Bridging the Gap Between the Analog and Digital Worlds
Most applications require the co-existence of analog and digital functionality, and the benefits of combining this functionality on a single chip are significant. Such mixed-signal integration, however, also presents significant challenges. Furthermore, digital and analog developments tend to evolve at differing rates, yet mixed-signal solutions for markets such as industrial, automotive and medical, must remain available over significant time periods. The latest mixed-signal semiconductor processes are helping to address some of these issues, and this article will look at some of the issues designers should consider when specifying integrated mixed-signal solutions.
Mixed-signal Solution for the Real World
System designers often partition the digital portion from the analog section of a given design for a variety of reasons: the availability of mixing components for the two technologies, the complexity of the digital design or again because of the existence of pure digital processing parts as standard products. Placing the analog elements in an integrated circuit definitively allows the system designer to optimize the costs of its entire module.
This integration approach is usually difficult for advanced markets such as telecommunications or computers, but makes sense for more mature or conservative markets such as automotive, medical and industrial. For most of these mature market's applications, digital functions are finding their way onto what once were pure analog designs. Adding digital functions to an analog design is helped in part by the development of new process technologies that can handle both short-channel, fast-switching digital transistors as well as high-voltage analog transistors. For example, AMI Semiconductor's latest mixed-signal technology offers digital and analog integration capabilities on the same design platform. The I3T technology family is based on standard CMOS 0.35µm, limiting the maximum gate voltage to 3.3V. Some consider this technology outdated, from a pure digital designer's point of view, but it is at the forefront for the automotive, industrial and medical markets.
This list of optional features that enables the design of real SoCs includes high voltage interfacing up to 80V, microprocessing capabilities up to 32 bits, wireless capabilities up to 2.8GHz, and dense logic design up to 15K gates/mm2. Beside these capabilities, NVM integration is possible: E2PROM up to 4 Kbytes, Flash memory up to half a megabit or On-Time-Programmable (OTP) cells for application calibrations. The ability to integrate all these features on a chip gives the customer the possibility to be independent from the obsolescence of the stand-alone NVM market, which is more or less driven by the computer market. This advantage is quite relevant when we consider the cost of re-qualifying a module for the OEMs in automotive, for instance. It also makes sense when considering the long lifespan of the applications embedded into cars, the industrial environment or medical self-treatment devices where patient cost is an important consideration.
Nevertheless bridging the gap from digital to analog on a single chip does not occur without issues. Clocking noise from high-speed digital circuits, for instance, often interferes with noise-sensitive analog functions. In addition, switching currents from high-power analog functions can interfere with low-voltage digital processors. The goal is to protect low-voltage transistors from the electric field effects of voltages that are 10 to 30 times higher.
These important issues are not without solutions. For example, one of the latest releases in the I3T family, the I3T50 DTI, uses a deep trench isolation technique. This technique uses a series of isolating trenches that bury deep into the IC substrate; effectively creating on-chip "pockets" where noise and power supply parameters are carefully controlled.
On top of its protection skills, the deep trench technology also helps to minimize die area by allowing dense packing of high-voltage analog pockets with low-voltage regions. You can obtain improvements in die area of 10 to 60 percent over designs that use standard junction isolation techniques.
As mentioned earlier, the reason that system designers are using deep sub-micron technologies in those markets is often linked to the availability of devices in those technologies, not the complexity of the application itself. The complexity can be handled in many cases by an 8-bit microcontroller, or 32-bit for high-end applications. Products such as the 0.35µm I3T are able to manage the integration at a reasonable cost. A typical application diagram of a real mixed-signal SoC is shown in Figure 1.
Figure 1: Mixed-signal SoC Diagram
Basically, the chip integrates the system functionality from the sensor to the actuator, going through some digital processing. Conventional mixed-signal technology allows analog control and signal processing functions such as amplifiers, analog-to-digital converters (ADCs) and filters to be combined with digital functionality such as microcontrollers, memory, timers and logic control functions on a single, customized chip. All signals that process an algorithm or arithmetic calculation are digital, so conversion of analog to digital signals is mandatory when submitting data for comparison or processing by via a microcontroller, while conversion from digital output signals to analog high-voltage signals is required to drive an actuator or a load. The most recent mixed-signal technology AMIS developed, significantly simplifies the implementation of such driver functionality by allowing much higher voltage functionality to be integrated into an IC alongside the relatively low voltages required for conventional mixed-signal functions. This high-voltage mixed-signal technology is particularly relevant to automotive electronics applications where higher voltage outputs—to drive a motor or actuate a relay—need to be combined with analog signal conditioning functions and complex digital processing.
A growing trend in mixed-signal circuit design is to add some type of central processing circuit to the analog circuits. For many applications the suitable choice of processing intelligence is an 8-bit microcontroller core such as an 8051 or 6502. 8 bits remains the most popular choice as this type of SoC is not intended to replace complex high-end central microcontrollers but more decentralized or slave applications such as sensor conditioning circuitry with local (as close to the sensor as possible) simple intelligence to control relays or motors. An automotive example would be the lateral actuation of a car's headlamps when the steering wheel is turned to improve the driver's safety and improve field of vision. The sensor input would come from the steering angle via a ****** link (most of the time with a LIN or I2C protocol) and the SoC would be close to the motor with an on-board set of algorithms to command the motor's movement.
For higher end applications that require more calculation power, the move to ARM processors is possible. This creates a high-end solution (up to date for the mature markets) which could last over the application's lifespan because the microcontroller would be a small part of an integrated circuit that emulates the module's functionalities.
In order to understand how larger geometries can be better suited for some mixed-signal applications, one needs to understand all of the characteristics involved. Below we will discuss seven key characteristics, however this is by no means comprehensive.
1.Gate and memory size in mixed-signal applications generally drive cost.
Gate and memory size drive cost because most mixed-signal devices are core limited. This can be quite different than an all digital circuit. Many times, the all digital device will have so many I/Os that the number of pads on the device determines the periphery and therefore the area. This is rarely the case for mixed-signal devices. For the most part digital cells scale pretty closely to the expected area savings. One would expect a 0.25-micron cell to be 51 percent smaller than a 0.35-micron cell of equivalent function. This is illustrated by the following formula:
While this holds for digital cells we will see that analog cells are quite a different story. Therefore the amount of digital content (including memory) is very key in determining the best technology for the application.