(E) GPRS BSS Optimization - S12

(E)GPRS BSS Optimization – S12
Gurgaon
06-08.10.2008.
1 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

Content
• Introduction
– GMSK and 8PSK modulation
– State, mobility, session and radio resource management
• Dimensioning and planning
– Network Analysis and planning inputs
– Deployment planning
– Air Interface capacity planning (CDEF, CDED)
– Connectivity capacity planning (BTS, EDAP, PCU, Gb, SGSN)
• Optimization
– Configuration and feature assessment
– BSS (KPI) and E2E performance assessment
– Signaling capacity & resource allocation improvement
– Data rate optimization
 Connectivity capacity (BTS, EDAP, PCU, Gb, SGSN)
 RTSL data rate improvement and multislot usage maximization (BSS)
 E2E data rate (applications)
– Mobility improvement

Introduction
GMSK and 8PSK modulation
Mobility, session and radio resource management

(E)GPRS Protocol Architecture
Relay
IP IP
GPRS Bearer L2
L1
User
Userinformation
informationtransfer
transfer GGSN
Um Gb Gn Gi
APP APP
TCP/UDP USER TCP/UDP
IP Relay PAYLOAD
IP
SNDCP Compression, segmentation SNDCP GTP GTP L2
LLC Ciphering and reliable link LLC UDP UDP L1
RLC RLC BSSGP BSSGP IP IP
MAC MAC NW sr NWsr L2 L2
GSM RF GSM RF L1bis L1bis L1 L1
MS BSS SGSN GGSN FIXED HOST
GPRS IP Backbone Internet

Bursts on the Air Interface – Mapping RLC blocks
RLC Blocks
RLC/MAC Blocks Note: Amount of RLC

blocks per radio block
1 TDMA frame = 4.615 ms depends on used
= BURST PERIOD TDMA Bursts (modulation) coding
scheme (M)CS
0 70 70 70 7
4 x TDMA Frames = 4 Bursts = 1 Radio block = 18.46 ms = 1-2 RLC block(s)
0 1 2 3 4 5 6 7 8 9 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4 5 5
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
P I P I
B0(0..3) B1(4..7) B2 (8..11) T B3(13..16) B4(17..20) B5(21..24) D B6(26..29) B7(30..33) B8(34..37) T B9(39..42) B10(43..46) B11(47..50) D
C L C L
C E C E
H H
52 TDMA Frames (240 ms)

12 x RLC/MAC Blocks = 1 x 52 PDCH MultiFrame = 240 ms
12 RLC/MAC Blocks / 0.240 s = 50 RLC/MAC Blocks / s

8-PSK Modulation
Phase states transitions
to avoid zero-crossing
• 8-PSK (Phase Shift Keying) has
3/8 been selected as the new
(d(3k),d(3k+1),d(3k+2))= modulation added in EDGE
(0,0,0) (0,1,0) • 3 bits per symbol
(0,1,1) • 22.5° offset to avoid origin crossing
(called 3/8-8-PSK)
(0,0,1) (1,1,1) • Symbol rate and burst length
identical to those of GMSK
• Non-constant envelope  high
requirements for linearity of the
(1,0,1) (1,1,0) power amplifier
(1,0,0) • Because of amplifier non-linearities,

a 2-4 dB power decrease (back-off)
is typically needed, Nokia
guaranteed a BO of 2 DB for BTS
EDGE GSM + EDGE

Modulation 8-PSK, 3bit/sym GMSK, 1 bit/sym
Symbol rate 270.833 ksps 270.833 ksps
Bits/burst 348 bits 114 bits
2*3*58 2*57
Gross rate/time slot 69.6 kbps 22.8 kbps

GMSK and 8PSK Bursts
dB
+4
+1
Phase state vector diagram

-1
-6
(**)
•Amplitude is not fixed

•Origin is not crossed - 30
(***)
•Overshooting
(147 bits)
(*) t
10 µs 8 µs 10 µs 7056/13 (542.8) µs 10 µs 8 µs 10 µs
GMSK Burst
dB
+4
+2,4
0
-2
-6
(***)
-20
(147 symbols)
-30
(**) 7056/13 (542,8)s
(*)
10 8 10 22 22 10 8 10 t (s)
8PSK Burst
8-PSK Modulation – Back-off Value
• Since the amplitude is changing in 8-PSK the transmitter non-linearities can be

seen in the transmitted signal
• These non-linearities will cause e.g. errors in reception and bandwidth spreading.
• In practice it is not possible to transmit 8-PSK signal with the same power as in
GMSK due to the signal must remain in the linear part of the power amplifier
• The back-off value is taken into Pout

account in link budget separately for Compression point
UL / DL and bands: 900/850,

1800/1900
• Too high MCA (8PSK) can lead to
unsuccessful TBF establishment, if the
Pin
MS is on the cell border with a low
Back Off= 4 dB
signal level and / or low C/I, because
of the back-off
Peak to Average of  3,2 dB

Burst Structure
• Burst structure is similar with current GMSK burst, but term 'bit' is
replaced by 'symbol'
• Training sequence has lower envelope variations
• Seamless switchover between timeslots
• In case of max output power only, back-off applied to 8-PSK
TSL0 TSL1 TSL2 TSL3 TSL4 TSL5 TSL6 TSL7

BCCH TCH TCH TCH TCH PDTCH PDTCH PDTCH
GMSK GMSK GMSK GMSK GMSK 8-PSK/ 8-PSK/ 8-PSK/
GMSK GMSK GMSK
P(dB)
t(us)

The EDGE Signal
1 2 3 4
1. Spectrum of Unfiltered 3pi/8 8psk modulation.

2. Filtered to fit GSM bandwidth.
3. Constellation after filtering: error vectors introduced.
4. Constellation after receiver Edge (equalised) filtering

GPRS Coding Schemes
Coding Payload (bits) Data Rate

Scheme per RLC block (kbit/s)
Nokia GPRS CS1 181 9.05
Correction
Error
Data
PCU CS2 268 13.4
Nokia GPRS CS3 312 15.6
PCU2 CS4 428 21.4
• CS1 & CS2 – Implemented in all Nokia BTS without

HW change More Data
=
•CS1 & CS4 – S11.5 (with PCU2) and UltraSite BTS Less Error
SW CX4.1 CD1 (Talk does not support CS3 and CS4) Correction

GPRS Coding Schemes
CS-1 CS-2 CS-3
CS-1
MAC RLC/MAC Block Size: 181 268 312
CS-2 Block Check Sequence: 40 16 16

USF BCS +4
Precoded USF: 3 6 6
CS-3
rate a/b convolutional coding 1/2 ~2/3 ~3/4
length: 456 588 676
puncturing 0 132 220
interleaving
456 bits
57 57 57 57 57 57 57 57 Data rate (kbit/s): 9.05 13.4 15.6
MAC RLC/MAC Block Size: 428
USF BCS BCS Size: 16

CS-4
Precoded USF: 12
Data rate (kbit/s): 21.4
20 ms
EGPRS Modulation and Coding Schemes
Scheme Code rate Header Modulation RLC blocks Raw Data Family BCS Tail HCS Data rate
Code rate per Radio within one payload kb/s
Block Radio Block
(20ms)
MCS-9 1.0 0.36 2 2x592 A 2x12 2x6 59.2
MCS-8 0.92 0.36 2 2x544 A 54.4
8PSK
MCS-7 0.76 0.36 2 2x448 B 44.8
MCS-6 0.49 1/3 1 592 A 29.6
544+48 8 27.2
MCS-5 0.37 1/3 1 448 B 22.4
12 6
MCS-4 1.0 0.53 1 352 C 17.6
MCS-3 0.80 0.53 1 296 A 14.8
GMSK 272+24 13.6
MCS-2 0.66 0.53 1 224 B 11.2
MCS-1 0.53 0.53 1 176 C 8.8
NOTE: the italic captions indicate the padding.
Ref: TS 03.64
EGPRS MCS Families
• The MCSs are divided into different families A,B MCS- 3
and C. Family A 37 octets 37 octets 37 octets 37 octets
• Each family has a different basic unit of payload: MCS- 6

37 (and 34), 28 and 22 octets respectively. MCS- 9
MCS- 3
• Different code rates within a family are achieved
34+3 octets 34+3 octets
by transmitting a different number of payload Family A
units within one Radio Block. padding MCS- 6
34 octets 34 octets 34 octets 34 octets
• For families A and B, 1 or 2 or 4 payload units
are transmitted, for family C, only 1 or 2 payload MCS- 8
units are transmitted MCS- 2
• When 4 payload units are transmitted (MCS 7, Family B 28 octets 28 octets 28 octets 28 octets
MSC-8 and MCS-9), these are splitted into two MCS- 5

separate RLC blocks (with separate sequence MCS- 7
BSN numbers and BCS, Block Check MCS- 1

Sequences) Family C 22 octets 22 octets
– The blocks are interleaved over two bursts only, for MCS- 4
MCS-8 and MCS-9.
– For MCS-7 the blocks are interleaved over four
bursts

EGPRS Channel Coding (MCS-9)
USF header FBI+E data 2 BCS tail
• EGPRS channel coding consists
1/3 tailbiting 1/3 convolutional coding
of separate data and header convolutional coding
coding, as shown in the figure
mother code
for MCS-9 downlink. block coding
puncturing puncturing
• Coding of data part: protecte

P1
encoded USF d
header P2 P3
• Data part includes user
data, two information from RLC
header, BCS (block check sequence)
and tail bits. 1st burst 2nd burst 3rd burst 4th burst
4 TDMA bursts = 20 ms
• Coded using 1/3 convolutional code.
• Punctured with a selectable puncturing scheme
(P1, P2 or P3).
• Two separate data parts for MCS-7...9. P1 P2 P3
puncturing
• Header part: mother code
• Includes RLC/MAC header information and 1/3 convolutional coding

information on the coding of the data part (like
used puncturing scheme). FBI+E data 1 BCS tail
• Convolutional coding + puncturing.

EGPRS Channel Coding (MCS-9)
3 bits 45 bits 612 bits 612 bits
Robust coding
for header USF RLC/MAC HCS FBI E Data = 592 bits BCS TB FBI E Data = 592 bits BCS TB
Hdr.
1 1 12 6 Adding
Rate 1/3 convolutional coding Rate 1/3 convolutional coding
redundancy
36 bits 135 bits 1836 bits 1836 bits
puncturing
puncturing puncturing Decreasing
redundancy
SB = 8 36 bits 124 bits 612 bits 612 bits 612 bits 612 bits 612 bits 612 bits
P1 P2 P3 P1 P2 P3
Interleaving over
2 bursts
(header: 4
Normal burst: Normal burst: Normal burst: Normal burst: bursts)
2x58x3 bits 2x58x3 bits 2x58x3 bits 2x58x3 bits
BP: 15/26 ms BP: 15/26 ms BP: 15/26 ms BP: 15/26 ms
1392 bits
20 ms
1392
 69,6kb / s 2  592
20ms Data  59,2kb / s
rate: 20ms
Ref: TS 03.64
GPRS Mobile States
Ready State Timer (RDY) – Default: 44 seconds
STANDBY state timer (STBY) – Default: 44 seconds
Force to Standby (FTS) – Default: N
GPRS Detach timer (DET) – Default: 00 hours – 00 minutes
Attach/ Detach MS Reachable Timer (MSRT) – Default: 120 minutes
MSRT, DET
READY Ready
Idle Timer Expiry
RDY
GPRS detach is either
STANDBY Packet MS, SGSN or HLR
Standby TX/RX
Timer Expiry initiated.
STBY
MS location known to
MS location MS location known to cell level.
not known. SGSN level. MS is transmitting or has
Subscriber is not MS can be paged for just been transmitting.
reachable by the data. MS is capable of receiving
GPRS NW. / transmitting data.

GPRS Mobility Management - Routing Areas
• Routing Areas are used for GPRS Mobility Management

• A RA is a subset of one, and only one, Location Area
• A RA is served by only one SGSN
• For simplicity, the LA and RA can be the same
• Combined RAU can be generated with NMO1 (Gs interface is needed)
Location
Area (LA) MSC/VLR
GS Interface
Periodic RA Update Timer (PRAU) – Default: 54

minutes Routing
RA Paging Area (RPA) – Default: 2 Area SGSN
RA Paging Repetition (RPR) – Default: 3.5
seconds (RA)

Session Management - Establishing a PDP
Context
• PDP Context (Packet Data Protocol): Network level
information which is used to bind a mobile station
(MS) to various PDP addresses and to unbind the
mobile station from these addresses after use
• PDP Context Activation
• Gets an IP address from the network
• Initiated by the MS
• Contains QoS and routing information enabling data
transfer between MS and GGSN
• PDP Context Activation and Deactivation should
occur within 2 seconds
nt e xt Request
PDP Co
.3 3.55
155.131

Data Transfer - Temporary Block Flow
A Temporary Block Flow (TBF):

is a one-way session for packet data transfer between MS and BSC (PCU)
uses either uplink or downlink but not both (except for associated signaling)
can use one or more TSLs
Compare with circuit-switched:
normally one connection uses both the uplink and the downlink timeslot(s)
for traffic
In two-way data transfer:
uplink and downlink data are sent in separate TBFs - as below
Uplink TBF (+ PACCH for downlink TBF)
BSC
Downlink TBF (+ PACCH for uplink TBF)
PACCH (Packet Associated Control Channel): Similar to GSM CSW SACCH

RR Operating Modes and MM States Relations
Non-DTM capable MS:

Packet Measurement
RR BSS No state No state
transfer mode report reception
Packet Packet
RR MS Packet idle mode
transfer mode idle mode
MM (NSS
Ready Standby
and MS)
DTM capable MS:

Measurement
RR BSS Dual Packet No state No state
report reception
transfer Dedicated transfer Dedicated
CS idle and
RR MS mode mode mode CS idle and packet idle mode
packet idle
GMM (NSS
Ready Standby
and MS)

Dimensioning and planning
Network Analysis and planning inputs
Deployment planning
Air Interface capacity planning (CDEF, CDED)
Connectivity capacity planning (BTS, EDAP, PCU, Gb, SGSN)

Network Analysis
Business Plan
The business plan shows the importance level of (E)GPRS services.
The networks need different setup in case of e.g. background service requirement
without guaranteed bit rate on selected BTSs only compared to the e.g. streaming
requirement for the whole network.
Therefore the network planning outcome should be in accordance with the business
plan.
Usually the business plan requirements are mainly related to:

• Number of users with traffic volume (with density map)
• Services with QoS both on CSW and PSW
• Controlled investment
All the expectations above must be fulfilled to operate successful (E)GPRS services.

Network Analysis
BSS Hardware and Software
Both the hardware types and software / feature releases used
by the operators must be analyzed to stress the importance of
limits in implementation.
BSS Hardware Considerations
 BSC types with PCU and PCU2 (if any), BTS types (with Baseband units),
TRX capability (mixture of GPRS and EGPRS TRXs)
GSM Feature Considerations
 MultiBCF and Common BCCH, RF and BB Hopping, FR/DR/HR and AMR,
IUO/IFH

Network Analysis
Air Interface
The coverage and interference The RLC/MAC TSL data rate used by
should be analyzed, because the the capacity dimensioning can be
TSL data rate is defined by coverage based on the graphs below:
and interference as well.
• The signal level and interference
RLC/MAC Data Rate (FTP Download on 2 TSLs)
level can be analyzed by:

120
100
– Planning tool plots 80

No Interference
– Drive test measurements C/I 25 dB
kbps
60
C/I 20 dB
– OSS measurements 40
C/I 15 dB
20
0
-65 -70 -75 -80 -85 -90 -95 -100 -105
Signal level (dBm)

Network Analysis
Configuration and Traffic Figures
The following items should be collected before capacity
dimensioning:
• Site configuration (segments with BTS, cells, TRXs)
• Existing traffic figures (CSW and PSW (if any) traffic volume
on BTS level with blocking expectations)
• Expected traffic figures with QoS (based on the operators'
business plan)

Deployment Plan
The aim behind the preparation of deployment plan:
•Maximize the TSL data rate (RLC/MAC) and multislot usage

•Minimize the impact of PSW services on CSW services (and vice versa)
•Take all the hardware and software considerations into account
•Controlled investment
Most of the networks can be described by few cell/segment options
The analysis of the different options can give exact picture about the network based on:
•Operators’ business plan
•Hardware types, software releases (features and parameters as well)
•Operators’ current network structure and functionality
•Coverage, quality and capacity characteristics of BSS

Deployment Plan
Cell / Segment Option Creation
Cell / segment option creation
•The options can cover most of the cell/segment configurations of the network
•These options can be analyzed in details, so the time consuming cell/segment based analysis is not
needed
•All the options are examples and can have different channel configuration
Cell / Segment option examples

Layer strategy BTSs TRXs TSL0 TSL1 TSL2 TSL3 TSL4 TSL5 TSL6 TSL7 PSW terr.
TRX1 BCCH TCH/D TCH/D TCH/D TCH/D TCH/D TCH/D TCH/D
Layer1 CSW, GPRS, EGPRS BTS1 TRX2 TCH/F TCH/F TCH/F TCH/F TCH/F TCH/F TCH/F TCH/F
Cell / Segment option 1
TRX3 TCH/F TCH/F TCH/F TCH/F TCH/F TCH/F Default Default (E)GPRS
Layer strategy BTSs TRXs TSL0 TSL1 TSL2 TSL3 TSL4 TSL5 TSL6 TSL7
TRX1 CBCCH SDCCH TCH/D TCH/D TCH/D TCH/D TCH/D TCH/D
Layer1 CSW, (E)GPRS BTS1
TRX2 TCH/F TCH/F TCH/F TCH/F Default Default Default Dedicated (E)GPRS
Layer2 CSW only BTS2 TRX3 TCH/F TCH/F TCH/F TCH/F TCH/F TCH/F TCH/F TCH/F
Layer strategy BTSs TRXs TSL0 TSL1 TSL2 TSL3 TSL4 TSL5 TSL6 TSL7
TRX1 CBCCH SDCCH TCH/D TCH/D TCH/D TCH/D TCH/D TCH/D
Layer1 CSW, GPRS BTS1
TRX2 TCH/F TCH/F TCH/F TCH/F TCH/F Default Default Dedicated GPRS
Layer2 CSW, EGPRS BTS2 TRX3 TCH/F TCH/F TCH/F TCH/F Default Default Default Dedicated EGPRS
Layer strategy BTSs TRXs TSL1 TSL2 TSL3 TSL4 TSL5 TSL6 TSL7 TSL8 PSW terr.
TRX1 CBCCH SDCCH TCH/F TCH/F Default Default Default Dedicated EGPRS
Layer1 CSW, EGPRS BTS1
TRX2 TCH/D TCH/D TCH/D TCH/D TCH/D TCH/D TCH/D TCH/D
TRX3 TCH/F TCH/F TCH/F TCH/F TCH/F TCH/F TCH/F TCH/F
Layer2 CSW, GPRS BTS2
TRX4 TCH/F TCH/F TCH/F TCH/F Default Default Default Dedicated GPRS

Air Interface Capacity
Available Capacity
The available air interface capacity can be calculated by the
following way:
Available air interface capacity for (E)GPRS (TSLs) = Amount_of_TRXs*8 -
signaling_RTSLs (KPI: ava_52* + ava_51) – Average CSW traffic (KPI: trf_202) -
free_RTSLs (Param.:CSD and CSU)
The air interface capacity (TSLs) is suitable, if it is >= 4,

because 4 TSLs are recommended for 4 TSL multislot usage
(CDEF=4 TSLs)
*CDED is not counted by ava_52, therefore ava_51 must be used if CDED is used

Air Interface Capacity
Available Capacity
If the Available air interface capacity for (E)GPRS is less than
four TSL, then the following changes can be made in the
configuration:
– Reduce the amount of signaling TSLs (signaling traffic volume or
congestion must be checked)
– Free TSLs can be reduced with CSD and CSU parameters (but the less
safety border can reduce QoS on CSW)
– Implement HR, AMR on CSW to reduce CSW TSL demand (HR TSLs
cannot be used for (E)GPRS)
– Recommend TRX expansion
Dedicated TSL(s) is (are) needed to provide:

– Continuous service for (E)GPRS even in very high CSW traffic
(minimum GoS) CDED = 1 TSL (KPI: ava_51)
– Continuous (E)GPRS availability if NMOI is used

Connectivity Capacity
CDEF Size
Define CDEF for all the BTSs in the BSC
– 4 RTSLs are recommended to provide the availability of 4 TSLs
capable MS without territory upgrade (Par.: CDEF)
– If there is low amount of PSW traffic on the cell, then the CDEF can
be less than 4 RTSLs (KPI: trf_74c/trf_213c, trf_167/trf_215a Par.:
CDEF)

Connectivity Capacity
DAP Size
Define the # of DAPs and DAP size for all the BCFs in the BSC
– The recommendation for DAP size (64 kbps TSL) is DAP size = CDEF
if one BTS/ EDAP configuration is used. If CDEF is less than 4, still 4
TSL DAP size is recommended.
– The recommendation for DAP size (64 kbps TSL) is DAP size =
average of CDEFs*1.3 if 2 BTS/ EDAP configuration is used. If average
CDEF*1.3 is less than 4, still 4 TSL DAP size is recommended.
– The recommendation for DAP size (64 kbps TSL) is DAP size =
average of CDEF*1.5 if 3 BTS/ EDAP configuration is used. If average
CDEF*1.5 is less than 4, still 4 TSL DAP size is recommended.
– If DAP size calculation gives less than 4 TSL DAP size the
recommendation is still 4 TSL DAP to support 4 RTSL capable MS with
MCS9 on all the RTSL
– Make the same calculations for all the required BCFs
– Make the same calculations for all the required BSCs

Connectivity Calculations
# of PCUs
Based on the CDEF and DAP size inputs the # of required
PCUs can be calculated
– Count the # of Segments, BTS objects, TRXs in the BSC (100 %
utilization is allowed)
– Define the EDAP strategy (one EDAP for the BCF, more than one
EDAP for a BCF)
– Calculate the # of EDAPs / PCU (it is recommended to have 1,2,4, or 8
EDAPs/ PCU1; 1-8 EDAPs with PCU2)
– Calculate the # of RTSLs and Abis sub-TSLs related to BSC (70 %
utilization is allowed)
– Calculate the # of logical and physical PCUs based on BSC type
– Make the same calculations for all the required BSCs

Connectivity in PCU
PCU Type BSC Types Network elements BSS10.5 BSS10.5 ED BSS11 BSS11.5
PCU BSCE, BSC2, BSCi, BSC2i BTS 64 64 64 64
TRX 128 128 128 128
Radio TSLs 256 256 256 128
Abis 16 kbps channels 256 256 256 256
Gb 64 kbps channels 31 31 31 31
PCU-S BSCE, BSC2, BSCi, BSC2i BTS 64 64 64 64
TRX 128 128 128 128
Radio TSLs 256 256 256 128
PCU-T BSCE, BSC2, BSCi, BSC2i BTS 64 64 64 64
TRX 128 128 128 128
Radio TSLs 256 256 256 256
PCU2-U BSCE, BSC2, BSCi, BSC2i BTS N/A N/A N/A 128
TRX N/A N/A N/A 256
Radio TSLs N/A N/A N/A 256
Abis 16 kbps channels N/A N/A N/A 256
Gb 64 kbps channels N/A N/A N/A 31
PCU-B BSC3i BTS 2 x 64 2 x 64 2 x 64 2 x 64
TRX 2 x 128 2 x 128 2 x 128 2 x 128
Radio TSLs 2 x 256 2 x 256 2 x 256 2 x 256
Abis 16 kbps channels 2 x 256 2 x 256 2 x 256 2 x 256
Gb 64 kbps channels 2 x 31 2 x 31 2 x 31 2 x 31
PCU2-D BSC3i BTS N/A N/A N/A 2 x 128
TRX N/A N/A N/A 2 x 256
Radio TSLs N/A N/A N/A 2 x 256
Abis 16 kbps channels N/A N/A N/A 2 x 256
Gb 64 kbps channels N/A N/A N/A 2 x 31

Gb over FR – Gb Size
• Calculate the Gb size based on # of EDAPs and size of
EDAPs connected to PCU
– The recommendation for Gb size is Min_Gb_1 = 1.25*Max_EDAP size
– The average traffic from several EDAP can be calculated with k-factor:
Min_Gb_2 = k * Average(EDAP_size_1 to EDAP_size_n)
– The k-factor is selected based on the estimate of the short term traffic
distribution among different EDAPs. If no specific information about the
distribution is available it is recommended to use the default values for
k.

Gb over FR – Gb Size
– The recommendation for Gb link size is Max(Min_Gb_1, Min_Gb_2)
– The final Gb link size might need to be adjusted to fit to available
E1/T1 links.
• Define the Gb size connected to all the PCUs in one BSC
• Define the Gb size for all the required BSCs

Optimization

(E)GPRS Network Optimization
Content
Assessment
• Configuration and feature assessment
• BSS (KPI) and E2E performance assessment
Signaling Capacity & Resource Allocation improvement

• Signaling
– RF, TRXSIG, MM and SM signaling
• Resource Allocation
– Cell (re)-selection, BTS selection, scheduling
Throughput optimization
• Connectivity Capacity (MS-SGSN)
– CDEF size, DAP size, # of PCUs and BCF allocation among PCUs, Gb size and PAPU (SGSN)
• TSL data rate improvement and multi-slot usage maximization (BSS)
– TSL utilization, TBF release delay and BS_CV_MAX, LA, UL PC, multiplexing, multislot usage
• E2E data rate (applications)
– Flow control, SGSN, TCP/IP, applications
Mobility improvement (outage reduction)

• PCU rebalancing
• LA/RA design
• NACC

GSM Network Optimization
The optimal GSM network from PSW services point of view has:
As high signal level as possible
• It means that even the indoor signal level should be high enough to have MCS9 for getting the highest
data rate on RLC/MAC layer.
As low interference as possible
• The aim of having high C/I is to avoid throughput reduction based on interference.
Enough capacity
• Enough BSS hardware capacity (interface and connectivity) is needed to provide the required capacity for
PSW services in time. Both CSW and PSW traffic management should be harmonized with the layer
structure and long term plans on signaling and traffic resources.
As few cell re-selection as possible
• The dominant cell coverage is important to avoid unnecessary cell-reselections in mobility. The prudent
PCU allocation can help to reduce the inter PCU cell reselections.
• Dominant cell structure can help to maximize the signal level and reduce the interference, too.
Features
• All the features should be used which can improve the PSW service coverage, capacity and quality in
general.
The GSM network is the physical layer of (E)GPRS, so the optimization of GSM network can improve the
performance of (E)GPRS, too. Pls. refer to GSM BSSPAR and RANOP

Assessment
Configuration and feature assessment
BSS (KPI) and E2E performance assessment

Configuration and Feature Assessment
Configuration Audit
• The configuration audit allows to have clear picture about the network
before optimization.
• The HW and SW releases must be known, as well as configuration and
capacity of air interface (deployment scenarios), BCF, EDAP, BSC/PCU,
Gb, SGSN/PAPU and HLR.
Parameter and Feature Audit

• The parameter and feature audit shows the parameters and features’
setup used by the operator.
• This setup must be collated with Nokia recommendations. The reason of
differences must be further investigated.
• The interworking of different features used by the operator must be
analyzed.
The OSS Radio Manager and MML commands are used to obtain the
configuration, parameter and feature information

Network Element and Configuration Assessment -
HLR
BSS
BTS
• QoS profile
• PCU variant & TC HLR/
• GPRS territory
dimensioning AC/ • GPRS settings
• BTS HW EIR
• PCU strategy
considerations in mixed
(TRX & BB- TCSM SGSN
configuration MSC/VLR
card) • Unit capacity
• BSS SW and (PAPU etc.)
• BTS SW
features
(EPCR) • BSS Gb Flow
MS/Client control
parameters
Gs
• GPRS/EDGE
capability and IP/MPLS/IPoATM
-
release backbone
GGSN
•Multislot BTS BSC 2G
support RF Abis Gb SGSN Gn
Gi
Applicatio
Application
RF interface Servers
n Servers
• Coverage (co-
Gb interface located
•C/I
Abis interface • Bearer size Server
• Capacity
• EDAP size / • IP v.s. FR • load
• Traffic volume
dimensioning • Dimensioning
• Mobility • settings
• # of E1/T1s (Linux/Win)
• GPRS/EDGE
traffic

BSS Feature Assessment
The following features must be checked and interworking issues must be

emphasized among features below:
– Idle mode settings

– BB and RF Hopping
– MultiBCF and Common BCCH
– FR / HR / DR / AMR
– HSCSD
– Extended Cell
– Intelligent Underlay Overlay
– Priority based QoS
– NMO1
– NCCR
– CS1-4
– NACC
– DFCA
– DTM

Performance Assessment
Performance measurements are needed to get clear picture

about network functionality and find the possibilities for
increasing the performance.
The (E)GPRS performance measurements and its analysis are

usually separated to
• OSS KPI Analysis
• Drive Test Analysis
• Real-time measurement analysis

Signaling Capacity
Accessibility

(E)GPRS – TCH Accessibility (Paging)
Introduction
Traffic volume and load NMOII
• CS pagings from Aif (c3000)… 358907057 (no Gs
interface)
• CS pagings from Gb (c3058)… 0
• PS pagings from Gb (c3057)… 1209319
c3050
• MAX_PAGING_BUFFER_CAPA (c3035)
Max Paging Buffer occupancy
• max_paging_gb_buf (c3050)
• Average free space of paging GSM buffer area – 12%
pgn_3a 10%
– AVE_PCH_LOAD (c3008) / 8%
RES_ACC_DENOM2 (c3009) Gb Pagings
6%
A Pagings
4%
Rejection
• delete_paging_command (c3038) (includes both 2%
PS and CS paging) 0% c3035

Op 1 Op 2 Op 3 Op 4 Op 5
Paging success ratio on PS and CS together

• delete_paging_command (c3038) /
(paging_msg_sent(c3000)+
cs_paging_msg_sent(c3058)+
ps_paging_msg_sent(c3057)) Example
Measurement results
from five different
networks
(E)GPRS – TCH Accessibility (Paging)
CS paging messages sent on PACCH – pgn_14a
• The ratio of CS paging messages sent on PACCH to all CS paging messages
• Object aggregation: BTS
Ratio of CS paging messages (via Gb) sent on PACCH,
• Time aggregation: hourly (BH) pgn_14
• Unit: %
0.030%
0.025%
0.020%
sum(b.pac_pag_req_for_cs_pag) 0.015%
100 * ------------------------------------------------------------------------------------------- 0.010%
sum(a.cs_paging_msg_sent + b.pac_pag_req_for_cs_pag) 0.005%
0.000%
Counters from table(s):
a = p_nbsc_res_access
b = p_nbsc_packet_control_unit
• Shows the benefit of using NMO I, as those pagings would have been lost if NMO II was
used.
• Formula will show zero if NMO II is used
GSM CS
voice calls A MSC
CCCH
BSC Paging
OR PACCH Gs Co-
ordination
via Gs-
Gb interface
GPRS/EGPRS 2G
packet data SGSN

(E)GPRS – TCH Accessibility (Imm. Assignment)
Introduction
Traffic volume on AGCH with rejection can MS BSC / PCU
be analyzed by the following KPIs: Cha nne l Req ue st – sin gle b lo ck pa cke t a cce ss
• Imm.assign (c3001) 4458795 Im m ediate Assignm e nt (UL assignm en t)
• Imm.assign rejected (c3002) 743 Packet Resource Re qu est + TLLI
• The immediate_assignment_reject message is sent to Pa cke t Uplin k Assignm en t + TLLI
MS in the following cases:

UL da ta block
• no free SDCCH or UL TBF re a dy
…
• Abis interface process has no internal resources to handle the
Packe t UL ACK/NACK
request.
• P-Imm.assign (c72084) 1411131

Packet Con trol ACK
UL Da ta Block
• P-Imm.assign rejected (c72087) 31
…
• When a PACKET_IMMEDIATE_ASSIGNMENT_REJECT
message is sent to the MS. This occurs when there are
no TCHs to be given to the MS in uplink TBF assignment.
• blck_13

(E)GPRS – TCH Accessibility (Imm. Assignment)
NACK ratio of p-imm assignment req – blck_21b
• NACK ratio of p-immediate assignment request
MS BSC / PCU
Ch an n el Requ est - on e pha se acce ss

• Time aggregation: hourly (pgn_13 BH) P-Im m ed iate Assign m en t (UL a ssignm e nt)
• Unit: % UL Data Blo ck + TLLI

Decision –
on e -ph ase vs.
tw o-ph ase
Packet UL ACK/NACK + TLLI
Packet Con trol ACK or

sum(packet_immed_ass_nack_msg) UL TBF rea dy
UL Data Block w /o TLLI
100 * ----------------------------------------------------------------------------------------
sum(packet_immed_ass_ack_msg + packet_immed_ass_nack_msg)
…
p_nbsc_packet_control_unit
• In case of 2-phase access, the immediate assignment for the first block is not ack’ed, so this
formula only measures the performance of the 1-phase access on the CCCH.
• A negative acknowledgement (NACK) is sent from BTS to BSC after all AGCH messages
which are deleted from TRX buffers due to - buffer overflow - maximum lead-time expiry -
expired starting time, and which are ordered by BSC to be acknowledged.
• The negative acknowledgement is sent immediately after the message has been deleted.

(E)GPRS – TCH Accessibility (DL CCCH usage)
PCH and AGCH Summary
DL CCCH Usage Traffic
100%
100% 80%
60%
80% Packet Imm Ass Reject
PS Erlangs
CS Erlangs
40%
Packet Imm Ass
60% Immediate Assignment Reject 20%
Immediate Assignment 0%
40% PS paging via Gb
CS paging via Gb
20% CS paging via A
0%
• Operator 3 & 4 has NMO = I

• PS generate many more Imm Ass per Erlang than CS
• Majority of DL CCCH load is caused by CS paging => If congestion, start to
optimize the CS part (VLR parameters, LA split). Use combined signaling
configuration

(E)GPRS – TCH Accessibility (RACH)
Average RACH load – rach_4
Average RACH load rach_4

PERIOD_START_TIME
PERIOD_STOP_TIME
• Indicates how high the RACH load is on average
• Time aggregation: hourly (BH)
• Unit: %
avg(ave_rach_busy/res_acc_denom3) 9/4/2007 9/4/2007 1:00:00 AM 0.5331%
100 * --------------------------------- 9/4/2007 1:00:00 AM 9/4/2007 2:00:00 AM 0.4194%
avg(ave_rach_slot/res_acc_denom1) 9/4/2007 2:00:00 AM 9/4/2007 3:00:00 AM 0.3771%
9/4/2007 3:00:00 AM 9/4/2007 4:00:00 AM 0.3523%
9/4/2007 4:00:00 AM 9/4/2007 5:00:00 AM 0.4055%
9/4/2007 5:00:00 AM 9/4/2007 6:00:00 AM 0.4365%
9/4/2007 6:00:00 AM 9/4/2007 7:00:00 AM 0.5224%
p_nbsc_res_access
9/4/2007 7:00:00 AM 9/4/2007 8:00:00 AM 0.7907%
9/4/2007 8:00:00 AM 9/4/2007 9:00:00 AM 0.9131%
9/4/2007 9:00:00 AM 9/4/2007 10:00:00 AM 0.8622%
• If the value is to the order of tens of per cent there 9/4/2007 10:00:00 AM 9/4/2007 11:00:00 AM 0.9091%
9/4/2007 11:00:00 AM 9/4/2007 12:00:00 PM 0.9385%
probably are access problems and MS users get, 9/4/2007 12:00:00 PM 9/4/2007 1:00:00 PM 0.9839%
more often than usual, 3 beeps when trying to 9/4/2007 1:00:00 PM 9/4/2007 2:00:00 PM 1.0063%
start calls. A probable reason is UL interference. 9/4/2007 2:00:00 PM 9/4/2007 3:00:00 PM 1.0223%
9/4/2007 3:00:00 PM 9/4/2007 4:00:00 PM 1.0618%
• The table shows a cell with heavy loaded 9/4/2007 4:00:00 PM 9/4/2007 5:00:00 PM 1.0933%
signaling on DL. The parallel traffic is not 9/4/2007 5:00:00 PM
9/4/2007 6:00:00 PM
9/4/2007 6:00:00 PM
9/4/2007 7:00:00 PM
1.0061%
0.8851%
high on UL (RACH) 9/4/2007 7:00:00 PM 9/4/2007 8:00:00 PM 0.8690%
9/4/2007 8:00:00 PM 9/4/2007 9:00:00 PM 0.9379%
9/4/2007 9:00:00 PM 9/4/2007 10:00:00 PM 0.7677%
9/4/2007 10:00:00 PM 9/4/2007 11:00:00 PM 0.6707%
9/4/2007 11:00:00 PM 9/5/2007 0.6743%

(E)GPRS – TCH Accessibility (RACH)
Ratio of PS RACH retries – rach_9
• Ratio of PS Channel Requests that the MS needs to RACH performance
repeat at least once
3.5%
• Object aggregation: BTS 3.0%
• Time aggregation: hourly (BH) 2.5%
• Unit: % 2.0% Average RACH load

1.5% Ratio of PS RACH retries
sum(ul_tbf_with_retry_bit_set) 1.0%
100 * ------------------------------ 0.5%
sum(packet_ch_req)
0.0%
p_nbsc_packet_control_unit RACH performance (Op 4)
14%
• There can be several reasons why a packet channel 12%
request needs to be repeated: 10%
– The first request collided with a request from another MS 8%
rach_9
in the same TDMA frame (ie a congestion issue) 6%
– The UL radio link is not good enough, so the BTS does 4%
not hear the request 2%
– The DL radio link is not good enough. The BTS hears the 0%
channel request, forwards it to the PCU, and the PCU 0% 2% 4% 6% 8% 10%
allocates a PDTCH, but the MS cannot hear the Packet rach_4
Immediate Assignment

(E)GPRS – TCH Accessibility (SDCCH)
• SDCCH is used by (E)GPRS, if NMOII is implemented. SDCCH should not be analyzed, if

NMOI is used (because of combined RAU).
• SDCCH analysis
– Average SDCCH traffic - trf_168
– Average SDCCH traffic - trf_11b
– SDCCH access probability, before FCS - csf_1
– SDCCH congestion time - cngt_2
– SDCCH real blocking - blck_5a
– SDCCH drop ratio without timer T3101 expiry % - sdr_4
– Ghosts detected on SDCCH and other failures - sd_1b
– SDCCH Drop % - sdr_1a
• Solution for reducing SDCCH load

– Increase of Periodic RA update timer (PRAU) / MS Reachable timer (MSRT)
 Drawback: more PAPU capacity is needed and more paging will be generated, if the MS is out of service
– More SDCCH TSL allocation and/or Dynamic SDCCH feature usage
– FACH call setup for compensating SDCCH blocking
– Combined RAU (NMO-I with Gs for (E)GPRS)
– (Resume feature decreases the amount of RAUs)
– LA and border re-planning

(E)GPRS – TCH Accessibility (TBF for signaling)
• DL/UL signaling TBFs usage ratio Ratio of signaling TBFs

100%
• Unit: % 80%
60%
DL
DL signaling TBFs usage ratio UL
40%
dl_tbf_for_signalling
tbf_62 = 100% * -------------------------------------------- 20%
dl_tbf_for_signalling + dl_tbf_for_data
0%
UL signaling TBFs usage ratio
Ratio of signaling TBFs (Thu, Fri, Sat)
ul_tbf_for_signalling
tbf_61 = 100% * --------------------------------------------
ul_tbf_for_signalling + ul_tbf_for_data 100%
Counters from table(s): 80%
p_nbsc_packet_control_unit 60%
tbf_61
tbf_62
40%
• Counters are measuring TBFs, not payload!
• UL formula probably gives more realistic view 20%
of usage 0%

TRXSIG Load
TRXSIG load analysis

Data collection
Processing (data conversion, data shorting)
Analysis with preparation of yellow and red list
• Unavailability
• Load (Layer1, layer3, alarms)
• Quality (bad physical channels, errors)
• Delays (RTT)
• Configuration (signaling and traffic structure on Air Interface)
Recommendation: Solutions for reducing TRXSIG load:

Reducing the Air Interface load
TRXSIG capacity setting modification proposals (case by case proposal)
Abis Signaling Link related links on IMS:

https://sharenet-ims.inside.nokiasiemensnetworks.com/livelink/livelink?func=ll&objId=368213370&objAction=Browse&viewType=1
https://sharenet-ims.inside.nokiasiemensnetworks.com/Open/368227253
https://sharenet-ims.inside.nokiasiemensnetworks.com/Open/368221050

Resource Allocation
PSW activation and territory settings
Cell selection
BTS selection
TSL selection (scheduling)

Resource Allocation
The aim of resource allocation improvement is to ensure not only the fast
access to the network but also the access to the right cells and TSLs, which
provide the fastest data rate.
Usually the following items should be checked in BSS resource allocation

improvement:
• PSW activation and territory settings
• Cell selection
• BTS selection
• TSL selection (scheduling)

Resource Allocation - Introduction
RLC/MAC
Basic Allocation related Topics Provide enough capacity to PSW
• PSW Activation traffic in general (find balance
• Territory settings between CSW and PSW)
• Channel pref.
Cell-(re)selection
• C1 Allocate the traffic to the most
• C2 appropriate resource
• NCCR
BTS Selection
Separate GPRS and EGPRS
• MultiBCF and CBCCH
and share the resources
• PCU algorithm
User
Channel Scheduling Service and user prioritization
• Priority based QoS

PSW Activation and Territory Setting
The TRX, cell, segment is used for:

TRX 1 BCCH SDCCH TS TS TS TS TS TS
• Signaling
• CSW traffic TRX 2 TS TS TS T S TS TS TS
TS
• Free TSLs
• PSW traffic = Free TSL for CSW
TS = (E)GPRS Traffic TS
TS = CSW Traffic
The size of territory for PSW traffic is BCCH = Signaling
defined by:
• Dedicated territory (CDED)
• Default Territory (CDEF)
• Max Territory (CMAX)

GPRS Territory
• Circuit Switched traffic has priority outside dedicated territory

• GPRS dedicated time slots (% of total cell capacity) can be defined. Only
(E)GPRS can use, no CSW
• Dedicated TSL is subset of Default TSL
• Territories consists of consecutive timeslots
• GPRS can be set to favour the BCCH Transceiver -> minimum interference
Extra GPRS capacity Default GPRS capacity threshold
Circuit
Switched
TRX 1 BCCH TS TS TS TS TS TS TS
Territory
TRX 2 TS TS TS TS TS TS TS TS Circuit /
Packet
Free time slots in Circuit Switched
Switched territory Territory
Default GPRS Dedicated

Territory upgrade in interval of GPRS
Territory Upgrade Guard Time Capacity (%)
Capacity (%)
Territory downgrade
forced by the Circuit
Switched traffic
PSW Activation, Territory and Allocation -
Parameters
PSW Activation Channel Allocation Parameters
GPRS Enabled (GENA) Prefer BCCH frequency GPRS (BFG)
EGPRS Enabled (EGENA) TRX priority in TCH allocation (TRP)
GPRS Enabled TRX (GTRX)
Adjacent GPRS Enabled (AGENA)
GPRS Cell Barred (GBAR)
Not Allowed Access Classes (ACC)
Territory Settings
Default GPRS Capacity (CDEF)
Dedicated GPRS Capacity (CDED)
MAX GPRS Capacity (CMAX)

Resource Allocation – PSW Activation and
Territory
TRP and BFG in S10.5
BFG status TRP status PSW CSW BFG status TRP status PSW CSW
prioritization prioritization prioritization prioritization
Y 0 no pri no pri N 0 no pri no pri
Y 1 BCCH BCCH N 1 Beyond BCCH BCCH
Y 2 Beyond BCCH beyond BCCH N 2 BCCH beyond BCCH
TRP and BFG in S11 (onwards)

TRP Effect BFG Effect
0 All TRXs are treated equally in TCH All TRXs are treated equally in TRX
allocation 0
(default) selection for GPRS use
(default)
1 A traffic channel is allocated prim arily

from the BCCH TRX 1 GPRS channels are allocated prim a rily
from the BCCH TRX

beyond BCCH TRX GPRS channels are allocated prim a rily
2
beyond the BCCH TRX

(n ew value) from the BCCH TRX for the non-AMR
users and for the AMR users beyond
the BCCH TRX

(E)GPRS – TCH Availability
• TCH availability measurements are needed to know whether the proper

capacity is available for PS services
• The proper setup of default and dedicated territory size can be analyzed
by these measurements, too
• It can be also used to find faulty or locked TRXs in the network
Dedic tsl
for PS
ave_non_avail_TCH ava_51
Unavail. Avail. TCH tsl (CS) Avail tsl for PS territory
TCH tsl ave_avail_TCH_sum/ave_avail_TCH_den ave_GPRS_channels_sum/
Area: uav_15a ava_52 ave_GPRS_channels_den
BTS: uav_13 ava_44
FTCH DualTCH HTCH
ava_32 ava_34 ava_30 GPRS territory can be established on TSLs that
are of type FCTH or DTCH. TSLs are taken to GPRS
territory in order from the end of TRX.
Signaling RTSLs
TCH RTSLs
Average PS territory – ava_44
• Average available (E)GPRS Channels (territory)
• Unit: TSL
sum(decode(trx_type,0,ave_GPRS_channels_sum))
------------------------------------------------------------------------------------------------------------------------------------------------------------------------
avg(decode(trx_type,0, decode(ave_GPRS_channels_sum,0,0,ave_GPRS_channels_den))) *count(distinct period_start_time)
where ave_GPRS_channels_sum > 0

p_nbsc_res_avail
• This KPI can be used for tuning the CDEF parameter together with CS territory utilization KPIs (e.g.
trf_1d, trf_365) and territory upgrade/downgrade related KPIs (e.g. blck_30, c1179)
• If the KPI shows less values than CDEF parameter setup, then PS territory is suffering from high CS
traffic volume
• If the KPI shows higher values than CDEF setup, then recommended to increase CDEF setup to avoid
frequent territory upgrade
• Trx_type,0, means that the TRX is a normal TRX

Average avail. dedicated GPRS channels – ava_51
• Average number of channels available only for CDEF
territory in
dedicated PS ((E)GPRS) traffic CDED
# of TSL territory in
• Object aggregation: BTS (BH) # of TSL
• Time aggregation: daily
• Unit: TSL
BTS ID ava_44 ava_51
sum(decode(trx_type,0,ave_permanent_GPRS_ch_sum))
--------------------------------------------------------------------------------------------------- 34 13.5 2.0
(avg(decode(trx_type,0,
decode(ave_permanent_GPRS_ch_sum,0,0,ave_permanent_GPRS_ch_de 85 9.0 2.0
n)))* count(distinct period_start_time)
55 9.0 2.0
where ave_permanent_GPRS_ch_sum > 0
135 8.8 2.0
Counters from table(s): 317 8.1 4.0
p_nbsc_res_avail 125 8.0 2.0

14 8.0 2.0
• This capacity is allocated by setting the parameter 366 8.0 4.0
CDED
308 7.9 4.0
• In the case that there are no dedicated GPRS
channels, throughput is not guaranteed, if CS 350 7.8 4.0
traffic needs all the capacity 18 7.7 2.0
• This KPI should be analyzed together with CS call
blocking (e.g. blck_8i) to avoid CS blocking due to 111 7.5 0.0
improper CDED setup

Cell (re) – selection with C1
C1 is used as a minimum signal level criterion for cell re-selection for GPRS in the same way
as for GSM Idle mode, but with specific GPRS parameters. (3GPP 05.08)
C1 = (A ‑ Max(B,0))
A = RLA_P - GPRS_RXLEV_ACCESS_MIN
B = GPRS_MS_TXPWR_MAX_CCH - P
RLA_P = received level average for each of the carriers in BA(GPRS)

GPRS_RXLEV_ACCESS_MIN = minimum received signal level at the MS required for access to the system
GPRS_MS_TXPWR_MAX_CCH = the maximum TX power level an MS may use when accessing the system
P = maximum RF output power of the MS
 GPRS_RXLEV_ACCESS_MIN & GPRS_MS_TXPWR_MAX_CCH parameters for

both serving and neighbouring cells are broadcast on PBCCH

Cell (re) – selection with C2
C1 "A"=30
C1 "B"=25
C1 "C"=5
C1 "D"=50
Macro900 “C”
Macro1800 “B”
C2 = C1 + cellReselectOffset - temporaryOffset*H(penaltyTime-T)
C2 "A"=30 + 0 serving cell H(x) = 0
C2 "B"=25 +20-30 * H(20-T)
Micro900 “D”
C2 "C"=5 + 0 - 0 * H(0-T)
C2 "D"=50+0-30*H(40-T)
Macro900 “A"
Time T: between 0 and 19 Time T: between 20 and 39 Time T: greater than 40

seconds seconds seconds
C2 "A"=30 C2 "A"=30 C2 "A"=30
C2 "B"=15 C2 "B"=45 C2 "B"=45
C2 "C"=5 C2 "C"=5 C2 "C"=5
C2 "D"=20 C2 "D"=20 C2 "D"=50
Cell (re) – selection with NCCR
NCCR (Network Controlled Cell Reselection) enables the network to control the
resource allocation when the MS performs the cell reselection.
NCCR is an optional feature in Nokia BSC. Operator can enable/disable the feature
on BSC level.
Network Control Mode (NCM) defines how cell re-selection is performed:
• Network Control Mode = 0 (NC0): the MS will perform an autonomous cell reselection.
• Network Control Mode = 2 (NC2): the MS sends neighbour cell measurements to the network and the network commands
the MS to perform cell re-selection.
• NCM is modified with MML command ZEEM.
NCCR does not affect cell reselection delays
Benefits with NCCR:

• Efficient allocation of resources
– GPRS MS can be prevented to access a EDGE TRX => better throughput for EDGE MS.
– EDGE MS can be held on EDGE TRX longer => better throughput for EDGE MS.
• More counter statistics
• Better control over mobile stations

Cell (re) – selection with NCCR - Criteria
Cell reselection criteria:

• Power budget will push EGPRS capable MSs to EGPRS cells and non-
EGPRS capable MSs to non-EGPRS capable cells.
• Quality Control trigger NCCR when the serving cell transmission quality
drops even if the serving cell signal level is good.
• Coverage based ISNCCR selects 3G network as soon as it is available or
when GSM coverage ends, depending on operator choice.
• Service based ISNCCR selects 3G network according to SGSN Service
UTRAN CCO BSSGP procedure if the serving cell signal level is good.
Feature candidate!
The NCCR functionality is based on:

• Messages
• Measurements
• Parameters

Commanding MS to network control (NC2 mode)
• NC mode is always included in SI messages

• The broadcast value is defined by parameter Broadcast network control mode
• When NC2 is broadcast, all MSs in MM Ready state will be under network control
• When NC0 is broadcast, MSs are commanded to NC2 mode by PACKET MEASUREMENT
ORDER message depending on the MS 3GPP release support (found from MS RAC) and
parameter NCCR control mode (NCM) setting (NCM->3)
– TN_BSC_SW_2006_839
• Network control includes whole MM ready state (MS NCCR context created)
Ne tw ork Control
TBF TBF
rea dy tim er tim e
RR Pa cket Tra nsfe r m ode RR Pa cket Idle m ode
• Overload Control: Number of controlled MSs (MS NCCR contexts) is limited by internal
PRFILE parameter Max number of MS NCCR contexts
– PACKET MEASUREMENT ORDER commanding extra MSs to NC0 mode is needed to keep service
for all MSs

Commanding MS to NC2 when NC0 is broadcast
New TBF
PCU orders MS to
MS RAC No
DL-UNITDATA
send measurement
received ?
during TBF reports by PACKET
Yes MEASUREMENT
Is there ORDER message on
No
MS RAC
included ?
PACCH.
Yes
Is PMO(NC2) needed based No

on mobile’s 3GPP release? STOP
Yes
Send PMO(NC2) measurement report
from MS
Execute NCCR algorithms for the MS

Receiving measurement reports
Serving and neighbor cell – MM Ready State
MS in RR Packet Idle mode
(on (P)CCCH)
MS PCU MS having downlink TBF Reporting periods:

(and no uplink TBF)
•NCCR idle mode reporting
PACKET CHANNEL REQUEST
MS PCU period (RR Packet Idle mode)
PACKET UPLINK ASSIGNMENT data block (NIRP)
PACKET ENHANCED MEASUREMENT REPORT data block (p) •Default: 15.36 s
PACKET DOWNLINK ACK/NACK
data block
•NCCR transfer mode
MS in having uplink TBF reporting period (RR Packet
data block
MS PCU data block (p)
Transfer mode) (NTRP)
data block PACKET ENHANCED MEASUREMENT REPORT
•Default: 0.48 s
data block (p)
data block PACKET DOWNLINK ACK/NACK
•Scale Ord (SCO)
PACKET ENHANCED MEASUREMENT REPORT
•Default: no (0)
data block

Sending PCCO, RR packet transfer mode
operation
• PCU orders MS to change cell by PACKET CELL CHANGE ORDER message on PACCH
when TBF exist.
– NACC procedure: PCU shall send target cell SI messages before sending PACKET CELL CHANGE
ORDER
– NACC procedure is done if:
• The MS supports NACC and
• The target cell belongs to the same BSC as source cell and
• The target cell is not 3G cell and
• NACC feature has been activated.
• After sending PACKET CELL CHANGE ORDER message to MS on PACCH, the MS is

polled for a PACKET CONTROL ACKNOWLEDGEMENT.
• After receiving PACKET CONTROL ACKNOWLEDGEMENT to PACKET CELL CHANGE
ORDER the PCU starts PCCO_timer (10s) to monitor if the MS leaves the cell
• During the timer the PCU establishes new uplink and downlink TBFs for the MS as normal.
– If the PCCO_timer expires then either MS has not obeyed PACKET CELL CHANGE ORDER or MS
has not received the message at all, or PCU has not received FLUSH-LL or PACKET CELL
CHANGE FAILURE message.
• After MS has camped on the new cell, the PCU receives FLUSH from the SGSN.

Sending PCCO, RR packet idle mode operation
• PCU orders MS to change cell by PACKET CELL CHANGE ORDER message.

– BSC sends IMMEDIATE ASSIGNMENT on CCCH assigning the downlink block to the MS.
– After sending PACKET CELL CHANGE ORDER message to MS on PACCH, the MS is not polled for
a PACKET CONTROL ACKNOWLEDGEMENT.
– PCU does not perform NACC because NACC is not possible in RR packet idle mode.
• PCU starts an PCCO_timer to monitor that the MS leaves the cell. During the timer the PCU
establishes new uplink and downlink TBFs for the MS as normal.
• After MS has camped on the new cell, the PCU receives FLUSH from the SGSN

NCCR Parameters – breakdown into algorithms
NCCR applicability NCCR criteria: Power budget NCCR criterion

• NCCR control mode (NCM) • NCCR EGPRS PBGT margin (EPM)
• WCDMA FDD NCCR enabled (WFNE) • NCCR GPRS PBGT margin (GPM)
MS reporting • NCCR other PCU cell offset (NOPO)
• NCCR idle mode reporting period (NIRP) • HCS signal level threshold (HCS)
• NCCR transfer mode reporting period (NTRP) • Priority class (PRC)
Averaging • GPRS temporary offset (GTEO)
• NCCR rxlev transfer mode window size (NRTW) • GPRS penalty time (GPET)
(default: 5)
• NCCR rxlev idle mode window size (NRIW)
(default: 5) NCCR criteria: QC triggered NCCR
• NCCR number of zero results (NNZR) (default: 2) • maximum BLER in acknowledged mode (BLA)
Return to old cell and penalty • maximum BLER in unacknowledged mode (BLU)
• NCCR return to old cell time (NOCT) • QC GPRS DL/UL RLC Ack Throughput Threshold
(default: 10s) (QGDRT, QGURT)
• NCCR target cell penalty time (NTPT) • QC EGPRS DL/UL RLC Ack Throughput
Threshold (QEDRT, QEURT)
(default: 10s) • QC reallocation action trigger threshold (QCATR)
• NCCR neighbor cell penalty (NNCP) • QC NCCR action trigger threshold (QCATN)
(default: 6s)
NCCR criteria: Service priority based ISNCCR
• ISNCCR FDD quality threshold (WFNP)
• ISNCCR FDD quality threshold (FQT)

Power Budget based NCCR
The averaged serving and neighbor cell RX LEV values are main input to Power
budget algorithm.
• Minimum path loss must be provided
• C31 and C32 related parameters are used for HCS in S11.5 onwards even if
PBCCH is not configured to a segment
• Can be used to force EGPRS MSs to reselect EGPRS cells and keep the GPRS
MSs in non EGPRS capable cells
• Can be used to select cells based on their CS3/4 capabilities
Due to separate thresholds for EGPRS capable and non-capable MSs, this criterion
cannot be used before the MS EGPRS capability is known.
GSM capable cell EDGE capable cell

E.g.
Average RXLEVfrom neigbouring cell

Average RXLEVfrom serving cell
Power
Negative
budget
margin
power
budget
margin
value

B A Distance
NCCR Power Budget Criterion GTEO and GPET
parameters.
The power budget criterion is evaluated each time the BSC receives a
measurement report from the MS (included in the HCS
The blue boxes explain how the parameters relate to the power budget and the PBGT(n)
criterion. formula)
Prevent moving MS
Measurement report processing unnecessary CR to
EPM and GPM cells they only briefly
parameters. pass by giving them a
temporary offset for a
Possible to allocate Compute HCS(n) & PBGT(n) values for all the
certain time.
reported neighbor cells
EGPRS and GPRS MSs
Select the cell that has the
on different cells. highest PBGT(n) value among
those cells that have the highest
PRIORITY_CLASS among those YES
Is HCS(n)  0 true for some
cells that fulfill HCS>=0 reported neighboring cell?
HCS threshold
Priority Class (PRC)
parameter.
parameter. NO
Search for the reported neighboring cell with A cell cannot be

Favour cells by highest PBGT(n) value
chosen as long as a
prioritizing them.
certain RXLEVEL,
NO
AV_RXLEV_NCELL(n) > defined by the HCS
Eg. Picocell= high RxLevMinCell
(n) + Max(0, aP ) NO
AV_RXLEV_NCELL(n) > threshold, is not
Priority RxLevMinCell (n) + Max(0, Pa) fulfilled.
Macrocell= low Priority
YES YES If no cell fulfill
HCS0 then all cells
NO
YES PBGT(n) > are included in the
Trigger NCCR to cell (n)
CellReselMarginPGBTforEG comparison.
PRSMS(n)
NO Erase the cell where the failure occurred from

NCCR successful? target candidate cell list until timer
T_NCELL_PENALTY* expires
YES

*NCCR neighbor cell penalty (NNCP)
Process for power budget NCCR criterion
• PBGT(n) = (AV_RXLEV_NCELL(n) - AV_RXLEV_SERV ) + (MS_TXPWR_MAX_CCH -

MS_TXPWR_MAX_CCH (n) ) - TO(n) * (1 - L(n))
– TO(n) = TEMPORARY_OFFSET_NCCR(n) * H(PENALTY_TIME_NCCRS(n) - T(n))
– L(n) = 0 if PRIORITY_CLASS(n) = PRIORITY_CLASS(s) , and 1 if PRIORITY_CLASS(n) <>
PRIORITY_CLASS(s)
• HCS(n) = AV_RXLEV_NCELL(n) - HCS_THR(n) - TO(n) * L(n)
• PRIORITY_CLASS defines the HCS (hierarchical cell structures) priority for the cells. 0 is the
lowest and 7 is the highest priority
• AV_RXLEV_NCELL(n) > RxLevMinCell(n) + Max(0, Pa)

– Pa = GPRS_MS_TXPWR_MAX_CCH(n) – P
– P is the MS maximum GMSK RF output power on the band of the cell n.
• PBGT(n) > PBGT_margin(n)

Power Budget – Example

EGPRS MS moving from serving GSM
cell to neighbouring EDGE cell and vice
Average RXLEV from EDGE cell

Average RXLEV from GSM cell
versa
Power
budget
margin
B A Distance between cells
Negative power budget margin value:

 Network will order the EGPRS MS to select neighbouring cell around the location A.
=> EGPRS MS would be pushed to EDGE capable cell.
 This is done although the average received signal level from the EDGE capable cell is lower
than on from the regular cell.
 In opposite direction: EDGE MS moves away from EDGE capable cell and approach to non
EDGE capable cell the cell-reselection would be triggered in location B.
=> EGPRS MS is held longer on EDGE capable cell.

Power Budget – Example

GPRS MS moving from serving GSM
cell to neighbouring EDGE cell and vice
Average RXLEV from EDGE cell

Average RXLEV from GSM cell
versa
Power
budget
margin
Distance between cells B A

Positive power budget margin value:
 GPRS MSs are kept in regular cells as long as possible to reserve EDGE resources for
EGPRS MSs.
 Network will order the GPRS MS to select neighbouring cell around the location A.
=> GPRS MS is kept in the GSM cell longer.
 When GPRS MS move away from EDGE capable cells and approach to non-EDGE-capable
cells the cell-reselection is triggered in location B.

NCCR Reference Measurements
• Idle Mode Serving Cell Areas • Serving cell

RXLEVEL
GPRS
BTS EDGE
BTS
EDGE

NCCR EGPRS PBGT margin (EPM) Results
Packet Transfer Mode Serving Cell Areas

Default values GPRS
Offset 6 dB when going from
EDGE to GPRS cell
GPRS to EDGE cell EDGE
BTS
Result: Cell reselection

locations same as in Idle
Mode.
EDGE

Packet Transfer Mode

Cell border
EPM = 12 dB when going from EDGE GPRS with default
to GPRS cell values
EPM = 0 dB when going from GPRS to

EDGE cell
BTS EDGE
Result:
EDGE MS stay slightly longer on EDGE
cells and re-enters EDGE slightly
Cell border
earlier. with default EDGE
values


GPRS
EDGE to GPRS cell Cell border
Offset -6 dB when going from GPRS with default
values
to EDGE cell
EDGE
Result: EDGE MS stay longer on BTS
EDGE cell and re-enters EDGE cell
earlier.
Cell border
with default
values
EDGE


Offset 24 dB when going from EDGE Cell border
to GPRS cell with default
values
Offset -12 dB when going from GPRS
to EDGE cell
EDGE
Result: EDGE MS does not enter BTS
GPRS cell.
Cell border
with default
values EDGE

Packet Idle Mode Serving Cell Areas

Offset 24dB when going from Cell border
EDGE to GPRS cell with default
values
Offset -12 dB when going from
GPRS to EDGE cell
EDGE
Result: NCCR functions also
BTS
properly in Packet Idle Mode. EDGE
MS does not enter GPRS cell.
Measurement reporting period:

• Packet Idle Mode = 3,84s Cell border
with default
• Packet Transfer Mode = 0,48s values EDGE
As long as MM Ready Timer has not
expired (44s default) MS will stay in
Packet Idle Mode.
When timer expires MS will move to Idle
Mode
=> MS is not longer controllable by Different Measurement reporting periods =>
NCCR Small difference in cell borders between Packet Idle and Packet
transfer mode. (compare with previous slide)

HCS Signal Level Threshold Results
Ping pong
Packet Transfer Mode Ping pong
• GPRS cell: HCS threshold = -60dBm
• EDGE cell: HCS threshold = -110dBm
Ping pong GPRS
Result:
EDGE
BTS
Use of the HCS threshold can
significantly increase the Ping-pong-
effect and decrease the network
performance.
Ping-pong occurs when the RXLEVEL

fluctuates around the HCS threshold => EDGE
cell is included in the NCCR calculations
on and off.
HCS threshold should not be used

together with the PRC parameter.

Priority Class (PRC) Results

• GPRS cell: PRC = 1
• EDGE cells: PRC = 7
EDGE
With Priority Class (PRC) parameter one can
prioritize one cell over another. The
parameter is used in cells which fulfil the HCS BTS
statement  0.
• Can e.g. be used to prioritize Pico- and
Micro cells. (indoor sites)
Result: GPRS cell is not selected since its

priority is lower than the EDGE cells.
Prioritization should be used very carefully EDGE
and only in cells with sharp borders (in
practice indoor sites).
Prioritizing can increase ping-pong and
decrease the network performance if wrongly
used.

GTEO and GPET Results
No Penalty Penalty
GPRS temporary offset (GTEO)
• Pico cell = 70 dB
GPRS penalty time (GPET)
• Pico cell = 120 s
Result:
Pico cell Pico cell
Pico cell is not selected during
the Penalty timer.
Pico cell is not
selected during the
GPET and GTEO should mainly Penalty timer.
be used for indoor sites to avoid => Unnecessary cell
unnecessary cell reselections reselection to Pico
from moving MSs to these sites. cell is avoided.

NCCR Power Budget Summary
With NCCR Power Budget parameters it is possible to:

• push EGPRS capable MSs to EGPRS cells
• push non-EGPRS capable MSs to non-EGPRS capable cells
• delay MSs entrance into a cell
• avoid moving MSs unnecessary entrance into cells they only briefly pass
NCCR EGPRS PBGT margin (EPM) and NCCR GPRS PBGT margin (GPM) can be used to
effectively allocate EDGE and GPRS capable MSs on different cells.
GPRS temporary offset (GTEO) and GPRS Penalty Time (GPET) can be used to avoid
unnecessary cell reselection from moving MS to cells they briefly pass. E.g. Pico cells.
Priority Class (PRC) and HCS signal level threshold (HCS) These two parameters should be
used carefully! Wrongly used they can increase ping-pong and decrease network
performance.

NCCR - Quality Control
The Quality Control (QC) shall maintain statistics about BLER for each TBF as well as bitrate per radio block
for each TBF in RLC ACK mode.
• BLER sample filtering
• Bitrate per radio block sample filtering
QC uses this information for monitoring radio link performance and delay.
The purpose of Quality Control is to monitor and detect degradation periods in service quality, and to perform
corrective actions to remove the service degradation.
• For BLER sample filtering, the threshold value is operator parameter maximum BLER in acknowledged
mode (BLA) or maximum BLER in unacknowledged mode (BLU), depending on the RLC mode of the
TBF.
• For bitrate per radio block sample filtering, the threshold value is one of the four operator parameters
– QC GPRS DL/UL RLC Ack Throughput Threshold (QGDRT, QGURT) or
– QC EGPRS DL/UL RLC Ack Throughput Threshold (QEDRT, QEURT),
The possible actions in BSS12 include TBF reallocation and network controlled cell reselection and triggered
by the following parameters:
• QC reallocation action trigger threshold (QCATR)
• QC NCCR action trigger threshold (QCATN)

BLER Filtering
The monitored samples are filtered. The filtering is compared to appropriate threshold values:
• For BLER sample filtering, the threshold value is operator parameter maximum BLER in
acknowledged mode (BLA) or maximum BLER in unacknowledged mode (BLU), depending
on the RLC mode of the TBF.
QC shall maintain the BLER degradation duration counter for each TBF according to the
following rules:
• The BLER degradation duration counter shall be incremented by 10, if ;
MAX _ BLER(%)  Measured _ BLER(%)
• the counter shall not be modified, if ;

MAX _ BLER(%)  Measured _ BLER(%)  MAX _ BLER(%)
• The counter shall be cleared (set to zero), if.

Measured _ BLER (%)  MAX _ BLER (%)
The QC thread shall monitor the BLER degradation duration counter, and if the counter is larger
than predefined triggering levels, the corresponding corrective action is tried.

BER Filtering
The monitored samples are filtered. The filtering is compared to appropriate threshold values:
For bitrate per radio block sample filtering, the threshold value is one of the four operator
parameters
– QC GPRS DL/UL RLC Ack Throughput Threshold (QGDRT, QGURT) or
– QC EGPRS DL/UL RLC Ack Throughput Threshold (QEDRT, QEURT),
depending on the type and mode of the TBF, multiplied with e.g. 1.2 in order to have a safety
margin of some degree during the first calculation cycles.
The bitrate per radio block degradation duration counter shall be maintained for each TBF
according to the following rules:
• The bitrate per radio block degradation duration counter shall be incremented by 10, when
the bitrate per radio block is below the threshold value.
• The counter shall be cleared (set to zero), when the bitrate per radio block is above or equal
to the threshold value.
The QC thread shall monitor the bitrate per radio block degradation duration counter. If the
counter is larger than predefined triggering levels, the corresponding corrective action is tried.

Corrective actions
• When any of the degradation duration counters monitored by QC gets larger than
a predefined action trigger threshold, QC shall try to corresponding corrective
action in background thread, running with low priority.
• Each action shall be triggered only once for a TBF in QC (once for a call (UL+DL
TBFs of the same phone) during one degradation period (200ms). If the
degradation period ends and a new starts, new actions can be tried).
• For example, if reallocation is already done, next action to be performed is NCCR,
triggered when a degradation duration counter exceeds the NCCR trigger
threshold.
• The flags of already performed actions shall be cleared when the degradation
ends, i.e. when all the degradation duration counters are cleared.
• The action trigger thresholds are expressed in block periods and the values can be
set by operator, see operator parameters QC Action Trigger Threshold:
– QC reallocation action trigger threshold (QCATR) (default: 25)
– QC NCCR action trigger threshold (QCATN) (default:100)
• It is possible to change the order of different actions by modifying the action trigger
threshold values (If the value is set to 0, then no action of that kind is tried).

ISNCCR
• Inter-System Network-Controlled Cell Re-selection uses two criteria: coverage and service priority.
• The coverage criterion enables the operator to set whether IS-NCCR is started as soon as WCDMA FDD
cell coverage is available or whether IS-NCCR is done only when GSM coverage is ending.
• In service priority criterion, the actual configuration of the GSM/EDGE vs. WCDMA FDD prioritization for
different services is done in the core network. The PCU follows the core network prioritization information
received in the Service UTRAN CCO information element.
• In target WCDMA FDD cell prioritization, the PCU ranks the target cell candidates based on the reported
Ec/No value.

ISNCCR criteria
If the BSC parameter WCDMA FDD NCCR Preferred (WFNP) is set on (Y),
• Y = ISNCCR is triggered as soon as an appropriate WCDMA cell is available.
• N = ISNCCR is triggered only in case there is not an appropriate GSM/EDGE cell available and an appropriate WCDMA cell
is available.
the Service UTRAN CCO IE value is ignored and IS-NCCR to WCDMA FDD cell is triggered
when any of the WCDMA FDD neighbor cells fulfils the following equation:
AV_3G_MES_QUAN(n) > CellReselQual3GRATThr(n)

– AV_3G_MES_QUAN(n) is the average of 3G measurement quantity
– CellReselQual3GRATThr(n) is the value that the AV_3G_MES_QUAN(n) parameter has to exceed before a cell change to cell 'n' can be
ordered. CellReselQual3GRAThr is determined by new adjacent WCDMA RAN cell parameter ISNCCR FDD quality threshold (FQT) CPICH
Ec/Io level measured by mobiles (default: -15 dB).
If the parameter WCDMA FDD NCCR Preferred is set off (No), IS-NCCR to WCDMA FDD cell is
triggered if
• SGSN has not denied IS-NCCR by Service UTRAN CCO IE
• serving and neighbor GSM cells do not fulfill the minimum RXLEV requirements
– AV_RXLEV_SERV > RxLevMinCell + Max(0, Pa), or
– AV_RXLEV_NCELL(n) > RxLevMinCell(n) + Max(0, Pa)
• any of the WCDMA FDD neighbor cells fulfill the equation AV_3G_MES_QUAN(n) >
CellReselQual3GRATThr(n)
If there are several WCDMA FDD neighbor cells that fulfill the equation, the cell that has the highest
AV_3G_MES_QUAN(n) is selected as the IS-NCCR target cell.

Network Controlled Cell Re-selection (NCCR)
Ratio of NC2 cell reselections to all cell – nccr_17a
• Shows the ratio between NC2 cell reselections and all cell reselections (NC0 +
NC2)
• Time aggregation: hourly
• Unit: %
sum(b.NCCR_SUCC_FLUSH_RECEIVED)
100 * -------------------------------
sum(a.FLUSH_MSGS_RECEIVED)

a = p_nbsc_packet_control_unit
b = p_nbsc_cell_reselection

Successful NCCRs ratio – nccr_13
• Share of successful NCCRs.
• Unit: %
nccr_succ_flush_received
100 * ----------------------------
nccr_1
a = P_NBSC_CELL_RESELECTION

# of NCCRs compared to data amount – nccr_12a
• Number of network controlled cell Nccr_4 (Total started NCCRs )
reselections compared to data amount .

• Object aggregation: BTS Sum(
• Time aggregation: hourly a.PCCO_TO_GPRS_MS_DUE_PWR_BDGT
+ a.PCCO_TO_EGPRS_MS_DUE_PWR_BDGT
• Unit: #/GB
+ a.PCCO_SENT_DUE_SERV_ISNCCR
+ a.PCCO_SENT_DUE_COVERAGE_ISNCCR
nccr_4 + a.PCCO_SENT_DUE_QUAL_CTRL
-------------------------------------------------------- )
(trf_213c+trf_212c+trf_215a+trf_214a)/(1024* 1024)
a = P_NBSC_CELL_RESELECTION

Average duration of successful NCCRs – nccr_14a
• Indicate average duration of successfully executed NCCR procedures.
• Unit: second
ave_nccr_duration_sum / 100
----------------------------
nccr_succ_flush_received

P_NBSC_CELL_RESELECTION

Resource Allocation - BTS Selection and TSL
Allocation
BTS selection and TSL allocation is
Segment
• Initial BTS Selection

• Reallocation of TBFs among the
BTS
1. BTS Load reallocation
2. Uplink Rx level reallocation
3. Downlink Rx level
reallocation
4. Downlink RX level received
first time reallocation

BTS Selection START READY
Select the BTS that

concurrent is already serving
TBF type ?
the MS
UL and DL
Select BTSs whose frequency band is included in

the MS RAC band information and save them in
BTS_LIST_1 Note1, Note2, Note3, Note4
Note1: If MS RAC band is not

known then BCCH band BTSs are RX-level ?
selected RX-level is
known
RX-level is not
known
Note2: BTSs that don’t have PSW
Select from BTS_LIST_1 the BTSs for Select from BTS_LIST_1 the BTSs
territory or channels for PSW are whom whose non_bcch_layer_offset is less
(RX-level - non_bcch_layer_offset) is than direct_gprs_access_ threshold
not selected. bigger than GPRS_non_BCCH_ and save them in BTS_LIST_2
layer_rxlev_upper_limit
Note3: UL: BTSs whose average and save them in BTS_LIST_2
TBF/TSL is not less than
MaximumNumberOfULTBF are Select from
BTS_LIST_1 the BTS
not selected BTS_LIST_2
Yes whose
non_bcch_layer_offset
Note4: DL: BTSs whose average empty ? is lowest
TBF/TSL is not less than No
READY
MaximumNumberOfDLTBF are MS is EGPRS capable
MS RAC ?
MS is only GPRS capable
not selected Is there any Is there any

EGPRS BTS in the GPRS BTS in the
Yes
BTS_LIST_2 ? BTS_LIST_2 ?
Yes No No
Remove Select the BTS who has the Remove

GPRS BTSs from lowest QoS load among the EGPRS BTSs
BTS_LIST_2 BTSs in the BTS_LIST_2 from BTS_LIST_2
READY

BTS Load
QoS load description

BTS_QoS_load: Average QoS load in the BTS.
 TSL_QoS_load of BTS' s all timeslot
number of BTS' s timeslots allocated for PS
TSL_QoS_load: Sum of TBF_QoS_load of TSL’s all TBF
TBF_QoS_load: Calculated value for load that TBF creates on a TSL. The calculated
value is weighted with QoS class of the TBF.
Reallocation triggering based on load checks
Periodical checks are based on counting block periods given to TBF. The amount of
block periods are defined using TBF_LOAD_GUARD_THRSHLD parameter
(PRFILE). (TBF specific counters were selected instead of timers because of expected lighter processing
load)
The BTS Load Reallocation check is triggered always when
TBF_LOAD_GUARD_THRSHLD blocks are transferred after previous triggering. This
is done for uplink, downlink and concurrent TBFs.

BTS Load – Penalties (Priority based QoS)
TBF weight
The heavier the TBF weight the smaller the number of TBFs sharing the RTSL, e.g the better the
service perceived by the user. The TBF weights for the service classes are derived from operator
parameters:
DHP (DL high priority SSS),
DNP (DL normal priority SSS),
DLP (DL low priority SSS),
UP1 (UL priority 1 SSS),
UP3 (UL priority 3 SSS)
There are different weightings for the TBFs based on their priority class:
Priority 1 (Gold(G))= 1 Gold
Priority 2 (Silver(S))= Gold/2
Priority 3 (Bronze(B))=Gold/3
Priority 4 (Best Effort(Be))= Gold/4

Resource Allocation - Priority based QoS
TBF1 with SSS=6
Latest Service Time = Current Time + Scheduling Step Size

TBF2 with SSS=1
1 2 3 4 5 6 7 8 9 10 11 12 (time) The
scheduling
127 12 12 12 12 12 13 18 18 18 18 18 is done
based on
7 8 9 10 11 12 12 13 14 15 16 17
latest service
Latest service time
time, one
TBF at a
time is
served by
the RTSL
t i t i
52 TDMA frames = 240 ms= 12 blocks

BTS Load - Penalties (PACCH, Type and
Multiplexing)
PACCH penalty
This penalty is added to the load of a UL timeslot for every DL TBF that increases UL polling traffic in the
slot.
UL TBF Penalty incremented by 0.00012 * Best Effort TBF weight
Timeslot type penalty
This is applied to additional and default timeslots, and it is more severe on the former. The purpose of this
penalty is to favour timeslots which are less likely to be removed from the PSW territory.
Default timeslot penalty=0.12 * Best Effort TBF weight
Additional timeslot penalty=0.24 * Best Effort TBF weight
GPRS/EGPRS multiplexing penalty
A multiplexing penalty is added to the load of a non-multiplexed EGPRS timeslot when resources are
searched for a GPRS TBF, and to the load of both multiplexed slots and non-multiplexed GPRS slots when
resources are searched for an EGPRS TBF.
Mux penalty= 2.5 * Gold TBF weight
When two allocations bear equal loads, the allocation farther from the CSW/PSW territory border is selected.

BTS and TSL Load - Checking
There are three load checks

BTS QoS load is too high when compared to other BTS (QoS load is described
on next slides)
The target is to balance BTS_QoS_load between BTSs.
For GPRS TBF:
If there is a GPRS BTS with 30% lower QoS load and with suitable MS RAC band and with proper signal levels then CHM
requests for a reallocation from MAC (or a concurrent reallocation if a concurrent RAT exist).
For EGPRS TBF:
If there is a EGPRS BTS with 30% lower QoS load and with suitable MS RAC band and with proper signal levels then
CHM requests for a reallocation from MAC (or a concurrent reallocation if a concurrent RAT exist).
GPRS TBF in EGPRS territory
This check is done for GPRS TBF in EGPRS BTS. The target is to reallocate a GPRS TBF away from EGPRS territory.
If there is GPRS BTS with proper UL/DL signal levels and the average TBF/TSL of the GPRS BTS is less than
maximum_number_of_UL_TBF / maximum_number_of_DL_TBF then CHM requests for a reallocation from MAC (or a
concurrent reallocation if a concurrent RAT exist).
BTS timeslot load too high when compared to BTS load

The target is to balance TSL_QoS_load inside the BTS.
If BTS_QoS_load / (average QoS load of timeslots where the TBF is allocated) is less than 0.7 then the CHM requests
reallocation from the MAC.

Throughput optimization
Connectivity Capacity (MS-SGSN)
TSL data rate improvement and multi-slot usage
maximization (BSS)
E2E data rate

Connectivity Planning – Maximized Capacity
The connectivity planning for maximum capacity is based on the proper set of CDEF and DAP size
• To provide enough capacity for territory upgrade the 75 % utilization of Abis channels is recommended by Nokia
Outputs Max Utilizatio Limit Unit

Limit n
Abis Channels (Radio and EDAP 256 75% 192* TSL
slave TSLs) s
EDAPs 16 100% 16** Pcs
BTS (cell, segment) 64 100% 64 Pcs
TRXs 128 100% 128 Pcs
• The CDEF is allocated to the cells (BTSs in segment), so the too big CDEF territory will need more PCUs.
• The Dynamic Abis Pool (DAP) is allocated to the sites (BCFs). Higher DAP size provides more MCS9 capable TSLs
on air interfaces, but on the other side, higher DAP size needs more capacity on E1s and more PCUs as well.
So the proper value of CDEF on cell (BTS) level and DAP on BCF level can help to be below the 192 (96*) radio TSL
limit with 75 % utilization to avoid connectivity bottlenecks even in case of territory upgrades
*It is important to know that the PCU and PCU-S have 128 radio TSL limit with S11.5, which can be a real bottleneck in GPRS only networks.
**Recommended number of EDAPs per PCU1 is 1,2,4 or 8, 1-8 for PCU2

(E)GPRS – Connectivity
Inadequate EDAP resources in DL limited by EDAP
• Time per gigabytes when adequacy of requested EDAP resources cannot be granted for a
scheduled DL TBF in BTSs where EGPRS is enabled.
• Object aggregation: BSC
• Unit: min/GB
dl_tbfs_with_inadeq_edap_res - dl_mcs_limited_by_pcu) / (50 * 60)
dap_7b = --------------------------------------------------------------------------
sum over BTS with EGENA = Y
(rlc_data_blocks_dl_cs1 *20 + rlc_data_blocks_dl_cs2 *30
+ sum over MCS-1 (xx)* 22 + sum over MCS-2 (xx)* 28
+ sum over MCS-7 (xx/2)*112 + sum over MCS-8 (xx/2)*136
+ sum over MCS-9 (xx/2)*148 + sum over MCS-11(xx)*36
+ sum over MCS-12(xx)*50 ) / (1024*1024*1024)
where xx = (dl_rlc_blocks_in_ack_mode + dl_rlc_blocks_in_unack_mode)
• CS3 & CS4 are not included.

• The formula may not give correct values if
– The object aggregation level is smaller than BSC level
– One or more BTSs in the area has more than one dynamic abis pool connected
• Target values:
– Good: < 75 min/GByte
– Bad: > 150 min/GByte
– Similar formula for UL (dap_8c)

Inadequate EDAP resources in DL limited by PCU
• Time per gigabytes when adequacy of requested EDAP resources cannot be granted for a scheduled DL
TBF in BTSs where EGPRS is enabled. Only PCU limitations are considered.
• Unit: min/GB
dl_mcs_limited_by_pcu) / (50 * 60)
dap_9 = --------------------------------------------------------------------------
sum over BTS with EGENA = Y
(rlc_data_blocks_dl_cs1 *20 + rlc_data_blocks_dl_cs2 *30
+ sum over MCS-7 (xx/2)*112 + sum over MCS-8 (xx/2)*136
+ sum over MCS-9 (xx/2)*148 + sum over MCS-11(xx)*36
+ sum over MCS-12(xx)*50 ) / (1024*1024*1024)
where xx = (dl_rlc_blocks_in_ack_mode + dl_rlc_blocks_in_unack_mode)
• CS3 & CS4 are not included.

• The formula may not give correct values if
– The object aggregation level is smaller than BSC level
– One or more BTSs in the area has more than one dynamic abis pool connected
• Target values:
– Good: < 15 min/GByte
– Bad: > 30 min/GByte
– Similar formula for UL (dap_10)

dl_mcs_limited_by_pcu and
dl_tbfs_with_inadeq_edap_res counter update
• PCU internal resource allocation (DSP channels) is done automatically
based on number of EDAPs connected to the PCU and based on RTSLs
(CDEF) and EDAP sizes connected to the PCU.
• These counters count events where the modulation and coding scheme
(MCS) is reduced due to the temporary congestion of PCU internal
resources that are allocated and available for the given cell.
• Certain amount of congestion is normal. If there is no congestion even
there is traffic the PCU resources may be over dimensioned (resources
are allocated for such cells which do not generate much traffic and these
resources may not be available for other cells).
• MCS limited by PCU (76019/76020 ) and TBFs with Inadequate EDAP

resources (076008/76009 ) are updated per RTSL and per RLC block
allocation (every 20 ms).
• 1 second resource bottleneck (all allocated RTSLs) for 4 RTSL MS can
lead to increasing counter value by 4 x 50 (=1000 ms/20 ms) = 200.
• In practice e.g. instead of MCS-9, MCS-8 or MCS-6 could be used.
Counters do not indicate what MCS was used during MCS downgrade.

Territory upgrade rejection due to PCU capacity
• Indicates the ratio of territory upgrade requests ending up in rejection due to PCU capacity.
• Unit: %
sum(gprs_ter_ug_rej_due_lack_pcu)
blck_32 = 100 * -------------------------------
sum(gprs_ter_upgrd_req)
p_nbsc_traffic

TSL Data Rate and Multislot Usage Improvement –
Content
•TSL Utilization
•Acknowledgement Request
•Delayed TBF Release (TBF Release Delay, TBF Release Delay Extended)
•BS_CV_MAX
•Link Adaptation (GPRS, GPRS (CS1-4), EGPRS)
•Multiplexing
•UL Power Control
•Multislot Usage

(E)GPRS – RLC/MAC traffic volume
GPRS RLC payload (DL and UL)
• GPRS RLC payload: trf_212c (UL) / trf_213c (DL) gives the payload data in Kbytes.
• Unit: Kbyte
trf_213c = sum(a.rlc_data_blocks_dl_cs1 *20 + a.rlc_data_blocks_dl_cs2*30 + sum over MCS-11
(b.dl_rlc_blocks_in_ack_mode + b.dl_rlc_blocks_in_unack_mode)*36 + sum over MCS-12
(b.dl_rlc_blocks_in_ack_mode + b.dl_rlc_blocks_in_unack_mode)*50 )/1024
trf_212c = sum(a.rlc_data_blocks_ul_cs1 *20 + a.rlc_data_blocks_ul_cs2 *30 + sum over MCS-11

(b.ul_rlc_blocks_in_ack_mode + b.ul_rlc_blocks_in_unack_mode)*36 + sum over MCS-12
(b.ul_rlc_blocks_in_ack_mode + b.ul_rlc_blocks_in_unack_mode)*50 )/1024
a = p_nbsc_packet_control_unit; b = p_nbsc_coding_scheme
• Payload is a clear quantity figure depending not only on data volume from/to the LLC layer,
but from the quality and capacity of the BSS chain, too
• Most of the GPRS/EDGE KPIs are on BTS level. E.g. the payload formulas can be used to
see how the different BTSs inside the segment are used

EGPRS RLC payload (DL and UL)
• EGPRS RLC payload: trf_214b (UL)/trf_215a (DL) give the payload data in Kbytes.
• Unit: Kbyte
trf_215a = (sum over MCS-1 (xx)* 22+ sum over MCS-2 (xx)* 28+ sum over MCS-3 (xx)* 37+ sum
over MCS-4 (xx)* 44+ sum over MCS-5 (xx)* 56+ sum over MCS-6 (xx)* 74+ sum over MCS-7
(xx/2)*112+ sum over MCS-8 (xx/2)*136+ sum over MCS-9 (xx/2)*148)) ) / 1024
where xx = (DL_RLC_BLOCKS_IN_ACK_MODE + DL_RLC_BLOCKS_IN_UNACK_MODE)
trf_214a = sum( decode(coding_scheme,1,(xx)*22,0)+decode(coding_scheme,2,(xx)*28,0)+

decode(coding_scheme,3,(xx)*37,0)+ decode(coding_scheme,4,(xx)*44,0)+
decode(coding_scheme,5,(xx)*56,0)+ decode(coding_scheme,6,(xx)*74,0)+
decode(coding_scheme,7,(xx/2)*112,0)+ decode(coding_scheme,8,(xx/2)*136,0)+
decode(coding_scheme,9,(xx/2)*148,0)) / 1024
where xx = (UL_RLC_BLOCKS_IN_ACK_MODE + UL_RLC_BLOCKS_IN_UNACK_MODE)

p_nbsc_coding_scheme

DL/UL combined GPRS and EDGE payload
• DL/UL combined GPRS and EDGE payload data
• Unit: MB
• Downlink RLC Payload

trf_244 = (trf_213c+trf_215a)/(1024*1024)
• Uplink RLC payload

trf_243 = (trf_212c+trf_214a)/(1024*1024)

DL GPRS traffic – trf_208b
• DL GPRS traffic
• Unit: TSL or Erlang
sum(a.rlc_data_blocks_dl_cs1 + a.rlc_data_blocks_dl_cs2 + a.retra_rlc_data_blocks_dl_cs1 +

a.retra_rlc_data_blocks_dl_cs2) + sum over MCS-11..12 (b.dl_rlc_blocks_in_ack_mode + b.dl_rlc_blocks_in_unack_mode +
b.retrans_rlc_data_blocks_dl) ;CS3 and CS4 + (a.rlc_mac_cntrl_blocks_dl - a.egprs_dl_ctrl_blocks)
------------------------------------------------------
sum(period_duration*60)*50; 50 blocks /sec /tsl

b = p_nbsc_coding_scheme
• The traffic measurement figures in TSL or Erlang show the used capacity of the
BTS for PS purposes. These figures depending on available capacity for PS in the
BSS chain and quality of air interface.
• Note that this formula does not contain the RLC data blocks, which carry the
dummy LLC frames used to keep the DL TBF open in the delayed release state.
• Therefore the formula gives too low values, and real values could be up to 50%
higher (this cannot be measured accurately with current counters, and the value
depends heavily on the traffic pattern, but up to 50% is a good guess)

UL GPRS traffic – trf_205c
• UL GPRS traffic [ignor rlc data bl ul due bsn gprs] =
trf_203
ignor_rlc_data_bl_ul_due_bsn * ---------------------------------
• Object aggregation: BTS trf_203 + trf_204a
• Time aggregation: hourly (BH) [rlc mac cntrl blocks ul gprs] =

trf_203
• Unit: TSL or Erlang rlc_mac_cntrl_blocks_ul * --------------------------------------------
trf_203 + trf_204a
Where:
sum(rlc_data_blocks_ul_cs1 trf_203b=Sum(a.rlc_data_blocks_ul_cs1 + a.rlc_data_blocks_ul_cs2
+ sum over MCS-11 (b.ul_rlc_blocks_in_ack_mode +
+ rlc_data_blocks_ul_cs2 b.ul_rlc_blocks_in_unack_mode) ;CS3
+ sum over MCS-12 (b.ul_rlc_blocks_in_ack_mode +
+ BAD_FRAME_IND_UL_CS1 b.ul_rlc_blocks_in_unack_mode)) ;CS4
+ BAD_FRAME_IND_UL_CS2 Trf_204a = ( -sum(a.RLC_DATA_BLOCKS_UL_CS1+
a.RLC_DATA_BLOCKS_UL_CS2)
+ BAD_FRAME_IND_UL_UNACK
+ sum over MCS1 to 9 (b.UL_RLC_BLOCKS_IN_ACK_MODE+
+ [rlc mac cntrl blocks ul gprs] b.UL_RLC_BLOCKS_IN_UNACK_MODE)
- sum(a.IGNOR_RLC_DATA_BL_UL_DUE_BSN) + sqrt (
+ [ignor rlc data bl ul due bsn gprs]) (sum(a.RLC_DATA_BLOCKS_UL_CS1+ a.RLC_DATA_BLOCKS_UL_CS2)
- sum over MCS1 to 9 ( b.UL_RLC_BLOCKS_IN_ACK_MODE
----------------------------------- + b.UL_RLC_BLOCKS_IN_UNACK_MODE)
sum(period_duration*60)*50 ; 50 blocks /sec /tsl + sum(a.IGNOR_RLC_DATA_BL_UL_DUE_BSN))
*
(sum(a.RLC_DATA_BLOCKS_UL_CS1+ a.RLC_DATA_BLOCKS_UL_CS2)
- sum over MCS1 to 9 (
b.UL_RLC_BLOCKS_IN_ACK_MODE +
Counters from table(s): b.UL_RLC_BLOCKS_IN_UNACK_MODE)
+ sum(a.IGNOR_RLC_DATA_BL_UL_DUE_BSN)) + 4 * sum over MCS1 to 9 (
b.UL_RLC_BLOCKS_IN_ACK_MODE +
p_nbsc_packet_control_unit b.UL_RLC_BLOCKS_IN_UNACK_MODE)
* sum(a.RLC_DATA_BLOCKS_UL_CS1+a. RLC_DATA_BLOCKS_UL_CS2))) ;end
of sqrt
----------------------------------------------------------------------------------
2

DL EGPRS traffic – trf_162f
• DL EGPRS traffic
• Unit: TSL or Erlang
sum over mcs1..6 of (

a.dl_rlc_blocks_in_ack_mode
+ a.retrans_rlc_data_blocks_dl
+ a.dl_rlc_blocks_in_unack_mode
)
+ sum over mcs7..9 of (
a.dl_rlc_blocks_in_ack_mode
+ a.retrans_rlc_data_blocks_dl
+ a.dl_rlc_blocks_in_unack_mode
)/2
+ (b.egprs_dl_ctrl_blocks)
--------------------------------------------------
sum(b.period_duration*60)*50 ;50 blocks /sec /tsl
• Note that this formula does not contain the RLC data blocks, which carry the dummy LLC
frames used to keep the DL TBF open in the delayed release state.
• Therefore the DL part of the formula gives too low values, and real values could be up to
50% higher (this cannot be measured accurately with current counters, and the value
depends heavily on the traffic pattern, but up to 50% is a good guess)

UL EGPRS traffic – trf_161h
• UL EGPRS traffic [ignor rlc radio bl ul due bsn egprs] =
• Object aggregation: BTS trf_204
• Time aggregation: hourly (BH) ignor_rlc_data_bl_ul_due_bsn * X * ----------------
• Unit: TSL or Erlang trf_203b + trf_204
sum over MCS0..6 of Where X = 0.5 * A/(A+B) + 1 * B/(A+B)

( a.ul_rlc_blocks_in_ack_mode
+ A = sum over MCS7...9 of (a.ul_rlc_blocks_in_ack_mode +
a.BAD_RLC_VALID_HDR_UL_ACK+a.bad_rlc_valid_hdr_ul_ a.ul_rlc_blocks_in_unack_mode)
unack+a.ul_rlc_blocks_in_unack_mode B = sum over MCS1...6 of (a.ul_rlc_blocks_in_ack_mode +
+ a.bad_rlc_bad_hdr_ul_ack + a.bad_rlc_bad_hdr_ul_unack) a.ul_rlc_blocks_in_unack_mode)
+sum over MCS7..9 of
( a.ul_rlc_blocks_in_ack_mode
+ a.bad_rlc_bad_hdr_ul_ack + a.bad_rlc_bad_hdr_ul_unack
+
a.BAD_RLC_VALID_HDR_UL_ACK+a.bad_rlc_valid_hdr_ul_
unack+a.ul_rlc_blocks_in_unack_mode)/2
+ egprs_ul_ctrl_blocks
+ [ignor rlc radio bl ul due bsn egprs]
------------------------------------------------------------------------------------
avg(period_duration*60) *count( distinct
period_start_time)*50 ; 50 blocks /sec /tsl

a = p_nbsc_coding_scheme
b = p_nbsc_packet_control_unit

Average DL/UL TBF duration – tbf_6c/tbf_5c
• Average DL/UL TBF duration
• Unit: second
sum(ave_dur_dl_tbf_sum)/100
tbf_6c = ----------------------------------------------------------------------------------------
sum(decode(ave_dur_dl_tbf_sum,0,0,ave_dur_dl_tbf_den))
sum(ave_dur_ul_tbf_sum)/100
tbf_5c = ----------------------------------------------------------------------------------------
sum(decode(ave_dur_ul_tbf_sum,0,0,ave_dur_ul_tbf_den))

• If the average TBF duration shows the service type on application layer. Short TBFs are
generated by e.g. push email, while long TBFs are used for e.g. FTP download/upload.
• The TBF duration has high impact on parameter setup for planning and optimization

TSL Utilization

Acknowledgement Request
Acknowledgement Request Description of EGPRS Downlink

• GPRS Uplink Penalty These parameters are used by the RLC
acknowledgement algorithm to determine
• GPRS Uplink Threshold how frequently the PCU polls the mobile
• GPRS Downlink Penalty station having a DL TBF in EGPRS
mode. The PCU has a counter, which is
• GPRS Downlink Threshold incremented by one whenever an RLC
• EGPRS Uplink Penalty data block is transmitted for the first time
or retransmitted pre-emptively. The
• EGPRS Uplink Threshold counter is incremented by (1 +
• EGPRS Downlink Penalty EGPRS_DOWNLINK_PENALTY)
whenever a negatively acknowledged
• EGPRS Downlink Threshold RLC data block is retransmitted. The
mobile station is polled when the counter
exceeds the threshold value of
EGPRS_DOWNLINK_THRESHOLD.

Acknowledgement Request Penalty Threshold
Increment Total
RLC data block retransmission
3 3
3 6
3 9
New RLC data Block 1 10
1 11
Time
1 12
PCU
1 13
1 14
MS
1 15
1 16
1 17
1 18
RLC data block with poll
Packet Downlink ACK/NACK

SSN 1010111……...111101001
Bitmap (64 Blocks)

Acknowledgement Request
Penalty Threshold
Increment Total
USF
RLC data transmission 1 1
3 4
1 5
1 6
3 9 PCU
Time
MS
1 21
1 22
1 23
Packet Uplink ACK/NACK

SSN 1010111……...111101001
Bitmap (64 Blocks)
Packet Control ACK/NACK

Delayed TBF Release and BS_CV_MAX
TBF Release Delay
UL TBF Release with Extended UL TBF mode
BS_CV_MAX
RTT

TBF Release Delay
If there is not any payload RLC/MAC block received, the TBF will not be
released immediately, but it can be kept alive for a given time period.
There are two modifiable parameters related to Delayed TBF feature among
PRFILE parameters:
DL_TBF_RELEASE_DELAY (0,1-5sec, def 1s)
This parameter is used to adjust the delay in downlink TBF release. An
appropriate delay time increases the system performance, since the possibly
following uplink TBF can be established faster, and frequent releases and
re-establishments of downlink TBF can be avoided.
UL_TBF_RELEASE_DELAY (0,1-3sec, def 0,5s)
This parameter is used to adjust the delay in uplink TBF release. An
appropriate delay time increases the system performance, since the possibly
following downlink TBF can be established faster.
Comment:
DL TBF Delay Release is improving PoC but the UL TBF Delay Release is harmful for PoC.

UL TBF Release with Extended UL TBF
mode MS does not continue TBF
Short description:
• Countdown procedure is ongoing. EUTM
MS BSC / PCU supporting mobile is allowed to
recalculate CV during procedure, if it gets
Data block with CV = 1 more data to send. PCU notices this by
monitoring Block Sequence Number
Data block with CV = 0
(BSN) and Countdown value (CV) sent
by MS.
• After receiving CV=0 block PCU starts
PACKET UL ACK/NACK (FAI=0, UL extended state. It sends Packet Uplink
Polling=NO) EUTM delay timer starts
Ack/Nack message to MS with no Final
UL TBF extended state
Ack Indicator (FAI) on, but

Schedule USF turn for MS acknowledging all received blocks.
UL dummy control block • During UL extended state PCU schedules
UL TBF Schedule Rate Ext USFs for MS according adjustable
Schedule USF turn for MS scheduling rate parameter. If MS has no
UL dummy control block new data to send it sends UL dummy
control blocks on its sending turn.
Schedule USF turn for MS • When UL extended state ends, according
UL dummy control block adjustable release delay parameter, PCU
sends Packet Uplink Ack/Nack message
PACKET UL ACK/NACK (FAI=1, EUTM delay timer expires to MS with Final Ack Indicator (FAI) on.
Polling=YES)
PACKET CONTROL
ACK UL TBF terminated

UL TBF Release with Extended UL TBF
mode
MS continues data transfer on TBF Short description:
• Countdown procedure is ongoing.
After receiving CV=0 block PCU
MS BSC / PCU starts UL extended state. It sends
Packet Uplink Ack/Nack message
to MS with no Final Ack Indicator
Data block with CV = 0
(FAI) on, but acknowledging all
received blocks.
PACKET UL ACK/NACK (FAI=0, • During UL extended state PCU
EUTM delay timer starts
UL TBF extended state
Polling=NO)
schedules USFs for MS according
Schedule USF turn for MS adjustable scheduling rate
parameter. If MS has no new data
UL dummy control block
to send it sends UL dummy control
UL TBF Schedule Rate Ext blocks on its sending turn.
• When MS gets new data to send
Schedule USF turn for MS during extended state, it sends UL
data block with new BSN, and also
Data block with new BSN and CV EUTM delay timer stopped, new CV value when needed. Due
TBF continues BSN PCU knows that new UL
LLC is to be sent by MS, and UL
TBF continues as normally.
Data block

BS_CV_MAX
Recommended value: 9
Basically the BS_CV_MAX parameter should define the RLC round-trip delay in block periods.
• If the BS_CV_MAX parameter has too high value (e.g. 15), then the mobile may ignore some nacks that
would require retransmissions. So in some cases a block has to be nacked twice before the mobile is
willing to make the retransmission. This may reduce the performance slightly.
• On the other hand, if the BS_CV_MAX parameter is too large or if the mobile is not able to do accurate
time stamping for the UL RLC blocks, then the mobile may ignore some negative acknowledgements that
were received in the Packet UL ACK/NACK message. This may distort the ARQ procedure slightly.
• If the BS_CV_MAX parameter is lower than the actual round-trip delay or if the mobile is not able to do
accurate time stamping for the UL RLC blocks, then the mobile may transmit needless retransmissions
after processing a Packet UL ACK/NACK message.
It is recommended to tune the BS_CV_MAX parameter so that the minimum value is searched with which the
mobile does not send needless RLC retransmissions right after the processing of a Packet UL ACK/NACK
message.
After modification of this parameter it takes about 5 minutes for processes to get the new values. After 5 minutes disable and then re-enable GPRS in
those cells where GPRS is active for the change to take effect.

Parameter Recommendations
Parameter Paramete BSS10.5 + BSS11+CD5 BSS11.5 BSS11.5 BSS12

r class / CD7 (recommenda (Default) Recommen Default
id (default) tion) dation
DL_TBF_RELEASE_DELAY 46 / 67 03E8 (1s) 03E8 (1s) 03E8 (1s) 03E8 (1s) 03E8 (1s)
UL_TBF_RELEASE_DELAY 46 / 68 01F4 (0.5s) 01F4 (0.5s) 01F4 (0.5s) 01F4 (0.5s) 01F4 (0.5s)
UL_TBF_RELEASE_DELAY_EXT 46 / 71 - 07D0 (2s) 07D0 (2s) 07D0 (2s) 03E8 (1s)
UL_TBF_SCHED_RATE_EXT 46 / 72 - 4 (80ms) 4 (80ms) 4 (80ms) 4(80ms)
BS_CV_MAX - 6 9 9 9 9
EGPRS_UPLINK_PENALTY 0001 0001 0001 0001 0001
EGPRS_UPLINK_THERSHOLD 0019 0019 0019 0019 0025
EGPRS_DOWNLINK_PENALTY 0001 0001 0001 0001 0001
EGPRS_DOWNLINK_THERSHOL 0019 0019 0019 0019 0025

D

S11.5 and S12 RTT Comparison
EDGE Idle & Active RTT Comparison S11.5 PCU2 (S11.5 SB 2.18-0 CD5.2) vs S12 (S12 1.40-
0 PCD0.10)
Idle Active Ping

32B 256B 1460B 32B 256B 1460B
Average Average Average Average Average Average
-50 dBm 593 685 985 -50 dBm 363 461 809
S11.5 -70 dBm 578 693 1016 S11.5 -70 dBm 353 579 753
EDGE -90 dBm 833 1342 2082 EDGE -90 dBm 893 1317 2220
-100 dBm 2782 3533 5281 -100 dBm 1542 2869 4455
-50 dBm 592 691 945 -50 dBm 224 309 597
S12 -70 dBm 603 681 1080 -70 dBm 233 311 598
EDGE FR
EDGE FR -90 dBm 662 943 1309 -90 dBm 295 443 825
-100 dBm 768 1187 2391 -100 dBm 401 690
-50 dBm 621 717 1011 -50 dBm 247 333 620
S12 EDGE
-70 dBm 603 733 1050 -70 dBm 246 332 634
EDGE GbIP -90 dBm 269 372 755
-90 dBm 614 788 1095
GbIP -100 dBm 324 493 1297
-100 dBm 1154 1218 1709

GPRS Link Adaptation Algorithm (CS1-2)
Introduction
The coding scheme will change based on set BLER Thresholds.
The BLER thresholds are defined by simulations and change from hopping to non
hopping networks
CS1 & CS2 Crosspoint

From CS1 to CS2
= 1- (X kbps/8 kbps)
CS1 FH 14%
CS1 NFH 69%
From CS2 to CS1 X
= 1 - (X kbps/12kbps)
CS2 FH 43%
CS2 NFH 79%

Parameters
GPRS Link Adaptation
GPRS Coding Scheme Hopping
GPRS Coding Scheme No Hopping
DL BLER Crosspoint for CS Selection Hopping (DLBH)
UL BLER Crosspoint for CS Selection Hopping (ULBH)
DL BLER Crosspoint for CS Selection Non Hopping (DLB)
UL BLER Crosspoint for CS Selection Non Hopping (ULB)
• S10.5 ED CD 7.1 documentation presented new default parameter

recommendations for GPRS LA

Measurements (KPIs)
Downlink GPRS RLC throughput trf_235b
Uplink GPRS RLC throughput trf_233c
GPRS RLC Throughput
12.0
Downlink GPRS CS1 ratio rlc_55b
11.9
Downlink GPRS CS2 ratio rlc_33
11.8
kbit/s/tsl
Downlink
Uplink
11.7
Downlink GPRS CS1 retransmission ratio rlc_12a
11.6
Downlink GPRS CS2 retransmission ratio rlc_13 11.5

Uplink GPRS CS1 ratio rlc_54b

Uplink GPRS CS2 ratio rlc_32
Uplink GPRS CS1 retransmission ratio rlc_10f

Uplink GPRS CS2 retransmission ratio rlc_11f

Introduction
• The new Link Adaptation algorithm replaces the LA algorithm used for
CS1 and CS2 and covers the coding schemes:
– CS-1 and CS-2 if the CS-3 and CS-4 support is not enabled in the territory
– CS-1, CS-2, CS-3 and CS-4, if the CS3 and CS-4 support is enabled in the
territory
• The Link Adaptation algorithm is applied to measure the signal quality for
each TBF in terms of RXQUAL, which refers to received signal quality.
• RXQUAL describes the channel quality with the accuracy of eight levels. It
is expressed with three bits. RXQUAL is measured for each received RLC
radio block being thus a more accurate estimate than the BLER, which
has two levels: 0 and 1.
• The PCU determines internally the average BLER separately for each
coding scheme and the reported RXQUAL value. This is done separately
for each segment by collecting continuously statistics from all the
connections in the corresponding territory. The PCU can estimate the
BLER if CS-1, CS-2, CS-3 or CS-4 coding schemes were deployed for
this particular TBF. Moreover, based on BLER estimates the PCU can
determine which coding scheme will give the best performance.

Parameters
• In PCU2 the coding schemes CS-1 - CS-4 are supported.
• The BTS level parameters DL coding scheme in acknowledged mode
(DCSA), ULcoding scheme in acknowledged mode (UCSA), DL coding
scheme in unacknowledged mode (DCSU) and UL coding scheme in
unacknowledged mode (UCSU) define whether the fixed CS value (CS-1 -
CS-4) is used or if the coding scheme is changed dynamically according
to the Link Adaptation algorithm.
• The BTS level parameter adaptive LA algorithm (ALA) defines whether
the Link Adaptation algorithm is adaptive or not. The new Link Adaptation
algorithm can be used both in RLC acknowledged and in unacknowledged
modes both in uplink and downlink direction.
• When the Link Adaptation algorithm is deployed, the initial values for the
CS at the beginning of a TBF can also be defined with the parameters DL
coding scheme in acknowledged mode (DCSA), UL coding scheme in
acknowledged mode (UCSA), DL coding scheme in unacknowledged
mode (DCSU) and UL coding scheme in unacknowledged mode (UCSU).

KPIs
• rlc_14a UL RLC blocks #
• rlc_15a DL RLC blocks #
• rlc_55b Downlink GPRS CS1 ratio
• rlc_33 Downlink GPRS CS2 ratio
• rlc_12a Downlink GPRS CS1 retransmission ratio
• rlc_13 Downlink GPRS CS2 retransmission ratio
• rlc_54b Uplink GPRS CS1 ratio
• rlc_32 Uplink GPRS CS2 ratio
• rlc_10f Uplink GPRS CS1 retransmission ratio
• rlc_11f Uplink GPRS CS2 retransmission ratio
• rlc_27 Uplink GPRS CS3 retransmission ratio
• rlc_28 DL CS3 retransmission ratio %
• rlc_29 Uplink GPRS CS4 retransmission ratio
• rlc_31 CS3 and CS4 UL block error ratio %

EGPRS Link Adaptation - Introduction
Link Adaptation
The task of the LA algorithm is to select the optimal MCS
for each radio condition to maximize RLC/MAC data rate,
so the LA algorithm is used to adapt to situations where
signal strength and or C/I level is low and changing within
time 60 MCS-1
MCS-2
RLC selects the data block and additionally selects the MCS-3
50 MCS-4
MCS depending on radio link quality and amount of MCS-5
available dynamic Abis channels MCS-6
MCS-7
40
Normally, LA adapts to path loss and shadowing but not MCS-8
MCS-9
fast fading. IR is better suited to compensate fast fading 30 LA
20
Link Adaptation and Incremental Redundancy
•The retransmission process is based on IR 10
•LA must take into account if IR combining is performed at

the receiver. 0
0 5 10 15 20 25 30
•LA must take into account the effect of finite IR memory.

Link Adaptation Basics
LA is done independently for

each UL and DL TBF on
RLC/MAC block level, but the DOWNLINK UPLINK
LA algorithm is same for uplink PCU receives
and downlink Uplink radio block
The MCS selection is not the

Uplink averaging
same in case of initial
transmission and retransmission
(IR) Averaged BEPs
PCU receives EGPRS PCU decides to send
LA algorithm works differently Packet Downlink Packet Uplink Ack/Nack
for acknowledged mode and Ack/Nack message. message.
unacknowledged mode The BEPs from the

message are delivered
The BEPs from
averaging are delivered
RLC control blocks are to adaptation algorithm. to adaptation algorithm.
transmitted with GPRS CS-1

coding Link Adaptation algorithm
Downlink case outputs: Uplink case output:

- MCS for initial transmission - MCS that is sent to MS in Packet
- MCS for retransmission Uplink Ack/Nack message

MCS Selection Management (1/2)
In DL case the MCS selection is based on EGPRS Channel Quality Report received in EGPRS
PACKET DOWNLINK ACK/NACK message sent from the MS to network using PACCH to
indicate the status of the downlink RLC data blocks received.
• In DL the MCS selection is based on using the BEP measurement data
In UL case the MCS selection is based on the respective BEP measurement values which are
received within the UL PCU frames
Channel Management
• Generally the CHM accepts the CS/MCS from the RLC
• During scheduling the CHM checks if lower coding scheme must be used than the RLC has
selected. Reasons for lower coding scheme:
 GPRS TBF and EGPRS TBF multiplexing
 MS synchronization
Dynamic Abis Management
• The DAM allocates Abis slave channels for the TRX’s TSLs based on the MCS that the
RLC/CHM has selected. If there are not enough slave channels available as it is required by
RLC and CHM, the DAM allocates less slaves and informs the RLC about next lower MCS
that fits on the allocated Abis capacity.

MCS Selection Management (2/2)
Generally the MCS decided for initial transmission by RLC is used.

Exceptions:
If MCS would be MCS-9 or MCS-8 but the RLC could generate only one data block
(Note1) then the RLC selects MCS-6
If MCS would be MCS-7 but the RLC could generate only one data block (Note1)
then the RLC selects MCS-5
The CHM or the DAM has required lower MCS. If the RLC asked 8PSK MCS but the
returned MCS is MCS-4 then the RLC decreases it to MCS-3
Operator defined initial MCS is used at the beginning of TBF
Operator defined initial MCS is used after TBF reallocation to other BTS in the SEG
Note1: The RLC can generate only one data block if there is no RLC data block to be used as the
second block. This can happen if the next RLC data block to be transmitted has a status of NACKED or
PENDING and cannot be retransmitted with the MCS selected for the first block.

Bit Error Probability (BEP)
The RLC predicts the MCS based on BEP

The BEP is measured at the receiver before the decoding
The receiver has to convert each symbol into bit(s) and during this process it
estimates the bit error probability, which is the BEP
The BEP is a decision which includes information about the reliability of the decision
(i.e. how sure the receiver is that the received bit is decided correctly) - BEP can be
calculated from that certainty information
1 2 3 4
BEP is invariant of the used coding scheme, but it depends on the modulation though
Both in ack and unack mode the BEP is used to estimate the BLER for each MCS,
but BEP does not take BLER into account (ACK/NACK information)

Incremental Redundancy - Puncturing
Puncturing TX convolution coded data

Puncturing is a form of coding where bits
(with known interval) are removed from
e.g. convolution coded data and therefore
the code rate can be "tuned" very PS1- blue symbols removed
accurately (with symbol accuracy) convolution coded data
Depuncturing
In reception depuncturing is done so that
the places that are punctured in
transmission are filled with zero symbols PS2- red symbols removed
(bits which will cause zero decision - no
decision what is the transmitted symbol) RX Punctured data to receiver
This way the punctured code can be
decoded with the same decoder as the
original (non punctured) code
PS1- blue "zero" symbols added

Incremental Redundancy - Storing
In IR-mode, channel coding D a ta B lo c k

(redundancy) is increased
gradually (Type II Hybrid O ne M C S
ARQ) P 1 P 2 P 3 T r a n s m itte r
If the first transmission of P 1 P 2 P 3
radio block fails, it is

retransmitted with different 1 s t tr a n s m is s io n 1 s t r e t r a n s m is s io n
u p o n r e c e p tio n
2 n d r e tr a n s m is s io n
u p o n r e c e p tio n
puncturing scheme (P1, P2, fa ilu r e fa ilu r e
P3 defined for each MCS) P 1
and soft combined with the N o d a ta
re c o v e re d
P r o te c tio n L e v e l 1
old data
P 1 P 2 R e c e iv e r
S to re d N o d a ta
re c o v e re d
C o m b in a tio n : P r o t e c tio n L e v e l x 2
P 1 P 2 P 3
S to re d S to re d
C o m b in a tio n : P r o t e c tio n L e v e l x 3

Incremental Redundancy - Soft Combining
The figure shows an example of IR transmission and combining with

different puncturing schemes for different transmission. The shown case
corresponds to MCS-4 or MCS-9, where the basic code rate is 1/1.
original data
1/3 coded data
1st xmission r = 1/1

r = 1/1
1st decoding attempt
2nd xmission r = 1/2

r = 1/1
2nd decoding attempt
3rd xmission r = 1/3

r = 1/1
3rd decoding attempt

EGPRS Link Adaptation – Parameters
EGPRS Link Adaptation

EGPRS Link Adaptation Enabled (ELA)
Initial MCS for Acknowledged Mode (MCA)
Initial MCS for Unacknowledged Mode (MCU)
Maximum BLER in Acknowledged Mode (BLA)
Maximum BLER in Unacknowledged Mode (BLU)
Mean BEP Offset GMSK (MBG)
Mean BEP Offset 8PSK (MBP)
Incremental Redundancy
Re-segmentation
Memory Out Flag Sum

(E)GPRS – RLC/MAC TSL data rate
Ave GPRS DL throughput/used TSL – trf_235b
• Average effective ACK GPRS DL throughput per used TSL
• Time aggregation: hourly Counters from table(s):
• Unit: kbps in a TSL a = p_nbsc_packet_control_unit
sum((a.rlc_data_blocks_dl_CS1
- a.rlc_data_blocks_dl_unack)*20
+ a.rlc_data_blocks_dl_CS2*30
+ sum over MCS-11 (b.DL_RLC_BLOCKS_IN_ACK_MODE)*36 ;CS3
GPRS ACK DL payload data in (kilobits) + sum over MCS-12 (b.DL_RLC_BLOCKS_IN_ACK_MODE)*50)*8 /1000 ;CS4
------------------------------------------------------ = -------------------------------------------------------------
GPRS ACK DL time for data transfer (sec) sum (a.rlc_data_blocks_dl_cs1
- a.rlc_data_blocks_dl_unack
+ a.rlc_data_blocks_dl_cs2
+ RETRA_RLC_DATA_BLOCKS_DL_CS1
+ RETRA_RLC_DATA_BLOCKS_DL_CS2
+ sum over MCS-11..12(b.DL_RLC_BLOCKS_IN_ACK_MODE
+ b.RETRANS_RLC_DATA_BLOCKS_DL)) /50
• The numerator does not contain the RLC header bytes neither the does the MAC header
because the aim is to count data volume from the users point of view as close as possible
• Retransmitted blocks due to other reasons than NACK are not counted in any of the RLC-
counters. In DL direction these retransmissions occur when the TBF release is delayed.
• Formula assumes that all GRPS unack RLC traffic is using CS1. With PCU2, GPRS unack
RLC traffic can use CS1-4. If there is a significant amount of GRPS unack RLC traffic with
CS2, the formula will give too high values.
• Typical value: 11-12 kbps

Ave GPRS UL throughput/used TSL – trf_233c
• Average effective ACK GPRS UL throughput per used TSL
• Object aggregation: BTS Counters from table(s):
• Unit: kbps in a TSL
sum((a.rlc_data_blocks_ul_cs1
- a.rlc_data_blocks_ul_unack)*20
+ a.rlc_data_blocks_ul_cs2*30
+ sum over MCS-11 (b.ul_rlc_blocks_in_ack_mode)*36 ;CS3
GPRS UL payload ACK data in (kilobits) + sum over MCS-12 (b.ul_rlc_blocks_in_ack_mode)*50)*8 /1000 ;CS4
----------------------------------------- = -----------------------------------------------------------------
GPRS UL time for ACK data transfer (sec) sum(a.rlc_data_blocks_ul_cs1
- a.rlc_data_blocks_ul_unack
+ a.rlc_data_blocks_ul_cs2
+ a.retra_data_blocks_ul_cs1
+ a.retra_data_blocks_ul_cs2
+ sum over MCS-11..12(b.ul_rlc_blocks_in_ack_mode
+ b.retrans_rlc_data_blocks_ul)) /50
• Typical value 10-12 kbps

Ave EGPRS DL throughput/used TSL – trf_236
• Indicates the impact of radio link quality on • The numerator does not contain the RLC
net data rate per used timeslot considering header bytes neither the does the MAC
retransmissions and coding scheme header because the aim is to count data
selection volume from the users point of view as
• Object aggregation: BTS close as possible
• Time aggregation: hourly • Other things than bad radio link quality can
• Unit: kbps in a TSL make this formula give lower values:
(sum over MCS-1 (xx)*22
– EDAP congestion forces downgrading of
coding schemes
+ sum over MCS-2 (xx)*28
+ sum over MCS-3 (xx)*37 – Low MCA setting combined with short TBFs
+ sum over MCS-4 (xx)*44 (because the link adaptation will not have time
to choose the optimum MCS). This effect is
especially visible in the UL
In case of timeslot sharing, some blocks may
+ sum over MCS-7 (xx)*56 be sent with MCS1 (for synchronization
+ sum over MCS-8 (xx)*68 purposes) and some with MCS3 (to transfer
+ sum over MCS-9 (xx)*74)*8/1000 USF to GPRS handsets)
--------------------------------------------- If the amount of data to be sent is so small,
sum over mcs1..6 of (yy)/50 that it cannot fill the radio block, the MCS will
+ sum over mcs7..9 of (yy)/2/50 in certain cases be downgraded to a more
robust MCS, which fits better to the data size
where • Typical value: 30-40 kbps
xx = b.dl_rlc_blocks_in_ack_mode
yy = b.dl_rlc_blocks_in_ack_mode
+ b.retrans_rlc_data_blocks_dl

Ave EGPRS UL throughput/used TSL – trf_234
• Average effective ACK EGPRS UL
throughput per used TSL
• Unit: kbps in a TSL
(sum over MCS-1 (xx)*22
+ sum over MCS-9 (xx)*74)*8/1000
---------------------------------------
sum over MCS1..6 of (yy)/50
+ sum over MCS7..9 of (yy)/2 /50
where
xx = b.ul_rlc_blocks_in_ack_mode Counters from table(s):
yy = b.ul_rlc_blocks_in_ack_mode a = p_nbsc_packet_control_unit
+ b.retrans_rlc_data_blocks_ul b = p_nbsc_coding_scheme

An example with heavy loaded cell
Cell level on hourly basis
GPRS Average Ave Eff ACK Ave Eff ACK
Average PEAK Available Maximum GPRS DL EGPRS DL EGPRS
Available GPRS TCH TCH Busy Throughput Throughput Payload
BTS_ID Period Channels CHANNELS channels Traffic (E) TCH (kbps/TSL) (kbps/TSL) (kbit)
184 2008.02.25 0:00 6.04 13 38 0.04 2 12.00 31.07 562.13
184 2008.02.25 1:00 6 12 38 0 1 11.99 30.40 288.16
184 2008.02.25 2:00 6.06 12 38 0.3 2 11.97 29.73 239.02
184 2008.02.25 3:00 6.05 8 38 0.04 1 11.99 45.12 2083.54
184 2008.02.25 4:00 6.01 12 38 0.03 1 11.97 41.92 621.63
184 2008.02.25 5:00 6.01 8 38 0.02 1 11.99 30.44 277.95
184 2008.02.25 6:00 9.59 37 34 0.13 2 12.00 38.03 28424.68
184 2008.02.25 7:00 32.16 37 12 1.31 5 11.99 33.70 106375.37
184 2008.02.25 8:00 34.92 37 9 4.78 13 11.97 33.68 168327.88
184 2008.02.25 9:00 32.04 35 12 10.39 25 12.00 34.16 203699.24
184 2008.02.25 10:00 29.84 32 14 14.44 27 11.97 38.82 338109.69
184 2008.02.25 11:00 29.24 32 15 15.43 27 11.96 36.24 325531.51
184 2008.02.25 12:00 27.15 32 17 18.67 31 11.99 35.77 252644.08
184 2008.02.25 13:00 26.14 29 18 21.18 35 11.96 37.52 314720.29
184 2008.02.25 14:00 26.97 32 17 18.81 35 11.66 36.26 296814.17
184 2008.02.25 15:00 27.8 30 16 17.96 29 11.90 35.40 251502.72
184 2008.02.25 16:00 27.59 30 16 18.95 29 11.99 36.32 280134.87
184 2008.02.25 17:00 28.26 32 16 17.22 32 11.95 35.63 201552.46
184 2008.02.25 18:00 29.22 34 15 11.63 22 11.96 36.69 101727.98
184 2008.02.25 19:00 18.19 35 26 5.77 13 12.00 37.14 42463.62
184 2008.02.25 20:00 12.04 35 32 3.32 9 11.96 43.59 52283.50
184 2008.02.25 21:00 8.88 20 35 2.66 7 12.00 36.79 13980.66
184 2008.02.25 22:00 7.63 20 36 1.62 7 11.96 34.89 7034.17
184 2008.02.25 23:00 6.52 13 37 0.99 6 12.00 32.62 3448.74

Multiplexing
Channel Allocation Algorithm tends to separate EDGE TBFs and GPRS

TBFs on different RTSL to avoid multiplexing, if only one PS Territory exists
in the cell or there is high load.
• Synchronization
• UL GPRS USF on DL EGPRS TBF (DA, DA with USF4, EDA)
• TSL sharing - TBFs’ multiplexing on a TSL
The algorithm checks the need for re-allocation every
TBF_LOAD_GUARD_THRSHLD, in order to separate sessions.
The max amount of TBFs per TSL can be limited
Param eter Nam e Abbreviation Description Range and Default
Maximum Number of DL TBF MNDL m axim um num ber of TBFs that a radio tim e slot ca n 1..9, default:9
have in average, in a GPRS territory, in the downlink
direction.
Maximum Number of UL TBF MNUL m axim um num ber of TBFs that a radio tim e slot ca n 1..7, default:7
have in average, in a GPRS territory, in the uplink
direction.

(E)GPRS – Multiplexing
DL/UL TBFs pr timeslot – tbf_38d/tbf_37d
BTS ID tbf_37d tbf_38d
• Indicates how many DL/UL TBFs are there on 4017331 1.29 2.71
average per timeslot.
4030101 1.58 2.48
4010111 1.59 2.43
• Time aggregation: hourly (BH) 4010672 1.35 2.09
• Unit: # 4040041 1.17 2.09
502386 1.24 2.03
sum(ave_dl_tbfs_per_used_tsl) 4011883 1.43 2.02
tbf_38d = -----------------------------------
4040042 1.32 2.00
sum(aver_tbfs_per_tsl_dl_den) * 100
4010923 1.18 1.92
4010191 1.30 1.89
sum(ave_ul_tbfs_per_used_tsl) 4011502 1.18 1.83
tbf_37b = ----------------------------------- 4010233 1.24 1.83
sum(aver_tbfs_per_tsl_ul_den) * 100
4011351 1.14 1.77
4011051 1.13 1.76
Counters from table(s): 830152 1.24 1.75
4010242 1.17 1.74
4030182 1.04 1.74
4010301 1.04 1.73
• BSC counts every second the average number of 4040291 1.15 1.72
GPRS or EGPRS DL TBFs per used time slot. 4011372 1.07 1.72
• Multiplexing leads to QoS degradation, so it is 4030233 1.03 1.70
recommended to avoid it. 4030032 1.13 1.69
4010231 1.19 1.69
4010053 1.06 1.67
4040182 1.22 1.67
4010151 1.15 1.66

Ave # of DL/UL simult. users – trf_232/trf_231
• Gives the average number of simultaneous users in DL/UL
• Unit: #
Sum(ave_dur_dl_tbf_sum)
tbf_232 = -----------------------------------------------------------------
100 * avg(period_duration)* count(distinct period_start_time)* 60
Sum(ave_dur_ul_tbf_sum)
tbf_231 = -----------------------------------------------------------------
100 * avg(period_duration)* count(distinct period_start_time)* 60

P_NBSC_PACKET_CONTROL_UNIT
• Does not include abnormally released TBFs. TBF release delay is included.

GMSK modulation used in DL – rlc_61
• This formula shows how many times GMSK modulation is used in DL instead of
8PSK modulation (normally 8PSK offers better performance)
• Unit: #
a.DL_8PSK_TO_GMSK_DUE_UL_GPRS
----------------------------------------------------
sum over mcs1...9 of (b.dl_rlc_blocks_in_ack_mode + b.dl_rlc_blocks_in_unack_mode)

• When an uplink GPRS TBF uses same timeslots as a downlink EDGE TBF, the
EDGE TBF is forced to use GMSK modulation. Number of forced modulation
method changes are shown with this formula.
• If EDA is implemented then the GPRS USF on EGPRS TBF is not that big issue anymore

DL/UL EDGE TBFs in GPRS territory – tbf_60/59
• Share of DL GPRS TBF allocations done for EGPRS request in GPRS territory because of
congestion in EGPRS area.
• Unit: %
ul_gprs_tbf_for_egprs_req dl_gprs_tbf_for_egprs_req
tbf_59 = 100 * ------------------------------------------ tbf_60 = 100 * ------------------------------------------
REQ_1_TSL_UL_FOR_EGPRS_MS REQ_1_TSL_DL_FOR_EGPRS_MS
+ REQ_2_TSL_UL_FOR_EGPRS_MS + REQ_2_TSL_DL_FOR_EGPRS_MS
EDGE TBFs in GPRS territory

Why does formulas not show 100%
when applied in non-EDGE network? 100%
=> Either counter problems or Should be 100%

80%
formula problems
=> Be careful with using these formulas 60%
UL (tbf_59)
DL (tbf_60)
40%
20%
0%
DL/UL GPRS TBFs in EDGE territory –tbf_57a/58a
• Share of UL GPRS TBF allocations done in EGPRS territory because of congestion in GPRS
area (or non-existence of pure GPRS territory)
• Unit: %
ul_gprs_tbf_in_egprs_terr dl_gprs_tbf_in_egprs_terr
tbf_57a = 100 * -------------------------------------------------- tbf_58a = 100 * --------------------------------------------------
(REQ_1_TSL_UL - REQ_1_TSL_UL_FOR_EGPRS_MS) (REQ_1_TSL_DL - REQ_1_TSL_DL_FOR_EGPRS_MS)
+ (REQ_2_TSL_UL - REQ_2_TSL_UL_FOR_EGPRS_MS) + (REQ_2_TSL_DL - REQ_2_TSL_DL_FOR_EGPRS_MS)

Uplink Power Control
PC parameters for MS are transmitted on BCCH
PCH = min(CH - C + 48),PMAX)
CH, sets the minimum power level

• Range 0…62
• Default 34 (GSM900) , 36 (GSM1800)
Uplink Power Control
, sets the slope for the uplink power level

35
• Range 0…10 equivalent 0.0….1.0 30
• Default 7 (GSM900) , 8 (GSM1800)

25
Ms Output Power (dBm)
C, received signal level 20
0.3 1
15
0, 39(GSM900), 36 (GSM1800) 10
PMAX, max MS power allowed in the cell

0
Signal Strength (dBm)

Multislot Usage and Limitations - Introduction
Territory Parameters
Default GPRS capacity (CDEF)
Dedicated GPRS capacity (CDED)
Max GPRS capacity (CMAX)
GPRS Territory Update Guard Time (GTUGT)
Free TSLs Rate reduction due to
Free TSL for CS Upgrade (CSU) territory occupancy
Free TSL for CS Downgrade (CSD)
CSW Traffic
Voice and HSCSD
Interface and Network Element Limitations
EDAP
PCU
GSM Features
MultiBCF/CBCCH, FH, IUO/IFH, NMO I, HR (DR), DFCA, IDD, HSCSD, Extended cell,
etc

Higher GPRS/EDGE throughput
High Multislot Classes increases GPRS/EDGE peak

downlink throughput to 296 kbit/s
Extended Dynamic allocation increases GPRS/EDGE peak
uplink throughput to 236.8 kbit/s
Together these two features increase the downlink and uplink
combined throughput
350 350
300
Peak downlink 300
Peak uplink
throughput throughput
250 250
200 200
S11.5 S11.5
kbit/s
kbit/s
S12 S12
150 150
100 100
50 50
0 0
GPRS GPRS CS3/4 EDGE GPRS GPRS CS3/4 EDGE

(E)GPRS – Multislot usage
Introduction
• Actual CS traffic can be measured by the following KPIs:
– Trf_1g
– Trf_109
max(decode(trx_type,0,peak_busy_tch))
p_nbsc_res_avail
• This PI is an important traffic load indicator on the cell level. By following this and
reacting proactively, blocking can be avoided in cells where the traffic growth is
smooth
– Trf_365
• Actual PS traffic can be measured by the following KPIs:
– ava_44
– Peak PS territory (c2063)
o Recommendation: Ava_44 and c2063 can be compared with the CDEF
settings. If too big difference, then CDEF should perhaps be changed,
or more capacity should be added to the cell.

Free TSL Size (after CS Upgrade and Downgrade)
When a downgrade or upgrade procedure is requested following parameters

can reduce or increase the border between CSW and PSW territories:
TSL number after CS downgrade

TRX number 1 2 3 4 5
70 0 0 0 1 1
free TSL for CS downgrade (%)
95 1 1 1 2 2
(CSD)
99 1 1 2 2 3
TSL number after CS upgrade

TRX number 1 2 3 4 5
1 0 1 1 1 2
free TSL for CS upgrade (sec) 4 1 2 2 3 4
(CSU) 7 1 2 3 4 5
10 2 3 4 5 6

GPRS Territory Upgrade
The Territory Upgrade/Downgrade procedure uses the following three

parameters to help redistribute TBFs across available TSLs:
 X1: 1.5
 X2: 1
 X3: 0.5
The PS RRM request an upgrade when the average number of TBF's per
TSL in the GPRS territory is greater than X1 (and Default territory is already
allocated)
The target average number of TBFs in the GPRS territory is defined by X2
When the average number of TBF per TSL in the GPRS territory is less than
X3, the PS RRM will request a GPRS downgrade. (but only as far as the
default boundary)

Share of EDGE requests – tbf_65
• Shows the ratio of timeslot requests where EDGE resources are required.
• Unit: %
req_1_tsl_ul_for_egprs_ms + req_2_tsl_ul_for_egprs_ms + req_3_tsl_ul_for_egprs_ms + req_4_tsl_ul_for_egprs_ms

+ req_5_tsl_ul_for_egprs_ms + req_6_tsl_ul_for_egprs_ms + req_7_tsl_ul_for_egprs_ms + req_8_tsl_ul_for_egprs_ms
+ req_1_tsl_dl_for_egprs_ms + req_2_tsl_dl_for_egprs_ms + req_3_tsl_dl_for_egprs_ms + req_4_tsl_dl_for_egprs_ms
+ req_5_tsl_dl_for_egprs_ms + req_6_tsl_dl_for_egprs_ms + req_7_tsl_dl_for_egprs_ms + req_8_tsl_dl_for_egprs_ms
tbf_65 = 100 * ------------------------------------------------------------------------------------------------------------------------
req_1_tsl_ul + req_2_tsl_ul + req_3_tsl_ul + req_4_tsl_ul + req_5_tsl_ul + req_6_tsl_ul + req_7_tsl_ul + req_8_tsl_ul
+ req_1_tsl_dl + req_2_tsl_dl + req_3_tsl_dl + req_4_tsl_dl + req_5_tsl_dl + req_6_tsl_dl + req_7_tsl_dl + req_8_tsl_dl
Works also in non-EDGE networks

Share of EDGE requests
100%
80%
60%
40%
20%
0%

Territory upgrade rejection due to CSW – blck_30
• Territory upgrade rejection due to CSW
• Object aggregation: BTS Default territory
• Time aggregation: hourly (BH) 0 1 2 3 4 5 6 7
• Unit: % 1174 1179
1181
gprs_ter_ug_rej_due_csw_tr
blck_30 = 100% * ------------------------------- 1180
gprs_ter_upgrd_req
• The PCU will not request an upgrade of the PS territory if the territory has been reduced below the default
settings. For example, if CDEF has been set so it corresponds to 4 timeslots, and there is so much CS
traffic in the cell that the actual PS territory has been reduced to 2 timeslots, the PCU will not request
territory upgrades, even if there is heavy sharing of the PS timeslots.
• Naturally, if there have been no requests, there will be no rejections, and that makes the value of this
indicator a bit unpredictable. Before interpreting the value, it should be checked if the actual GPRS
territory is at least the size of the default territory.
Territory upgrade rejection due to CSW , blck_30 0% 16% 1% 5% 7%
Territory upgrade requests from PSW, c1174 28534 321 9602 9606 1706
Territory downgrade due to increasing CSW, c1179 49 3680 355 617 1849
Territory downgrade due to decreasing PSW traffic, c1181 23497 132 8792 8077 1589

DL/UL mslot allocation blocking – tbf_16a/15a
• Hard blocking of multislot allocation
• Unit: %
sum(NO_RADIO_RES_AVA_DL_TBF)
100 * -------------------------------------------------------------------------------
sum(req_1_TSL_DL + req_2_TSL_DL + req_3_TSL_DL + req_4_TSL_DL
+ req_5_TSL_DL + req_6_TSL_DL + req_7_TSL_DL + req_8_TSL_DL)
• If the blocking is met regularly, there is need either to expand the territory (CS traffic low) or
TCH capacity (CS traffic high).
• If the statistics show that there is blocking but no upgrade requests yet, the reason may be
that the territory has been smaller
than the default setting defines (CS use). The PCU will not make an upgrade request. This is
because the CS side will return
the default channels back to the PS territory as soon as the CS load allows that.

DL multislot soft blocking – blck_33a
• Indicates the ratio of unserved timeslots requests to all requests in downlink.
• Unit: %
sum(alloc_1_TSL_DL+2*alloc_2_TSL_DL+3*alloc_3_TSL_DL +4*alloc_4_TSL_DL
+ 5*alloc_5_TSL_DL+6*alloc_6_TSL_DL+7*alloc_7_TSL_DL +8*alloc_8_TSL_DL)
100 - 100 * --------------------------------------------------------------------------------------------------------------------
sum(req_1_TSL_DL+2*req_2_TSL_DL+3*req_3_TSL_DL +4*req_4_TSL_DL
+ 5*req_5_TSL_DL+6*req_6_TSL_DL+7*req_7_TSL_DL +8*req_8_TSL_DL)

GPRS triggered handovers – ho_61
• Number of GPRS triggered handovers that a voice user can expect per minute.
• Unit: #
Nbr of GPRS-triggered handovers sum(a.ho_att_due_to_gprs)

------------------------------------------------- = -------------------------------------------------------------------------------------
voice talk time sum(period_duration * b.ave_busy_tch / b.res_av_denom14)

a = P_NBSC_HO
b = P_NBSC_RES_AVAIL

Territory upgr. caused by decreasing CS traffic
• Indicates the ratio of territory upgrades caused by decreased CSW traffic. Can e.g. be used
in analyzing the reason for many GPRS-triggered handovers
• Unit: %
sum(GPRS_TER_UG_DUE_DEC_CSW_TR)
trf_240 = 100 * --------------------------------
sum(GPRS_TER_UG_DUE_DEC_CSW_TR
+ GPRS_TER_UPGRD_REQ
- GPRS_TER_UG_REJ_DUE_CSW_TR
- GPRS_TER_UG_REJ_DUE_LACK_PSW
- GPRS_TER_UG_REJ_DUE_LACK_PCU)

p_nbsc_traffic

Territory upgrades caused by increased PS traffic
• Indicates the ratio of territory upgrades caused by increasing PSW traffic. Can e.g. be used
in analyzing the reason for many GPRS-triggered handovers
• Unit: %
sum(GPRS_TER_UPGRD_REQ - GPRS_TER_UG_REJ_DUE_CSW_TR
- GPRS_TER_UG_REJ_DUE_LACK_PSW - GPRS_TER_UG_REJ_DUE_LACK_PCU)
Trf_241 = 100 * ---------------------------------------------------------------
sum(GPRS_TER_UG_DUE_DEC_CSW_TR + GPRS_TER_UPGRD_REQ
- GPRS_TER_UG_REJ_DUE_CSW_TR - GPRS_TER_UG_REJ_DUE_LACK_PSW
- GPRS_TER_UG_REJ_DUE_LACK_PCU)

p_nbsc_traffic

E2E data rate
Gb interface

Flow Control Parameter Setting
Default Flow Control Parameters
Num. Parameter Unit Before TN819 After TN819 (12
(12 May 2005) May 2005)
0020 FC_R_DIF_TRG_LIMIT % (/10) 250 (25%) 100 (10%)
0021 FC_MS_R_DEF B 3000 (24 Kbps) 4500 (36 Kbps)
0022 FC_MS_R_DEF_EGPRS B/s 3000 (24 Kbps) 15000 (120 Kbps)
0023 FC_MS_R_MIN B/s 100 (800 bps) 40 (320 bps)
0024 FC_MS_B_MAX_DEF B 5000 (5 kB) 10000 (10 kB)
0025 FC_MS_B_MAX_DEF_EGPRS B 5000 (5 kB) 25000 (25 kB)

0026 FC_R_TSL B/s 1500 (12 Kbps) 1500 (12 Kbps)
0027 FC_R_TSL_EGPRS B/s 1500 (12 Kbps) 4500 (36 Kbps)
0028 FC_B_MAX_TSL B 25000 (25 kB) 25000 (25 kB)
0029 FC_B_MAX_TSL_EGPRS B 25000 (25 kB) 25000 (25 kB)
•4TSL capable mobile at good radio conditions (MCS=9), theoretical max throughput would be
4X60kbps=240kbps and consider, therefore both Bearer Channel Access Rate, and
Committed Information Rate (CIR) should be at minimum 384 kbps.
• Recommended to have one 8 TSL Gb link having CIR is 512 kbps (8 E1/T1 TSLs)
• The CIR is usual set at both ends, i.e. BSC and SGSN. The BSC setting is the deciding one

BSSGP Flow Control - Parameter View
FC_MS_B_MAX_DEF_EGPRS
FC_MS_R_DEF_EGPRS FC_R_TSL_EGPRS
(default leak rate – at startup) (default leak rate / TSL) This together with
territory size gives the default BVC level
Um
Abis PC leak rate.
Gb PAP
U U
BTS RLC/MAC LLC frame
MS bucket
frame MS (MS allocation) NS/FR BSSGP/LLC
1
FC_B_MAX_TSL_EGPRS
MS (Buffering capacity / TSL)
2 This together with territory
IP packet
MS BVC size gives the actual BVC
1
Cell 1
Leak rate allocation allocation.
Access rate
MS 2
CIR BVC MS check

Actual Cell
check check
Leak Rate
PCU DL buffer (FR) Gb buffer
Gb BVC Flow control

FC_R_DIF_TRG_LIMIT
(triggering limit for updates of the
reported Leak rate)
Gb MS Flow control

Gb over FR – Availability
Average bearer unavailability ratio – uav_12
• Average bearer unavailability ratio
• Object aggregation: bearer_id
• Time aggregation: weekly
• Unit: %
100*sum(time_bear_oper_unoper)/sum(period_duration*60)

p_nbsc_frame_relay

Gb over FR – Traffic volume
Max. received load – frl_8a
• Indicates the load of the frame relay bearer for incoming data from SGSN.
• Unit: %
sum(8*(dlci_1_bytes_rec
+ dlci_2_bytes_rec
+ dlci_3_bytes_rec
+ dlci_4_bytes_rec
+ dlci_5_bytes_rec)/(period_duration*60))
100 * ---------------------------------------------
sum per frbc over all unlocked child nsvc
(c_nsvc.committed_info_rate*16);
frbc object_instance = bearer_id in p_nbsc_frame_relay

p_nbsc_frame_relay
• The access rate is taken from configuration data and represents only the current
setting. This may cause errors when used in historical perspective if the settings
have been changed

Gb over FR – Traffic volume
Max. sent load – frl_7a
• Indicates the load of the frame relay bearer for outgoing data to SGSN
• Unit: %
sum(8*(dlci_1_bytes_sent
+ dlci_2_bytes_sent
+ dlci_3_bytes_sent
+ dlci_4_bytes_sent
+ dlci_5_bytes_sent)/(period_duration*60))
100 * -------------------------------------------------
sum per frbc over all unlocked child nsvc
(c_nsvc.committed_info_rate*16);
frbc object_instance = bearer_id in p_nbsc_frame_relay

p_nbsc_frame_relay
• The access rate is taken from configuration data and represents only the current
setting. This may cause errors when used in historical perspective if the settings
have been changed

Gb over FR – QoS
Discarded frames
• Discarded sent frames • Discarded received frames
• Object aggregation: bearer_id • Object aggregation: bearer_id
• Time aggregation: hourly • Time aggregation: hourly
• Unit: % • Unit: %
sum(dlci_1_disc_sent_frms sum(dlci_1_disc_rec_frms
+ dlci_2_disc_sent_frms + dlci_2_disc_rec_frms
+ dlci_5_disc_sent_frms) + dlci_5_disc_rec_frms)
Counters from table(s): Counters from table(s):;

p_nbsc_frame_relay p_nbsc_frame_relay

Gb over IP
Introduction
• The increased demand for packet switched traffic
transmission cost efficiency can be met by deploying IP in the
transmission network
• IP offers an alternative way to configure the subnetwork of the
Gb interface:
– the subnetwork is IP-based and the physical layer is Ethernet
IP: 10.10.10.50
NSEI=1 NS_VC(NSEI+IP+UDP Port) NSEI=1
UPD port UPD port
50000 50000
UPD port UPD port

50001 50001
IP: 10.10.10.1
NSEI=2
IP: 10.10.10.60
NSEI=2
UPD port UPD port
50003 50000
UPD port UPD port

50004 50001
PAPU in SGSN PCUs in BSC

Gb over IP – Traffic volume
PCU Gb DL/UL load – gbip_6/gbip_5
• This KPI is used to monitor the PCU load in terms • This KPI is used to monitor the PCU load in terms
of Gb processing capacity of Gb processing capacity
• Object aggregation: bearer_id • Object aggregation: bearer_id
• Time aggregation: hourly • Time aggregation: hourly
• Unit: % • Unit: %
sum(payload_rcvd_from_sgsn_in_kb) * 8 sum(payload_sent_to_sgsn_in_kb) * 8
100 * -----------------------------------------------------------------
100 * ------------------------------------------------------------------
2000 * sum(period_duration*60)
2000 * sum(period_duration*60)
p_nbsc_gb_over_ip p_nbsc_gb_over_ip
• Note that this does not indicate Gb link load
as the actual bandwidth is not known. For • Comments are the same as case of DL load
the PCU1 and the first release of PCU2 the
processing capability in terms of Gb traffic
is 2Mbits/s.
• The 2Mbits/s is allowed at the same time in
DL and UL and it is not a hard limit. The
load may be higher for short time without
any internal overload, however, 2Mbits/s is
good limit for daily monitoring purposes.
The allowed load over one hour shall then
be smaller than 100%. How much smaller is
a good question. Basically depends on the
traffic burstiness.

Gb over IP – QoS
Packet drop ratio – gbip_2
• Packet drop ratio
• Unit: %
sum(ns_alive_ack_msgs_from_sgsn)
100 * (1- ---------------------------------------------------------------------------------------------------------)
sum(ns_alive_msg_sent_by_pcu + ns_alive_msgs_sent_pcu_retry)

p_nbsc_gb_over_ip
• The Packet drop ratio is one of the most useful KPIs to monitor Gb connection quality in Gb over IP
environment. This is actually a sample of dropped packets.
• NS-Alive message is sent frequently by the PCU and the purpose of the message is to monitor the NS level
connection availability. If the PCU do not receive the response after predefined number of retries it marks the
NS-VC unavailable.
• The acceptable threshold shall be less than 1%. Already 2% drop ratio end to end (MS to server) would be
would be seen for same applications (e.g. VoIP) some are more sensitive (1% end to end or less). Drops
occurs some where else as well so drops related to Gb should be much lower. (may be 0.2% to 1.0%).

(E)GPRS – E2E data rate
Volume weighted LLC throughput
• Volume-weighted LLC throughput (4 tsl capable EDGE
handsets, DL only, all QoS classes)
• Unit: kbps
sum(AVG_VOL_WGHT_LLC_TROUGHPUT_NUM) * 8
llc_3a = -------------------------------------
sum(AVG_VOL_WGHT_LLC_TROUGHPUT_DEN) * 10
• Also when MS gets less than 4 TSLs in DL, because of small GPRS territory or
concurrent UL&DL allocation => Congestion makes KPI show smaller value
• Short TBFs not taken into account based on byte threshold ( >1560 bytes)
• The counters in the formulas are only updated for 4-timeslot or more capable
EDGE handsets - but they are updated also in the case when less than 4 timeslots
are allocated to the MS. So radio interface congestion can make this KPI show low
values. Also EDAP congestion, bad radio link quality and bad transmission link
quality will make this KPI show low values.

(E)GPRS – E2E data rate
Volume weighted LLC throughput
Throughput calculation principle
• Traffic burst (burst)= Continuous data transfer from data arrives to PCU
buffer till PCU buffer gets empty
• VW LLC throughput = Sum of (Payloads per traffic burst * Throughput
per traffic burst) / Payload
Counters
• 72109 (AVG_VOL_WGHT_LLC_TROUGHPUT_NUM)= Sum of
(Payloads per burst * Throughput per burst) of EDGE 4 TSL MS (Units :
100B* B/10ms) (AVG_VOL_WGHT_LLC_TROUGHPUT_NUM)
• 72110 (AVG_VOL_WGHT_LLC_TROUGHPUT_DEN)= Payload of
EDGE 4 TSL MS (Unit: 100B)
• counters are used to minimize impacts in BSC - NetAct interface.
• Similar formulas pr QoS class for the following handsets:

– Volume weighted LLC throughput for GPRS: llc_4a
– Volume weighted LLC throughput for EDGE with other than 4 timeslot capable MSs:
llc_5a
– Volume weighted LLC throughput for EDGE with 4 timeslot capable MSs: llc_6

Mobility
Cell outage
PCU allocation
LA/RA design
NACC

Mobility Optimization
The aim of mobility optimization is to reduce the cell outage time during cell
re-selection.
Cell outage can be reduced by
• Providing enough signaling capacity for cell re-selection (the RACH, PCH,
AGCH and SDCCH channel are not limiting the signaling flow)
• Rebalancing BCFs among PCUs properly (the important neighbors are
allocated to the same PCU)
• Reallocating LA/RA borders properly
• Enabling Network Assisted Cell Change (NACC) feature

Outage Definition Used in Measurements
Three delays can be calculated from

logs:
MS GERAN
Cell outage:
• In one-phase access: the time
between the last EGPRS Packet
Downlink Ack/Nack message and
the first Packet Uplink Ack/Nack. Last IP packets Application Outage
• In two-phase access: the time EGPRS Packet Downlink Ack/Nack
between the last EGPRS Packet Routing Area Update Request
Downlink Ack/Nack message and Cell
the first Packet Uplink Data Packet Uplink Assignment Outage
Assignment. Outage
Routing Area Update Accept
Routing Area Update Complete

Data outage: the time between the EGPRS Packet Downlink Ack/Nack
last and the first EGPRS Packet
Downlink Ack/Nack message. First IP packets
Application outage: the time

between the last and the first
successfully received FTP-packet.

Cell Outage Time with intra/inter PCU -
Measurements
Test cases
Event name Time Channel Message
RLC/MAC Uplink
Layer 3 Downlink
20:42.0 PACCH "EGPRS_PACKET_DOWNLINK_ACK/NACK"
20:42.0 BCCH "SYSTEM_INFORMATION_TYPE_1"
Intra PCU cell-reselection outage
Layer 3 Downlink 20:42.0 BCCH "SYSTEM_INFORMATION_TYPE_2"
Layer 3 Downlink 20:42.0 BCCH "SYSTEM_INFORMATION_TYPE_3" Inter PCU cell-reselection outage
…
…
Cell Reselection 20:42.8 from CI 5032 to CI 5033
…
…
Layer 3 Uplink 20:43.1 RACH "CHANNEL_REQUEST"
Layer 3 Downlink 20:43.2 CCCH "IMMEDIATE_ASSIGNMENT"
From EGPRS PACKET DOWNLINK ACK/NACK Diff. Between BSS and
Layer 3 Downlink 20:43.2 CCCH "PAGING_REQUEST_TYPE_1"
till Packet Uplink Assign. (ms) till Packet Downlink Assign. (ms) full cell-outage (ms)
Layer 3 Downlink 20:43.2 CCCH "PAGING_REQUEST_TYPE_1"
Layer 3 Downlink 20:43.3 CCCH "PAGING_REQUEST_TYPE_1" 2.07 2.341 0.27
Layer 3 Downlink 20:43.3 BCCH "SYSTEM_INFORMATION_TYPE_2" 3.07 3.349 0.28
2.09 2.354 0.26
…
2.10 2.358 0.26

Layer 3 Downlink 20:43.8 BCCH "SYSTEM_INFORMATION_TYPE_13" 2.09 2.375 0.28
RLC/MAC Uplink 20:43.8 PACCH "PACKET_RESOURCE_REQUEST" 2.11 2.393 0.28
RLC/MAC Downlink 20:44.0 PACCH "PACKET_UPLINK_ASSIGNMENT" 2.10 2.658 0.56
RLC/MAC Downlink 20:44.0 PACCH "PACKET_DOWNLINK_DUMMY_CONTROL_BLOCK" 2.09 2.355 0.26
…
2.10 2.38 0.28

RLC/MAC Downlink 20:44.2 PACCH "PACKET_DOWNLINK_ASSIGNMENT" 2.12 2.379 0.26
RLC/MAC Uplink 20:44.3 PACCH "EGPRS_PACKET_DOWNLINK_ACK/NACK" 2.094 2.629 0.54
2.09 2.631 0.54
2.14 2.7 0.56
2.07 4.011 1.94
2.10 2.379 0.28
2.10 2.385 0.28
2.37 2.80 0.43

Cell Outage Time with LAU/RAU - Measurements
Event name Time Channel Message
…
Layer 3 Downlink 8:44:05.801 BCCH "SYSTEM_INFORMATION_TYPE_1"
…
Cell Reselection 8:44:10.906 from 5691 to 5753
…
Layer 3 Uplink 8:44:11.997 RACH "CHANNEL_REQUEST"
Layer 3 Downlink 8:44:12.101 CCCH "IMMEDIATE_ASSIGNMENT"
Layer 3 Uplink 8:44:12.313 SDCCH "LOCATION_UPDATING_REQUEST"
Layer 3 Downlink 8:44:12.353 SACCH "SYSTEM_INFORMATION_TYPE_6"
Layer 3 Uplink 8:44:12.388 SACCH "MEASUREMENT_REPORT"
Layer 3 Uplink 8:44:12.548 SDCCH "CLASSMARK_CHANGE"
Layer 3 Downlink 8:44:12.764 SDCCH "CIPHERING_MODE_COMMAND"
Layer 3 Uplink 8:44:12.784 SDCCH "GPRS_SUSPENSION_REQUEST"
Layer 3 Uplink 8:44:13.020 SDCCH "CIPHERING_MODE_COMPLETE"
Layer 3 Downlink 8:44:13.224 SDCCH "IDENTITY_REQUEST"
Layer 3 Uplink 8:44:13.490 SDCCH "IDENTITY_RESPONSE"
Layer 3 Downlink 8:44:13.697 SDCCH "LOCATION_UPDATING_ACCEPT"
Layer 3 Downlink 8:44:14.168 SDCCH "MM_INFORMATION"
Layer 3 Downlink 8:44:14.399 SDCCH "CHANNEL_RELEASE"
…
…
Layer 3 Uplink 8:44:16.258 PDTCH "ROUTING_AREA_UPDATE_REQUEST"
…
…
Layer 3 Uplink 8:44:16.752 RACH "CHANNEL_REQUEST"
Layer 3 Downlink 8:44:16.829 CCCH "IMMEDIATE_ASSIGNMENT"
…
…
Layer 3 Uplink 8:44:16.258 PDTCH "ROUTING_AREA_UPDATE_REQUEST"
RLC/MAC Uplink 8:44:17.401 PACCH "PACKET_RESOURCE_REQUEST"
RLC/MAC Downlink 8:44:17.607 PACCH "PACKET_UPLINK_ASSIGNMENT"
…
…
RLC/MAC Downlink 8:44:17.886 PACCH "PACKET_DOWNLINK_ASSIGNMENT"
…
Layer 3 Downlink 8:44:18.950 PDTCH "ROUTING_AREA_UPDATE_ACCEPT"

Layer 3 Uplink 8:44:18.964 PDTCH "ROUTING_AREA_UPDATE_COMPLETE"
RLC/MAC Uplink 8:44:19.119 PACCH "EGPRS_PACKET_DOWNLINK_ACK/NACK"
…

Cell Outage Time with LAU/RAU - Measurements
Test cases
LAU/RAU cell-outage
Time to LAC [sec.] Time for RAC [sec.] Full LAU/RAU [sec.]
4.056 3.37 10.227
2.808 3.15 8.785
3.886 2.99 9.61
3.813 2.96 9.573
3.814 3.53 10.146
2.888 2.925 8.602
2.922 2.949 8.617
3.042 2.953 8.618
2.868 2.637 8.326
3.048 2.955 8.832
3.82 2.10 8.635
2.865 3.32 8.8
2.811 2.95 8.618
1.045 3.301 11.088
2.918 2.953 8.616
3.11 3.00 9.14

Signaling
The capacity bottlenecks on signaling channel can increase TBF

establishment, hereby cell re-selection outage
Therefore the following signaling channels must be optimized:
• Air interface
• TRXSIG
• BCSU with PCU
For more information, refer to Signaling Capacity Optimization and Resource

Allocation module of this material

PCU Allocation Optimization
The proper allocation of the cells among PCUs can help to maximize the number of intra PCU
cell re-selections, which is the most stable and shortest cell re-selection event.
• RLC/MAC layer: The intra PCU cell re-selection takes less time compared with inter PCU cell
reselection
• LLC layer: In case of intra PCU cell re-selection the untransferred data is moved to new cell
(BVCI) and the transfer can be continued on new cell without packet loss on higher layer,
while in case of inter PCU cell re-selection the untransferred data is not moved to new cell
(BVCI).
The following main rules must be followed:

• The cells of a BCF should be connected to the same PCU
• The neighbor relations with high re-selection traffic should be connected to the same PCU
• The neighbor relations in very bad signal and quality environment should be connected to the
same PCU
NACC and NCCR can be used if there is not any possibility to connect the neighbor cells to the
same PCU (NACC is working inside BSCs only)

PCU Allocation Optimization Hints
Connectivity related (optimizing PCU usage)

• Take the live traffic figures into account, so the less traffic sites can have less CDEF, while
sites with more traffic can have more CDEF
• Try to allocate min 4TSL DAP to all the BCFs
• Try to allocate max 8DAP pools to one PCU (For PCU1 the 1, 2 or 4 are even better)
• Try to keep balanced traffic allocation among PCUs (and inside DSPs, too)
• Try to keep the 75 % of connectivity limits for acceptable (E)GPRS data rate
• Try to connect less than 64 radio TSLs to one PCU to maximize EGPRS coverage in case of
GPRS-EGPRS traffic mix
Mobility related (reducing interPCU cell re-selection)
• Take the neighbor traffic figures into account in BCF allocation among PCUs to reduce
interPCU cell re-selection
• Allocate all the BTSs of one BCF into one PCU
• Modify (if it is needed) the LA/RA border for reducing the interPCU and interBSC cell re-
selection
• Do not connect GPRS and EGPRS BCFs to separated PCU in case of island coverage,
because of mobility issues and PCU and PCU-S limits with S11.5 in GPRS only networks

LA/RA Design – Radio Aspects
Important to avoid LA/RA border allocation between cell with high neighboring traffic
Usage of NMO I, where the combined RA reduces the cell re-selection time
Radio Aspect of LA/RA Design

• The too big LA/RA will increase the paging, while the too small LA/RA will increase the
LA/RA Update. So the balance should be found between too big and too small RA/RAs.
• The not so appropriate LA/RA border design can significantly increase the signaling on air
interface signaling channels and TRXSIG on LA/RA border cells, so the cell-reselection
outage can be longer in this case because of congestion on signaling.
• The LA/RA border should be moved from those areas where the normal CSW and PSW
traffic is very high.
• The combined RAU (NMO I with Gs) is shorter compared to MNO II
• In S11 backwards the GPRS resume always can cause a lot of RAs if GPRS MS has high
CS call activity, but this behavior cannot be avoided by proper LA/RA design
– In S11.5 the Resume is working without LA/RA update

LA/RA Design – Large Routing Area Support
This feature makes it possible to define PAPU groups where multiple PAPUs can
serve a single routing area or network service entity (NSE). It is also possible to
configure more than one RA/NSE to a group of PAPUs.
The Large Routing Area Support feature has affect on:

• PAPU dimensioning
• BSS LA/RA dimensioning
Operator benefits
• With Large Routing Area Support feature the LA or RA can be handled in several
PAPUs  subscribers per RA can be up to 320 000.
• Network management is made easier with automatic cell distribution and you can
achieve evenly distributed load between the units.
• With the cell distribution improvement, the downlink data forwarding between the
PAPU units decreases.

Network Assisted Cell Change (NACC)
NACC is in Rel 4 of 3GPP GERAN , mandatory for R4 mobiles. Nokia’s

S11.5 implementation is based on Rel4.
Both, autonomous and network controlled cell reselections are supported.
Support is for intra-BSC cell changes (Support of inter-BSC NACC
specification is available in Rel 5)
NACC support is for MSs in RR Packet Transfer Mode only
NACC shortens the cell reselection in two ways:
•Sending neighbour cell system information on PACCH to MS in packet

transfer mode while it is camped on the serving cell
•By supporting PACKET SI STATUS procedure in a target cell

Introduction
Outage measurements were done in Nokia SyVe live test network in June 2005 and
Nokia Test Network (NTN) in July 2005.
The impact on cell reselection delay of following features are measured:
• Network Assisted Cell Change (NACC) enabled/disabled
• EGPRS Packet Channel Request (EPCR) on/off

– EPCR is available with UltraSite (CX3.3 or newer SW).
• Network Controlled Cell Reselection (NCCR)
The impact of Routing Area Update (RAU) on cell reselection delay.

Measurement Setup 1
Outage tests were done in Nokia SyVe and NTN live test networks.
Following cases were tested:
• Cell reselection + NACC disabled
• Cell reselection + NACC enabled
In these three conditions:

• Intra-BSC, intra-RA, intra-PCU, EPCR = enabled (one-phase access).
• Intra-BSC, intra-RA, intra-PCU, EPCR = disabled (two-phase access).
• Intra-BSC, inter-RA, inter-PCU, EPCR = disabled (two-phase access).
Using these mobile phones: EGPRS Packet Channel Request (EPCR) is an enhancement
to speed up the EGPRS TBF establishment.
• Nokia 5140 and Nokia 6230i.
With this feature MS can get EGPRS TBF with one-phase
access.
Continuous FTP download. HW/SW requirements: UltraSite with CX3.3 or newer SW.
No operator parameter to enable/disable.
10 cell reselections for each test case.

NC0 = “normal” cell reselection
Cell Reselection Delays
One-phase access, Intra-RA, Intra-PCU
Nokia 5140
2.5
Application Outage
Time (s)
1.5 Data Outage

Cell Outage
0.5 36% drop in

outage times
0
NC0 NC0 + NACC
NC0 without NACC and with NACC

Two-phase-access, Intra-RA, Intra-PCU
Nokia 6230i
3.5
2.5
2
Application Outage
Time (s)
Data Outage
1.5
Cell Outage
47% drop in
0.5 outage times
NC0 NC0 + NACC

Network Control Mode

Two-phase access, Inter-RA, Inter-PCU
Nokia 6230i
4
Time(s)
Application Outage
3 Data Outage
Cell Outage
1
38% drop in
outage times
0
NC0 NC0 + NACC

Measurement Setup 2
The following slides show the measurement test results collected for analyzing the impact of
NCCR on cell reselection outage time
Following cases were tested:
• NCCR + NACC disabled
• NCCR + NACC enabled
In these three conditions:

• Intra-BSC, intra-RA, intra-PCU, EPCR = enabled (one-phase access).
• Intra-BSC, intra-RA, intra-PCU, EPCR = disabled (two-phase access).
• Intra-BSC, inter-RA, inter-PCU, EPCR = disabled (two-phase access).
Using these mobile phones:

• Nokia 5140 and Nokia 6230i.
Continuous FTP download.

10 cell reselections for each test case.

Nokia 6230i
NCCR does not
3.5 increase the cell
reselection delay
3
2.5
2 Application Outage
Time (s)
Data Outage
1.5
Cell Outage
0.5
0
Normal Cell Reselection NCCR

NC2 ~800ms longer than
NC2 = NCCR
Cell Reselection Delays normal cell reselection
due to N5140 software
One-phase access, Intra-RA, Intra-PCU bug
Nokia 5140
3.5
2.5
2
Application Outage
Time (s)
Data Outage
1.5 Cell Outage
57% drop in
0.5 outage times
0
NC2 NC2 + NACC

NC2 = NCCR
Two-phase-access, Intra-RA, Intra-PCU
Nokia 6230i
3.5
2.5
2
Application Outage
Time (s)
Data Outage
1.5
Cell Outage
46% drop in
0.5 outage times
NC2 NC2 + NACC


NC2 = NCCR Cell Reselection Delays
Two-phase access, Inter-RA, Inter-PCU
Nokia 6230i
Application Outage
Time(s)
3 Data Outage
Cell Outage
34% drop in
1
outage times
0
NC2 NC2 + NACC

NC2 = NCCR
Data Outage in Cell Reselection
Nokia 6230i
4.5
One-Phase access, Intra-RA, Intra-PCU

4
Two-Phase access, Intra-RA, Intra-PCU
Two-Phase access, Inter-RA, Inter-PCU
3.5
3
41%
2.5
Time (s)
1.5
Data Outage
time drops 1
41% with one-
phase access.
(=EPCR in 0.5
use)
0
NC0 NC0 + NACC NC2 NC2 + NACC
This was measured with an old N6230i
where the SW had the bug. In reality time
should equal NC0s time for the same case.
Network Assisted Cell Change (NACC)
Ratio of NACC usage for NC0 cell reselections

• Shows the ratio between NC0 cell reselections that have used NACC and all NC0 cell
reselections
• Unit: %
NACC_WITH_NC0
nccr_15 = 100% * ------------------------------------------------------
FLUSH_MSGS_RECEIVED - NCCR_SUCC_FLUSH_RECEIVED
Ratio of NACC usage for NC2 cell reselections

• Shows the ratio between NC2 cell reselections that have used NACC and all NC2 cell
reselections
• Unit: %
NACC_WITH_NC2
nccr_16 = 100% * ------------------------------------
PCCO_TO_GPRS_MS_DUE_PWR_BDGT
+ PCCO_TO_EGPRS_MS_DUE_PWR_BDGT
+ PCCO_SENT_DUE_SERV_ISNCCR
+ PCCO_SENT_DUE_COVERAGE_ISNCCR
+ PCCO_SENT_DUE_QUAL_CTRL

(E) GPRS BSS Optimization - S12

Uploaded by

(E) GPRS BSS Optimization - S12

Uploaded by

(E)GPRS BSS Optimization – S12

1 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

2 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

3 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

MS BSS SGSN GGSN FIXED HOST

GPRS IP Backbone Internet

4 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

RLC/MAC Blocks Note: Amount of RLC

4 x TDMA Frames = 4 Bursts = 1 Radio block = 18.46 ms = 1-2 RLC block(s)

52 TDMA Frames (240 ms)

5 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

(1,0,0) • Because of amplifier non-linearities,

EDGE GSM + EDGE

6 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

Phase state vector diagram

•Amplitude is not fixed

• Since the amplitude is changing in 8-PSK the transmitter non-linearities can be

• The back-off value is taken into Pout

UL / DL and bands: 900/850,

8 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

TSL0 TSL1 TSL2 TSL3 TSL4 TSL5 TSL6 TSL7

9 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

1. Spectrum of Unfiltered 3pi/8 8psk modulation.

10 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

Coding Payload (bits) Data Rate

Nokia GPRS CS1 181 9.05

• CS1 & CS2 – Implemented in all Nokia BTS without

11 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

CS-2 Block Check Sequence: 40 16 16

length: 456 588 676

puncturing 0 132 220

57 57 57 57 57 57 57 57 Data rate (kbit/s): 9.05 13.4 15.6

MAC RLC/MAC Block Size: 428

USF BCS BCS Size: 16

Data rate (kbit/s): 21.4

• Each family has a different basic unit of payload: MCS- 6

units are transmitted MCS- 2

MSC-8 and MCS-9), these are splitted into two MCS- 5

BSN numbers and BCS, Block Check MCS- 1

14 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

• Coding of data part: protecte

• Header part: mother code

• Includes RLC/MAC header information and 1/3 convolutional coding

15 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

17 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

• Routing Areas are used for GPRS Mobility Management

Periodic RA Update Timer (PRAU) – Default: 54

18 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

19 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

A Temporary Block Flow (TBF):

Uplink TBF (+ PACCH for downlink TBF)

20 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

Non-DTM capable MS:

DTM capable MS:

21 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

22 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

Usually the business plan requirements are mainly related to:

23 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

24 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

level can be analyzed by:

– Planning tool plots 80

– Drive test measurements C/I 25 dB

25 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

26 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

The aim behind the preparation of deployment plan:

•Maximize the TSL data rate (RLC/MAC) and multislot usage

Most of the networks can be described by few cell/segment options

27 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

Cell / Segment option examples

28 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

The air interface capacity (TSLs) is suitable, if it is >= 4,

29 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

Dedicated TSL(s) is (are) needed to provide:

30 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

31 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.

32 © Nokia Siemens Networks (E)GPRS BSS Optimization – S12 / 06-08.10.2008.