.png){kind=icon}
.png){kind=icon}
Return
Resource Optimization in Polyphase-Filter STFT Based on Time-Multiplexed Constant Multiplication
.png){kind=icon}
PRE
AI
DOI：10.1109/TVLSI.2025.3608756.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Real-time spectrum analysis is widely used in applications such as embedded audio analytics, digital acquisition instruments, electronic warfare, and micro-Doppler radar. These applications heavily rely on the short-time Fourier transform (STFT) engine, which delivers high throughput and leads to strict area and power budgets. However, current STFT architectures implemented on hardware still rely on massive arrays of generic multipliers or on-chip memories that store precomputed window coefficients, thereby limiting flexibility and increasing latency and resource consumption. The optimized windowing unit of the polyphase-filter STFT architecture proposed in this article is redesigned by time-multiplexed single and multiple constant multiplication (Tm-SCM/MCM) that are created from directed acyclic graphs (DAGs). The Tm-SCM-based design can replace the original generic multiplier. The Tm-MCM-based design can reduce the adders and multiplexers (MUXs) by analyzing the window function coefficients among different channels. Both methods can minimize the redundant adders and MUXs compared with the existing algorithms and remove all DSP generic multipliers. Experiments on field-programmable gate array (FPGA) and ASIC show that the proposed Tm-SCM/MCM Hann window core reduces area and delay and lowers dynamic power. When the same cores are integrated into a 32-point STFT, overall slice count drops, all BRAM blocks in the windowing unit are removed, latency stays low, output SQNR matches recent polyphase and fast Fourier transform (FFT)-based designs, and figure-of-merit analysis confirms that the proposed architecture delivers the best overall balance of resources, speed, and power among the compared solutions. These findings show that optimization of windowing units by constant multiplication offers a cost-effective solution for upcoming high-overlap, real-time STFT applications by concurrently reducing resource footprint and maintaining spectral fidelity.
Keywords:
Fast Fourier transform (FFT)
short-time Fourier transform (STFT)
time-multiplexed constant multiplication (Tm-CM)
windowing unit
Journal
I
IF:
3.1
Papers:
440
Citations:
7.3K