CDMA

In the mid-1990s CDMA cellular networks began to be deployed in the United States,
Korea, and parts of Southeast Asia. Effectively, CDMA takes many of the traditional RF
tasks (the achievement of selectivity, sensitivity, and stability) and moves them to baseband.
The objective is to deliver processing gain that can in turn deliver coverage
and/capacity advantage over the coverage and/capacity achievable from a TDMA air
interface. Endless arguments ensued between the TDMA and CDMA camps as to
which technology was better.
In practice, because of political and regulatory reasons and other factors such as timing,
vendor, and operator support, GSM became the dominant technology in terms of
numbers of subscribers and numbers of base stations deployed, which in turn conferred
a cost and market advantage to GSM vendors. However, the technology used in
these early CDMA networks has translated forward into 3G handset and network
hardware and software. It is easier to qualify some of the design options in 3G handsets
if we first cover the related design and performance issues highlighted by CDMA
implementation to date.
The original principle of CDMA, which still holds true today, is to take a relatively
narrowband modulated signal and spread it to a much wider transmitted bandwidth.
The spreading occurs by multiplying the source data with a noise like high-rate
pseudorandom code sequence�"the pseudorandom number (PN). The PN as a digital
number appears to be random but is actually predictable and reproducible having
been obtained from a prestored random number generator. The product of the source
data and the PN sequence becomes the modulating signal for the RF carrier.
At the receive end, the signal is multiplied by the same prestored PN sequence that
was used to spread the signal, thereby recovering the original baseband (source) digital
data. Only the signal with the same PN sequence despreads. Effectively, the PN
sequences characterize the digital filter, which correlates or captures wanted signal
energy, leaving unwanted signal energy down in the noise floor.
Multiple users can exist simultaneously on the same RF channel by ensuring that
their individual spreading codes are sufficiently different to be unique. To control
access and efficiency on a CDMA network, the spreading code is a composite of several
digital codes, each performing a separate task in the link. It is usual to refer to each
sequence or code as a channel.
IS95 defines the dual-mode AMPS/CDMA technology platform, IS96 the speech
coding (currently either 8 kbps or 13 kbps), IS97 and 98 the performance criteria for
base stations and handsets, and IS99 data service implementation. What follows is
therefore a description of the IS95 air interface, which then served as the basis for
CDMA2000.
In IS95, there is one pilot channel, one synchronization channel, and 62 other channels
corresponding to 64 Walsh codes. All 62 channels can be used for traffic, but up to
7 of these may be used for paging. The 64 Walsh codes of length 64 bits are used for
each of these channels. Walsh Code W0 is used for the pilot, which is used to characterize
the radio channel. Walsh Code W32 is used for synchronization. Other Walsh
codes are used for the traffic. The Walsh codes identify channels on the downlink,
which means they provide channel selectivity.

Walsh codes are a sequence of PN codes that are orthogonal in that, provided they
remain synchronized with each other, the codes do not correlate or create co-code or
adjacent code interference. Orthogonal codes are codes of equal distance (the number
of symbols by which they differ is the same). The cross correlation�"that is, code interference�"
is zero for a perfectly synchronous transmission.
On the uplink, the channel bits are grouped into 6-bit symbols. The 6-bit group
(higher-order symbol) generates a 64-chip Walsh code. The orthogonality of the 64
codes gives an increased degree of uniqueness of data on the uplink�"that is, it provides
selectivity.
The resultant Walsh code is combined with a long code. The composite channel rate
is 1.228 Mcps; in other words, the code stream is running at a rate of 1.228 MHz. The
long code is a PN sequence truncated to the frame length (20 ms). On the uplink, the
long code provides user-to-user selectivity; on the downlink, one long code is used for
all base stations but each base station has a unique PN offset (a total of 512 time PN offsets
are available). So within a relatively wideband RF channel, individual user channels
are identified on the downlink using Walsh codes�"with long codes providing
cell-to-cell selectivity�"individual user channels are identified on the uplink by use of
the 6-bit symbols, and long codes are used to provide user-to-user selectivity.
From a handset design point of view, digital filters have replaced the time slots and
RF filters used in the TDMA networks. Although RF filtering is still needed to separate
multiple 1.25 MHz RF carriers, it is intrinsically a simpler RF channel plan, and it can
be implemented as a single-frequency network if traffic loading is relatively light and
evenly distributed between cells. 14