Circuit-level considerations for mixed signal programmable components

Abstract

IEEE Design & Test of ComputersIN THE PAST DECADE, researchers have paid significant attention to the design of heterogeneous embedded systems. However, most of their work focused on digital hardware and associated software, 1,2 which remain firmly rooted in an all-digital environment. In contrast, an embedded system must generally interface with the real world. At this level, most signals are continuous time and can assume only analog values. Most sensors (such as those for temperature, pressure, humidity, resistance, and so on) generate small voltages or currents, and systems must preprocess (amplify and filter) these signals before using them.Currently, many applications use digital signal processing because of its fast design cycle, available tools, and integration with current design methodologies. Designers can use this technique in any problem that does not require a gigahertz frequency bandwidth or extremely low power dissipation. Nevertheless, although digital processors can implement many analog functions, some analog functions must be developed in the analog domain. For example, amplifiers, low-pass antialias filters, and an A/D converter can be developed only in the analog domain; the digital domain has no equivalent. So although most of a mixed-signal system might be digital, such systems still require analog functions.For linear behavior, a system should in principle use linear components. However, linear capacitors or resistors in MOS technology take up too much area. Moreover, programmability generally comes in the form of MOS switches. These transistors introduce some extra pole-zero pairs, and worse, they have nonlinear voltage-current characteristics. We will show that, even when using analog technologies that allow components like linear capacitors or resistors, the resulting circuit will display nonlinear behavior because of the programmability requirement. Moreover, this approach will still consume significant circuit area.For these reasons, designers need a new paradigm for designing analog circuits. Here, we propose the use of nonlinear analog components with digital compensation. Such components can be area effective, allowing designers to easily build important circuits. It is also possible to directly develop this class of circuits in a digital technology. Moreover, the proposed circuit technique is easily programmable using switches and incurs only a slight performance degradation. Some of our results, such as those for amplifiers and integrators, use only a small area and demonstrate a perfect balance between the possibility of programming an analog or a digital device in the same chip. This methodology allows the development of linear or sampled circuits like switched capacitors.Although some work has addressed analog fieldprogrammable gate arrays, 3-5 most of it focuses on the prototyping of analog functions like filters and amplifiers. However, the embedded-systems market seems to require mixed-signal programmable systems, not Circuit-Level Considerations for Mixed-Signal Programm...

A FPGA version of a non-linear adaptive filter

“…h [1] h [2] h [3] x [3] x [2] x [1] x [4] h[0] h [1] h [2] h [3] x [3] x [2] x [1] x [4] h[0] h [1] h [2] h [3] x [3] x [2] x [1] x [4] h[0] h [1] h [2] h [3] x [3] x [2] x [1] x [4] h[0] h [1] h [2] h [3] x [3] x [2] x [5] x [4] h[0] h [1] h [2] h [3] x [3] x [2] x [5] x [4] time time As shown in table 2, the number of logic cells for the adder and the multiplier is 160. With a total of 518 logic cells for the filter with sequencial multiplier, the main conclusion is that the 16 bit multiplexers needed in the data path are the great responsible for the area occupation.…”

“…This way, this study checked possible multiplier implementations that could reduce filter area, while still maintaining throughput. Since the non-linear filter will be used in a larger system [1], it is important to verify area requirements, so that the filter and a microprocessor could fit in the same chip. This paper is organized as follows.…”