MCUs and DSPs: Confluence is the Key

Digital Signal Processors and microcontrollers are the building blocks of embedded systems. They have been functional in the embedded systems for their own benefits and utilities, though operative in totally isolated environments, for long. But now the trends are changing. We have many products already in the market that are hybrids of DSPs and MCUs (microcontrollers). Siemens Semiconductors (now Infineon Technologies), for instance, announced its entry into the field of hybrid DSP/microcontrollers in late 1997 with TriCore chip. Such hybrids are high-performance microcontrollers and use high-end CPUs for execution. In addition, they incorporate considerable DSP-oriented hardware and instructions. Such hybrid embedded systems find their use in a variety of applications such as

high-performance disk drives, real-time systems, communication systems and automotive engine management systems.

It's an MCU, it's a DSP, it's... 



To begin with, microcontrollers are primarily used in interrupt-driven control-oriented applications to sense and control external events. But they have moved in to the domain of DSPs that traditionally performed calculations on digitized signals that were originally analog (eg voice and sound). Same has happened with the DSPs as they have grown within themselves, the capabilities of MCUs. With the convergence of the two parallel technologies, MCUs can now process signals and DSPs are capable of undertaking real-time executions, while both perform their primary functions. 

To be able to do all this, the hybrid chips couple an MCU-like load/store design with a DSP-like memory architecture. The AISC technology is used to customize memory configuration in such systems, as they do not have a memory configuration of their own.

The digital revolution 



Over the past few years, we've seen several trends emerge in the way the embedded systems are tailored. First, DSPs have become a hot market because digital media and algorithmic calculation are almost synonymous. Second, the MCUs have been successfully running the tasks assigned to them, that is, more performance and more peripherals at lower power.

Another aspect of MCU performance is signal-processing capability, which is being designed into MCUs by creating hybrid architectures. Microchip Technology's digital signal controllers (DSCs) are one example. Texas Instruments and Freescale Semiconductor have taken a similar route by adapting their DSP architectures to better cope with control-oriented applications.

An alternative to adding signal processing capabilities to the MCUs, is having a stack architecture and bit manipulation capability for control-oriented operations and multiply-and-accumulate (MAC) and looping capability for signal-processing operations. In this case, the functionalities of the microcontrollers are embedded into the

DSPs.

Applications



The DSC revolution is just beginning. And the coming years will witness an uptake in DSC applications beyond the most obvious-motor control-into virtually anything that moves and is controlled electronically. The trend toward more efficient, smaller, and quieter motors for white goods and other cost-sensitive, feature-rich applications is increasingly putting the design spotlight on a hybrid processor architecture that combines DSP computation power with the control efficiencies of

MCUs.

Consumer applications are a strong driver of semiconductor trends because this is increasingly where communications, computing, and entertainment converge. Digital media, in particular, is creating new and fast growing markets for embedded systems. Consumer applications offer great potential for both the MCU and DSP semiconductor product segments. 

In fact, while the Semiconductor Industry Association has forecast about a 1% growth rate for the entire industry, the DSP market is expected to grow at about 20%. The vast microcontroller market, on the other hand, is expected to grow at a more modest rate, in part because of its enormous size.

So by the end of this year, DSCs will have a much larger market share in cost-sensitive consumer applications such as appliances, washing machines and other white goods. The applications for such hybrids will become visible in verticals such as HVAC blowers and fans; factory automation; industrial air conditioning and refrigeration; and elevators and cranes. Vibration control, power management in products such as uninterruptible power supplies, and even large LED lighting displays are prime DSC markets beyond motor control and consumer goods.

Tool development-and integration of DSC-specific tool features into the existing embedded development environment-has been another primary focus for companies that have fielded DSCs over the past few years. The general idea is to make signal-processing development look as much like an MCU as possible and this typically means automatic code generators. 

This is because it is difficult to generate a conversation between the DSP and the MCU programmers. Simply because, the DSP is programmed directly in assembly language unlike the microcontroller programming that's done in C.

On the other hand, DSCs are not particularly capable with digital media in large part due to their relatively low clock rates and the application specific nature of their signal processing architectures. For example, the basic requirements of streaming media differ quite a bit from those of motor control. Meanwhile, DSP vendors who do have their eye on digital media will increasingly read from the MCU playbook by integrating more peripherals adopting a third-party tools strategy similar to the one MCU vendors have used for a decade or more.

DSPs: now and tomorrow



The strength of DSP architectures will continue to be hardware accelerators implemented as relatively small circuit blocks that provide a performance boost by executing specific algorithms. This adds a lot of value to the embedded systems by conserving power.

Traditionally, increase in clock speed is directly proportional to the performance. And this implies that a good part of the entire chip has to run faster when an algorithm is executed. So one can save power by adding a good number of optimized gates for executing that specific algorithm.

Another example of the convergence of the MCU and DSP worlds is the introduction of a design tool and compiler for signal processing accelerators that can be designed right into ARM-core-based ASICs —OptimoDE by ARM.

Wireless is, of course, the primary market driver for DSPs and the ramp up of 3G phones will keep it that way for the times to come. From an applications perspective, video is destined to be even more important for DSPs than it has been in past years. New products such as video IP set-top boxes, videophones (including 3G phones), and networked video security are other applications that will see a lot of action.

Talking of the security applications, not only new surveillance products that are going digital, but digitized videos also have powerful advantages over analog video. This is because, with digitized images, it is easier for computers to recognize what's going on and they can respond accordingly in an intelligent manner. And none of this is possible without the DSPs being strengthened.

MCUs aren't dead



As far as the microcontroller technology is concerned, nothing extraordinary is happening on this front. But beyond this, the microcontrollers offer the much needed performance on the control-oriented functions.

So move-them-along is the theme currently. While 32-bit and 64-bit applications are in, the market still hasn't forsaken 8-bit and 16-bit chips. Moreover, an ARM-based MCU plus a small ASIC can give an SoC a run for its money in some applications.

In a nutshell



Thanks to higher performance, higher levels of integration and a changing, challenging mix of applications, both MCUs and DSPs are competing with each other in broad application areas.

So, expect to see this competition intensify. Don't be surprised to see more 32-bit MCUs with RISC cores and built-in signal processing capability. This is not the end of it. The future of these DSP/microcontroller hybrids has much to be seen. For one, you may expect the floating-point

version of these hybrid cores soon.¨

ASIC and FPGA

Application-Specific Integrated Circuit or ASIC is another key component of embedded systems. True to its name, ASIC is an integrated circuit, which is a specialty chip that is used for more specific than general-purpose uses. While the term refers to both analog and digital devices, it is conventionally used for the digital circuits. 

An ASIC can now squeeze in more than a hundred million gates and often includes the 32-bit processors and other larger units. In that case, the ASIC is referred to as SoC (System-on-Chip) since it is more or less a complete system on one chip. The programming for ASICs is done in HDLs (Hardware Description Languages) such as Verilog or

VHDL.

While ASICs are suitable for large volumes of production and big-size designs, they are not otherwise. Therefore, the entry of FPGAs (Field-Programmable Gate Arrays), which are more capable. FPGAs can be hardware programmed to do specific tasks. They are slower than ASICs and consume more power. They are also programmed using HDL. While the FPGAs do not outperform ASICs either in performance or power drawn, when it comes to speed they run neck to neck with their ASIC counterparts.