Digital Design of Communication Systems - Digital Design of Signal Processing Systems: A Practical Approach - page 536

The FPGA for many streaming applications implements MPSoC-based architecture. The
components that are mapped on FPGA are already developed. An ideal SDR framework generates
all the interfaces and makes an executable waveform mapped on a hybrid platform.
In the case of an SCA (software communications architecture) there may be CORBA-compliant
and non-CORBA-compliant components in the design. ACORBA-compliant resource runs ORB as
a middleware and provides an abstraction for inter-component communication among CORBA-
compliant resources.
13.3 Typical Digital Communication System
A digital communication system is shown in Figure 13.16. A digital transmitter primarily consists
of a cascade of sub-systems. Each sub-system performs a different set of computations on an input
stream of data. Most of the blocks in the system are optional. Depending on the application, a
transmitter may not source encode or encrypt data. Similarly the system may not multiplex data
from different users in a single stream for onward transmission. The system at the receiver usually
has the same set of blocks but in reverse order. As the receiver needs to correct impairment
introduced by the environment and electronic components, its design is more complex. In mobile/
cellular applications a base station or an access point may be simultaneously communicating
with more than one end terminal, so it does a lot more processing than individual mobile
end terminals.
13.3.1 Source Encoding
The input stream consists of digital data or a real-time digitized voice, image or video. The raw data
is usually compressed to remove redundancy. The data, image and video processing algorithms are
regular in structure and can be implemented in hardware, whereas the voice compression algorithms
are irregular and are usually performed on a DSP.
For lossless data compression, LZWor LZ77 standard techniques are used for source encoding.
These algorithms for high data rates are ideallymapped inHW. The images are compressed using the
JPEG-2000 algorithm. This is also an ideal candidate for HWmapping for high-resolution and high-
speed applications. Similarly for video, a MPEG-4 type algorithm is used. Here, motion estimation
is computationally the most intensive part of the algorithm and can be accelerated using digital
techniques. Low bit rate voice compression, on the other hand, uses CELP-type algorithms. These
have to compute LPC coefficients followed by vector quantization. The quantization uses an
adaptive and fixed code book for encoding. These algorithms are computation- and code-intensive
and are better mapped in software on DSPs.
13.3.2 Data Compression
As a representative component for source encoding, lossless data compression algorithms and
representative architectures are discussed in this section.
LZ77 is a standard data compression algorithm that is widely used in many applications [27]. The
compression technique is used also to preserve bandwidth while transmitting data on wireless or
wired channels. An interesting application is in computer networks where, before data for each
session is ported to the network, it is compressed at the source and then decompressed at the
destination.