The trend of MCUs becoming increasingly integrated into everyday applications is undeniable. In particular, the integration of DSP digital signal processors or FPU floating-point arithmetic units into MCUs for the purpose of functional optimization or market segmentation has greatly expanded the application scope of MCUs.
Microcontrollers (MCUs) have become deeply integrated into people's daily lives, and their presence can be seen in almost every device, large or small. With the introduction of DSP digital signal processors and FPU floating-point arithmetic units, MCUs have greatly expanded the range of components they can be used with. In recent years, many major MCU manufacturers have launched a variety of integrated solutions for their products, and both product strategies and market segmentation have made the MCU market more diverse and richer.
The trend of MCUs becoming increasingly integrated into everyday applications is undeniable. In particular, the integration of DSP digital signal processors or FPU floating-point arithmetic units into MCUs for the purpose of functional optimization or market segmentation has greatly expanded the application scope of MCUs.
Integrating an MCU with an FPU can significantly improve the precision of advanced numerical calculations and enhance processing performance.
For MCU solutions developed for IoT applications, integrating DSP can optimize sensor data acquisition quality and improve signal processing performance.
More
If the purpose is to incorporate FPU or DSP, the FPU or DSP integrated architecture is usually added to the MCU. The main purpose is to consider the design direction under the consideration of cost. Especially in the early semiconductor components, there was a price difference between SOC (System-on-Chip) and MCU. If only SDP or FPU is needed for computing acceleration, and you do not want to choose a high-priced SOC, then MCU products that integrate DSP or FPU hardware acceleration units can not only provide better operating performance, but also perform better in cost control.
The cost of MCU integrated chip packaging has dropped sharply, increasing MCU functionality and expanding application scope.
Looking at early SOC products, the inclusion of DSP and FPU hardware accelerators was an important feature of SOC products. The application of DSP and FPU was mainly for accelerating the processing of audio and video. As process technology continued to be optimized, the cost of SOC gradually narrowed the gap with that of MCU. MCUs, in 32-bit and even 64-bit architectures, also began to have solutions that combined DSP or FPU hardware acceleration units.
Let's first look at the advantages of adding a hardware acceleration unit to an MCU. The most direct benefit of adding an FPU to an MCU is that in the early days, when using an MCU to process similar FPU operations, the time to obtain the results would be relatively long due to the limitations of the MCU's own computing architecture. However, when a hardware accelerator is introduced to process floating-point operations, the data can be calculated through hardware calls or data transfers. The memory resources consumed by the MCU for floating-point operations can be reduced by at least 10% due to the hardware acceleration integration.
Of course, from a purpose perspective, regardless of whether the MCU integrates an FPU hardware acceleration unit, floating-point operations can still be performed using the MCU's existing computing power. However, this is only because the calculation process will consume more computing time and hardware resources. For systems that can wait and do not require immediate response, an MCU solution integrating an FPU can be disregarded. But for integration requirements with high system performance and fast response speed, the benefits of combining an MCU with an FPU are not just about optimizing computing resource consumption and energy saving. Rather, the most direct and important functional improvement of combining an MCU with an FPU hardware acceleration unit is the faster system response and improved performance. This also allows the MCU to handle more complex integration tasks.
High-order numerical computation utilizes hardware acceleration to meet design requirements.
In the early days, MCUs were mainly based on 8-bit architecture. Due to the architecture, MCUs had limitations in data processing and computation. For example, when MCUs performed decimal and fraction operations, because of the limited number of bits (4 or 8), they had to use a limited number of values to process the data. This limitation of numerical results was used to simplify the processing complexity and meet performance requirements. The error caused by this numerical processing is called "truncation error". Truncation error can also be amplified by the limitations of using MCUs for data processing.
When an MCU integrates an FPU hardware acceleration unit, it can handle similar data processing tasks, such as in IoT or terminal sensor applications. There is often a need to acquire and process data by converting external analog sensing data into digital data. In this case, the FPU/DSP hardware acceleration unit integrated into the MCU can not only process the sensing data faster and speed up the system response, but also introduce advanced computing to reduce data calculation errors.
In practical applications, the FPU hardware accelerator itself cannot completely solve the error amplification problem. Therefore, there are different application considerations under different hardware acceleration integration architectures such as FPU and DSP. For example, through the DSP hardware accelerator, faster and more reliable operation and output can be processed for special data types. For example, the DSP can use instructions to perform a variety of operations, such as handling important and resource-intensive operations in advanced operations such as Fast Fourier Transform or Finite Impulse Response. It can even handle multiple data operation requirements of a single instruction in a single cycle. The MCU can also obtain advanced enhancement benefits in advanced numerical processing.
The FPU/DSP hardware acceleration units complement each other.
Although integrating an FPU or DSP differs in architecture and application, they are actually designed for data processing and signal processing, respectively, corresponding to various algorithm applications. Their functions are complementary and difficult to separate independently. Taking the ARM Cortex-M4 as an example, providing only a DSP hardware acceleration processor without an FPU floating-point accelerator would actually limit its applications. In Cortex-M4 applications, if there is only digital signal processing acceleration hardware support and no floating-point support, it will limit the design flexibility when developers need to accelerate advanced numerical calculations. It may also require external functional chips or utilize existing computing resources to meet the needs of advanced numerical calculations. In this case, the numerical processing performance limits the application possibilities of the Cortex-M4.
The same situation occurs in microcontroller application solutions that only have an FPU and no DSP. The application functions of DSP or FPU are complementary. Independent integration does not produce synergy for the configuration of microcontrollers, but instead becomes a limitation on the development path.
Furthermore, taking the development direction of next-generation IoT products, through the application of sensor fusion as an example, if the concept of Sensor Fusion is to integrate multiple sensors into a single system to work together, the system needs high-level numerical and signal processing capabilities to extract key numerical signals from complex data.
As for sensor fusion, it can be combined with real-time adjustment, control and calibration processing. The DSP and FPU work together to achieve high precision and high efficiency in the precise analysis of the acquired data. In particular, the existing SensorFusion has integrated gyroscopes, accelerometers, temperature, pressure and even touch sensors into the same module. The DSP and FPU must be used to pre-screen relatively precise signal acquisition and pre-processing of the sensor data, which takes into account the processing efficiency, to provide a more efficient system and a more efficient sensor data processing mechanism.
DSP digital filtering applications can improve the quality of sensor signal acquisition.
Furthermore, another advantage of integrating an FPU into an MCU is the ability to leverage its computational capabilities within the system. For example, digital filtering applications can utilize digital algorithms to extract numerical values. Further processing of the processed signals can be achieved using hardware-accelerated digital algorithms to reprocess waveforms or data, resulting in a convenient way to improve the signal-to-noise ratio (SNR). Digital filters can also utilize computational mechanisms to optimize and provide filtering effects of varying degrees. This is useful in applications where microcontrollers are used to sense popular physiological information such as heart rate, blood oxygen levels, and exercise data, or in digital meters and smart meters. It addresses the issue of signal distortion caused by noise or environmental interference at the terminal, optimizing and compensating for the distortion, and improving the quality of the signal waveform obtained by the terminal, making it more suitable for subsequent processing or data use.
To optimize end-user applications, integrating hardware acceleration units into microcontrollers has become a trend. This includes not only DSPs or FPUs, but also, for example, some microcontrollers have incorporated VMUs into their architecture to handle trigonometric function numerical calculations that are crucial for motor applications, or data analysis and computation support that supports radio communication needs. This is distinct from the existing FPU floating-point hardware acceleration function and adopts a collaborative approach to accelerate the overall application performance of the microcontroller.
Interestingly, in order to target different markets and computing needs, microcontrollers not only differentiate themselves by computing clock speeds to distinguish different application scenarios and market segments based on the most practical computing performance, but also integrate the hardware acceleration units required for different applications, which has become an important dividing line for product market positioning. For example, for microcontrollers targeting the wearable computing application market, they can only integrate FPU and DSP hardware acceleration to position themselves in the market based on requirements for power consumption, sensor fusion, and component costs. In high-end microcontroller applications, there are even solutions that directly integrate hardware graphics engines to directly target the application needs of industrial human-machine interface terminals. In addition, for different market needs such as automotive electronics and IoT, there are also a variety of hardware acceleration unit configurations and combinations to meet application architectures with different integration requirements.
Another purpose of integrating DSP and FPU hardware acceleration units into a microcontroller is to add hardware acceleration units without adopting external solutions to meet hardware acceleration computing needs. Its biggest advantage lies in extreme cost optimization, because the electronic circuit board can save more board space, and the overall computing efficiency can be improved by using a single chip. At the software development level, the integrated architecture can use simple call and data transfer reprocessing to meet the data computing output performance requirements of application services. Even the finished product can use consistency debugging analysis tools to directly perform comprehensive analysis and debugging of the system, which can improve the efficiency and speed of development and design.
