You are on page 1of 18

WCDMA coverage Voice Quality improvement

Contents
Background..................................................................................................................................................3
May 2015 WB-AMR Gain Factor BetaC tuning............................................................................................4
July 2015 WRFD-140201 AMR Voice Quality Improvement Based on PLVA deployment............................6
September 2015 WRFD-160204 CS Voice Precise Power Control deployment...........................................9
April 2016 WRFD-170201 Seamless Crystal Voice and WRFD-171201 Crystal Voice in Deep Coverage
deployment...............................................................................................................................................12
Conclusion.................................................................................................................................................18
Background

1. Figure 5-1 Codec operations after the TrFO function is enabled

After both Narrow Band (NB) and Wide Band (WB) Advanced Multi Rate (AMR) features activations
together with Transcoder Free Operation (TrFO) scenario, then the customer noticed an overall Voice
Quality degradation by both RAN Voice Quality Indicators counters and Drive Test sessions Mean
Opinion Score (MOS) results.

So they asked us for support in order to improve the situation.

By our analysis we found out a complex problem, including also other vendor CN provider issues, but
for sure we proposed a road map of UTRAN improvements (parameters optimization, new specific Voice
Quality features activations) that lead the customer to be happy with Huawei product performances.
May 2015 WB-AMR Gain Factor BetaC tuning
After troubleshooting activity (involving GTAC support, too), we found out that after WB-AMR feature
deployment with 12.65 kbps maximum bit rate codec, the specific WB-AMR RAB Gain Factor BetaC
parameter required a tuning (from 12 to 10), in order to guarantee the best uplink Voice quality.
Together with Gain Factor BetaD, BetaC acts on the UE uplink power distribution between the two
WCDMA dedicated physical channels DPCCH and DPDCH respectively carrying Dedicated Physical
Control information and Dedicated Physical Data information.

Considering the definitions of 3GPP 25.214:


Computed gain factors

The gain factorsc and d may also be computed for certain TFCs, based on the signalled settings for a
reference TFC.

Let c,ref and d,ref denote the signalled gain factors for the reference TFC. Further, let c,j and d,j denote
the gain factors used for the j:th TFC. Also let L ref denote the number of DPDCHs used for the reference
TFC and L,j denote the number of DPDCHs used for the j:th TFC.

Define the variable

K ref =∑ RM i⋅N i
i ;

where RMi is the semi-static rate matching attribute for transport channel i (defined in [2] subclause
4.2.7), Ni is the number of bits output from the radio frame segmentation block for transport channel i
(defined in [2] subclause 4.2.6.1), and the sum is taken over all the transport channels i in the reference
TFC.

Similarly, define the variable

K j=∑ RM i⋅N i
i ;

where the sum is taken over all the transport channels i in the j:th TFC.

The variable Aj, called the nominal power relation is then computed as:

A j=
√ √
β d , ref Lref

β c , ref L j
Kj
K ref .

VF-IT original settings for both NB-AMR and WB-AMR reference TFC bit rate were:
BetaC=12 and BetaD=15.

But according to Huawei RNC, the reference TFCs Max Rates are different:

1) NB-AMR Reference TFC Max Rate=12.20 kbps


2) WB-AMR Reference TFC Max Rate =23.85 kbps

So for NB-AMR at 12.2 kbps codec, Kj = Kref =12200, but for WB-AMR at 12.85 kbps codec, Kj = 12650 and
Kref =23850

1) NB-AMR at 12.20 kbps codec Aj=1.25


2) WB-AMR at 12.65 kbps codec Aj ≈ 0.91

This WB-AMR Aj low value caused high uplink BLER, heavily affecting the resulting Voice quality.

Hence by setting WB-AMR Reference TFC BetaC=10:

for WB-AMR at 12.65 kbps codec


A j=
β d , ref
β c , ref √ √

Lref
Lj
Kj
K ref ≈ 1.09

So the resulting 12.65 kbps TFC Aj partially compensated the original nominal power relation loss.

This nominal power relation compensation increased the uplink DPDCH UE transmitted power, leading
to a reduction of the average uplink received WB-AMR BLER from 1.8% to 0.4%. Because of this BLER
reduction, WB-AMR uplink MOS improved of about 0.3.
July 2015 WRFD-140201 AMR Voice Quality Improvement Based on PLVA
deployment

To further improve both NB and WB AMR Voice quality, we proposed to deploy PLVA feature

Figure 3-3 Working mechanism of the Viterbi algorithm

The Viterbi algorithm selects the optimal path based on the maximum likelihood theory and exports the
data decoded on the optimal path. If the data decoded on the optimal path fails the CRC, the AMR
speech codec discards the data. As a result, voice quality deteriorates.

Figure 3-5 Working mechanism of the PLVA

Instead of selecting only the most optimal path, the PLVA selects the top N optimal paths and performs
CRC on the data decoded on these paths. The PLVA only exports data that passes the CRC. If the data
decoded on these paths fails the CRC, the PLVA exports the data decoded on the optimal path, which is
the same path selected by the Viterbi algorithm. The PLVA outperforms the Viterbi algorithm because it
chooses the data decoded on multiple paths, which include the optimal one selected by the Viterbi
algorithm. Therefore, when the data decoded by the Viterbi algorithm is correct, the data decoded by the
PLVA is also correct. However, when the data decoded by the PLVA is correct, the data decoded by the
Viterbi algorithm is not necessarily correct because there are occasions when the data decoded on the
optimal path is incorrect whereas the data decoded on other paths selected by the PLVA is correct. In
simulations where the PLVA selects four paths, the signal-to-noise ratio (SNR) is 0.2 to 0.8 dB better than
that produced by the Viterbi algorithm.

Figure 3-6 Narrowband AMR speech decoding using the PLVA

As shown in Figure 3-6, CRC is performed on the AMR voice data of class A and therefore the data can
be decoded using the PLVA. The mean opinion score (MOS) of class A AMR voice data is improved by
reducing the BLER. CRC is not performed on classes B and C AMR voice data and therefore classes B
and C AMR voice data can only be decoded by using the Viterbi algorithm. The PLVA exports and sends
the decoded data, Viterbi CRCI, and PLVA CRCI to the RNC. The Viterbi CRCI is used for OLPC. The
PLVA CRCI serves as the BFI, indicating whether a speech frame is erroneous.
This dual-CRCI mechanism does not affect power control. In addition, the performance of the AMR voice
data of class A is improved, maximizing the speech quality. Note that this feature decreases the uplink
BLER and improves the MOS of AMR voice service by using PLVA, and it has no impact on the downlink.

This feature sets the following requirements on NodeB hardware:


 The 3900 series base station must be configured with the WBBPd, WBBPf or UBBPd board.

This feature activation lead to a further 0.2 increase of MOS (on top of the previous gain step) and to a
remarkable AMR_UL_VQI improvement.
September 2015 WRFD-160204 CS Voice Precise Power Control deployment

Figure 3-1 Principle of the CS Voice Precise Power Control feature

The outer-loop power control based on BLER and BER uses both the BLER-based outer-loop power
control algorithm and BER-based outer-loop power control algorithm. The following figure illustrates the
principle of outer-loop power control.
BLER-based Outer Loop Power Control
The BLER-based outer-loop power control algorithm calculates the BLER measurement value according
to the CRC result. It then determines whether to adjust the QE target value (QEtarget) by comparing the
BLER measurement value (BLERmeas) with the BLER target value (BLERtarget).
QEtarget after adjustment = QEtarget before adjustment + QEtarget adjustment value

 QEtarget adjustment value = (BLERtarget - BLERmeas x QeFactor(BSC6900,BSC6910)) x


QeStepSize(BSC6900,BSC6910)
 For the first frame, the QEtarget before adjustment is the initial QEtarget.
Initial QEtarget = int (InitQeFactor(BSC6900,BSC6910) x BLERtarget +
InitQeConst(BSC6900,BSC6910))
BER-based Outer Loop Power Control
The BER-based outer-loop power control function determines whether to increase or decrease the SIR
target value (SIRtarget) by comparing the QE (BER quantized value) reported by the NodeB with the
QEtarget after adjustment.
SIRtarget after adjustment = SIRtarget before adjustment + SIRtarget adjustment value

 SIRtarget adjustment value = (QE reported by the NodeB - QEtarget after adjustment) x
QeStepSize(BSC6900,BSC6910)
 For the first frame, the SIRtarget before adjustment is the initial SIRtarget.
DTX Power Control
When performing a voice service in the DTX state, a UE transmits only control information (such as pilot
information and TPC) and does not transmit data on the uplink DPCCH. DTX power control appropriately
reduces the target SIR while ensuring the PDCCH quality, thereby reducing transmit power of UEs.
DTX power control determines whether a UE performing a voice service is in the DTX state as follows:
 When the NodeB receives an SID frame from this UE, it determines that the UE is in the DTX state.
 When the NodeB receives a voice frame from this UE, it determines that the UE is not in the DTX
state.
When a UE is switched from the non-DTX state to the DTX state, the target SIR is automatically
decreased depending on the value of SirDtxConst(BSC6900,BSC6910). When the UE is switched from
the DTX state to the non-DTX state, the target SIR is restored to the original value.
Optimized Multipath Searching
The multipath searching algorithm calculates the accumulated power of all signals in a timeslot based on
pilot information and the accumulated power is used to determine the multipath status. However, when
the SIR is low, the signals in some paths are so weak that the accumulated power of these signals may
be surpassed by noise and the signals fail to be detected. The optimized multipath searching algorithm
accumulates all pilots of multiple timeslots for assessing the multipath signal quality. In this way, the
impact of noise on signal quality is reduced and the performance of searching for weak signals in low-SIR
scenarios is enhanced, thereby improving the accuracy of ILPC.
Optimized SIR Estimation
The SIR estimation algorithm estimates the SIR based on all pilots of a timeslot. However, the pilot
information of a single timeslot is insufficient for accurate SIR estimation. The optimized SIR estimation
algorithm uses all information of the previous timeslot and pilot information of the current timeslot to
obtain the average SIR of the two timeslots, which facilitates accurate ILPC.
After CS Precise Power Control feature deployment, by drive tests sessions the customer experienced a
MOS increase of both NB and WB AMR, depending on Nodeb HW configuration:

-about 0.1 for Nodebs not equipped with WBBPf boards (72% of the total amount of Nodebs).

-about 0.2 for Nodebs equipped with WBBPf boards (the other 28% of the total amount of Nodebs).

In fact only WBBPf board can support Optimized SIR Estimation.

The customer has no UBBPd board, so no Optimized Multipath Searching is supported.


April 2016 WRFD-170201 Seamless Crystal Voice and WRFD-171201 Crystal
Voice in Deep Coverage deployment

The RNC performs selective combinations or softer combinations for different voice frames and then
reports combination results to the MSC, as shown in the following figure. The selective combination
process is implemented through the original soft handover process and the softer combination process is
enabled by this feature.
This feature is implemented on the RNC and NodeB, and the RNC can perform softer combinations only
when the NodeB reports pre-decoding information. In addition, if flow control occurs over the Iub or Iur
interface, this feature does not take effect. The following figure shows the architecture of this feature.
Figure 2-1 Architecture of Seamless Crystal Voice

This feature improves MOS in case of inter-NodeB soft handover scenario. Without it, in soft handover
areas (that is, active set with more than one cell) only the tributary with the best quality is decoded
whilst the information belonging to the other tributaries are discarded. By Seamless Crystal Voice, a new
channel decoder in the RNC (on DEUa board) performs a softer combination of the tributaries of the
active set not correctly decoded with the usual criterion. In such a way, the original soft handover
mechanism is transformed into a softer handover one and this results in an improvement of the uplink
voice quality (UL MOS) in the soft handover areas.

Architecture of Crystal Voice in Deep Coverage

By this feature the RNC applies a new high-order channel decoding algorithm on voice frames that have
experienced a decoding failure with the usual low-order Viterbi or “Parallel List Viterbi Algorithm”
(PLVA) algorithms. Since by the low-order decoding algorithms, in case of poor-quality scenarios the
cyclic redundancy check (CRC) may fail, the usage of this new high-order decoding algorithm will help in
improving UL MOS above all in deep coverage scenarios.

Hardware

Hardware Dependency

 Dependency on the RNC


The RNC must be configured with the DEUa board. If RNC in Pool is enabled for user-plane load
sharing and the master RNC is configured with the DEUa, the overflow RNC must also be
configured with the DEUa.
 Dependency on the NodeB
The 3900 series macro base stations must be configured with the WBBPb, WBBPd, WBBPf, or
UBBPd board.
Impacted Features

 WRFD-140201 AMR Voice Quality Improvement Based on PLVA


With the target BLER remaining unchanged, the proportion of error frames reported by the NodeB to
the RNC decreases when the WRFD-140201 AMR Voice Quality Improvement Based on PLVA
feature is enabled. Gains provided by Crystal Voice in Deep Coverage decrease.

 WRFD-160204 CS Voice Precise Power Control


With the target BLER remaining unchanged, the proportion of error frames reported by the NodeB to
the RNC decreases when the WRFD-160204 CS Voice Precise Power Control feature is enabled.
Gains provided by Crystal Voice in Deep Coverage decrease.
 WRFD-170201 Seamless Crystal Voice
The number of error frames in soft handover areas significantly decreases when the WRFD-170201
Seamless Crystal Voice feature is enabled. Gains provided by Crystal Voice in Deep Coverage
decrease.

During the live Network trial on a whole RNC (BARNC50) coverage area, the customer complained about
the drive test poor gains in terms of UL AMR MOS: no significant MOS gain was actually observed.

We remarked that the expected features gain was reduced by the other already deployed
functionalities.

Anyway by trial RNC, the overall UL AMR MOS increased (of about 0.04 step).
Moreover we proposed a different MOS gain evaluation criteria, basing on the geographical coverage
area guaranteeing a good average MOS (EVQI=3.0 for NB-AMR and 3.25 for WB-AMR):

Assuming poor-quality EVQI values the ones less than 3 for NB-AMR and less than 3.25 for WB-AMR,
after the activation of both features the percentage of cells belonging to BARNC50 with mean UL EVQI
less than these thresholds improved (that is, decreased):

NB-AMR: cells with UL EVQI < 3.00 improved from 37 % to 25 %

WB-AMR: cells with UL EVQI < 3.25 improved from 52 % to 37%


Conclusion
This long and complex troubleshooting and RF optimization process lead the customer to be more
confident with Huawei product performances and with Huawei Network Performance Service efficiency.

By P3 comparative testing sessions, the customer could obtain benefits of the resulting voice quality
KPIs, improving the final E2E customer quality of the service.

I could personally develop both analysis and RF network optimization design skill, that helped me to
become a customer reference point when dealing about Voice Quality topics: in fact the customers
involved me to support troubleshooting and optimization with other vendor CN configuration, in order
to find out the solution that can guarantee the best Voice Quality.

You might also like