guard time

Transmission in each time slot is terminated with a guard time of 8.25 bit periods(8.25 x 3.69 µs ≈ 30 µs) during which no information bits can be sent. During thistime, the burst level must be reduced by up to 70 dB to avoid the next timeslot beingaffected. The "owner" of the subsequent timeslot uses this time to increase histransmitter power to nominal. This means that the guard time is used twice for powerramping (the transmitter power must be increased and reduced within narrowtolerances).



Fig. 14: Guard time at the end of each timeslot

Delay correction

The integrity of a timeslot depends on subscribers transmitting only during the timeassigned to them and otherwise keeping their transmitters off the air. This is onlypossible when all subscribers are in strict sync. For practical reasons, the clock signalis generated by the base station and all the mobiles synchronize to it. There will beno problems on the downlink, i.e. when (one) base station sends to (several) mobiles.Mutual interference may, however, occur in the uplink, where up to eight subscribersmust share one radio channel, if the mobiles are not accurately synchronized to thetimeslots.Where can problems with synchronization occur?Under the given conditions, radio signals propagate at almost the speed of light.Even at a speed of 300 000 km/s they still take about 33 µs to cover a distance of10 km. A mobile station 10 km away from the base station synchronizes to thereceived signal which has already travelled for about 33 µs before it arrives at themobile. If the mobile station now transmits back to the base station (without any delaycorrection), this signal will require the same travel time. The base station, therefore,receives a signal which is delayed by about 66 µs in its own time frame. Consideringthat a timeslot is 577 µs wide (including the guard time of about 30 µs) and cells havea max. radius of. 35 km (limited by the delay correction factor described in thefollowing) it is obvious that the neighbouring timeslot will be compromised if there isno compensation.



Fig. 15: Signal delay and its effect

The above example with the numbers:

Signal travel time over 10 km is ≈ 33 µs (distance / speed of light)The (time) sync signals from the base station require this time to arrive at the antenna of the mobile station, i.e. mobile synchronization is delayed by 33 µs.From the point of view of the mobile station, the mobile sends a burst to the base station with correct timing (without delay correction) but from the point of view of the base station the signal is sent 33 µs too late.The signal covers the same distance on its way to the base station antenna and so is delayed by another 33 µs.Relative to the base station’s time frame, the received signal is delayed by 66 µs.The signal is not sent in the assigned timeslot and compromises neighbouring timeslots.

The guard time at the end of each burst is only about 30 µs and is certainly not longenough in the above example (apart from it being required for power ramping of thetransmitters). The greater the distance between mobile station and base station, thegreater the effect of the signal delay. The only way to solve this problem is to makethe mobile send the burst earlier. To do so, the base station has to measure thesignal delays and send the appropriate correction factor to the mobile. A special bursttype (access burst, see further down) is used for this purpose. This burst has a muchwider guard time and is used by the mobile to attach to the base station when itwants to establish a connection.

Effect of signal propagation time

Base StationMobile Station

Propagation time ~33 µsecto phone

Propagation time ~33 µsecto base station



Because of the longer guard time of this burst, interference with neighbouringtimeslots is prevented. The base station can determine a correction factor (referred toas TA, Timing Advance in GSM, it represents the number of bit periods) and send itto the mobile station. The mobile station advances the sending time of its burstaccordingly and the signal from the mobile station arrives at the base station in sync.If the mobile station is moving while a call is in progress, the distance between themobile station and the base station generally changes and so also the signal delay.For this reason, the timing advance is checked about twice every second while a callis going on.

A few technical limits of the GSM system (GSM 900 and GSM 1800) can be derivedfrom the timing advance specifications:The timing advance is transmitted as a 6-bit word. With 6 bits, the numbers 0 to 63can be represented. Increasing this number by one means that the mobile stationhas to advance transmission by one bit duration, i.e. by 3.69 µs. This means a delayof 63 x 3.69 µs = approx. 232.5 µs can be corrected, which corresponds to a distanceof almost 70 km. Therefore, the mobile station cannot be more then about 35 kmaway from the base station. This also means that the distance between the basestations in the network cannot be more than 70 km. The timing advance also explainswhy, for instance, a mobile on a ship near to shore can find a GSM network(propagation conditions and coverage on water are optimal) but cannot register to thenetwork because the base station is more than 35 km away. These extreme ambientconditions make it clear that the timing advance may also be a criterion for callhandover or disconnection. On shore, this scenario is only of theoretical interestbecause, due to the traffic volume encountered, none of the base stations has tocover a cell radius of 35 km.

The timing advance can be used to determine the distance between the mobile andthe base station because each increment in the TA (1 bit duration = 3.69 µs)corresponds to a signal travel time of 1.1 km. If the base station uses anomnidirectional antenna, a valid timing advance indicates that the subscriber is on a550 m wide circle centered on the base station. On the other hand, a valid timingadvance is only available when a link has been set up, i.e. only when a call is inprogress or during a location update. At any other time, the timing advance may beinvalid because the mobile is not transmitting.

Burst types

In addition to the normal burst described previously, which is used in most of thecases, other burst types are available for special purposes (see Fig. below). All theseburst are exactly 1 timeslot (577 µs) in duration.



Fig. 16: Burst types (burst_ty.dsf)

Frequency correction burst

The 142 "fixed bits" of the frequency correction burst are all set to logic 0. GMSK(Gaussian minimum shift keying), the type of modulation used for GSM, produces astationary carrier-frequency deviation of approx. 67.7 kHz with this burst. This burst issent by the base station only and used by the mobiles for initial synchronization tothe carrier frequency and for compensating any Doppler shift caused by a mobilemoving at speed. It is sent by the base station every 10 frames (i.e. approx. every 46ms) but only in timeslot 0 and only on carrier C0 (sometimes referred to as the BCCHcarrier).

Abbreviations :

T=Tail Bit (3 Bits/ 8 Bits in leading part of Access BurstF=Flag (1 Bit)G=Guard Period (8,25 Bits/ 68,25 Bits in Access Burst)

T3

T3

T3

T3

T3

T3

T3

T3

T8

57 data bits 57 data bits 26 BitsTraining Sequence

F1

F1 G

8,25

G8,25

G8,25

G8,25

G68,25

142 fixed Bits

64 BitsExtended Training Sequence39 data bits 39 data bits

142 fixed Bits

36 data bits T3

41 BitsTraining Sequence

Normal Burst

Frequency Correction Burst

Synchronisation Burst

Dummy Burst

Access Burst



Synchronization burst

The synchronization burst is also transmitted in timeslot 0 in the frame after thefrequency correction burst. This burst, too, is only sent by the base station on the C0carrier. The considerably longer training sequence is one significant differencebetween the synchronization burst and the normal burst. Like the 26-bit type, thistraining sequence is also used for bit synchronization but, because it is longer,synchronization is more accurate.The two 39-bit data blocks contain the timers T1, T2 and T3 in coded form and also abase station identification code (incl. the training sequence No.). When this "GSMtime" is received, the mobile station is in sync with the base station.

Dummy burst

A base station must continuously transmit at nominal power on its C0 carrier in alltime slots as this carrier is used by the mobile stations to find the nearest basestation and to evaluate reception quality. If normal bursts are not available fortransmission in a timeslot, dummy bursts are sent by the base station insteadbecause an unmodulated carrier cannot be transmitted. These bursts, too, are usedonly by the base station on the C0 carrier but may be sent in any of the timeslots.

Access burst

As already pointed out, the access burst is sent when the mobile station calls thebase station for the first time. The base station uses this burst for a delaymeasurement, determines the associated timing advance and informs the mobilestation accordingly. This means that delay correction is performed for the next callfrom the mobile which can now use a normal burst with a much shorter guard time.



4. The Concept of Channel in GSM

Every base station sends on at least one frequency in 8 timeslots. It has becomecommon practice to refer to physical channels that are defined by frequency andtimeslot. Several "types" of data are sent on these physical channels, e.g. speech,test reports, instructions, etc. For these data types the term "logical channel" is used.Logical channels are considered to be "parallel" channels which are serially mappedby the hardware onto the physical channel (which must not always be the same;frequency and/or timeslot may be changed as required). For instance, the FCCH(Frequency Correction Channel), which is used for correcting the frequency of themobile station, the SCH (Synchronization Channel) with the initial information on thebase station, the BCCH (Broadcast Control Channel) acting as a kind of "noticeboard" with further information, and many other logical channels are transmitted intimeslot 0 (first time slot) of carrier C0. Some of these logical channels are onlytransmitted in specific contexts and their position in the physical data stream is notalways the same.

Physical channels and their definition:

- ARFCN + TNThe number of the carrier frequency channel (absolute radio frequency channelnumber) together with the timeslot number defines the simplest version of a physicalchannel.

- several ARFCNs + TN + HSN + MAIOWhen frequency hopping is activated, the hopping sequence number (HSN) and themobile allocation index offset value MAIO must also be specified.

- ARFCN + TN + SSNIf the half-rate speech codec is used for communication, two calls can share a full-rate channel. The subsequence number SSN is used to distinguish the calls. Itindicates whether the half-rate link uses even or odd frame numbers.

- several ARFCNs + TN + HSN + MAIO + SSNsame as above, but the frequency of a half-rate channel is also assigned andfrequency hopping is activated at the same time.



Main logical channels and their functions

Traffic channels

Traffic channels are used to carry digitized speech or other user data. They arenormally classified according to transmission speed. For voice transmission, thefollowing is defined:

- Traffic channel using full-rate data transmission, a full-rate channel operatingat 22.8 kbit/s. 13 kbit/s are used for speech transmission, the rest is basicallyused for error protection.

- Traffic channel using half-rate transmission, a half-rate channel operating at11.4 kbit/s. 6.5 kbit/s are available for speech transmission.

The subscriber can also choose between half-rate and full-rate transmission for data.Available bit rates:

Designation Explanation

TCH/FS Full-rate speech traffic channelTCH/HS Half-rate speech traffic channelTCH/F14.4 14.4 kbit/s full-rate data traffic channelTCH/F9.6 9.6 kbit/s full-rate data traffic channelTCH/F4.8 4.8 kbit/s full-rate data traffic channelTCH/H4.8 4.8 kbit/s half-rate data traffic channelTCH/F2.4 2.4 kbit/s full-rate data traffic channelTCH/H2.4 2.4 kbit/s half-rate data traffic channel

Fig. 17: GSM traffic channels and their bit rates

Control channels

Even if no call is in progress (traffic channel), the resources required for signalling areconsiderable. Information has to be continuously exchanged via the air interface (e.g.location update). The control channels allow the mobile station to receive informationfrom the base station any time or to send information to the base station.

There are three main groups of control channels:



Broadcast Channels

This channel group is used by the base station to send relevant information to allactive mobile stations (unidirectional transmission to mobile)

- Frequency Correction Channel FCCHfor the frequency synchronization already described. It is transmitted inframes 0, 10, 20, 30, 40 and 50 within the 51-frame multiframe

- Synchronization CHannel SCHwith "GSM time" and a code for base station identification. This channel issent in the frame directly after the FCCH.

- Broadcast Control Channel BCCHwith information on the radio channel configuration of the home cell and ofneighbouring cells, on the location area code for a location update and on theorganization of the common control channels CCCH (described below). Thischannel also contains other important signalling information. The BCCHcomprises four normal bursts which are sent in frames 2 to 5 of the 51-framemultiframe.

- Cell Broadcast CHannel CBCHThis is a kind of open information channel and comparable to teletext in TVbroadcasting.

Common Control Channels CCCH

This group is used for information exchange between base station and mobile station(bidirectional) - mainly for access management

- Paging CHannel PCH,used by the base station for paging mobile stations, e.g. for a mobile-terminated call (a call is made to the mobile station).

- Random Access CHannel RACHused by the mobile for a first call to the base station to request an exclusivecontrol channel. In the case of a mobile-originated call (mobile station calls asubscriber), the mobile sends an access burst on this channel.

- Access Grant CHannel AGCHis practically the response to the RACH. After having received the accessburst, the base station tells the mobile the traffic channel.

- Notification CHannel NCHenables the base station to notify incoming group calls.

Dedicated Control Channels

A bidirectional dedicated control channel performs signalling tasks independently orassigned to a traffic channel.

- Slow Associated Control CHannel SACCH,a slow, dedicated control channel which is coupled to a traffic channel andused, for instance, for power control, setting the timing advance and for testreports (receive field strength and quality). The SACCH uses frame 12 of the26-frame multiframe and is 4 bursts long. This means that it is sent in 4consecutive 26-frame multiframes.



- Fast Associated Control CHannel FACCHThis fast, dedicated control channel is coupled to a traffic channel andperforms signalling tasks that cannot be postponed. Example: preparing ahandover. This channel has to notify the "stealing" of bits for the FACCH viathe two flags of the normal burst (traffic transmission). For this reason, thetwo flags for the normal burst immediately before and after the trainingsequence are called "stealing flags". The FACCH therefore "steals"transmission capacity from the traffic channel.

- Stand-alone Dedicated Control CHannel SDCCHThis independent control channel is used for exchanging informationbetween the base station and the mobile station when no call is in progress.Example: location update, authentication and link setup up to the point whena call goes through.



Mapping of logical channels

a) Traffic Channels

b) Control Channels

T T

T = Traffic

T T T T T T T T T AT T T T T T T T T T T T T I0 12 25

A = SACCH

I = Idle

Mapping for Traffic Channels

Multiframe Structure (26 MF)

Multiframe Structure (51 MF)

F S B CB B B C CC CC CCC C C CF S F S C I F S CCCC B B B B C F S CCC0 10 5020 100

Channel C0

Channel C0Time Slot 0

F S B B B

TS0 TS7TS0 TS7

51 - Multiframe

t(down link)

(down link)

~ 235 ms

~ 4.62 ms

F = FCCHS = SCHB = BCCHC = CCCHI = Idle



R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R

F S F SF S F S F SB C C C C C C C C C I

- - -

D1 D2 D4D3 D5 D6 D7D0

D0 D1 D2 D3 D4 D5 D6 D7 A0 A1 A2 A3

A4 A5 A6 A7 - - -

A5 A6 A7 D0 D1 D2 D3 D4 D5 D6 D7 A0

D0 D1 D2 D3 D4 D5 D6 D7 A4A1 A2 A3 - - -

- - -

DOWNLINK

UPLINK

DOWNLINK

UPLINK

FCH +SCH + BCCH + CCCH

SDCCH/8(0..7) + SACCH/C8(0..7)

0 10 20 30 40 50

0 10 20 30 40 50

0 10 20 30 40 50

CHANNEL MAPPING (1)

CHANNEL MAPPING (2)

RR RR RRR RR RRRRRRRR RR RR RRRR

RR RR RRR RR RRRRRRRR RR RR RRRR RR

RRD3

A0 A1

A2 A3 D0 D1

D0 D1

D2

D2D3

F S F SF S F S F SB C C C I

F S F SF S F S F SB C C C I

D0 D1 D2 D3 A0 A1

A2 A3D0 D1 D2 D3

DOWNLINK

UPLINK

0 10 20 30 40 50

0 10 20 30 40 50

FCCH +SCH + BCCH + CCCH + SDCCH/4(0..3) + SACCH/4(0..3)

F = FCCH = Frequency Correction ChannelS = SCH = Synchronization ChannelB = BCCH = Broadcast Control ChannelC = CCCH = Common Control Channel (= PCH + RACH + AGCH)D = SDCCH = Stand-alone Dedicated Control ChannelA = SACCH = Slow Associated Control ChannelI = Idle

PCH = Paging ChannelRACH = Random Access ChannelAGCH = Access Grant Channel



List of Abbreviations Used:

BSIC Base station identification code

BTS Base transceiver station

BxT Bandwidth/bit duration product

CDMA Code division multiple access

CIR Channel impulse response

DCS1800 Digital Communication System 1800 (new: GSM1800)

FCB Frequency correction burst

FDMA Frequency division multiple access

FN Frame number

GMSK Gaussian minimum shift keying

GSM Global system for mobile communications

GSM900 GSM at 900 Mhz

GSM1800 GSM at 1800 MHz

MS Mobile station

NMT Nordic mobile telephone

RX Receiver

SB Synchronization burst

S/N Signal/noise ratio

TA Timing advance

TACS Total access communication system

TDMA Time division multiple access

TSC Training sequence code

TX Transmitter

GSM By Alcatel

Documents

expected training sequence

ts training sequence

dedicated control channel

channel impulse response

rate channel operating

valid timing advance

base station sends

signal travel time