
2 1 Introduction
Companding Nonlinear ADC architectures allow the representation of a high
dynamic range signal with less information or bits than otherwise required by a lin-
ear converter. However, to perform this function some information about the original
signal is lost in the conversion process. Companding can be useful or even a require-
ment in several areas, such as image acquisition or hearing aid apparatus. In the
case of image acquisition, companding can be useful in helping to cope with large
variations in brightness in a scene, such as zones of shadow and bright light [7]. In
the case of hearing aid apparatus, companding is a requirement. The reason is that it
is absolutely necessary to condition t he sound from the environment to conform to
the channel dynamic range allowable by the biological systems [10]. Other examples
of nonlinear converters used in medical applications, and where the converters have
been tailored specifically for that application, include advanced closed-loop neuro
stimulation applications [11, 12].
Other types of nonlinear converters perform a piecewise linear conversion, which
can be used to compensate nonlinearities in sensors, such as presented in [13–15].
These types of converters are also employed in nuclear physics [16]. Other types
of converters using pulse width modulation (PWM) can also compensate sensor
nonlinearities [17]. In this case, it is desirable that the complete system uses as little
power as possible to enable long operation times if running on battery power, or make
use of energy harvested from the environment. Thus, in order to minimize the energy
requirements, it is desirable that the analog-to-digital conversion is performed with
the lowest number of bits needed. In case a linear ADC architecture is used to convert
the analog output of nonlinear sensors, the conversion result will require additional
processing cycles to linearize, thus making the complete system require more power.
It might also require the usage of an ADC with a higher resolution than otherwise
needed to acquire the full signal range with the desired precision. This is the case
when, for any zone of interest, the sensor exhibits a small variation in the output
when the input stimulus varies widely.
Another type of nonlinear converters performs an amplification of the input signal
before making the analog-to-digital conversion, as depicted in Fig. 1.1. This results
in a conversion characteristic that can be approximated by a straight line. Due to the
signal amplification before quantization, small amplitude input signals will seem to
have been quantized with a higher resolution than higher amplitude signals. There
are examples of application of these types of converters in nuclear physics [18],
where the collision results can have very low and very high energies, and, also in
communication systems [19, 20], where the received signal amplitude varies widely
due to attenuation with distance or obstacles.
Some examples of nonlinear analog-to-digital converters found in the literature
and described in the f ollowing sections have been implemented with commercial of
the shelf (COTS) parts, however, only a small number of converters have been imple-
mented in recent technologies and operating at speeds comparable with the current
state-of-the-art for linear converters. In summary nonlinear conversion techniques
are clearly an open research area for many of today’s applications.