IFFT - 2022.2 English

Vitis Model Composer User Guide (UG1483)

Document ID
UG1483
Release Date
2023-01-13
Version
2022.2 English


Library

AI Engine/DSP/Window IO

Description

This block implements the Inverse FFT targeted for AI Engines which use the rounding method and saturates the output samples on overflow.

Parameters

Main
Input/Output data type
Describes the type of individual data samples input to and output from the filter function. Supported types are cint16, cint32 and cfloat.
IFFT size
This is an unsigned integer which describes the point size of the transformation. This must be 2^N, where N is in the range 4 to 11 inclusive. However, for cint16 datatype, the IFFT size can be 2^12, provided the IFFT receives and outputs data to/from kernels on the same processor.
Note: To understand more on achieving the 4096 point size FFT, refer to the AI Engine examples in GitHub.
Input Window Size (Number of Samples)
Describes the number of samples used as an input to the IFFT.
Scale output down by 2^
Describes the power of 2 to scale the result by prior to output.
Advanced
Target input throughput
Specifies the rate at which data samples should be processed. The default value is 200.
Specify the number of cascade stages
When this option is not enabled, the tool will determine the filter configuration that best achieves the specified input sampling rate. When this option is enabled and the 'Number of cascade stages' is specified, the tool will guarantee the same. In such cases, however, the specified sample rate constraint may not be achieved.
Number of cascade stages
This determines the number of kernels the FFT will be divided over in series to improve throughput.