<<

. 94
( 132 .)



>>

4ms
CPS
payload to fill
header

3 bytes 8 bytes

Figure 7.32 AAL2 interaction with voice channel

Consider the following example of six voice channels being transported from the base
station to the RNC over one virtual circuit using AAL2, as shown in Figure 7.31.
For simplicity, in this example, each channel is 16 kbps adaptive differential pulse code
modulation (ADPCM). Further consider that there is a requirement for a packetization
delay of only 4 ms. A payload size of 8 bytes is chosen, since it will take 4 ms to ¬ll the
payload at 16 kbps, as illustrated in Figure 7.32. Note that ATM is only used over the
transmission network, and not the air interface.
CPS
CPS 8 CPS 8 CPS 8 CPS 8 CPS 8 CPS 8
Packet

11 bytes 11 bytes 11 bytes 11 bytes 11 bytes 11 bytes

SFD

CPS 8 CPS 8 CPS 8 CPS 8 CPS
CPS
PDU 48 bytes
SFD
7.7 THE ATM ADAPTATION LAYER (AAL)




padding
8 CPS 8

48 bytes



CPS 8 CPS 8 CPS 8 CPS 8 CPS
ATM
Cell 1

5 byte Payload
header

8 CPS 8 padding
ATM
Cell 2

5 byte Payload
header

Figure 7.33 AAL2 overheads
455
456 UMTS TRANSMISSION NETWORKS


The voice channels will pack into the ATM as shown in Figure 7.33. First, each channel
¬lls a CPS packet, adding the 3-byte CPS header, with each channel identi¬ed by its own
CID. Then, the CPS packets are packed into CPS PDUs with the 1-byte start ¬eld added,
and padding where necessary. This is then inserted into ATM cells, adding the 5-byte ATM
header. The padding may seem excessive, but delay requirements are of highest priority.
In summary, the advantages of using AAL2 connections are:

• AAL2 is particularly suitable for the transport of voice packets produced by advanced
speech CODECs.
• AAL2 enables up to 248 channels to be multiplexed on a single virtual circuit. This is
extremely advantageous if the virtual circuit is externally owned, since it can then be
fully utilized. This may frequently be the case in a 3G network.
• The packetization delay introduced by ¬lling a 48-byte ATM cell can further be reduced
by using a small CPS payload.
• The delay can be kept ¬xed as the CODEC changes by allowing the size of the AAL2
packet to vary.

Typically in a UMTS environment, there will be many users, requiring different resources
and data rates. All of these must be carried by AAL2 between the base station and the RNC
on the Iub interface. At the user equipment (UE), all applications pass their data to the
RLC/MAC layer where the data is formatted into TBs and then is forwarded to the physical
layer for coding and radio frame segmentation. At the BTS, the data is brought back to
TBs. If the application is sending IP packets, these will be segmented at the RLC layer to
¬t into an appropriate TB size. For example, a 32 kbps connection may generate a 336-bit
TB every 10 ms (320 bits + 16 bit RLC overhead). The total SDU size used at the AAL2
layer will be the TB(s) size plus the overheads of the frame protocol. This protocol includes
such overheads as transport formats and relevant air interface parameters.
The CODECs refer to voice data samples in terms of bits, whereas at the AAL2 layer,
packet size is de¬ned in multiples of bytes. The padding to byte boundaries is done by
the frame protocol.
AAL2 presents no problem in the transport of these different sized TBs. Consider the
following example of, again, six users, but this time with six different packet lengths. The
user packets are as listed in Table 7.13.

Table 7.13 Example AAL2
user packet length
User Packet length
User 1 45
User 2 15
User 3 22
User 4 8
User 5 24
User 6 40
7.7 THE ATM ADAPTATION LAYER (AAL) 457


As shown in Figure 7.34, the payload from user 1 is too large to ¬t into one CPS PDU
and the last octet is placed in the second CPS PDU after the start of frame delimeter.
AAL2 connections may be established and released using the AAL2 signalling protocol,
Q.2630, de¬ned by the ITU-T. This signalling mechanism is outlined in Section 7.13. Note
that AAL2 channels inherit the QoS of the virtual circuit in which they are carried and
there is no standardized way to provide different QoS to individual CID streams.



7.7.4 Service-speci¬c convergence sublayer (SSCS)
As was seen in Figure 7.28, the CS allows for an SSCS to provide any additional features
to an application that are not directly supported in AAL2. Considering again a UMTS
network, the Iub interface is required to transport traf¬c from multiple users, multiplexing
them at the AAL2 layer. Although part of that traf¬c will be generated by small voice
CODECS, some of it will consist of data traf¬c, such as IP packets. The payload for
AAL2 is the frame protocol, and its size is dependent on the TB size and the number
of TBs that may be sent simultaneously. Particularly for higher data rates, the size and
number of TBs generated will create a frame protocol block that is larger than the AAL2
packet size of 45 bytes. Consider that a 384 kbps connection may generate a TB of 3840
bits (480 bytes) every 10 ms. Therefore, what is needed above the AAL2 is a layer that
can provide segmentation and reassembly of these larger packets. The ITU-T has de¬ned
such a layer (I.366.1), referred to as the service-speci¬c segmentation and reassembly
convergence sublayer (SSSAR). It sits on top of the AAL2 CS layer and consists of three
sublayers, as shown in Figure 7.35.
At a minimum, the service offered is merely segmentation and reassembly of large user
data packets, and it is this function that is utilized in UMTS. This is performed by the
SSSAR. The SSSAR will accept a packet of up to 64 kbytes in size (the maximum size
of an IPv4 packet) from the upper layer, segment it and reassemble it at the far end. On
a connection, there is no opportunity for cells to get out of order; therefore sequencing of
the segmented portions is not necessary as this feature is inherited from the lower layers.
What is required is a noti¬cation that segmentation and reassembly is being used, and an
indication of the last segment received. This is performed using the UUI bits of the CPS
packet header. Recall that a UUI value of 0“27 indicates the use of a SSCS layer. A UUI
value of 27 is used to indicate that there is more data needed to complete the SSSAR
SDU, while 0“26 indicate that the ¬nal segment has been received. Since there may be
other SSCS layers implemented, normally a value of 26 is recommended to indicate the
last piece. The format of the SSSAR PDU is shown in Figure 7.36.
It is usual that when segmentation occurs, all segments excluding the ¬nal one are the
same length as determined by the maximum payload size of the CPS packet, i.e. 45 bytes.
Consider a payload of 1200 bytes that needs to be segmented. The segmentation will
consist of 26 segments of 45 bytes, with the CPS UUI ¬eld set to 27, and one segment
of length 30, with the CPS UUI ¬eld set to 26.
In addition to the SSSAR, this SSCS also provides two further optional functions:
SSTED and SSADT. The SSTED provides a mechanism to detect errors in the pay-
load. It does this by adding a trailer of 8 bytes. The format of the trailer, shown in
458




user 1 user 2 user 3 user 4 user 5 user 6


CPS
45 15 22 8 24 40
Packet




CPS
CPS
CPS
CPS
CPS
CPS
48 bytes 18 bytes 25 bytes 11 bytes 27 bytes 43 bytes

SFD

CPS 1 9 padding
44 15 22 8 24 31




CPS
CPS
CPS
CPS
CPS
CPS
PDU
48 bytes 48 bytes 48 bytes 48 bytes



Cell 1 ATM 44




CPS
5 byte Payload
header
1
Cell 2 ATM 15 22




CPS
CPS
CPS
5 byte Payload
header

Cell 3 ATM 8 24 9




CPS
CPS
5 byte Payload
header

padding
Cell 4 ATM 31

5 byte Payload
header


Figure 7.34 AAL2 with unequal payload
UMTS TRANSMISSION NETWORKS
7.7 THE ATM ADAPTATION LAYER (AAL) 459


SAP SAP SAP


Service Specific Assured Data Transfer
(SSADT)


primitives


Service Specific Transmission Error Detection (SSTED)


primitives


Service Specific Segmentation and Reassembly (SSSAR)




Common Part Sublayer (CPS)



Figure 7.35 SAR service-speci¬c convergence sublayer



UUI SSSAR-PDU Payload: 1 - 65 568 bytes


CPS-packet header CPS-packet payload

Figure 7.36 SSSAR PDU

<<

. 94
( 132 .)



>>