TDMA/CDMA System Planning Comparisons

In GSM or US TDMA networks, we have said that the handset produces a measurement
report that is then sent to the BSC to provide the basis for power control and handover.
The handset does radio power measurements typically every half second
(actually, 480 ms) and measures its own serving base station and up to five other base
stations in neighboring cells. This is the basis of Mobile-Assisted HandOff (MAHO).
This measurement process has had to be modified as GPRS has been introduced.
GPRS uses an adaptive coding scheme—CS1, 2, 3, or 4, depending on how far the
handset is away from the base station. The decision on which coding scheme to use is
driven by the need to measure link quality. Link quality measurement can only be performed
during idle bursts. In voice networks, the measurement has traditionally been
done every 480 ms. The fastest possible measurement rate is once every 120 ms (once
every multiframe), which is not fast enough to support adaptive coding.
IN E-GPRS (GPRS with EDGE), measurements are taken on each and every burst
within the equalizer of the terminal resulting in an estimate of the bit error probability
(BEP). The BEP provides a measure of the C/I on a burst-by-burst basis and also provides
information on the delay spread introduced by the channel and the velocity
(mobility) of the handset—that is, how fast the handset/mobile is traveling through
the multipath fading channel. The variation of BEP value over several bursts also provides
information on frequency hopping. A mean BEP is calculated per radio block
(four bursts), as well as the variation (the standard deviation of the BEP estimation
divided by the mean BEP) over the four bursts. The results are then filtered for all the
radio blocks sent within the measurement period.

For the higher coding schemes within GPRS (MCS7 to 9), the interleaving procedure
is changed. If frequency hopping is used, the radio channel is changing on a per-burst
level. Because a radio block is interleaved and transmitted over four bursts for GPRS,
each burst may experience a completely different interference environment. If, for
example, one of the four bursts is not properly received, the whole radio block will be
wrongly decoded and have to be retransmitted. In E-GPRS, the higher coding schemes
MCS7, MCS8, and MCS9 transmit two radio blocks over the four bursts. The interleaving
occurs over two bursts rather than four, which reduces the number of bursts that
need to be transmitted if errors are detected. This is shown in Figure 11.14. The process
is known as link adaptation.
These higher coding schemes work better when frequency hopping is used, but you
lose the gain delivered from deep interleaving.
Link adaptation uses the radio link quality measurement by the handset on the
downlink or base station on the uplink to decide on the channel coding and modulation
that should be used. This is in addition to the power control and handover decisions
being made. The modulation and coding scheme can be changed every frame
(4.615 ms) or four times every 10 ms (the length of a frame in IMT2000).
E-GPRS is effectively adapting to the radio channel four times every 10 ms (at a 400
Hz rate). IMT2000 adapts to the radio channel every time slot, at a 1500 Hz rate.
Although the codecs and modulation scheme can theoretically change every four
bursts (every radio block), the measurement interval is generally slower.
The coding schemes are grouped into three families: A, B, and C. Depending on the
payload size, resegmentation for retransmission is not always possible, thus determining
which family of codes are used.
Table 11.10 shows the nine coding schemes used in E-GPRS, including the two modulation
schemes (GMSK and 8PSK).

The performance of a handset sending and receiving packet data is therefore
defined in terms of throughput, which is a product of the gross throughput less the
retransmissions needed. The retransmission will introduce delay and delay variability,
which will require buffering. If the delay variability exceeds the available buffer bandwidth,
then the packet stream will become nonisochronous, which will cause problems
at the application layer.
We can see that it becomes much harder to nail down performance in a packetrouted
network.
As we will see in the next section, link budgets are still established on the basis of
providing adequate voice quality across the target coverage area. Adaptive coding
schemes if well implemented should mean that when users are closer to a base station,
data throughput rates can adaptively increase. When a user is at the cell edge, data
throughput will be lower, but in theory, bit error rates and retransmission overheads
should remain relatively constant.
Having convinced ourselves that this may actually happen, we can now move on to
radio system planning.