. 92
( 132 .)


name suggests, this sublayer also reassembles the cells received, and passes them to the
CS. The overheads added by these sublayers are speci¬c to the protocols, but generally
cover such functions as message framing and error detection. The uses of the different
AALs are shown in Figure 7.20.

7.7.1 AAL1

AAL1 is designed primarily for the transmission of traf¬c that requires a constant bit rate
connection-oriented service. It provides the following services:

real-time non real-time


ATM Cell Switching

Physical Medium

Figure 7.20 ATM adaptation layers

• transport of traf¬c with a constant data rate and guarantees maintenance of that data rate;
• exchange of timing information between communicating systems;
• if required, the transfer of structure information such as message boundaries;
• indication of errors.

The prime example of this is voice traf¬c. Since delay is of key concern, adding a lot of
error checking would introduce too many delays so error detection and correction schemes
are kept to a minimum, and rather mechanisms for detecting missing or misinserted cells
are employed. Voice traf¬c is pretty robust and can handle errors in traf¬c.
Typically, AAL1 is used for transporting PDH connections, e.g. E1/T1, over an ATM
network. The principal application of this in UMTS is for the combined transport of GSM
and UMTS traf¬c on the same ATM backbone, thus simplifying the transport infrastruc-
ture. In many cases, this is advantageous, and cost-effective, for co-siting of GSM and
The primary consideration for this protocol is that bits come from the application
real-time, at a constant data rate, which must be maintained so that the application at
the destination receives the data at the same rate with a minimum amount of delay. To
minimize packetization delays, the header overhead added is kept as small as possible,
hence the simplicity of the error detection functions. Since timing is crucial, empty cells
will be transmitted, even if there is no traf¬c to send. Segmentation and reassembly (SAR) sublayer
The function of the SAR sublayer is to provide a mapping between the CS and the ATM
layer. At transmission, it receives a 47-byte block and a sequence number from the CS.

1 bit 3 bits 3 bits 1 bit 47 bytes

CSI SN SNP P Payload

Figure 7.21 AAL1 PDU format

It then adds a 1-byte header to make a 48-byte block, inserting the sequence number into
this header. This header provides sequencing and header error protection. At the receiver
end, the SAR accepts a 48-byte block from the ATM layer and extracts the header byte,
leaving the remaining 47-byte block of data. It checks for header errors and then presents
both the payload and the sequence number to the CS. The format of the SAR PDU is
shown in Figure 7.21.
CSI is the convergence sublayer indicator provided by the CS, and is described below.
SN is a 3-bit sequence number, used by the CS to check for missing or misinserted cells.
It has a cycle of 8 (i.e. 0 to 7). The SNP is the sequence number protection ¬eld. It is
again a 3-bit number, which is a CRC on the SN and CSI ¬elds. It uses the polynomial
x 3 + x + 1 to detect double-bit errors and correct single-bit errors. A further error check
is provided by the P or parity bit, which uses even parity. Convergence sublayer (CS)
During transmission, the CS takes data from an application and passes 47-byte blocks
to the SAR sublayer, and at the receiver, takes 47-byte blocks and passes them to an
application. It is at this layer that any variation in cell delay is handled to smooth the
cells and maintain the constant bit rate required by the application. This is performed
using a buffer. It may also deal with packetization delay, minimizing it by partially ¬lling
payloads. Since variation in delay is critical to real-time applications, methods to provide
clock recovery are also provided.
The CS will process the sequence number received from the SAR and based on this,
it can check for missing or misinserted cells. Should the application be video or high-
quality audio, optional forward error correction (FEC) may be performed. However, as
mentioned, the most common application is for PDH circuit emulation.
AAL1 can function in one of two modes of operation: structured and unstructured.

Structured data transfer (SDT)
Often, the data stream to be processed by the AAL is unbroken, or unstructured, meaning
that there is no requirement for the ATM to provide any information with regard to
framing, for example, an E1/T1 line. However, it can be structured, where boundaries
between messages need to be preserved. An example of a structured application is the
emulation of n — 64 k circuits, such as support for fractional E1 circuits, where only
some of the voice channels are being extracted for transport across the ATM network.
Here the ¬rst byte of the 47-byte SAR payload is used as a pointer to indicate where the
next message starts within the payload, measuring from the end of the pointer ¬eld. This
would be another voice channel. The CSI bit of the SAR header is set to 1 to indicate that

sublayer pointer 46 bytes 47 bytes 47 bytes 47 bytes

Segmentation & 0 47 byte payload 1 47 byte payload 2 3
47 byte payload 47 byte payload
header header header header

Figure 7.22 AAL1 structured data

this pointer is present. Only cells where the SN is an even number may contain pointers,
so the pointer must be a number from 0 to 92 to cover two payloads, 46 bytes for this
cell and 47 bytes for the next cell. The CSI in odd cells may be used for clock recovery,
as described below for unstructured mode. Enabling data to start at any point in the cell
through the use of this pointer means that message streams are not required to align to
cell boundaries. The pointer occurs once and only once in every eight cells, since the
3-bit sequence number (SN) has a wraparound of 8. It may therefore be present only in
cells with SN 0, 2, 4 and 6, occurring at the ¬rst possible chance. Note that the payload
must align to an octet boundary. An example is shown in Figure 7.22.
If the block size is 1 byte long, for example, where only a single 64 kbps channel is
being extracted, the pointer is not required. However, for larger block sizes it is needed
to point to the start of the next full block in the payload. The block size equates to the
number of channels being carried. For example, if three channels are used, then it will be
3 octets long. The block size being used is determined during the signalling establishment
of the connection.

Unstructured mode
For non-structured cells, the pointer ¬eld is not used so the full 47 bytes are available for
each cell. In even SN cells, the CSI has the default value of 0. However, for all cells with
an odd SN, the CSI bits form a 4-bit number over a cycle of 8 cells, referred to as the
residual timestamp (RTS), which encodes the difference between the clock of the sender
and that of a common reference. This enables the receiver to synchronize to the sender.
An example of the use of unstructured mode is the emulation of a full E1 circuit.

7.7.2 Circuit emulation service (CES)
The circuit emulation service application uses AAL1 to emulate a number of PDH circuits
over the ATM network. It speci¬es mechanisms for transporting circuits with and without
channel-associated signalling (CAS; see Section 3.7). As an example, consider that many
operators of cellular networks will be looking at co-siting their new UMTS base stations
with their existing 2G GSM base stations. If this is the case, it is useful to be able to use
the same infrastructure for both 2G and 3G traf¬c (see Figure 7.23).
One option that is available is to transport the 2G traf¬c over the ATM connection that
is being used to carry the 3G traf¬c to the RNC, via the 3G base station. To achieve

Antenna array


shared connection

Figure 7.23 2G/3G co-siting

this, the TDM circuit switched traf¬c of the GSM network, carried on E1/T1, is sent
to the UMTS base station where it is carried over AAL1 using the circuit emulation
service (CES). Both structured and unstructured mode can be used depending on whether
full or fractional E1/T1 lines are being used by the 2G network. At the RNC/2G base
station controller (BSC), some device, such as an ATM cross-connect, is used to separate
the different traf¬c streams, and then the circuit emulation process is reversed, with the
original E1/T1 stream being retrieved from AAL1 and forwarded to the 2G BSC. Note that
the UMTS base station traf¬c is carried over the same ATM connection, but uses AAL2,
which is discussed in the next section. An example of this is illustrated in Figure 7.24.
This can be extremely cost-effective if the operator owns or has access to an existing
ATM network that it wishes to use for both GSM and UMTS traf¬c. Consider that the
links from base station to BSC/RNC are where the bulk of the infrastructure investment
costs lie, since they are geographically separated, usually by a considerable distance. In
the core network, much of the equipment may be located in the same building or even
the same room, so cost of interconnection is trivial.
Consider the following use of structured mode. A GSM BTS consists of a three-sectored
site with one TRX per sector. Each call requires 16 kbps of transmission, and with eight
channels per TRX, this equates to 384 kbps of bandwidth (8 — 3 — 16 kbps). The GSM
BTS is connected to the UMTS BTS using an E1 channel. Of the E1, only six channels
are being used (6 — 64 kbps = 384 kbps). The CES function needs to extract these six
channels and pack them into the ATM layer using AAL1. The block size will be six
octets, using the ¬rst six channels from the E1 frame. This block is shown in Figure 7.25.
The six octets from each frame are extracted and placed in the AAL1 payload with the
pointer indicating the start of the ¬rst block. Figure 7.26 illustrates this concept.

Operator Premises
GSM Base
Station n x E1/T1
(Wired connection n x E1/T1
or microwave link)


Figure 7.24 AAL1 circuit emulation service


Octet from timeslot 1

Octet from timeslot 2
from current frame

Octet from timeslot 3

Octet from timeslot 4

Octet from timeslot 5

Octet from timeslot 6


. 92
( 132 .)