Next-Generation CDMA Technologies

Next-Generation CDMA Technologies
As mentioned at the beginning of this chapter, the CDMA technologies used in all 3G wireless
communication systems share the same core IPRs as proposed by the IS-95A standard, which was
initially designed by Qualcomm Inc. in the early 1990s. Those core IPRs include closed and openloop
power control, RAKE receiver, DS spreading, soft-handover technique, and so on. Therefore, it
is sad to say that CDMA technology has not been innovated in 3G wireless communications. For this
reason, we call the CDMA technology adopted in 2- to 3G wireless systems The First-Generation
CDMA technology, or simply 1G-CDMA technology.
1G-CDMA technology has to undergo a thorough technological innovation to suit the needs of
B3G wireless communications. Several technical aspects that limit further performance enhancement
in the 1G-CDMA system should be addressed. In particular, a CDMA system should not always
be considered as an interference-limited system, whose capacity can never reach half of the PG. To
design the next-generation CDMA technology, we should bring several fundamental changes to the
existing 1G-CDMA on CDMA sequence generation, spreading and carrier modulation schemes, and
overall system architecture designs, and so on. In the following text, we discuss these issues based
on our previous research experience.
7.3.1 Importance of Using Good CDMA Codes
As mentioned in Section 2.3.3, the characteristic features of a CDMA system can be basically determined
by the spreading codes used by the CDMA system. Thus, it is of ultimate importance to use
the right codes to make sure the performance of the overall CDMA system will not be limited by
the inherent defects of the codes. For instance, the use of the Walsh-Hadamard sequences in the
IS-95A/B and the CDMA One systems for down link channelization gives a very poor performance
when Multipath Interference (MI) is present. Even the use of a RAKE receiver will not help much
in such a scenario, as the output from each finger of the RAKE consists not only of the useful signal
components, but also of plenty of unwanted interference components caused by multipath returns. In
this case, the choice of the Walsh-Hadamard sequences has predetermined the basic performance of
the IS-95A/B system, whose cell-wise capacity never exceeds one third of the Walsh-Hadamard code
length of 64/3.
It is also noted that the need for many complex subsystems required by two to three wireless
systems based on 1G-CDMA technology are due to the poor spreading codes. These complex  subsystems include closed- and open-loop power control, CDMA multiuser detection, smart antenna
with adaptive beam-forming, RAKE receiver, and so on. The closed- and open-loop power control
is required because of the near–far effect caused by nontrivial CCFs of the CDMA codes. The multiuser
detection should be applied to all 1G-CDMA-based wireless systems because of the strong
correlation existing in user signature codes due to their imperfect correlation properties. The reason
that we need to use smart antenna and adaptive beam-forming techniques in a cell site is to suppress
strong cochannel interference produced by poor CCFs among user signature codes. A RAKE
receiver is used in all 1G-CDMA systems to overcome MI, which will never be harmful if the
signature codes offer ideal correlation properties (zero autocorrelation side lobes for any code and
zero CCFs for any pair of codes). Therefore, the use of more desirable spreading codes in a CDMA
system will not only improve the overall performance, but also greatly simplify the hardware complexity,
without the need for all those complex subsystems designed in particular for 1G-CDMA
technology.
7.3.2 System Model and Assumptions
Several assumptions should be made to facilitate our discussions carried out in this section.
• First, we consider a generic CDMA system that uses short codes (with chip width being Tc) to
spread data bits. The code length is exactly equal to the data bit duration (Tb).
• Second, the wireless system under consideration consists of mobile terminals and a base station
(BS). We will focus on the intracell physical layer architecture of a new CDMA system and
will not address any upper layer issues of a mobile network, nor those involving different cells
in a mobile network.
• Third, in the discussions given in this section all CDMA codes are classified into two categories,
unitary codes and complementary codes. The former includes almost all traditional spreading
codes, such as Gold, Kasami, Walsh-Hadamard, and OVSF codes, and so on, which work
on a one-code-per-user basis. The latter forms another group of CDMA codes working on a
flock of codes basis. Each user in such complementary code–based CDMA systems should
be assigned a flock of M element codes, which ought to be sent via different channels (either
in frequency or time) to a specific receiver for complementary auto- and CCF reconstruction.
Obviously, the unitary codes are only a special case of the complementary codes with M = 1.
• Fourth, our discussion on the CDMA codes will not be limited to any specific chip value, being
either complex, real, or binary, to make the discussions as general as possible.
Let us consider a generic K-user CDMA system, where each user is assigned one unique flock
of M codes (ck,1, ck,2, . . . , ck,M) for CDMA purpose. Each code ck,m consists of N chips, where
1 ≤ k ≤ K and 1 ≤ m ≤ M. Assume that the signal of interest is from the user 1. If M = 1, then the
system model is equivalent to a traditional unitary code based CDMA system; otherwise, ifM >1 it
makes a complementary code–based CDMA system. Therefore, the discussions given in this section
will make sense in general for new CDMA systems using either unitary or complementary codes.
7.3.3 Spreading and Carrier Modulations
The most important role of spreading modulation is to achieve a PG for some operational advantages
over a non-spread-spectrum communication system. On the other hand, carrier modulation functions
as a vehicle to send user data to a receiver through RF transmission. They work closely together and should thus be jointly considered in a CDMA system design. The major concern in the design of
spreading and carrier modulations for the next generation CDMA, similar to all traditional CDMA, is
centered on bandwidth efficiency and power efficiency. Obviously, they form a dual, which often work
in a counteractive way. Therefore, to achieve a good trade-off between the two becomes extremely
important. In this subsection, we discuss two different spreading modulation schemes, DS spreading,
and offset-stacking (OS) spreading, both of which could possibly be used in the next generation
CDMA systems, and their impact on the carrier modulations.
DS spreading versus OS spreading
Both DS and OS spreading modulation schemes can be applied to a CDMA system based on complementary
codes (here, any unitary code is treated as a special case only), resulting in either a
DS-CDMA or an OS-CDMA scheme. The DS spreading modulation has been widely used in 2- to
3G CDMA mobile cellular standards, whereas the OS spreading scheme was only introduced recently
[210]. The basic idea behind the OS spreading is that a new bit will start immediately after n-chip
shift relative to its previous bit, and thus the consecutive bits are stacked over one another with n
relative offset chips, where n can take any integer from 1 to N (N is the element code length). For
more detailed information of the OS spreading technique, the readers may refer to [210], in which,
however, only the case with n = 1 is discussed. Clearly, if we allow arbitrary relative offset chips n,
where 1 ≤ n ≤ N, between two consecutive data bits in an OS spreading modulator, a DS-CDMA
system becomes only a special case of the OS-CDMA scheme with its relative offset chips being
equal to the element code length or N chips. Thus, the study on an OS-CDMA scheme makes
sense in general. The use of more relative offset chips between two consecutive bits will result in a
slower transmission rate. Figure 7.1 illustrates the variable numbers of relative offset chips between
two consecutive bits in an OS-CDMA system, where only two short element codes, (+++−) and
(+−++), which are assigned to the same user and sent via different carriers (f1 and f2), are shown
in Figure 7.1 for the simplicity of illustration.
Therefore, we can introduce a merit figure of spreading efficiency (SE) to describe how many
data bits can be conveyed in one chip duration. If a fixed chip width is considered, the SE figure
only gives the bandwidth efficiency of a CDMA system. The greater the SE is, the higher the bandwidth
efficiency of a CDMA system will be. Obviously, we have the relation between the number
of relative offset chips n and SE as SE = 1
n, or 1
N ≤ SE ≤ 1. Therefore, the use of n = 1 relative
offset chip in an OS-CDMA scheme leads to the highest SE figure equal to one, which is exactly N
times higher than that of a traditional DS-CDMA system, and thus a substantial gain in bandwidth
efficiency can be achieved. Unfortunately, all unitary codes, such as Gold, Kasami, Walsh-Hadamard,
OVSF codes, and so on, are not suitable for the OS spreading due to the excessively high CCF
between any two codes if modulated by an OS spreader. Only orthogonal complementary (OC)
codes can be successfully applied to an OS-CDMA system, giving a multiple access interference
(MAI) free operation, because of the ideal CCFs between any two OC codes sent in either synchronous
or asynchronous channels. We name this superior property of the OC codes as isotropic
MAI-free operation. Therefore, an OC code–based OS-CDMA system has a great advantage over
its counterparts in terms of cell-wise average capacity, in addition to its extremely high-bandwidth
efficiency.
Binary versus M-ary carrier modulations
If a transmitter is implemented with a spreading modulator followed by a carrier modulator, the
choice of the latter greatly depends on the former’s operation mode (i.e., either DS or OS spreading).
For instance, if DS spreading is considered, the simplest form of the carrier modem can be BPSK
(although other more complex modems, such as QPSK, quadrature amplitude modulation (QAM), and
so on, can also be applied). However, if using an OS spreader, we have to deal with a multileveled baseband signal, as shown by the shaded blocks in Figure 7.1. The dynamic range of the composite
multileveled signal depends on relative offset chips n between two neighboring bits and the element
code length N. In general, if each chip takes only binary values, the dynamic range, DOS, of the
output signal from an OS spreader can be DOS = N
n
 + 1, where 1 ≤ n ≤ N and the notation x
 stands for the smallest integer greater than x. The output signal from an OS spreader has to be
modulated by a multileveled or M-ary digital modem, such as M-QAM, M-PSK or M-PAM, and so
on, where M ≥ DOS and DOS is always an odd number, as shown in Figure 7.1. Also note that the
appearance frequencies of different levels in the composite multileveled signal are different, with zero
appearing most frequently if n < N. The greater the absolute value of a level is, the less frequently
it appears. Thus, the histogram of the different levels gives a Gaussian-shaped envelope with its
center or the mean value being zero. If an M-QAM modem should be used to map each level to
a particular constellation point in its two-dimensional constellation plan, then there is an interesting
design problem as to how to minimize the average symbol or the bit error rate (BER) with respect
to the ways to map all the DOS levels onto the same number of constellation points, which can be
placed onto any points in an X-Y plan.
Therefore, the use of OS spreading can enhance bandwidth efficiency due to its very high SE,
approaching to one for n = 1. However, the realization of this high-bandwidth efficiency is conditioned
on the relatively low-power efficiency of an M-ary modem. If the latter fails to perform
satisfactorily in a channel where many impairing factors such as MI and noise may exist, its bandwidth
efficiency will be a question. On the other hand, DS spreading can never be comparable to the
OS spreading in terms of bandwidth efficiency with its SE being only 1
N, where N is the element
code length. However, the output signal from DS spreading forms a binary bit stream, which can be
modulated via an extremely power efficient BPSK or QPSK scheme. Therefore, the overall performance
difference between a DS spreader cum BPSK/QPSK modem and an OS spreader cum M-QAM modem in varying circumstances is an interesting topic of study. Therefore, the final selection for
either DS spreading or OS spreading should be exercised very carefully, depending on the nature of
the operation environment.