Page 109 - From GMS to LTE

General Packet Radio Service (GPRS) and EDGE 95

USF in each downlink block, the PCU can dynamically schedule who is allowed to send
in the uplink. Therefore, this procedure is also referred to as ‘dynamic allocation’.
Mobile devices that support high multislot classes are not able to listen in downlink
direction on all the timeslots assigned to it for uplink transmission opportunities. In
such cases the ‘Extended Dynamic Allocation’ scheme is used, which uses a single USF
to assign resources on several timeslots for a user for uplink data transfers.
Note that the USF information in the header and data portion of a downlink block is
usually not intended for the same user. This is because the assignments of uplink and
downlink resources are independent. This makes sense when considering web surfing,
for example, where it is usually not necessary to already assign downlink resources at
the time the universal resource locator (URL) of the web page is sent to the network.
For mobile devices that have an uplink TBF established, the network needs to send
control information from time to time. This is necessary to acknowledge receipt of
uplink radio blocks. The logical PACCH that can be sent in a radio block, instead of a
PDTCH, is used to send control information. The mobile device recognizes its own
downlink PACCH blocks because the header of the block contains its TFI value.
The PCU will continue to assign uplink blocks until the mobile device indicates that
it no longer requires blocks in the uplink direction. This is done with the so‐called
‘countdown procedure’. Every block header in the uplink direction contains a four‐bit
countdown value. The value is decreased by the mobile device for every block sent at
the end of the data transfer. The PCU will no longer assign uplink blocks for the mobile
device once this value has reached 0.
While coordinating the use of the uplink in this way is quite efficient, it creates high
latency if data is only sent sporadically. This is especially problematic during a web‐brows-
ing session, for two reasons. As shown at the end of this chapter, high latency has a big
impact on the time it takes to establish Transmission Control Protocol (TCP) connections,
which are necessary before a web page can be requested. Furthermore, several TCP connec-
tions are usually opened to download the different elements, such as text, pictures, and so
on, of a web page, so high latency slows down the process in several instances. To reduce this
effect, the GPRS standard was enhanced by a method called the ‘extended uplink TBF’. If
both network and mobile device support the functionality, the uplink TBF is not automati-
cally closed at the end of the countdown procedure but is kept open by the network until the
expiry of an idle timer, which is usually set in the order of several seconds. While the uplink
TBF is open, the network continues to assign blocks in the uplink direction to the mobile
device. This enables the mobile device to send data in the uplink direction quickly without
requesting for a new uplink TBF. The first mobile devices and networks that supported
extended uplink TBF appeared on the market in 2005 and a substantial improvement of web
page download and delay times can be observed, as discussed at the end of the chapter.

Temporary Block Flows in the Downlink Direction
If the PCU receives data for a subscriber from the SGSN, it will send a Packet Downlink
Assignment message to the mobile device similar to the one shown in Figure 2.20 in the
AGCH or the PAGCH. The message contains a TFI of a TBF and the timeslots
the mobile device has to monitor. The device will then immediately start monitoring
the timeslots. In every block it receives, it will check if the TFI included in the header
equals the TFI assigned to it in the Packet Downlink Assignment message as shown in
Figure 2.21. If they are equal, it will process the data contained in the data portion of the