TISP Real-Time Signal Chain: Cascade Biquad Parametric EQ Example

Table of Contents

• Introduction
  • Example Workflow
  • Signal Chain Details
    • Processing Stages
    • Key Features
    • Filter Characteristics
• Supported Combinations
• Steps to Run the Example
  • Build the Example
  • Run the Example
  • Configuration Parameters
• Sample Output

Warning

EXPERIMENTAL FEATURE DISCLAIMER: TISP (Texas Instruments Signal Processing) library and associated examples are currently in experimental versions. These are provided for evaluation and development purposes only. Texas Instruments does not offer official support for TISP at this time. Use at your own discretion.

Introduction

This example demonstrates a simple real-time parametric equalization signal processing chain using TISP (Texas Instruments Signal Processing) library running on the DSP Core (C75). The example showcases a 3-stage cascade biquad filter implementation for parametric EQ processing.

The signal chain applies parametric equalization to multi-channel audio data in real-time, using a cascade biquad filter, optimized for DSP execution.

Example Workflow

1. Signal Chain Initialization:
   • The create_graph() function initializes all processing nodes in the signal chain.
   • Memory buffers are allocated for intermediate data storage.
   • Filter coefficients for the 3-stage parametric EQ are configured.
2. Real-Time Processing:
   • Input audio data (multi-channel, int32_t format) is received.
   • The execute_graph() function processes data through the entire signal chain.
   • Each processing node operates on the data sequentially.
   • Output audio data (multi-channel, int32_t format) is generated.
3. Processing Pipeline:
   • Type Conversion: Converts input from int32_t to float for DSP processing.
   • Cascade Biquad Parametric EQ: Applies 3-stage parametric equalization filter.
   • Type Conversion: Converts output from float back to int32_t.

Signal Chain Details

The parametric EQ signal chain consists of 3 processing nodes executed sequentially:

Signal chain

Processing Stages

1. TypeConversion (int32 → float): Converts multi-channel input data from int32_t to float format for processing.
2. CascadeBiquad (Parametric EQ): 3-stage cascade biquad filter implementing parametric equalization:
   • Uses Direct Form I (DF1) biquad implementation
   • Each stage can implement different EQ characteristics (peak, notch, shelving, etc.)
   • Coefficients define the frequency response, Q factor, and gain for each stage
   • Processes all channels independently
   • Frequency response is as shown below

Frequency response of cascade biquad filter

1. TypeConversion (float → int32): Converts processed multi-channel output from float back to int32_t format.

Key Features

• Real-time processing: Optimized for low-latency audio processing on C75 DSP core
• Parametric EQ: A 3-stage parametric equalization
• Multi-channel support: Processes multiple audio channels simultaneously
• Floating-point precision: Uses floating-point processing for high-quality audio

Filter Characteristics

The example includes pre-configured 3-stage parametric EQ with:

• Stage 1: Low-shelf filter
• Stage 2: 1kHz notch filter
• Stage 3: High-shelf filter

Supported Combinations

Parameter	Value
CPU + OS	c75ss0-0 freertos
	mcu-r5fss0-0 freertos
Toolchain	ti-arm-clang and ti-C7000-CGT
Board	am62dx-evm
Example folder	examples/tisp/sigchain_dsp_rt/sigchain_biquad_cascade

Steps to Run the Example

Build the Example

• Using CCS Projects: Import the system CCS project and build it using the CCS project menu (see Using SDK with CCS Projects). This will build all dependent CPU projects as well.
• Using Makefiles: Build the system makefile using the make command (see Using SDK with Makefiles). This will build all dependent CPU makefiles as well.

Run the Example

1. Load and run the application:
   • Load and run the compiled binary on the DSP Core (C75).
   • Load and run the compiled binary on the R5F0 Core.
   • The application will initialize the signal chain and begin processing audio frames.
2. Monitor the output:
   • The example processes audio data in real-time through the 3-stage parametric EQ pipeline.
3. Audio Flow:
   • Input: Eight channels of audio data (int32_t format)
   • Processing: Real-time execution through 3-stage parametric EQ
   • Output: Eight channels of equalized audio data (int32_t format)

Configuration Parameters

The example includes configurable parameters:

• Filter Coefficients: 3-stage cascade biquad coefficients for parametric EQ
  • Each stage has 5 coefficients (b0, b1, b2, a1, a2)
  • Total of 15 coefficients for the 3-stage filter
• Number of Channels: Configurable based on application requirements
• Block Size: Number of samples per frame to process

Sample Output

Input is a chirp waveform from Audacity with the following parameters:

• 200Hz to 10kHz with linear interpolation of frequency
• Amplitude of 0.6

The output spectrum of the audio stream is as shown below, which was captured using Audacity.

Output spectrum