EGPRS: GPRS/EDGE
8.6 GPRS/GSM Territory in a Base-Station Transceiver
116 Introduction to Mobile Network Engineering Table 8.1 EGPRS modulation coding schemes [1].
Scheme Coderate Modulation
Data rate per PDCH, kbps
MCS-9 1 8PSK 59.2
MCS-8 0.92 8PSK 54.4
MCS-7 0.76 8PSK 44.8
MCS-6 0.49 8PSK 29.6
MCS-5 0.37 8PSK 22.4
MCS-4 1 GMSK 17.6
MCS-3 0.85 GMSK 14.8
MCS-2 0.66 GMSK 11.2
MCS-1 0.53 GMSK 8.8
BCCH SDCCH CS TS CS TS CS TS CS TS CS TS CS TS
CS TS CS TS CS
reserved PS TS PS TS PS TS PS TS
CS/PS territory border Dedicated (E)GPRS time slots TRX 1
TRX 2
Default (E)GPRS territory PS TS
CS Territory
PS Territory
Figure 8.13 CS-PS borders in the BS transceiver.
All timeslots in one BS controlled by the PCU belong to the PS or EGPRS territory, all other to the CS territory. The border between CS and PS territory moves dynamically, depending on parameters set and on the required traffic, Figure 8.13.
Timeslots in the PS territory are classified into dedicated, default and additional time slots. Dedicated time slots, when present, are reserved for EGPRS traffic exclusively.
8.6.1 PS Capacity in the Base Station/Cell
Voice traffic always has priority over data traffic in EGPRS. For that reason, in the early stages of deployment it was possible to implement GPRS in an existing GSM network without adding extra capacity. The concept is illustrated in Table 8.2, which shows that for any given number of carriers there exists an associated amount of Erlang capacity available for a ‘best effort’ GPRS where blocking probability can be disregarded.
In this case, is possible to implement ‘best effort’ GPRS with no additional resources in base-station TRXs in the absence of a requirement on packet data service level and performance indicators.
Apparently, when there is a need to support a defined amount of traffic and tar- geted performance level, additional capacity resources need to be introduced in all network nodes. The dimensioning inputs are: traffic volume, type and performance
Table 8.2 The relationship between carriers, time slots, voice traffic and Erlang capacity [2].
Number of carriers
Number of time slots
Signalling time slots
Traffic time slots
Offered voice
traffic (Erlang), 1% GOS
Capacity available for GPRS (Erlang)
1 8 1 7 2.5 4.5
2 16 1 15 8.1 6.9
3 24 2 22 13.7 8.3
4 32 2 30 20.3 9.7
5 40 2 38 27.3 10.7
6 48 3 45 33.4 11.6
7 56 3 53 40.6 12.4
8 64 3 61 47.9 13.1
9 72 3 69 55.2 13.8
10 80 4 76 61.7 14.3
requirements. The dimensioning output is the required amount of traffic-dependent hardware and the associated software configurations.
The primary technique for dividing resources between the circuit-switched (CS) and packet ((E)GPRS) traffic is known as the Territory Method. In this, timeslots within a cell are dynamically divided into the CS and (E)GPRS territories. This means that a certain number of consecutive traffic timeslots are reserved for CS GSM calls with the remainder being available for the (E)GPRS traffic.
The dynamic variation of the territory boundary (and hence the number of timeslots in each territory) are controlled by territory parameters. This enables the system to adapt to different load levels and traffic proportions, thus offering optimized performance under a variety of load conditions. The Figure 8.13 shows an example of how a traffic resource within a cell (2 TRX in this case) can be divided into CS and (E)GPRS territories.
With a service requirement for dedicated (E)GPRS capacity, the number of timeslots are allocated on a permanent basis to (E)GPRS. These timeslots are always configured for (E)GPRS and cannot be used by the circuit-switched traffic. This ensures that the (E)GPRS capacity is always available in a cell. The drawback with this approach is that, for a given cell configuration, blocking levels for the CS traffic will increase in the case of a respective increase in the voice traffic load above the committed level.
The decision on whether to assign the dedicated (E)GPRS territory is a trade-off between providing a minimum level of (E)GPRS service and increasing the blocking for CS services. This decision needs to take into account operator priorities, network performance and predicted (E)GPRS usage levels. The commitment by the mobile operator to support a certain class of multislot terminals will determine the minimum number of time slots allocated for dedicated EGPRS territory.
In addition to dedicated EGPRS territory, which may or may not exist, a default (E)GPRS capacity can be allocated, as shown in Figure 8.13. The default (E)GPRS territory is an area that will always be included in the instantaneous (E)GPRS territory provided that the current CS traffic levels permit. With the exception of the dedicated (E)GPRS area, CS services always take priority over (E)GPRS services and so, if
118 Introduction to Mobile Network Engineering
circuit-switched traffic levels dictate, the circuit-switched traffic will occupy as much (E)GPRS default territory as is needed. If the CS level decreases, while previously occupying some of the (E)GPRS default territory, these timeslots will automatically be re-allocated back to (E)GPRS irrespective of the actual (E)GPRS load. In this approach, allocation to (E)GPRS will only occur if the (E)GPRS load reaches a predefined (see later).
The outcome of the EGPRS base-station dimensioning could be a supported mean data rate per cell. Throughput performance is of the EGPRS cell is affected by many fac- tors such as an optimization of link adaptation, application of frequency hopping, coding schemes supported and so on. Some details can be found in [3]. The accepted measure of cell design performance is a mean cell data rate that can be reasonably estimated to be in the range of 30 kbps per time slot averaged over the cell area.
Once the average throughput per time slot has been defined in kbps, the correspond- ing EGPRS Erlang value can be obtained by dividing the total GPRS traffic load (kbps) by the average throughput per time slot (kbps). This will give the total channel usage per unit of time; that is, ‘carried traffic’ in Erlangs.
Dimensioning of the PS core network is far more complicated. In addition to the lim- itations of radio interface that acts more as a data pipe, PS core dimensioning should take into account various overheads for various protocols associated with different traf- fic types, such as non-real-time versus real-time traffic and respective quality indicators.
Dimensioning is based on the internet activity model and queuing theory [2].