Published on Gabe on EDA (http://www.gabeoneda.com)

Next generation analog design

By Tets Maniwa
Feb 19 2007 - 6:43pm

February 12, 2007—San Francisco—The International Solid State Circuits Conference started with a plenary talk from Lewis Counts, vice president of Analog Devices. His talk titled "Analog and mixed signal innovation: the process-circuit-system-application interaction", looked at some of the trends in analog design.

The ITRS '05 noted that increased functionality supersedes the capabilities available from scaling. Analog and mixed signal designs are areas for expansion, especially since the focus is not on making chips but on improving users' experience. At the interface levels, system performance is independent of technology, which should lead towards more system in package solutions.

Many application areas such as medical and automotive require more inspiration and creativity to meet the time and economic constraints. Consumers are looking for lower costs for a function. It would seem that an SoC would be the solution to deploy greater functionality per die, but scaling has limits especially when the analog, mixed-signal interfaces are involved.

The five greatest challenges for these interface functions are: dynamic range, bandwidth, power efficiency, complexity, and the absolutes of voltage, current, time, and temperature. The challenges apply and interact across multiple dimensions and affect more than just the transistor count.

Analog circuits have a long history of innovation, starting with the first differential amplifier which generated other intellectual property, including feedback within a circuit. Feedback amplifiers are essential for many types of circuits. One of the first applications for integrated circuits beyond simple amplifiers was the voltage reference. First there were differential voltage references which evolved into band gap references. The innovation in these references was to use a junction equation as a function of collector current. The result is a predictable circuit behavior or wide range of temperature and process variations.

The designs depended upon the planar bipolar process with its oxide providing both the passivation and masking layer. New applications drove the proliferation of process technologies and eventually led to complementary processes. Now, these types of designs require a super silicon—SiGe, SoI, metal gates—and these new process will be viable for a long time, possibly more than 20 years.

Consumer-type products require much more capability than standard CMOS can supply. Therefore, the silicon process will need multiple variations, and designers will need to match the process to the application. The availability of complementary P and N materials enables reduced power consumption and new circuits. For example, an ADC used 5 nJ/step at 10 bit resolution in I2L. Now, an eight megasample per second 10-bit converter use less than 1pJ/step. Overall, the standard functions in analog circuitry have seen a three orders of magnitude reduction in power over the last 30 years.

Analog circuitry always must make a trade between speed and bandwidth. This tradeoff is a result of fundamental limits in materials. The Johnson limit, a function of electron mobility and electrical field strength, predicts a maximum ft of 200 gigahertz per volt for silicon. Bipolar devices can exceed this limit while NMOS tops out at around 50 gigahertz. The designer cannot get more bandwidth out of a circuit in a 90 nanometer process because some of the difficulty in achieving high bandwidth at less than 2 V supply. The PMOS devices exacerbate the situation due to the lower hole mobility. One alternative to complementary processes is to use open drain or collector structures.

Data converters are moving closer to the front-end of the systems. Their enhanced performance is enabled by the processes, but is driven by innovation. Early converter architectures used resister ladders or dual slope integration. Now, over sampling (∑-∆) converters are operating at radio base-band frequencies. In successive approximation converters, sensitive and error-prone resistor ladders have been replaced by ratios of capacitors.

The next challenge for data converters is to become the basis for a universal radio appliance. For example, a cellphone may have more than six radios within it, and most of these radios work on multiple bands. Designers must reduce power and cost to enable a reconfigurable software defined radio. Interference is a major design constraint especially in conjunction with requirements to reduce size. Three-dimensional packaging will enable replacement of heterodyne receivers with direct conversion. To reduce the second order effects, which showed as a DC or baseband errors, these converter will need error correction and cancellation.

The frequency-adaptive radios will also need adaptive antennas, driven by high-voltage varactors, coupled to reconfigurable and adaptive filters. One alternative to CMOS for these circuit elements is to look at MEMS (micro-electro-mechanical structures). Although there are many efforts to increase active devices as replacements for passive devices, inductors are not likely to go away in the near future. Since phase noise is a function of 1/Q2, an LC circuit will outperform a ring oscillator by a factor greater than 50.

As more electronics becomes consumer oriented, the requirements for applications requires continual improvements in bandwidth and ft. These parameters are related to process advances, but innovation in other areas is also required. As systems incorporate other sensors over optical it becomes more difficult to find electrical variables to map to the sensors. The need for greater power efficiency will drive to the use of more super silicon and complementary processes. These technologies in turn will drive innovation in circuit protector and system designs. Design complexity will increase as a function of the number of available transistors and the system environment those transistors will be used in.

There are four dimensions in technology. They are process, circuit, system, and application. Because all dimensions interact, designs will, by nature, become more multidisciplinary. Designers will use pairs—the up-down symmetry of N and P, the left- right of differential circuits, and a mix of analog and digital to optimize both domains. In addition, a new dimension may be added for consumer products—MEMS.

To comment on this article send email to:gmoretti@gabeoneda.com [1]


Source URL:
http://www.gabeoneda.com/node/68