Improving Codec Execution With ARM Cortex-M Processors
by from IEEE Spectrum on (#4X9PQ)
Digital Signal Processing (DSP) has traditionally required the use of an expensive dedicated DSP processor. While solutions have been implemented in microcontrollers using fixed-point math libraries for decades, this does require software libraries that can use more processing cycles than a processor capable of executing DSP instructions.
In this paper, we will explore how we can speed up DSP codecs using the DSP extensions built-in to the Arm Cortex-M processors.
You will learn:
- The technology trends moving data processing to the edge of the network to enable more compute performance
- What are the DSP extensions on the Arm Cortex-M processors and the benefits they bring, including cost savings and decreased system-level complexity
- How to convert analog circuits to software using modeling software such as MathWorks MATLAB or Advanced Solutions Nederlands (ASN) filter designer
- How to utilize the floating-point unit (FPU) with Cortex-M to improve performance
- How to use the open-source CMSIS-DSP software library to create IIR and FIR filters in addition to calculating a Fast Fourier Transform (FFT)
- How to implement the IIR filter that utilizes CMSIS-DSP using the Advanced Solutions Nederlands (ASN) designer