<<

. 93
( 132 .)



>>

Figure 7.25 n — 64 kbps block
6x64kbps channels

Frame 1 Frame 2 Frame 3 Frame 4

E1
7.7 THE ATM ADAPTATION LAYER (AAL)




CS
pointer


SAR

1 0 0 1 0 2
CSI SN SNP CSI SN SNP CSI SN SNP

Figure 7.26 E1 encapsulation in ATM using AAL1
449
450 UMTS TRANSMISSION NETWORKS


Note that the connection signalling establishes the required parameters such as mode
(structured/unstructured), clock recovery mechanism, structured mode block size and
required cell rate to support the constant bit rate.
For an n — 64 kbps structured service, the cell rate is derived from the bit rate according
to the following formula:

Cell rate = Round-up ((8000 — N)/46.875)

where 8000 is the cycle time for the frame, N is the block size, and the 46.875 indicates
the payload (i.e. 47 bytes but with 1 byte in 8 lost to the pointer.
In the E1 example, the cell rate would therefore be:

Cell rate = ((8000 — 6)/46.875) = 1024 cells/s

Limitations to AAL1
Even though AAL1 is designed for transport of real-time traf¬c, there are a number of
limitations to its use in the context of meeting the requirements of real-time traf¬c in a
modern network:

1. The AAL supports only one user over a virtual circuit. This requires a separate virtual
circuit for each user call, necessitating a large amount of signalling or circuit setup.
2. Since cells are sent even if there is no traf¬c, bandwidth is wasted. If an operator is
paying for bandwidth, this can be of great signi¬cance.
3. The AAL is designed for 64 k or n — 64 k voice channels. This is not particularly
suitable for the advanced CODECs used in cellular communications.
4. Currently, there is no mechanism for supporting these advanced CODECs which
provide for compressed voice or voice with silence suppression, etc.

These limit the uses of AAL1 in the 3G network to those described above and render
it unsuitable for the transport of 3G real-time user data.



7.7.3 AAL2
The AAL2 protocol was the last to be standardized, and the standard remained unde¬ned
for a number of years. It ¬nally saw the light of day in late 1997 after a close working
relationship between the ITU-T and the ATM Forum™s working group on voice telephony
over ATM (VToA). The standard has principal applications in trunking of narrowband
services, such as compressed audio with silence suppression. In UMTS networks, AAL2
is the main transport of user data. The AAL2 adaptation layer is used to transport user
traf¬c between the circuit switched core and RANs. Figure 7.27, shows the protocol stack
for this traf¬c across both the Iub and Iu interfaces. Along the Iu interface, AAL2 carries
only circuit switched traf¬c. However, along the Iub interface between the base station
7.7 THE ATM ADAPTATION LAYER (AAL) 451


Radio Access Network

Core Network
Iub Iu
Circuit
Switched


Base Station RNC


AAL2 AAL2 AAL2 AAL2


ATM ATM ATM ATM

Iub Iu
Physical Physical Physical Physical


Figure 7.27 UMTS circuit switched user data transport



Service specific SSCS-PDU SSCS-PDU
Payload
part header trailer
AAL2 Convergence Sublayer




(SSCS)
SSCS PDU


CPS packet
Payload
header

CPS packet
Common Part
(CPS)
Start Field Payload

CPS PDU




ATM Cell
Cell payload
layer header



Figure 7.28 AAL2 structure

and the RNC, all traf¬c, voice, video and data, is carried by AAL2 and then segregated
to/from the circuit- and packet switched cores at the RNC.
Unlike the other AALs, AAL2 has no SAR sublayer, but rather introduces a number
of sublayers at the CS. The structure of AAL2 is shown in Figure 7.28.
The service-speci¬c part of the CS (SSCS) sublayer is not part of the AAL de¬nition,
but rather may be speci¬ed by the application above the AAL. There can be multiple
452 UMTS TRANSMISSION NETWORKS


8 6 5 5 bits

CID LI UUI HEC CPS info

3 bytes 1-64 bytes

Figure 7.29 CPS packet format

SSCS layers de¬ned, and indeed there need not be a SSCS layer present at all. However,
the most commonly used SSCS is the SSSAR, de¬ned by the ITU-T (I.366.1), described
in Section 7.7.4.
Recall from Chapter 6 that at the BTS, transport blocks (TBs) will be encapsulated
inside the frame protocol (FP) for the interface before being passed to the AAL2 layer.
The common part of the convergence sublayer (CPS) has two components, a CPS
packet and a CPS PDU. The SSCS PDU is dependent on the particular service using
AAL2 and speci¬cs are de¬ned in the standard for that service. The role of the CPS
packet is to allow and identify a number of bidirectional AAL circuits, multiplexed over
a single virtual circuit. This minimizes packetization delay, critical for voice and video
applications, to reduce problems associated with echoing. The format of the CPS packet
is shown in Figure 7.29.
The ability to multiplex several channels over a single virtual circuit is an extremely
useful feature of AAL2, particularly in the context of 3G. The 8-bit channel ID (CID)
¬eld is used to identify the different AAL channels. There are up to 248 AAL channels
allowed over a single VC, with CID 0“7 used for management functions or reserved for
future uses. These are de¬ned in Table 7.11.
Since a CID represents a bidirectional channel, the same CID is used in both directions.
The length indicator (LI) gives the size, in bytes, of the payload, which is variable in
length. The maximum length can be 45 bytes. Optionally it can be 64 bytes for ISDN
compatibility. The user-to-user interface (UUI) is a means of identifying the particular
SSCS layer being used, and to pass information to this layer. Values 0“27 are for different
SSCS layers, 30“31 are for layer management and 28“29 are reserved for future use. The
UUI ¬eld may be null if the application does not de¬ne an SSCS layer. Finally, the HEC
provides header error control in the form of a CRC check over the rest of the header.
The CPS PDU ¬lls its payload with 48 bytes worth of CPS packets, i.e. 47 plus the CPS
packet header (Figure 7.30). Note that since the CPS packet is variable in length, there
may be more than one packet in the CPS PDU payload, or indeed a CPS packet may span
more than one PDU payload. The 1-byte start ¬eld (STF) consists of three components:
the offset ¬eld (OSF) indicates the start of the next CPS packet header within the payload.
This allows packets to span PDUs without wasting payload space, or requiring alignment
to the PDU structure. A value of 47 indicates that no CPS packet start is present in the

Table 7.11 AAL2 CID designations
CID Use
0 Unused
1“7 Management and future use
8“255 CPS user identi¬cation
7.7 THE ATM ADAPTATION LAYER (AAL) 453


6 1 1 bits
Offset field Sequence Parity
CPS packet(s) Pad
(OSF) number (SN) (P)

1 byte start field (STF) 1-47 bytes 0-46 bytes

Figure 7.30 CPS PDU format

payload. The sequence number (SN) and parity (P) bits provide some error detection on
the header. For instance, when there is no data received, the payload is padded out to ¬ll
the 48-byte ATM payload; this is to maintain real-time delivery.

Example of AAL2 application
In the context of UMTS, AAL2 provides a number of advantages in the transport of
real-time applications. Considering voice CODECs, two major aspects have changed, and
this is particularly evident in cellular systems:

1. CODECs used compress the voice to very small data rates in comparison to the
legacy 64 kbps channels of the standard telephone network.
2. CODECs take advantage of silence detection and suppression mechanisms.

In GSM, for example, most networks now use the 12.2 kbps enhanced full rate (EFR)
CODEC for voice. This enables a more ef¬cient use of bandwidth across the air and
BSS transmission network, where resources are ˜expensive™ either in quantity or mon-
etary terms. However, GSM is still based on a circuit switched network where time
slot resources are allocated for call duration. Therefore, even though the EFR CODEC
ceases generating voice samples or moves to a low rate silence descriptor once the sub-
scriber stops talking, the network cannot reuse that bandwidth. Its major advantage is in
reduction of power consumption, and hence battery use, in the mobile device. However,
AAL2/ATM allows this saving to be realized since the bandwidth during silence can be
reused. Recall that a typical voice activity factor is about 50%, whereas standard circuit
switched networks allocate resources on a full-duplex basis.
One of the original design features of ATM was that the payload size was chosen to pro-
vide a ¬xed, minimized packetization delay. With current coding schemes sampling voice
at much lower data rates, ¬lling a 48-byte payload is now considered too much delay.
In UMTS, the voice CODEC used is the adaptive multirate codec (AMR), which is
really a suite of coding schemes. In the course of a call, the network can vary the coding
scheme used as often as every 20 ms to optimize the use of resources. This means that if
the packet/cell size is ¬xed, while the CODEC rate changes, a variable delay is introduced.
AAL2 resolves this problem by implementing a small size, variable length packet. Each
of the different data rates required by the CODEC generates a certain size of SDU for the
AAL2 layer, so there can be a direct relationship between the coding scheme used, the
delay requirements and the AAL2 packet size. Table 7.12 shows the AMR coding rates
and data sizes (3GPP 26.102).
The different source rates in the AMR draw on many existing cellular coding rates.
For example, 12.2 kbps is the GSM EFR coding scheme. The AMR SID is a low rate
(1.8 kbps) used for transporting background noise.
454 UMTS TRANSMISSION NETWORKS


Table 7.12 AMR CODEC rates
AMR source SDU Size
rate (kbps) (bits)
AMR SID 39
4.75 95
5.15 103
5.9 118
6.7 134
7.4 148
7.95 159
10.2 204
12.2 244




Iub


1 VC connection


Base Station RNC
6 connections
on air interface


Figure 7.31 Six voice calls over a single virtual circuit

16kbps


<<

. 93
( 132 .)



>>