Convolution and Correlation

Convolution and Correlation

Convolutional encoding is a key mechanism for delivering coding gain, or sensitivity,
in 2G and 3G cellular handsets. A further development of convolutional encoding,
called turbo coding, is proposed for 3G handsets. Here two or more convolutional
encoders are used together with interleaving and puncturing to increase coding distance.
Convolutional encoders are effectively implemented as shift registers. They are
similar in terms of implementation to the PN code generators used to create long codes
for IMT2000DS and IMT2000MC.
In IMT2000DS long codes are used to provide node B-to-node B selectivity on the
downlink and user-to-user selectivity on the uplink (covered in detail in Chapter 3).
Both techniques exploit digital domain processes to deliver distance. Convolutional
encoders deliver distance between 0s and 1s (sensitivity). PN code generation delivers
distance between parallel code streams (selectivity).

There is thus commonality in terms of processing between realizing convolutional
encoders/decoders and CDMA code generation (long codes and OVSF codes). Both
are designed to be adaptive. You can move in real time from a 1/2 encoder to a 2/3
encoder, and you can move in real time between multiple long codes and variablelength
orthogonal variable spreading factor (OVSF) codes, depending on the information
to be transmitted and the channel conditions.
Fourth-generation handsets may also use trellis coding. Trellis coding is used
presently in some satellite systems using higher-level modulation where the pulse and
amplitude states are close together�"for example, 64-level QAM or higher., in trellis
coded modulation schemes where modulation states are close together (A), they are
channel coded for maximum distance; where they are far apart (B), they are channel
coded for minimum distance. This delivers a significant increase in Eb/No performance
but at significant cost in terms of DSP processor power and power budget; this was
practical in 4G handsets but is not practical today.
We have already described the transition to higher-level modulation methods. As
the number of modulation states increases, the requirement for linearity increases. In
subsequent chapters we explore the role of the digital signal processor in delivering
linearity and power efficiency by adapting and predistorting waveforms prior to modulation.
DSPs therefore allow us to deliver performance gains both in terms of
throughput (higher-level modulation), robustness (channel coding), voice, image and
video quality (source coding), and flexibility (the bandwidth on demand and multiple
per-user traffic streams available from CDMA).
Summary
Over the past 100 years, a number of key enabling technologies have helped deliver
year-on-year performance gains from wireless devices�"the development of valve
technology and tuned circuits in the first half of the twentieth century, the development
of transistors and printed circuit boards from the 1950s onward, the development
of microcontrollers in the 1970s (making possible the first generation of frequencyagile
analog cellular phones), and more recently, the development of ever more powerful
digital signal processors and associated baseband processing devices.
In terms of RF performance, as RF component selectivity, sensitivity, and stability has
improved, we have been able to move to higher frequencies, realizing a reduction in the
size of resonant components and providing access to an increased amount of bandwidth.
Two-way radio design in the latter half of the twentieth century moved to progressively
narrower RF channel spacing. Conversely, cellular networks have moved progressively
from 25 or 30 kHz spacing to 1.25 MHz or 5 MHz, with selectivity
increasingly being delivered at baseband. This has resulted in simpler RF channelization,
though the need to support backward compatibility has resulted in some significant
design and implementation challenges, encompassing not only multiple modes
(multimode AMPS/TDMA, AMPS/CDMA, GSM/IMT dual-mode processing) but
also multiple bands (800, 900, 1800, 1900, and 2100 MHz), both paired and unpaired.

There are various evolutionary migration paths for existing TDMAand CDMAtechnologies,
but at present 5 MHz RF channel spacing is emerging as a reasonably common
denominator for 3G handset design. The choice of 5 MHz has made possible the
design of handsets delivering variable bit rate�"supporting a ratio of 64 to 1 between
the highest and lowest bit rates�"multiplexed as multiple traffic streams and modulated
onto a relatively constant quality radio channel. Variable-rate constant-quality
channels provide the basis for preserving information bandwidth value.
Bandwidth quality can be improved by exploiting digital coding and digital processing
techniques. This technique can be used to increase throughput, to improve resilience
against errors introduced by the radio channel, and to improve bandwidth flexibility.
In the next two chapters, we discuss the RF hardware requirements for a cellular
handset and how RF hardware determines bandwidth quality. 31