Fusion of MCU and DSP capabilities for block and stream processing

Infineon / Mitsubishi / Fuji / Semikron / Eupec / IXYS

Fusion of MCU and DSP capabilities for block and stream processing

Posted Date: 2024-01-17

Microcontrollers (MCUs) excel when it comes to controlling our user interfaces. They can initialize other circuits, monitor inputs, respond to interrupts, and drive outputs for displays and controls. However,While integrated analog functions (such as A/D converters, D/A converters, PWMs, voltage references, op amps, and comparators) enable MCUs to function in the analog realm - this is true for many relatively simple analog processing tasks. Common - but in general, microcontrollers generally don't do well at handling analog signals.

Take a thermostat, for example, which almost any mixed-signal microcontroller can handle. It monitors the temperature, compares it to a set point, and then flips the switch. You can add programming lag, timed operations, cloud connectivity, and global connectivity, but ultimately the MCU performs its task by throwing or releasing a switch.

However,In the real world, most control loops operate much faster than room temperature can change and monitor more than just one input. Here, fundamental attributes such as speed, performance and flexible architecture are very important, especially when faced with block or streaming processing needs, but equally important is the type of DSP functionality,Whether integrated into the core of the microcontroller, or processing analog functions in parallel with the microcontroller.


Certain aspects of mixed-signal design are better achieved by cleverly dividing functionality rather than placing all the heavy lifting on a single control block. The two approaches are capturing the raw data and doing all the processing in the digital world, and doing some basic processing before the signal reaches the A/D. If you have enough processing resources, the easiest way is to extract everything in its raw form and go from there.

Most linear signals require some external components to match the signal range and the range of the A/D converter. This maximizes the dynamic range of the data, providing the highest resolution.microWeak signals need to be amplified, while strong signals need to be attenuated. If we can do some simple processing of the signal in the analog world before going into the A/D converter, we can reduce the processing power required depending on the algorithm the processor is running.

Hardware basic signal processing

A simple example is,An attenuator can be turned into a low-pass filter simply by adding a capacitor.This shows that the balanced method of hardware-based signal processing is very low-cost, takes up very little space, and can simplify subsequent design.

While filtering may still be necessary in the digital realm, by cleverly partitioning functionality, you may be able to reduce the amount of processing required, thereby lowering costs, saving energy, and reducing code development.

DSP level

In digital form, high-pass, low-pass, band-pass and notch filters can be used as input waveforms to the filtering process, continuously implemented in precise digital form.The same goes for gain, inversion, attenuation, averaging, peak detection, low value detection, integral, derivative, etc.

Once data enters the processor, the specifics of the processor type, architecture, speed, and special features determine whether it is suitable for the task at hand. While the old von Neumann CISC architectures were suitable for simple control tasks, they were not ideal when real-time aspects were introduced.One limitation is that in CISC machines, different types of instructions can use different numbers of cycles, multiple clock cycles per instruction. Now the effectiveness of your code depends on which directive you use and where you use it. Interrupt response also affects real-time performance.

The Harvard architecture using RISC is more suitable for stable pipeline processing of data flows. These are typically executed in a single clock cycle or a single instruction cycle. RISC machines typically run at higher speeds, adding another performance-boosting feature.For system designers, a benefit is that the data area is separated from the code space. This provides more flexibility in dividing and using blocks of memory for captured data processing and temporary RAM.

The key to deciding which processor to use for mixed-signal tasks is the instruction set, specifically fast multiplication, multiply-accumulate, and fast division. Fixed or floating point requirements are also important here. These are particularly important in filter and analog processing blocks implementing algorithms, especially if they require real-time solving of algebraic and quadratic equations.

Many processor cores add one or two multiplication instructions and claim to have DSP functions, but in fact they also have other processors dedicated to signal processing.But not every DSP-capable processor has a full 32-bit architecture and runs at hundreds of MHz. Many applications only require 16-bit functionality and 32-bit extensions.

Whether the selected microcontroller is suitable for real-time signal-intensive designs depends on the processor architecture, performance level, peripheral mix, and computing resources.Block and stream processors that operate on changing waveforms. Combining MCUs and DSPs in block or stream simulation is possible when the appropriate components are chosen and the most efficient design partitioning is done at the top level of the design phase. Signal advantages are brought into play.

#Fusion #MCU #DSP #capabilities #block #stream #processing