Você está na página 1de 10

Document type

RECOMMENDATION TO DEPLOY AMR-FR

RSQ

GT/NAD/RSQ/NT/2005-0013
Version: V0.1
RECOMENDATION TO DEPLOY AMR-FR
Date: 07/04/2005
Page: 10
LEVEL: RECOMMENDATION
Executive summary
This document describes the high level operation of AMR-FR, and summarises the
results that have been obtained from testing by Orange UK and FT Div RD.
It then discusses problems that have been encountered in 1:1 re-use networks and why
these are not applicable to Orange networks. An overview is also given on Orange UKs
experience of deploying AMR-FR on the live network.

IT IS RECOMMENDED THAT AMR-FR IS DEPLOYED ON GSM


NETWORKS.
It has been demonstrated that AMR-FR does provide an improvement of up to 0.5
MOS points in radio environments that would typically be encountered in Orange
networks. Therefore AMR-FR will be of benefit in the following areas:
(i)
Those areas of the network where the speech quality of FR/EFR suffers due
to poor C/I (e.g. approximately <11dB), or where the interference
environment is rapidly changing.
(ii)
To improve speech quality where there is marginal network coverage.
(iii)
Where a higher frequency re-use is sought to increase network capacity.
Depending on the cost of deployment in individual networks it may not be cost effective
to roll out AMR-FR over the whole network. Instead it may be better to target those
areas where there are known problems with speech quality.
Written by : Ric Bailey
Signature
Date: 1/4/2005
Checked by:
Signature
Date:
Approved by:
Date:

Signature

IMPORTANT
These documents are Orange confidential and it is the responsibility of each Member
Company not to transfer, nor copy, even partially, these documents to individuals not
belonging to Orange Group without the written approval of Radio Strategy and Quality

Group Technical/Network Architecture and Design/Radio Strategy and Quality


AMR-FR recommendation v1.0a.doc

Page 1/10

Document type
RECOMMENDATION TO DEPLOY AMR-FR

RSQ

TABLE OF CONTENTS

1.

INTRODUCTION ........................................................................................................................... 3

2.

AMR-FR OPERATION ................................................................................................................... 3

3.

PERFORMANCE EVALUATION ..................................................................................................... 4

3.1.
3.2.
4.

Test set-up.................................................................................................... 4
Results.......................................................................................................... 5

REAL NETWORK PERFORMANCE ............................................................................................... 6

4.1.
4.2.

Highly loaded 1:1 re-use networks ................................................................ 6


Orange UK experience .................................................................................. 6

5.

RECOMMENDATION .................................................................................................................... 7

6.

DOCUMENT HISTORY.................................................................................................................. 8

7.

REFERENCES............................................................................................................................... 9

8.

ANNEXES ................................................................................................................................... 10

Group Technical/Network Architecture and Design/Radio Strategy and Quality


AMR-FR recommendation v1.0a.doc

Page 2/10

Document type

RSQ

RECOMMENDATION TO DEPLOY AMR-FR

1. INTRODUCTION
This document provides a recommendation for deploying the AMR-FR speech codec
across the Orange Group. It will describe the fundamentals of AMR-FR operation and
the benefits of the codec over existing codecs in operation. A summary of some issues
encountered in highly loaded networks in the US is presented together with a
discussion of whether these relate to Orange networks, together with Orange UKs
experience. Finally a recommendation is given concerning the features suitability for
wide scale deployment.

2. AMR-FR OPERATION
With the older speech codecs, such as FR, EFR and HR, the speech coding rate and channel
coding rate is fixed, regardless of the error conditions of the radio channel. Consequently under
poor C/I conditions neither the speech coding nor channel coding is robust enough for optimum
speech quality.
Adaptive Multi-Rate Codec (AMR) introduces the ability to adapt the speech coding rate as a
function of the radio channel quality and error conditions. AMR uses a variable balance between
speech coding and channel coding. When radio conditions are good, it increases the speech
coding rate to improve the speech quality at the expense of reducing the channel coding rate
when less error protection is required. When radio conditions are bad, it reduces the speech
coding rate thus reducing the speech quality, but increases the channel coding thus adding
extra error protection. Overall this provides better speech quality than FR/EFR in poorer C/I
conditions.

Medium radio
conditions

Bad radio
conditions

Good radio
conditions

Speech coding = speech information


Channel coding = speech protection
AMR codec consists of a family of codecs (source and channel codecs) operating in the GSM
FR (and HR) channels. The table below illustrates the various codec modes supported by AMRFR, all of which operate within a 22.8 kb/s FR timeslot on the air interface. It clearly shows the
different split in source coding rate and channel coding rate CH0-FS to CH7-FS, where CH0
provides the best speech quality with reduced error protection, and CH7 has the worst speech
quality but most robust protection.

Group Technical/Network Architecture and Design/Radio Strategy and Quality


AMR-FR recommendation v1.0a.doc

Page 3/10

Document type

RSQ

RECOMMENDATION TO DEPLOY AMR-FR

Channel
mode

Channel codec Source coding


bit-rate, speech
Mode

Net bit-rate,
in-band
channel

Channel
coding
bit-rate,
speech

Channel
coding
bit-rate, inband

Total
channel bit
rate

TCH/FR

CH0-FS

12.20kbit/s
(GSMEFR)

0.10 kbit/s

10.20 kbit/s

0.30 kbit/s

22.8 kb/s

TCH/FR

CH1-FS

10.20 kbit/s

0.10 kbit/s

12.20 kbit/s

0.30 kbit/s

22.8 kb/s

TCH/FR

CH2-FS

7.95 kbit/s

0.10 kbit/s

14.45 kbit/s

0.30 kbit/s

22.8 kb/s

TCH/FR

CH3-FS

7.40 kbit/s

0.10 kbit/s

15.00 kbit/s

0.30 kbit/s

22.8 kb/s

TCH/FR

CH4-FS

6.70 kbit/s

0.10 kbit/s

15.70 kbit/s

0.30 kbit/s

22.8 kb/s

TCH/FR

CH5-FS

5.90 kbit/s

0.10 kbit/s

16.50 kbit/s

0.30 kbit/s

22.8 kb/s

TCH/FR

CH6-FS

5.15 kbit/s

0.10 kbit/s

17.25 kbit/s

0.30 kbit/s

22.8 kb/s

TCH/FR

CH7-FS

4.75 kbit/s

0.10 kbit/s

17.65 kbit/s

0.30 kbit/s

22.8 kb/s

Fast link adaption operates based on received channel quality estimations in the MS
and BTS and changes the channel codec mode accordingly. Not all codec modes are
available when a call is established, but only a subset will be defined, which can be
specified by the operator. (For example the codec mode set may consist of CH0, CH3,
CH5 and CH7). The codec mode selected on uplink and downlink can be different.
The benefits of AMR-FR are improved speech quality in poor C/I conditions, improved
robustness in sparsely covered areas, and the potential to increase network capacity by
allowing a higher frequency re-use without sacrificing too much in speech quality.

3. PERFORMANCE EVALUATION
Detailed testing has been performed by Orange UK and FT Div RD and the results are
summarised below. The full report and results can be found in [1].
3.1. TEST SET-UP
A test bench simulating the radio path (uplink and downlink) was set up with interference and
fading applied to the uplink direction using a signal generator and Fading Simulator.
Voice quality was estimated with equipment using the PESQ algorithm, which estimates the
Mean Opinion Scores (MOS) using real speech samples. Evaluation is done on an end-to-end
basis between the GSM terminal and a fix ISDN base access. In each case the MOS for AMRFR was compared against EFR under identical conditions. The MOS vs speech quality is
defined by the following table:
Mean Opinion Scores
5: Excellent
4: Good
3: Fair
2: Poor
1: Bad

Group Technical/Network Architecture and Design/Radio Strategy and Quality


AMR-FR recommendation v1.0a.doc

Page 4/10

Document type

RSQ

RECOMMENDATION TO DEPLOY AMR-FR

In the context of mobile radio the voice quality remains acceptable as long as the MOS value is
greater or equal to 3.
TU3 and TU50 fading profiles were applied to the uplink, and static and dynamic interference
profiles were applied according to the following table. With the static interferer the interference
was injected at a constant power level. With the dynamic profiles the interference level varied
according to a defined pattern in each test.
Static profiles
No interference

C/I = 15 dB

C/I = 9 dB

Dynamic profiles
Regular C/I slope from
12 to 2 dB by steps of
1s
Alternate C/I value
between 15 and 5 dB
every 10 seconds
Eq 11 (ETSI defined
profile for AMR testing)

C/I = 5 dB
C/I = 2 dB

The AMR settings used for the testing were the default Nokia C/I threshold and
hysteresis values (details in Appendix). These parameters define the limit value for
switching AMR codecs.
3.2. RESULTS
The table below summarises the MOS results obtained.
TU3

C/I Profile
Static

Dynamic

Inf
15 dB
9 dB
5 dB
2 dB
Slope 12 2 dB
Alt 15/5 dB
EQ11

AMR FR
3.95
3.86
3.20
2.76
2.30
2.98
2.91
2.63

TU50
EFR
3.95
3.74
2.78
2.1
1.61
2.27
2.60
2.22

AMR FR
3.95
3.95
3.85
3.49
3.26
3.67
3.50
3.13

EFR
3.94
3.92
3.68
2.72
1.65
2.88
3.21
2.49

An analysis of the results shows that:


(i)
Under static interference conditions there is negligible difference in speech
quality when the C/I is > 15 dB. (In fact with the default Nokia codec
parameter set used it should dictate that there is negligible difference down
to 11 dB as the same EFR codec is used for both AMR-FR and EFR).
(ii)
Under static interference conditions when the C/I is < 15 dB there is
improved speech quality with AMR-FR, and the benefit is greatest for very
low C/I.
Group Technical/Network Architecture and Design/Radio Strategy and Quality
AMR-FR recommendation v1.0a.doc

Page 5/10

Document type
RECOMMENDATION TO DEPLOY AMR-FR

(iii)

(iv)
(v)

RSQ

Under the dynamic interference conditions the speech quality for AMR-FR is
consistently better than EFR when the C/I is rapidly changing, which will be
encountered in live networks.
In conditions that will be encountered in Orange networks it is anticipated
that improvise of up to 0.5 in MOS could be expected.
There are no circumstances where AMR-FR is inferior to EFR.

4. REAL NETWORK PERFORMANCE


4.1. HIGHLY LOADED 1:1 RE-USE NETWORKS
A potential problem exists in networks that operate with very low C/I as the speech
coding is now more robust than the associated SACCH and FACCH. When AMR was
being developed the performance of the SACCH and FACCH remained the same as
used for FR and EFR, with acceptable performance down to approximately 9 dB C/I.
However these signalling channels were never intended to operate in the low C/I
environments that AMR can operate in with the lower rate speech coding rates.
Therefore in networks operating with very poor C/I the speech can be maintained at
reasonable quality, but the associated signalling channels are so badly errored that the
call will drop due to missed measurement reports, handover and power control
commands and Radio Link Timeout, etc.
Real network data and simulation results submitted to 3GPP, and analysed by FT Div
RD in [2], have indicated that problems are most likely to arise in networks that are
planned to operate with almost 100% AMR mobile penetration, 1:1 re-use and high
frequency load resulting in consistent C/I of less than 7 dB. In these situations the most
likely reason for call drops was the high level of handover failures due to failed message
transactions.
3GPP is currently investigating several methods of improving SACCH and FACCH
robustness to overcome this problem, which generally rely on sending a second
SACCH/FACCH, but the final implementation has yet to be agreed.
Orange networks are designed to operate with a far more relaxed frequency re-use and
far higher C/I, and high call drops are not anticipated due to implementation of AMRFR.
4.2. ORANGE UK EXPERIENCE
Orange UK first rolled out AMR-FR to one BSC in October 2004 and has subsequently
rolled it out to two more BSCs. AMR-FR traffic accounts for up to 25% of all speech
calls carried on these sites, and there has been no degradation in call statistics since its
introduction.

Group Technical/Network Architecture and Design/Radio Strategy and Quality


AMR-FR recommendation v1.0a.doc

Page 6/10

Document type
RECOMMENDATION TO DEPLOY AMR-FR

RSQ

Assuming the Nokia default parameter set, four speech coding rates can be selected
depending on the C/I. The two highest rates of 12.2 kb/s and 7.4 kb/s operate down to
a C/I of 11 dB and 7 dB respectively, and consequently these rates will be used over
the vast majority of the network.
The only anecdotal difference reported between AMR-FR and EFR is in the call
quality/behaviour in poor radio conditions when the call is close to dropping. With EFR
codecs the speech quality typically degrades badly at a similar point as the
SACCH/FACCH begins to fail. With AMR-FR there is little or no indication that the call is
close to dropping as the speech quality does not suffer to the same degree.
Consequently the user is less likely to make attempts to improve the signal quality (e.g.
by moving closer to a window). However it must be stressed that in both cases it is the
failure of the SACCH/FACCH signalling that causes the call to drop, and that the
signalling has identical performance in both cases. Therefore overall AMR-FR does have
the better performance.

5. RECOMMENDATION
It has been demonstrated that AMR-FR does provide an improvement of up to 0.5
MOS points in radio environments that would typically be encountered in Orange
networks. The largest benefits are to be found when the C/I is rapidly varying or is less
than approximately 11 dB. For areas with good C/I of 11 dB or better there is negligible
difference in speech quality. (The precise threshold will depend on the C/I thresholds
selected for changing the codec rate).
Therefore AMR-FR will be of benefit in the following areas:
(i)
Those areas of the network where the speech quality of FR/EFR suffers due
to poor C/I (e.g. approximately <11dB), or where the interference
environment is rapidly changing.
(ii)
To improve speech quality where there is marginal network coverage (e.g.
indoor or rural).
(iii)
Where a higher frequency re-use is sought to increase network capacity.
Depending on the cost of deployment in individual networks it may not be cost effective
to roll out AMR-FR over the whole network. Instead it may be better to target those
areas where there are known problems with speech quality.
It is not anticipated that any Orange networks will suffer from the high rate of call drops
experienced in 1:1 re-use networks that are highly loaded with a high penetration of
AMR terminals, as Oranges C/I will greatly exceed the 4-5 dB that is typically found in
those circumstances. However a fix is being specified by 3GPP which will improve the
robustness of the SACCH/FACCH channels to greatly reduce this problem.

Group Technical/Network Architecture and Design/Radio Strategy and Quality


AMR-FR recommendation v1.0a.doc

Page 7/10

Document type
RECOMMENDATION TO DEPLOY AMR-FR

RSQ

6. DOCUMENT HISTORY
Version

Author

0.1
1.0

Ric Bailey
Ric Bailey

Modifications/Comments
First draft
First release

Group Technical/Network Architecture and Design/Radio Strategy and Quality


AMR-FR recommendation v1.0a.doc

Page 8/10

Document type
RECOMMENDATION TO DEPLOY AMR-FR

RSQ

7. REFERENCES
Document
[1] Adaptive Multi Rate Codec (AMR) Test
Report
[2] AMR Robustness Issue

Author
Peter Sutton

Date
6/8/2004

Version
1.1

Samuel Guillon

10/3/2005

1.0

Group Technical/Network Architecture and Design/Radio Strategy and Quality


AMR-FR recommendation v1.0a.doc

Page 9/10

Document type
RECOMMENDATION TO DEPLOY AMR-FR

RSQ

8. ANNEXES
Nokia AMR default parameter set
AMR FR CODEC MODE SET
AMR FR START MODE
AMR FR INITIAL CODEC MODE INDICATOR
AMR FR THRESHOLD 1
AMR FR THRESHOLD 2
AMR FR THRESHOLD 3
AMR FR HYSTERESIS 1
AMR FR HYSTERESIS 2
AMR FR HYSTERESIS 3

(FRC) 12.2 7.40 5.90 4.75 kbit/s


(FRS) 00
(ICMI)(FRI) 0
(FRT1).. 4.0 dB
(FRT2).. 7.0 dB
(FRT3).. 11.0 dB
(FRH1).. 1.0 dB
(FRH2).. 1.0 dB
(FRH3).. 1.0 dB

Group Technical/Network Architecture and Design/Radio Strategy and Quality


AMR-FR recommendation v1.0a.doc

Page 10/10

Você também pode gostar