Method and system for fixed point fast fourier transform with improved SNR
AssigneeAnalog Devices, Inc.
InventorsYosef Stein, Haim Primo
Life of this patent
Estimating the term…
Abstract
A system and method of improving signal to noise ration (SNR) in a fixed point fast Fourier transform (FFT/IFFT) generates from sample inputs and a twiddle factor butterfly outputs for each stage; scales the butterfly outputs of this stage from a predicted normalization scale factor to obtain the maximum butterfly output without overflow from this stage; determines from the butterfly outputs of this stage the minimum normalizing exponent for the butterfly outputs of this stage and predicts a normalization scale factor of the next stage from the minimum normalizing exponent of this stage and a stage guard scale value to obtain the maximum butterfly output without overflow from that next stage.
Claims19
A fixed point fast Fourier transform (FFT/IFFT) system with improved SNR comprising:
- a compute unit responsive to sample inputs and a twiddle factor for generating butterfly outputs for each stage;
- a scale unit responsive to a predicted normalization scale factor to scale the butterfly outputs of this stage to obtain the maximum butterfly output without overflow from this stage;
- a comparator unit responsive to the butterfly outputs of this stage to determine the minimum normalizing exponent for the butterfly outputs of this stage;
- and
- an arithmetic logic unit responsive to said minimum normalizing exponent of this stage and a stage guard scale value to predict a normalization scale factor for the next stage to obtain the maximum butterfly output without overflow from that next stage.
A method of improving signal to noise ratio (SNR) in a fixed point fast Fourier transform (FFT/IFFT) comprising:
- generating from sample inputs and a twiddle factor butterfly outputs for each stage;
- scaling the butterfly outputs of this stage from a predicted normalization scale factor to obtain the maximum butterfly output without overflow from this stage;
- determining from the butterfly outputs of this stage the minimum normalizing exponent for the butterfly outputs of this stage;
- and
- predicting a normalization scale factor of the next stage from said minimum normalizing exponent of this stage and a stage guard scale value to obtain the maximum butterfly output without overflow from that next stage.
Dependencies are inferred from the claim text.