Amr Trial Final Report_v2

7/30/2019 Amr Trial Final Report_v2

http://slidepdf.com/reader/full/amr-trial-final-reportv2 1/51

Done by: Document Date Page

Walied SAMIE AMR 29/01/2008 1

ADAPTIVE MULTI RATE

Field Trial and Implementation





Table of Contents:

1. INTRODUCTION ..................................................................................................................3

2. WHY AMR? .............................................................................................................................3

3. FUNCTIONALITY .................................................................................................................3

4. MECHANISM ..........................................................................................................................4

5. HO CAUSES ............................................................................................................................8

6. AMR PARAMETERS AND DEFAULT VALUES .......................................................13

7. BLIDA TRIAL.......................................................................................................................14

A. NETWORK STATUS ................................................................................................................14 B. R ADIO STATISTICS ................................................................................................................14 C. QVOICE STATISTICS...............................................................................................................16 D. QOS DEGRADATION I NVESTIGATIONS...................................................................................17 E. CHANNEL ADAPTATION ACTIVATION ...................................................................................21

8. ALGIERS AMR IMPLEMENTATION ..........................................................................23

A. ALGIERS NETWORK STATUS .................................................................................................23 B. QVOICE STATISTICS...............................................................................................................24 C. IMPLEMENTATION SCENARIOS ..............................................................................................27

9. NETWORK AMR IMPLEMENTATION .......................................................................28

A. ALCATEL NETWORK STATUS ................................................................................................28 B. NSS ACTIVATION..................................................................................................................29 C. AMR TARGET FOR 2008 ........................................................ ............................................... 29

10. HO FAILURES PROBLEM ...............................................................................................30

A. PROBLEM DESCRIPTION ........................................................................................................30 B. DRIVE TEST I NVESTIGATION .................................................................................................30

11. FURTHER TRIALS AND RECOMMENDATIONS...................................................32

A. E NABLING A 4TH CODEC .......................................................................................................32 B. AMR_FR DISABLE ...............................................................................................................32 C. E NABLING CHANNEL ADAPTATION.......................................................................................32 D. RLT AND AMR.....................................................................................................................33 E. AMR FOLLOW UP: ....................................................... ......................................................... 33

12. ANNEX A-B9 NEW FEATURES ....................................................................................34

13. ANNEX B-PARAMETERS ................................................................................................36

14. ANNEX C ................................................................................................................................45







4. MECHANISM

o CODEC mode adaptation (ADAPTIVE)

• By in-band signalling interchange between the MS and theBTS, the BTS can dynamically adapt the CODEC mode(between four modes in the course of a call)

• The different CODEC modes use different size speech and

channel encoding fields (moving boundary according toradio conditions).

• Codec mode used are:1. Full Rate:

a. 12.2 Kbpsb. 10.2 Kbpsc. 7.95 Kbps

d. 6.7 Kbpse. 5.9 Kbps

f. 5.15 Kbpsg. 4.75 Kbps

2. Half Rate:a. 6.7 Kbpsb. 5.9 Kbpsc. 5.15 Kbps

d. 4.75 Kbps

Speech Redundancy

Speech Encoding Channel Encoding

Source code according to bit rate: 12.2, 7.95… according to codec mode

22.8 Kbps (FR) or 11.4 kbps (HR)

Moving boundary according to Codec mode





• The AMR principle is to have a set of codecs to use the one

with the best speech quality. Under good radio conditions, a codec with a high bit-rate is used. Speech is

encoded with more information so the quality is better. Inthe channel coding, only little place is left for redundancy.Under poor radio conditions, a codec with a low bit-rateis chosen. Speech is encoded with less information, butthis information can be well protected due to redundancyin the channel coding. The quality of the radio link isestimated by evaluating the carrier to interference ratio

C/I.

• A different codec can be used in the uplink and downlink.This permits to use an optimal codec for each C/I value of each direction.

• Uplink codec mode adaptation: For each SACCH frame, the BTS compares C/I value

to the threshold corresponding to the current codec

(belonging to the codec subset defined by theoperator)

• Downlink codec mode adaptation: Same process as uplink adaptation. Nevertheless, the

BTS remains the master





• For each pair of adjacent codecs, a threshold and ahysteresis is given. The figure below shows the case wherefour codecs are included in the subset.





o Channel mode adaptation (MULTI-RATE)

• The FR or HR mode is selected by the BSC depending on

the traffic load of the cell:

o Dynamic adaptation (FR<->HR handover) istriggered as intracell HO; cause 26 and 27.

o It is based on the result of measurementsperformed by the MS on the uplink anddownlink radio channels between the MS and

the BTS; RxLev and RxQual.





5. HO Causes

For the Channel Mode Adaptation, we have the following HO causes:

HO family HO Cause

HO cause

Reference

Emergency HO Too high interference intracell Uplink Cause = 15

Emergency HO Too high interference intracell Downlink Cause = 16

Channel adaptation HO HR to FR channel adaptation due to bad quality Cause = 26

Channel adaptation HO FR to HR channel adaptation due to good quality Cause = 27

o Cause 15 and 16

Emergency handovers Causes 15 and 16 are triggered forintracell application when the radio link is deemed to suffer a highlevel of interference. In this case, the channel assigned to the callis changed for another channel in the same cell, on which themeasured interference level is the smallest possible. Since AMRcalls can be performed over worse carrier-to-interference ratiosthan non AMR calls, the parameter setting for Causes 15 and 16 is

different for non AMR and AMR calls.

CAUSE = 15 (too high interference level on the uplink)

AV_RXQUAL_UL_HO > THR_RXQUAL_CAUSE_15 + OFFSET_RXQUAL_FH

and AV_RXLEV_UL_HO > RXLEV_UL_IHand EN_CAUSE_15 = ENABLE

and (no previous intracell handover for this connection failed orEN_INTRACELL_REPEATED = ENABLE )

Two sets of parameters are defined to control Cause 15whether the current call is AMR or not:

If the current call is not an AMR call,EN_CAUSE_15 = EN_INTRA_UL,THR_RXQUAL_CAUSE_15 = L_RXQUAL_UL_H.

If the current call is an AMR call,EN_CAUSE_15 = EN_INTRA_UL_AMR ,

THR_RXQUAL_CAUSE_15 = L_RXQUAL_UL_H_AMR .





o Cause 26 and 27

At call setup, if the MS supports AMR HR and the cell is enabled for AMRHR then the AV_LOAD for the cell is calculated and compared with the threshold

values THR_FR_LOAD_L_SV3 and THR_FR_LOAD_U_SV3 to determine thevariable LOAD_SV3, otherwise the LOAD_SV3 is set to FALSE.

If LOAD_SV3 is found to be TRUE (High Load and AMR HRenabled), then an AMR HR channel is allocated.

If LOAD_SV3 is found to be FALSE (Low Load or AMR HRdisabled), then an AMR FR channel is allocated.

Two new handover causes are introduced for AMR channel mode adaptation

(changing from FR to HR and vice versa) for AMR calls: Cause 26: HR to FR channel adaptation due to low radio quality Cause 27: FR to HR channel adaptation due to high radio

qualityThese two causes are enabled by a single flag: EN_AMR_CA. These two causesof intracell handover belong to a new family called "Channel adaptation HO".

CAUSE = 16 (too high interference level on the downlink)

AV_RXQUAL_DL_HO > THR_RXQUAL_CAUSE_16 +OFFSET_RXQUAL_FH

and AV_RXLEV_DL_HO > RXLEV_DL_IH

and EN_CAUSE_16 = ENABLEand (no previous intracell handover for this connection failed or

EN_INTRACELL_REPEATED = ENABLE )Two sets of parameters are defined to control Cause 16 whether thecurrent call is AMR or not:

If the current call is not an AMR call

EN_CAUSE_16 = EN_INTRA_DL,THR_RXQUAL_CAUSE_16 = L_RXQUAL_DL_H.

If the current call is an AMR call EN_CAUSE_16 = EN_INTRA_DL_AMR ,THR_RXQUAL_CAUSE_16 = L_RXQUAL_DL_H_AMR .





As such, They are checked after intercell emergency handover causes. While cause 26 (HR-FR) is checked before intracell interference handovers,

Cause 27 is checked after them.

That is because while Cause 27 is considered a handover to optimize theresource usage. Cause 26 is considered as urgent in order to save the callfrom the bad speech quality.

Cause 26 (HR to FR channel adaptation) must satisfy the intra-cellinterference HO level condition.

Cause 26 (HR to FR channel adaptation due to bad quality):

There are two ways to trigger Cause 26:

The first way consists in triggering Cause 26 only if a previous intracell

handover Cause 15 or 16 has been previously detected in the serving cell forthe current MS. This way is intended to non-hopping channels for which an

intracell handover Cause 15 or 16 is sometimes sufficient to improve thequality of the call. If the quality is not sufficient after an intracell HO cause 15or 16 due to a too high interference level, instead of continuing triggeringintracell handover Cause 15 or 16, a HR-to-FR channel adaptation is triggeredthanks to Cause 26.

The second way applies when the intracell handover Causes 15 and 16 are

both disabled for AMR calls(EN_INTRA_DL_AMR = EN_INTRA_UL_AMR = DISABLE). If a too highlevel of interference is detected in the serving cell for the current MS, Cause26 is then triggered directly. This second way intends to improve the quality of hopping channels which quality is generally not much improved after anintracell handover Cause 15 or 16. Since AMR is performing well in interferenceconditions, the thresholds for cause 15 & 16 are replaced especially for AMR

(L_RXQUAL_UL/DL_H replaced by L_RXQUAL_UL/DL_H_AMR )





AV_RXQUAL_UL_CA_HR_FR and AV_RXQUAL_DL_CA_HR_FR are theaverages of, respectively, the uplink and downlink RxQual measured bythe MS/BTS, averaged over A_QUAL_CA_HR_FR measurements

(Default:8).

Cause 27 (FR to HR channel adaptation due to good quality):

Channel adaptation handover aims to reduce the number of busy FR TCH whenthe radio quality is very good and the serving cell becomes loaded. Thechannel adaptation consists in changing the current full rate TCH (AMR FR) toa half rate TCH (AMR HR).

Current channel is Full Rate and EN_AMR_HR = enabledAND Channel type is “dual rate, changes allowed” AND AV_RXQUAL_UL_CA_FR_HR <= THR_RXQUAL_CA + OFFSET_RXQUAL_FHAND AV_RXQUAL_DL_CA_FR_HR <= THR_RXQUAL_CA + OFFSET_RXQUAL_FHAND EN_AMR_CA = enabled

The current channel is dual rate and changes allowed

AND EN_AMR_FR = enableAND{{AV_RXQUAL_UL_CA_HR_FR > THR_RXQUAL_CA+ OFFSET_CA +OFFSET_RXQUAL_FHand AV_RXLEV_UL_HO > RXLEV_UL_IH }OR{AV_RXQUAL_DL_CA_HR_FR > THR_RXQUAL_CA + OFFSET_CA +OFFSET_RXQUAL_FH

and AV_RXLEV_DL_HO > RXLEV_DL_IH }}

AND EN_AMR_CA = enableAND

{(a previous intracell HO Cause 15 or 16 has been raised for this call in theserving cell) OR(EN_INTRA_UL_AMR = disable and EN_INTRA_DL_AMR = disable)}





Similarly,

AV_RXQUAL_UL_CA_FR_HR and AV_RXQUAL_DL_CA_FR_HR are theaverages of, respectively, the uplink and downlink RxQual measured by

the MS/BTS, averaged over A_QUAL_CA_FR_HR measurements(Default: 61).

Having 2 different averaging window parameters for causes 26 and 27makes the algorithm more dynamic and makes it possible to facilitate theQuality HO cause 26 while delaying the Better Cell HO cause 27 to avoid aping pong of channel modes.

According to the load of the serving cell, the variables THR_RXQUAL_CA and OFFSET_CA are set as follows:

If LOAD_SV3(0) = falseTHR_RXQUAL_CA = THR_RXQUAL_CA_NORMALOFFSET_CA = OFFSET_CA_NORMAL

If LOAD_SV3(0) = trueTHR_RXQUAL_CA = THR_RXQUAL_CA_HIGH

OFFSET_CA = OFFSET_CA_HIGH





6. AMR Parameters and Default Values

CELL Parameter Default Value Comments

EN_AMR_FR Enable

EN_AMR_HR Enable Only for Cells with HR enabled

EN_INTRA_UL_AMR EN_INTRA_UL EN_INTRA_UL

EN_INTRA_DL_AMR EN_INTRA_DL EN_INTRA_DL

L_RXQUAL_UL_H_AMR 45 (x 0.1)

L_RXQUAL_DL_H_AMR 45 (x 0.1)

DTX_Indicator_FR_AMR SHALL USE

DTX_Indicator_HR_AMR SHALL NOT USE

AMR_FR_THR_1 13 (x 0.5)

AMR_FR_THR_2 23 (x 0.5)

AMR_FR_THR_3 23 (x 0.5)

AMR_HR_THR_1 22 (x 0.5)

AMR_HR_THR_2 25 (x 0.5)

AMR_HR_THR_3 25 (x 0.5)

AMR_FR_HYST 4 (x 0.5)

AMR_HR_HYST 4 (x 0.5)

EN_AMR_CA Disable

A_QUAL_CA_HR_FR 8 A_QUAL_CA_HR_FR <= A_QUAL_HO

A_QUAL_CA_FR_HR 61

W_QUAL_CA 1THR_RXQUAL_CA_NORMAL 0

THR_RXQUAL_CA_HIGH 0

OFFSET_CA_NORMAL 45 (x 0.1)

OFFSET_CA_HIGH 45 (x 0.1)

THR_FR_LOAD_L_SV3 VALUE ONAIR THR_FR_LOAD_L_SV1

THR_FR_LOAD_U_SV3 VALUE ONAIR THR_FR_LOAD_U_SV1

BSC Parameter Default Value Comments

Downlink_DTX_enable_AMR_FR DISABLE

Downlink_DTX_enable_AMR_HR DISABLE

Forbid_AMR_NS ENABLE

AMR_FR_SUBSET 12.2, 7.95, 5.9

AMR_HR_SUBSET 7.4, 5.9, 5.15

AMR_START_MODE_FR 0 lowest bit rate

AMR_START_MODE_HR 0 lowest bit rate

Parameters descriptions are in Annex A.





7. BLIDA Trial

a. Network Status

AMR was activated in Blida on 3 BSCs with B8 MR7 ed2 release. The Trialis carried in Blida Down Town area which is characterized by its high traffic cells

with dense sites.

b. Radio Statistics

The statistics showed an increase of DL Quality HO while AMR wasenabled. DL_DTX_AMR was enabled after the AMR activation. It was noticed that

by deactivating the DL_DTX_AMR, the quality HO reduced, also Externalinterference was detected in the network in this time.

Quality HO:

Quality handovers

0

50000

100000

150000

200000

250000

300000

350000

0 1 / 0

1 / 2 0

0 7

0 1 / 0 3 / 2 0

0 7

0 1 / 0 5 / 2 0

0 7

0 1 / 0 7 / 2 0

0 7

0 1 / 0 9 / 2 0

0 7

0 1 / 1

1 / 2 0

0 7

0 1 / 1 3 / 2 0

0 7

0 1 / 1 5 / 2 0

0 7

0 1 / 1 7 / 2 0

0 7

0 1 / 1 9 / 2 0

0 7

0 1 / 2

1 / 2 0

0 7

0 1 / 2 3 / 2 0

0 7

0 1 / 2 5 / 2 0

0 7

0 1 / 2 7 / 2 0

0 7

0 1 / 2 9 / 2 0

0 7

0 1 / 3

1 / 2 0

0 7

0 2 / 0 2 / 2 0

0 7

0 2 / 0

4 / 2 0

0 7

0 2 / 0 6 / 2 0

0 7

0 2 / 0 8 / 2 0

0 7

0 2 / 1 0 / 2 0

0 7

0 2 / 1 2 / 2 0

0 7

0 2 / 1

4 / 2 0

0 7

0 2 / 1 6 / 2 0

0 7

0 2 / 1 8 / 2 0

0 7

0 2 / 2 0 / 2 0

0 7

0 2 / 2 2 / 2 0

0 7

0 2 / 2

4 / 2 0

0 7

0

5

10

15

20

25

UL

DL

% Quality

AMR with DL_DTX non AMR with DL_DTX

AMR without DL_DTX + Externa l interference

AMR without DL_DTX

o AMR with DL_DTX_AMR (1/1 – 19/1): 19.43%

o Non AMR with DL_DTX (20/1 – 17/2): 16.39%

o AMR without DL DTX+ External Interference (18/2 – 23/2): 17.79%

o AMR without DL_DTX_AMR (24/2 – 25/2): 17.08%

Quality degraded by 0.7% (comparing period 2 and 4)





Call Drop:

In terms of CDR a small degradation was noticed with AMR enabled.

Call drop

0

5000

10000

15000

20000

25000

0 1 / 0

1 / 2 0

0 7

0 1 / 0 3 / 2 0

0 7

0 1 / 0 5 / 2 0

0 7

0 1 / 0 7 / 2 0

0 7

0 1 / 0 9 / 2 0

0 7

0 1 / 1

1 / 2 0

0 7

0 1 / 1 3 / 2 0

0 7

0 1 / 1 5 / 2 0

0 7

0 1 / 1 7 / 2 0

0 7

0 1 / 1 9 / 2 0

0 7

0 1 / 2

1 / 2 0

0 7

0 1 / 2 3 / 2 0

0 7

0 1 / 2 5 / 2 0

0 7

0 1 / 2 7 / 2 0

0 7

0 1 / 2 9 / 2 0

0 7

0 1 / 3

1 / 2 0

0 7

0 2 / 0 2 / 2 0

0 7

0 2 / 0

4 / 2 0

0 7

0 2 / 0 6 / 2 0

0 7

0 2 / 0 8 / 2 0

0 7

0 2 / 1 0 / 2 0

0 7

0 2 / 1 2 / 2 0

0 7

0 2 / 1

4 / 2 0

0 7

0 2 / 1 6 / 2 0

0 7

0 2 / 1 8 / 2 0

0 7

0 2 / 2 0 / 2 0

0 7

0 2 / 2 2 / 2 0

0 7

0 2 / 2

4 / 2 0

0 7

0

0.2

0.4

0.6

0.8

1

1.2

Preemption

Drop - BSS

Drop - HO

Drop - Radio

% RTCH drop

% Call Drop

AMR with DL_DTX non AMR with DL_DTX

AMR without DL_DTX + External interference

AMR without DL_DTX

o AMR with DL_DTX_AMR (1/1 – 19/1): 0.86%

o Non AMR with DL_DTX (20/1 – 17 0.78%

o AMR without DL_DTX+ External Interference (18/2 – 23/2): 0.84%

o AMR without DL_DTX_AMR (24/2 – 25/2): 0.81%

Call Drop degraded by 0.03% (comparing period 2 and 4).

Speech version allocations

The average distribution of Allocations:

Speech Version Average % of Allocation

FR 0.90%

HR 9.45%

EFR 34.45%

AMR_FR 45.76%

AMR_HR 13.47%

Data 0.01%





c. Qvoice Statistics

We made Qvoice campaigns in the concerned area to evaluate the effect of the AMR on the speech quality.

Three different campaigns were done with the following features:

October 2006 : Reference campaigns November 2006 : AMR activated with DL_DTX_AMR disabled January 2007 : AMR activated with DL_DTX_AMR enabled

Voice Quality per Codec

Voice Quality Evolution per Codec

0 0

3.583.68

3.59

3.7

3.39

3.71

3.41

3.723.59

3.4

0

0.5

1

1.5

2

2.5

3

3.5

4

EFR HR AMR FR AMR HR

Oct_06 Nov_06 Jan_07

It is obvious from the graph above that the average pace for the AMR callsis better than the non AMR calls with and without DL_DTX_AMR.

In EFR and AMR_FR ; we had 4% increase in average PACE while in HR

and AMR_HR ; we had 5.4% increase in average PACE.





Ascom Class Distribution

Ascom Class Evolution

0.53% 0.16% 0.02%0.44% 0.27% 0.06%

24.08%

75.20%

0.03%0.24%0.41%

17.46%

81.86%

14.84%

84.40%

0.00%

10.00%

20.00%

30.00%

40.00%

50.00%

60.00%

70.00%

80.00%

90.00%

100.00%

Excellent Good Fair Poor Bad

Oct_06 Nov_06 Jan_07

We had increase in the excellent number of samples (PACE > 3.5) by10% with the AMR.

d. QoS Degradation Investigations

The issue of increased CDR when activating AMR is explained by 3GPP with the end user behaviour in bad radio conditions, Document in Annex B.

The investigation was focused on the Quality HO increase due to DL DTX

on AMR.

As DL DTX influences the way the mobiles measure the RxQual; only asubset of frames is used for measurements as described by ETSI below.





Aspects of Discontinuous Transmission (DTX)

When DTX is employed on a TCH, not all TDMA frames may betransmitted, however, the following subset shall always be transmitted, and

hence can be employed to assess quality and signal level during DTX.

On any TCH this subset of TDMA frames is always used for transmissionduring DTX. For speech, when no signalling or speech is to be transmitted these

TDMA frames are occupied by the SID (Silence Descriptor) speech frame.

Measurement Reporting for the MS on a TCH

For a TCH, the reporting period of length 104 TDMA frames (480 ms) isdefined in terms of TDMA frame numbers (FN) as follows:

For a multislot configuration, the reporting period and SACCH Messageblock for each timeslot is defined as for TCH/F for TN = 0. When on a TCH, the

MS shall assess during the reporting period and transmit to the BSS in the next

SACCH message block the following:

RXLEV for the BCCH carrier of the 6 cells with the highestRXLEV among those with known and allowed NCC part of BSIC.

RXLEV_FULL and RXQUAL_FULL (non DTX): RXLEV and

RXQUAL for the full set of TCH and SACCH TDMA frames. Thefull set of TDMA frames is either 100 (i.e. 104 - 4 idle) frames

for a full rate TCH or 52 frames for a half-rate TCH.

RXLEV_SUB and RXQUAL_SUB (DTX): RXLEV and RXQUALfor the subset of 4 SACCH frames and the SID TDMA frames.





Our investigations were done on the reported RxQual in the below three

cases using the RMS matrix for a cell with degraded QoS.

Case 1: AMR disabled, DL_DTX enabled

Case 2: AMR enabled, DL_DTX_AMR disabled

Case 3: AMR enabled, DL_DTX_AMR enabled





In case 2 we can observe very slight increase in the reported RxQual,which is mainly explained by 3GPP, as the User is able to complete a call in a bad

radio condition. In case of decodable SACCH this will be translated as badreported RxQual, and possible Quality Handover triggering or Call Drop.

In case 3, the situation is worse with BER estimation error introduced byDTX.

Effect on TCH Duration for the area

AMR Allocation and TCH Duration

0

200000

400000

600000

800000

1000000

1200000

1400000

1600000

0 6 / 0

1 / 2 0

0 7

1 1 / 0

1 / 2 0

0 7

1 6 / 0

1 / 2 0

0 7

2 1 / 0

1 / 2 0

0 7

2 6 / 0

1 / 2 0

0 7

3 1 / 0

1 / 2 0

0 7

0 5 / 0 2 / 2 0

0 7

1 0 / 0 2 / 2 0

0 7

1 5 / 0 2 / 2 0

0 7

2 0 / 0 2 / 2 0

0 7

2 5 / 0 2 / 2 0

0 7

21

21.5

22

22.5

23

23.5

24

RTCH_assign_AMR_allocated (TCNAAMN) RTCH_duration_avg (TCTRMHT)

TCH_AV_DURATION = 22.3TCH_AV_DURATION = 22.6

TCH_AV_DURATION = 23

Average TCH Duration is as follow:

AMR with DL DTX: 22.3 sec Non AMR: 22.6 sec AMR without DL DTX: 23 sec

This means that the user is kept connected to the network for longer time.





Investigations Conclusion

When introducing AMR, an increase in Call Drop as well as Quality

Handover was observed. 3GPP published an explanation to part of thisphenomenon. It states that due to good speech quality of AMR, users don't hang-up in bad radio conditions, this result in:

o Un-decodable SACCH frames and radio link timeout, whichexplicitly explains increase in call drop rate and was observed inmany networks.

o Decodable SACCH frames with bad radio conditions, this resultin triggering Quality Handover.

In case of DL_DTX_AMR, the situation could be worse. The mobileestimates the RxQual value based on SACCH and SID radio frames, compared tothe whole SACCH multiframe in case of non DTX. Moreover this estimation ismobile implementation dependant, thus this could contribute significantly to

increase of Quality Handover triggering.

The conclusion is that DL Quality HO and call drop increases first of all dueto AMR, explained by 3GPP with the end-user behaviour, and secondly due tounreliable RxQual reporting of the mobiles in DL_DTX_AMR.

The field evaluation (Qvoice) confirmed that the customer perception isnot affected but has enhanced.

e. Channel Adaptation Activation

We enabled the AMR adaptation on BSC Blida_3 to test the feature and itseffect on the capacity.

Cause 26 (HR to FR):

HO cause 26_HR_FR

0

50

100

150

200

250

300

03/01/2007 03/06/2007 03/11/2007 03/16/2007 03/21/2007

0

0.005

0.01

0.015

0.02

0.025

0.03

0.035

0.04

0.045

BadQualHR %BadQualHR





Cause 27 (FR to HR):

HO cause27_FR_HR

0

5000

10000

15000

20000

25000

30000

03/01/2007 03/06/2007 03/11/2007 03/16/2007 03/21/2007

0

0.5

1

1.5

2

2.5

3

3.5

4

GoodQualFR %GoodQualFR

From the above graphs we see that the handovers were all due to good

quality from Full rate to Half rate.

The below graph shows the increase of the intra cell handovers count onthe BSC with no increase in failures or drop.

Intracell handovers

0

5000

10000

15000

20000

25000

30000

35000

40000

45000

50000

03/01/2007 03/06/2007 03/11/2007 03/16/2007 03/21/20070

0.5

1

1.5

2

2.5

3

Congestion Prep fail Success ROC Drop - BSS Drop - Radio % ROC % Drop

SDCCH_TCH_CONGESTION

0

0.5

1

1.5

2

2.5

3

03/01/2007 03/06/2007 03/11/2007 03/16/2007 03/21/200

TCH_CONG_RATE SDCCH_CONG_RATE

No effect on the BSC congestion.





8. Algiers AMR Implementation

a. Algiers Network Status

Algiers consists of 46 BSCs with 2085 cells. It has 920 cells Half Rateenabled. It is subdivided into 8 zones.

ZONE CELLS TRXs

ZONE7 585 1771

ZONE8 337 1130

ZONE1 260 911

ZONE5 254 871

ZONE2 252 900

ZONE6 221 775

ZONE3 133 466

ZONE4 43 146

AMR Penetration in this area is 58%.





b. Qvoice Statistics

Qvoice campaign was done in Algiers after the AMR activation on all the

cells, below is the comparison with and without the AMR.

Speech Type Distribution:

0.10%0.33%

7.09%

92.47%

25.93%

74.07%

0.00% 0.00%0.00%

10.00%

20.00%

30.00%

40.00%

50.00%

60.00%

70.00%

80.00%

90.00%

100.00%

HR EFR AMR HR AMR FR Feb_07

Mar_07

Speech Type Distribution

Feb_07 Mar_07

92% of the samples in the campaign were AMR_FR





Ascom Class Distribution

87.34%

10.76% 0.86% 1.01% 0.04%

78.30%

20.30%0.74% 0.62% 0.05%

0.00%

10.00%

20.00%

30.00%

40.00%

50.00%

60.00%

70.00%

80.00%

90.00%

Excellent Good Fair Poor Bad Feb_07

Mar_07

ASCOM Class

Feb_07 Mar_07

We had increase in the number of excellent samples by 12%

Speech Quality Distribution

3.713.27

3.59 3.74 3.70

3.803.44

0.00 0.00

3.70

0.00

0.50

1.001.50

2.00

2.50

3.00

3.50

4.00

EFR HR AMR HR AMR FR Total

F e b _ 0 7

M ar _ 0 7

Speech Quality Distribution

Feb_07 Mar_07

The average PACE for the HR calls has increased by 4.3%.





RxQual Distribution

0 to 4

5 to 7

Feb_07

Mar_07

90.29%

9.71%

87.49%

12.51%

0.00%

10.00%

20.00%

30.00%

40.00%

50.00%

60.00%

70.00%

80.00%

90.00%

100.00%

RxQual DL

Feb_07 Mar_07

The percentage of good quality has increased by 3.2%

The Total PACE per zone is as follow:

ZONE PACE Total ZONEPACETotal

ZONE4 3.6 ZONE3 3.73

ZONE6 3.62 ZONE8 3.74

ZONE7 3.62 ZONE1 3.8

ZONE5 3.66 ZONE2 3.81

The west part of the capital has the least average PACE.





c. Implementation Scenarios

Due to some limitations, we had to study different scenarios to activate

the AMR in Algiers. It was as follow:Scenario One:

We target small important areas in Algiers as follow:

o VIP Cells; contains airports, important Hotels and OTA offices inthe capital.

o Auto route; North and Southo Universities

o Down Town; represented by Zone 7.

o Major Stadiums in the capital; 5 Juliet and Omar Hamadi

Scenario Two:

We target in this scenario the zones with the least Average pacecompared to our competitors to enhance it.

Scenario Three:

In this scenario we have the average PACE per cell; enabling the AMR onthe cells with the least PACE and carrying high traffic.

Scenario Four:

Activate the AMR on the HR cells with high traffic in Algiers.

We combined these different scenarios and applied the AMR on 1554 Cellswhich represent 75% of the cells in Algiers, below are the distribution per zone.





9. Network AMR Implementation

From our study we found that the fourth scenario will be the better to

globalize the AMR on the network. We will apply the AMR on the cells with activeHR on it.

a. Alcatel Network Status

Alcatel Network has 176 on air BSCs with 8881 cells divided on 26 Wilayaswith 50% Half Rate Enabled Cells.

HALF RATE SumOfTRXs

DISABLED 14137ENABLED 14641

TOTAL TRXs 28778

Distributed among the wilayas as below:

WILAYA Sum Of HR TRXs Sum Of HR AMR TRXs

BOUIRA 575 65

TIZIOUZO 1062 167

MEDEA 556 127

BAYADH 76 20

SOUKAHRAS 205 54

GUELMA 336 100

BEJAIA 1004 318

BBA 383 122

MSILA 588 189

TEBESSA 407 140

SETIF 997 377

BLIDA 1155 522

GHARDAIA 188 92

LAGHOUAT 218 109

KHENCHELA 180 91

BECHAR 171 90

TIPAZA 488 263

CONISTANTINE 804 496OUARGLA 373 271

ILLIZI 16 13

NAAMA 97 79

TINDUOF 88 76

TAMANRASSET 36 32

BOUMERDAS 745 676

ALGER 3836 3605

ADRAR 57 57

TOTAL 14641 8151





b. NSS Activation

We need to activate the AMR on all the BSCs for all the RCPs of the network.

The BSCs-RCPs to be activated are attached.

BSC-RCP-ACTIVATE.xls

c. AMR Target for 2008

Our target is to apply the AMR on 13000 TRXs with 45% of Alcatel Network asfollows:

WILAYASum Of HR

TRXs

Sum Of HR AMR

TRXs

% of HR AMR

TRXsALGER 3836 3836 100.00%

SETIF 997 997 100.00%

BLIDA 1155 1155 100.00%

TIZIOUZO 1062 1062 100.00%

CONISTANTINE 804 804 100.00%

BEJAIA 1004 1004 100.00%

MSILA 588 588 100.00%

BOUMERDAS 745 745 100.00%

MEDEA 556 556 100.00%

BOUIRA 575 575 100.00%

BBA 383 122 31.85%

TIPAZA 488 263 53.89%

TEBESSA 407 140 34.40%

OUARGLA 373 271 72.65%

GUELMA 336 100 29.76%

LAGHOUAT 218 109 50.00%

BECHAR 171 90 52.63%

KHENCHELA 180 91 50.56%

GHARDAIA 188 92 48.94%

SOUKAHRAS 205 54 26.34%

NAAMA 97 79 81.44%

BAYADH 76 20 26.32%

ADRAR 57 57 100.00%

TAMANRASSET 36 32 88.89%

ILLIZI 16 13 81.25%

TINDUOF 88 76 86.36%

TOTAL 14641 12931 88.32%

This scenario is based on the highest Radio Busy Hour traffic for eachWilaya except for Adrar as it is a very small one.

The cells to activate the AMR on are attached.

AMR ACTIVATEQ1_2008.xls





10. HO Failures Problem

A problem of handover failures appeared during the Qvoice and drive testcampaigns all over the network. There was no significant increase in thehandover failures noticed from the QoS indicators on the network.

a. Problem Description

Handover failures appear with cause 09 (CHANNEL MODEUNACCEPTABLE), after investigating it was found that all the HO failures withthis cause was having Source cell with Speech Version2 and the target cell waswith Speech Version3 (from non_AMR to AMR cell).

b. Drive Test Investigation

o Case One(Normal Case):

It appears from the HANDOVERCOMMAND shown that the targetcell has AMR enabled on it and

the assigned TCH isAMR FULL RATE also theMULTIRATE CONFIGURATION

exists in the command.





o Case Two(Abnormal Case):

HO Command Message:

It appears from the handovercommand shown that the targetcell has AMR enabled on it andthe assigned TCH isAMR FULL RATE and theMULTIRATE CONFIGURATION

is missing in the command.

HO Failure Message:

The cause of the HO failure is of Value 9 (Channel Mode

unacceptable).





According to the ETSI recommendations:

If the HANDOVER COMMAND message instructs the mobile station to:

o use a Channel Description or Mode that it does not supporto if the Channel Mode to use is not defined for all channel sets

Then the MS shall return a HANDOVER FAILURE message with cause "channelmode unacceptable", and the MS shall remain on the current channel(s) and usesthe old Channel Description or Mode(s).

All the mobiles used in the drive test supports the AMR with all its codecs,the problem is that the MULTIRATE CONFIGURATION is missing in theHANDOVER COMMAND.

This problem is still pending with ALU.

11. Further Trials and Recommendations

We can have several trials on the Network as follows:

a. Enabling a 4th Codec

A trial can be made to add a fourth codec to the Active Set of Codec used.

b. AMR_FR Disable

According to the trial and implementation results, we didn’t havesignificant improvement in the AMR_FR pace compared to the AMR_HR. we canhave a trial to only enable the AMR_HR and check the Qvoice results and theKPIs.

c. Enabling Channel Adaptation

The results of this trial in Blida encourage us to start a trial for theChannel adaptation in Algiers; a great care should be taken as this may affect thenetwork capacity. This trial contradict the AMR_FR Disable Trial as this one boththe AMR_FR and AMR_HR should be enabled.





d. RLT and AMR

In order not to conflict with other trials in the network; Radio Link Timeout

trial, I suggest the following:If the RLT will be reduced from 28 to 20, we should disable the AMR_FR on

the network. By this will get the benefits from the AMR_HR and minimized the calldrop due to AMR and reached the RLT target.

e. AMR Follow up:

o Every month, the AMR parameters should be revised and checked in orderto follow up the HR activation and deactivation on the network.

o Every Quarter, the cells and sites selection should be made from the startin order to follow the rollout plan of the new sites in the concerned areas.

KNOW-HOW was transferred to the team for the AMR follow-up.





12. Annex A-B9 New Features

6.5.3 Radio quality counters

RMS counters for AMR Monitoring (1/4)To provide a better tool to dimensioning the AMR thresholds, B9 introduce

a new set of RMS counters to verify the use of different speech codecs:

For Full Rate and Uplink:

AMR_FR_UL_BAD= RMS44a that has 8 cells (1 for each FR codec) withthe relative number of bad speech frames received in uplink.

MAX_AMR_FR_UL_BAD= RMS44b that indicates the maximum number

of bad speech frames received in uplink in one FR codec.

AMR FR codec used in uplink (TRX based)

6.5.3 Radio quality countersRMS counters for AMR Monitoring (2/4)

For Half Rate and Uplink:

AMR_HR_UL_BAD= RMS45a that has 8 cells (1 for each HR codec) withthe relative number of bad speech frames received in uplink.

MAX_AMR_HR_UL_BAD= RMS45b that indicates the maximum numberof bad speech frames received in uplink in one HR codec.

AMR HR codec used in uplink (TRX based)

For Full Rate, UL & DL:

AMR_FR_UL_RXLEV_UL= RMS46a that has a table (8x10) with relative

number of correct speech frames received in uplink in each AMR FR codec(8 codecs) and each level band (10 level bands).

MAX_AMR_FR_UL_RXLEV_UL= RMS46b that has the 10 maximumresults. Each cell Ci of the table indicates the greatest value of the Vik fora i given in RMS46a.





AMR_FR_DL_RXLEV_DL= RMS47a that has a table (8x10) with relative

number of correct speech frames received in downlink in each AMR FRcodec (8 codecs) and each level band (10 level bands).

MAX_AMR_FR_DL_RXLEV_DL= RMS47b that has a table of 10maximum results. Each cell Ci of the table indicates the greatest value of the Vik for a i given in RMS47a.

6.5.3 Radio quality countersRMS counters for AMR Monitoring (4/4)

For Half Rate, UL & DL:

AMR_HR_UL_RXLEV_UL= RMS48a that has a table (5x10) with relativenumber of correct speech frames received in uplink in each AMR HR codec

(5 codecs) and each level band (10 level bands).

MAX_AMR_HR_UL_RXLEV_UL= RMS48b that has a table of 10maximum results. Each cell Ci of the table indicates the greatest value of the Vik for a i given in RMS48a.

AMR_HR_DL_RXLEV_DL= RMS49a that has a table (5x10) with relative

number of correct speech frames received in downlink in each AMR HRcodec (5 codecs) and each level band (10 level bands).

MAX_AMR_HR_DL_RXLEV_DL= RMS49b that has a table of 10maximum results. Each cell Ci of the table indicates the greatest value of the Vik for a i given in RMS49a.





13. Annex B-PARAMETERS

GENERAL AMR PARAMETERS

EN_AMR_FR EN_AMR_HR

Instance Cell

These 2 flags control whether AMR full rate or AMR half rate are allowed inthe cell.

Enable Allows AMR full rate/half rate in the cell.

Disable Forbids AMR full rate/half rate in the cell.

NotesFor AMR activation, this parameter should be enabled.

EN_INTRA_UL_AMR EN_INTRA_DL_AMR

Instance Cell

These flags enable / disable the detection of intracell uplink HO (cause 15)and intracell downlink HO (cause 16) respectively for AMR calls.They replace the parameters EN_INTRA_UL and EN_INTRA_DL only for AMRcalls.

Enable

Enable the detection of intracell UL/DL handover (Causes

15/16).

DisableDisable the detection of intracell UL/DL handover (Causes15/16).

NotesThese parameters are put in place specifically to be able to have differentintracell handover strategies for AMR and non AMR calls [E.g. Allow intracellhandover causes 15 and 16 for non AMR calls and Disable them for AMRcalls].

L_RXQUAL_UL_H_AMR

L_RXQUAL_DL_H_AMR

Instance Cell

Uplink [Downlink] quality thresholds for handover cause 15 [16].

They are used instead of the parameters L_RXQUAL_UL_H andL_RXQUAL_DL_H only for Handover causes 15 and 16 and strictly for AMRcalls.

These parameters do not affect the algorithms of quality handover causes 2and 4.

0 – 7 [Step size 0.1]

Default value: 4.5

Notes

These parameters are put in place in order to have the freedom of settingdifferent intracell handover strategies for AMR and non AMR calls.





Downlink_DTX_enable_AMR_FR

Downlink_DTX_enable_AMR_HR Instance BSC

This flag Enables / disables the use of downlink DTX for AMR FR (or HRrespectively) speech calls.

EnableEnables the use of downlink DTX for AMR FR (or HRrespectively) speech calls.

DisableDisables the use of downlink DTX for AMR FR (or HR respectively)speech calls.

Notes

___

DTX_INDICATOR_FR_AMR DTX_INDICATOR_HR_AMR

Instance Cell

This flag Controls uplink DTX usage for GSM phase 2 MS. Applies only to AMR

FR and AMR HR respectively.

It replaces the parameter DTX_INDICATOR_FR / DTX_INDICATOR_HR for AMR calls.

0 MAY use DTX

1 SHALL use DTX 2 SHALL NOT use DTX

Default value: 2

Notes

The following combinations between DTX_INDICATOR_FR / DTX_INDICATOR_HR are

possible (FR/HR): may/shall, may/shall not, may/may, shall/shall, shall/shall not,shall/may, shall not/shall and shall not/shall not.

The only combination that is NOT allowed is: DTX_INDICATOR_FR = shall not andDTX_INDICATOR_HR = may, I.e. shall not/may.

FORBID_AMR_NS Instance BSC

Enable/Disable Noise Suppression in the MS with the AMR codec

Enable The MS can use AMR NS [Default]

Disable The MS can not use AMR NS

Notes

___





CODEC MODE ADAPTATION PARAMETERS

AMR_FR_SUBSET Instance BSS (Omc-R changeable)

Bitmap (8 bits) defining the codec subset that shall be used for AMR FR calls(1 to 4 codecs to be specified from 8 available codecs in AMR FR).

Bit 8 (most significant)=1: 12,2 kbit/s is part of the subset;

Bit 7=1: 10.2 kbit/s is part of the subset;

Bit 6=1: 7,95 kbit/s is part of the subset;



Bit 3=1: 5,90 kbit/s is part of the subset;Bit 2=1: 5,15 kbit/s is part of the subset;

Bit 1 (least significant)=1: 4,75 kbit/s is part of the subset

A minimum of 1 codec and a maximum of 4 codecs must be setmandatory.

Notes

The default values of the codec subset should depend on the best speechquality performance viewed under all radio conditions.

As viewed in Voice Quality platform tests, the default value is [12.2 kbit/s ,7.95 kbit/s , 5.90 kbit/s]

AMR_HR_SUBSET Instance BSS (Omc-R changeable)

Bitmap (6 bits) defining the codec subset that shall be used for AMR HR calls(1 to 4 codecs to be specified from 5 available codecs in AMR HR).

Bit 8 = Bit 7=0;

Bit 6=1: 7,95 kbit/s is part of the subset; [7.95 kbit/s is notsupported by Alcatel]


Bit 4=1: 6,70 kbit/s is part of the subset;Bit 3=1: 5,90 kbit/s is part of the subset;


Bit 1 (least significant)=1: 4,75 kbit/s is part of the subset

A minimum of 1 codec and a maximum of 4 codecs must be setmandatory.





Notes

Mandatory rule: the list should not contain only the 7.40 kbit/s codec modealone, other codecs should be added in order to have more robustalternatives in case of degraded radio conditions.

The default value of the codec subset should depend on the best speechquality performance viewed under all radio conditions.Default value: [7.40 kbit/s , 5.90 kbit/s , 4.75 kbit/s]

AMR_START_MODE_FR AMR_START_MODE_HR

Instance BSS (Omc-R changeable)

This parameter indicates the codec used at the beginning of a FR

(respectively HR) AMR call, until codec mode adaptation is started by theBTS.

0 Implicit rule is used for initial codec mode

1 Lowest codec mode of the subset

2Second lowest codec mode (used only if the subset contains more than

one codec mode)

3Third lowest codec mode (used only if the subset contains more thantwo codec modes)

4 Highest codec mode (used only if the subset contains four codec modes)

Notes

By lowest codec mode, it is meant the lowest in terms of speech coding bitrate

(e.g. the 5.90 kbit/s codec is lower than the 12.2 kbit/s codec)For AMR_START_MODE_HR, 7.40 kbit/s is not permitted as a start codec mode,also 6.7 kbit/s should be avoided as a start codec mode. [3]

The implicit rule is defined as follows:

If the Active Codec Set contains:

1 mode, then this shall be the Initial Codec Mode;

2 or 3 modes, then the Initial Codec mode shall be the most robust mode of theset (lowest bit rate);

4 modes, then the Initial Codec Mode shall be the second most robust mode of the set (with second lowest bit rate).





AMR_FR_THR_1 AMR_FR_THR_2

AMR_FR_THR_3Instance Cell

Thresholds for AMR FR codec mode adaptation These thresholds are in C/I(the C/I is estimated in the demodulator by comparing the received Trainingsequence with the expected Training sequence)

0 - 31.5 dB (step size 0.5 dB)Default value: corresponds to the optimal voice quality performance foreach codec mode in the FR codec subset.The defaults are: AMR_FR_THR_1 = 6.5 dB , AMR_FR_THR_2 = AMR_FR_THR_3= 11.5 dB

Notes

Mandatory rule: AMR_FR_THR_1 <= AMR_FR_THR_2 <= AMR_FR_THR_3

AMR_FR_THR_1 is used between the lowest and the second lowest codec modes(ex: 5.90 and 7.40 kbit/s). It is not used if the number of codec modes in the FR

codec subset < 2

AMR_FR_THR_2 is used between the second lowest and the third lowest codecmodes (ex: 7.40 and 7.95 kbit/s). It is not used if the number of codec modes inthe FR codec subset < 3

AMR_FR_THR_3 is used between the third lowest and the fourth lowest codec

modes (ex: 7.95 and 12.2 kbit/s). It is not used if the number of codec modes in

the FR codec subset < 4





AMR_HR_THR_1 AMR_HR_THR_2AMR_HR_THR_3

Instance Cell

Thresholds for AMR HR codec mode adaptation These thresholds are in C/I(the C/I is estimated in the demodulator by comparing the received Trainingsequence with the expected Training sequence)

0 - 31.5 dB (step size 0.5 dB)Default value: corresponds to the optimal voice quality performance foreach codec mode in the HR codec subset. The setting of the AMR HRthresholds based on Voice Quality performance requires further tests.

Notes

Mandatory rule: AMR_HR_THR_1 <= AMR_HR_THR_2 <=AMR_HR_THR_3

AMR_HR_THR_1 is used between the lowest and the second lowest codecmodes (example: 4.75 and 5.15 kbit/s).

It is not used if the number of codec modes in the HR codec subset < 2.

AMR_HR_THR_2 is used between the second lowest and the third lowestcodec modes (example: 5.15 and 5.90 kbit/s).


AMR_HR_THR_3 is used between the third lowest and the fourth lowest

codec modes (example: 5.90 and 7.40 kbit/s).


AMR_FR_HYSTAMR_HR_HYST

Instance Cell

Hysteresis for AMR FR/HR codec mode adaptation.

0 – 7.5 dB [step size 0.5 dB]Default value: 2 dB

Notes

These hysteresis are only applied when adapting from the lower codec mode(more robust) to the higher codec mode (less robust). [E.g. when adaptingthe codec mode from 5.15 kbit/s to 5.90 kbit/s the hysteresis is added].

That is to avoid fluctuating back and forth between 2 codecs when thecalculated C/I is close to the threshold separating them.





CHANNEL MODE ADAPTATION PARAMETERS

EN_AMR_CA Instance Cell

This flag Enables/Disables intracell HO for AMR channel adaptation(Handover Causes 26, 27).

EnableEnables intracell HO for AMR channel adaptation.Handover Causes 26, 27 are checked by HOP (HOPreparation process).

DisableDisables intracell HO for AMR channel adaptation.Handover Causes 26, 27 are not checked by HOP.

Notes

This parameter can only be enabled if both EN_AMR_FR and EN_AMR_HR areenabled.

Handover causes 26 and 27 (AMR HR-AMR FR and AMR FR-AMR HR) are onlyvalid for AMR calls.

A_QUAL_CA_HR_FR A_QUAL_CA_FR_HR

Instance Cell

These two parameters are the Quality Averaging window sizes for HR-FR(and FR-HR respectively) channel adaptation. These parameters are used tocalculate the AV_RXQUAL_UL/DL_CA in Handover causes 26 and 27.

1 – 61. Unit: SACCH multi frame.

Default: A_QUAL_CA_HR_FR = 8 , A_QUAL_CA_FR_HR = 61

Notes

Mandatory rule: A_QUAL_CA_FR_HR >= A_QUAL_CA_HR_FR

It is recommended that A_QUAL_CA_HR_FR <= A_QUAL_HO in order tohave a faster reactivity on the channel adaptation handover and avoidmaking unnecessary interference/quality handovers.

The value of 61 (~29 sec) for A_QUAL_CA_FR_HR is a relatively high valueand could be optimized during field trials according to the requiredaggressiveness of usage of HR.





W_QUAL_CA Instance Cell

Averaging weight applicable for Full Quality for channel adaptation.

1 – 3 Default value: 1

Notes

The averaging weighting algorithm works the same as for other weightingfactors (E.g. W_LEV_HO , W_QUAL_HO , W_PBGT_HO,…).

THR_RXQUAL_CA_NORMALTHR_RXQUAL_CA_HIGH

Instance Cell

Threshold for channel adaptation under normal/High load.

0 – 7 (step size 0.1)Default: 0

Notes

Mandatory rule: THR_RXQUAL_CA_HIGH >= THR_RXQUAL_CA_NORMAL

THR_RXQUAL_CA_XX is used for FR to HR (Cause 27) adaptation.While THR_RXQUAL_CA_XX + OFFSET_CA_XX is used for HR to FR (Cause 26)adaptation.

OFFSET_CA_NORMALOFFSET_CA_HIGH

Instance Cell

Offset for channel adaptation hysteresis under normal/high load (cause 26).

0 – 7 (step size 0.1)Default: 4.5

NotesMandatory rule: OFFSET_CA_HIGH >= OFFSET_CA_NORMAL

THR_FR_LOAD_L_SV3

THR_FR_LOAD_U_SV3Instance Cell

Load thresholds for allocation policy of (Speech Version 3) AMR HR in a dual ratecell supporting AMR HR.

0 – 100%Default value: Depends on the required degree of HR usage and the number of TRX





Notes

Mandatory rules: THR_FR_LOAD_L_SV3 <= THR_FR_LOAD_U_SV3 ,

THR_FR_LOAD_L_SV3 <= THR_FR_LOAD_L_SV1.

The load evaluation algorithm for SV3 works the same way as for the loadevaluation algorithm for SV1, only with these different thresholds.





14. Annex C

Source: Nortel Networks

AMR activation – field results

Introduction

AMR has proven many benefits in terms of coverage, voice quality and capacity,but field monitoring studies have revealed increase of the dropped calls ratios,which variations depend upon environment, mobile penetration and traffic

mobility.

This document provides a few figures from field analysis and tries to illustrate thedegradation due to robustness imbalance between lower AMR codec modes and

signalling channels. Some possible work-around solutions are also suggested,while waiting for ACCH channel coding to be improved.

Call drop analysis

2.1 Introduction

With the AMR activation over different clusters, all the usual indicators are

performing within normal ranges, with the exception of call drop indicators. It isinteresting to identify the root cause of these drops. Within the RF drops, it isnecessary to analyze the different elements that contribute to the drop callfluctuations. The main ones studied in this document are:

Radio Link failure or RLT – Radio Link Timeout drops: happen when SACCH blocksare not being decoded anymore. RLT drops mostly occur in coverage limitedareas or under very aggressive interference situation.

Inter-cell handover drop: these drops occur when neither a HO failure nor a HO

complete message is being received in time after a HO command. They areusually occurring when the radio interface is not clear enough to perform the HOsuccessfully because of the weak signal strength or because of a highly interferedarea.

Other elements that have not been detailed in this document but are worthmonitoring are LAPDm drops, which occur whenever a procedure sustained byLAPDM protocol fails, and intra-cell handover drops.

The drop ratio used in this analysis was in Drop per Erlang. But since it ispreferable to work on ratios rather than absolute numbers, only relative ratios areprovided below.

It should be noted that the results and trends presented hereafter are based onmonitoring made on different live areas. Each area having specific characteristics

in terms of:





design and engineering strategy,

spectrum available,traffic profile (numbers, mobility, offered services…),terrain morphology,

AMR mobile penetration.

For this reason, this document does not claim to define strict rules with regards toAMR, but the purpose is rather to provide some general trends of the impact onnetwork performances of AMR activation, and try to identify what are thecontributors of the degradations that could be observed.

Observations have been made in a network where penetration of AMR is around95%.In order to make statistics over this network, different types of clusters havebeen selected: rural, suburban and urban.The frequency bands used in this network are 850MHz and 1900MHz.The monitoring was done using daily statistics, which smooth the traffic effect byaveraging events spread over 24hours, but provide a general view of the

performance of the network.

2.2 Cluster 1: rural

Cluster 1, of 14 sites, is in a rural area, with low density of urbanized areas andwhere coverage, more than interference is the limiting factor.Here are the drop call rates observed on that cluster.

Total drop call rateincrease with AMRactivation

+ 21%

Contribution on total dropcall rate

Before AMRactivation

After AMRactivation

- RLT expiry 42% 45%

- HO timer expiry 21% 22%

Table 1. Drop call values in rural area

There was a peak of drop call rate immediately after AMR activation: RLT timerwas adjusted after that peak, giving the values provided in the table.

In this type of environment, RLT drop contribution is usually much higher becausecalls are being dragged at cell edges and the overlap between neighboring sites is

less important than in urban areas. HO timer drops are less significant than inareas where the overlap is more dense, since they are directly related to usersmobility.

2.3 Cluster 2: suburban

Cluster 2, of 13 sites, is in a suburban area with spots of dense traffic and where

interference is definitely the limiting factor.





Here are the drop call rates observed on that cluster.

Total drop call rateincrease with AMRactivation

+ 38%

Contribution on total drop

call rate

Before AMR

activation

After AMR

activation



Table 2. Drop call values in suburban area

Here also there was a peak of drop call rate immediately after AMR activation:RLT timer was adjusted after that peak, giving the values provided in the table.

In this kind of highly interfered environment, HO drops are dominant as shown in

the table, while RLT is around 25%.

2.4 Cluster 3: urban

Cluster 3, of 20 sites, is in an urban area where interference and indoor coveragelimitations are mainly the constraining factors. Traffic (Erlang and mobility) is alsoan important constraint.

Here are the drop call rates observed on that cluster.

Total drop call rate

increase with AMRactivation

+ 6%

Contribution on total dropcall rate

Before AMRactivation

After AMRactivation



Table 3. Drop call values in urban area

Here also RLT timer was adjusted a few days after AMR activation, giving thevalues provided in the table.

The respective contributions of RLT and HO timer expiry to the total drop call ratein urban environment are very similar to the cluster 2.

In this type of environment, user profile is responsible of wider variations of theobserved metrics because of:density of traffic (resource allocation),

resource usage (Erlang),users mobility (handovers per calls),

indoor usage (RLT and Handover failures).





There are very important variations between week and week-end traffic, thefigures provided here are averaged over both periods.

Solutions?

The problems encountered when activating AMR are intrinsic to itsstandardization, since AMR lower modes voice channels have better performancesthan signalling channels. Improving the channel coding of these 2 signallingchannels is certainly the best solution (see [1]), but it will take some time beforethe new coding is implemented on both terminal and network sides.

Therefore it seems necessary to find work-around solutions in the meantime tolimit as much as possible these drawbacks of AMR. There are several possibilities:

adjusting RLT value, to reduce impact on SACCH erroneous decoding,limiting the use of lower codecs in some environments,maximising use of AMR HR, which robustness is closer to that of signallingchannels,

repeating FACCH as suggested by Ericsson already.

3.1 Possible solutions in rural environment

With AMR activation, since signalling channels are less robust than AMR lowcodecs, the phenomenon of user behavior leads to an increase of call drop ratio:at cell edge, while signalling is not correctly received and thus RLT counter is

decreasing due to radio conditions, AMR low codecs are still able to give a fairvoice quality. In similar conditions, an EFR user would realize its voice qualitybecoming poor, and release the call normally. But an AMR user does not hear thisradio condition degradation since AMR low codecs allow a fair voice quality at lowC/I. AMR user drags the communication until RLT counter reaches 0, or when ittries to hang up, one of the links is lost.Therefore, in rural environments, coverage limited with low traffic, a solution is toincrease RLT timer, which maximizes probability to recover radio conditionswithout degradation of AMR user perception or leave enough time to clearproperly the call when it starts to be very poor in terms of voice quality. It can

compensate, to some extent the user behavior.This modification of the timer must however be done carefully, since very large

RLT timer values would result in situations where calls are still active, but SACCHcannot be decoded properly.

Another work-around is to limit the use of lower codecs in rural environments.Since increasing capacity is not the objective in these areas, and the use of lowercodecs has shown to increase the drop ratio, modifying the link adaptation table

in such a way to limit the lower codecs usage can be a solution.This method has proved to bring the drop ratio down to normal levels.

3.2 Possible solutions in suburban/urban environment





As seen in tables above, suburban/urban environments suffer more from HO

timer drops.It is however difficult to see exactly where the problem is: in HO commandreception, in acknowledgement reception…?

Traces analysis is in process but conclusions could not be ready in time forGERAN#21.

Looking at the comparative performances of AMR HR and EFR, one can see thatAMR HR with link adaptation and EFR provide the same level of voice quality,slightly better though for EFR. AMR-HR provides fair quality down to 9dB C/Iwhile EFR can go down to 7dB C/I for the same quality (MOS). But both, AMR-HR

and EFR are quite poor below 7dB while FACCH and SACCH performance are in arange where the BLER is not very good but still can be handled. This means thatan AMR-HR user and an EFR user would have roughly the same behavior at celledge, either will drop or will attempt to release the calls as the quality willdegrade.

This leads to the conclusion that maximizing HR penetration in areas where the

limiting factor is not the coverage but the C/I is a good solution. This has beentried on the field and the drop indicator did go down to a level similar to what it

was before AMR activation as shown in the figure below. From a voice qualityperspective, there has been no real improvement here as the objective was tolower down the drop rate.

RLT_DLC T200_DLC T3103_DLC T3107_DLC RF_DLC

EFR Only

A

MR FR AMR

AMR HR maximised

RLT change

Fig1: Urban area: effect of AMR HR maximisation

Repeating the FACCH is a solution that was proposed in earlier meetings, but test

results of this method are not available yet.





Conclusion

This paper has presented some field measurements, showing the drop call

increase when AMR is activated in a network with almost 100% AMR penetration.Both contributions from RLT timer and HO timer expiry were detailed, to showimpact on both SACCH and FACCH.Some work-around solutions are proposed, before improvement of ACCH channelcoding is agreed in the standards. These are: adjusting RLT value, limiting theuse of lower codecs, and maximising use of AMR HR.More studies are undergoing on the field, in order to have a better understanding

of the problems.

References

[1] “Enhanced ACCH for AMR”, source: Nortel Networks, T-Doc GP-041927, 3GPPTSG GERAN#21



End of Document

Date post:	04-Apr-2018
Category:	Documents
Upload:	kaidiamar
View:	222 times
Download:	0 times

Amr Trial Final Report_v2

Documents