Processor Architecture for High Performance Video Decode
Processor Architecture for High Performance Video Decode
Vendor:
Noesis Technologies
Category:
Filters Transforms
Ultra High Speed FFT/IFFT processor
Noesis Technologies ntFFT_UHS IP implements a customized FFT/IFFT programmable fixed point (Decimation in Frequency - DIF) transf…
Overview
Noesis Technologies ntFFT_UHS IP implements a customized FFT/IFFT programmable fixed point (Decimation in Frequency - DIF) transform processor, supporting low latency, streaming, ultra-parallel complex samples per clock cycle in natural order. Input, internal and output 2’s complement fixed point precision are fully configurable before IP synthesis.
Radix-2, Radix-4 or Mixed-Radix design may be selected with parallel butterflies deployment, depending on the implemented transform sizes. Each stage features its own permutation network buffer, implemented optionally as either Register File or BRAM primitives. Twiddle factors fixed-point precision is selected via parameterization and the values are precalculated and stored in small distributed LUTs next to the respective butterflies, using a scalable design methodology. The permutation of each buffer stage is necessarily custom-made, since it is dependent on the parallel samples per clock cycle configuration and the supported FFT transform sizes. An optional Circular Shift buffer can be instantiated for those applications that need to correct a detected Carrier Frequency Offset in the frequency domain (FFT), with range of circular shifts correction relevant to both the FFT transform size and the parallel samples per clock cycle. Additional Overlap-Save (OLS) method wrappers may be provided to support real time high bandwidth filtering applications.
Radix-2, Radix-4 or Mixed-Radix design may be selected with parallel butterflies deployment, depending on the implemented transform sizes. Each stage features its own permutation network buffer, implemented optionally as either Register File or BRAM primitives. Twiddle factors fixed-point precision is selected via parameterization and the values are precalculated and stored in small distributed LUTs next to the respective butterflies, using a scalable design methodology. The permutation of each buffer stage is necessarily custom-made, since it is dependent on the parallel samples per clock cycle configuration and the supported FFT transform sizes. An optional Circular Shift buffer can be instantiated for those applications that need to correct a detected Carrier Frequency Offset in the frequency domain (FFT), with range of circular shifts correction relevant to both the FFT transform size and the parallel samples per clock cycle. Additional Overlap-Save (OLS) method wrappers may be provided to support real time high bandwidth filtering applications.
Key features
- Highly pre-synthesis design configurability with detailed generic/parameter values, in order to meet desired implementation trade-offs.
- Matlab 100% bit-true reference model with performance metrics and quality markers.
- Natural order input and output.
- Programmable FFT/IFFT transform selection and active transform size.
- Deployed Radix-2, Radix-4 or Mixed-Radix data-path units as needed. Radix selection also provides a trade-off for implemented multipliers.
- Uniform signal energy by scaling in every FFT/IFFT processing stage. For Radix-2 stages optional sqrt(2) scaling with multipliers or with add/sub approximation trade-off.
- Optional programmable Cyclic Shifter for Carrier Frequency Offset (CFO) correction.
- Optional Overlap-Save wrapper logic for both FFT and IFFT operation.
- Example architecture #1: 256-parallel programmable 1024/256 FFT low latency, pipelined streaming implementation [I/O FI S7.6, INT. FI S11.8].
- Example architecture #1 is synthesized & proven in 5nm process node, achieving ~960 MHz clock with throughput rate ~245 IQ Gsps (complex Giga-samples per second) for ntFFT.
- Example architecture #2: 32-parallel programmable 256/128/64 FFT low latency, pipelined streaming implementation [I/O FI S7.6, INT. FI S11.8].
- Synchronous clock design.
Applications
- Communication Systems.
- Spectrum Analysis.
- OFDM modems.
- Image processing.
- Streaming Filtering applications.
- Defense Receivers and Signal Monitoring.
- Medical and Scientific Instruments.
What’s Included?
- Fully commented synthesizable VHDL / System Verilog source code or FPGA netlist.
- VHDL / System Verilog test bench and example configura-tion files.
- Matlab fixed point model.
- Comprehensive technical documentation.
- Technical support.
Files
Note: some files may require an NDA depending on provider policy.
Silicon Options
| Foundry | Node | Process | Maturity |
|---|---|---|---|
| TSMC | 28nm | 28nm 280 nm | Silicon Proven |
Specifications
Identity
Part Number
ntFFT_UHS
Vendor
Noesis Technologies
Type
Silicon IP
Provider
Noesis Technologies
HQ:
Greece
Noesis Technologies specializes in design,development and marketing of high quality, cost effective communication IP cores and provides expert ASIC/FPGA design services in telecom DSP area.
Our solutions are key components to the most sophisticated telecom systems. Backed-up by our leading-edge expertise on forward error correction, encryption and networking technology as well as on DSP algorithm development we provide robust solutions that are used to improve data quality, increase bandwidth or reduce the overall system cost of end-application.
Learn more about Filters Transforms IP core
No size fits all for signal processing on FPGA (RF Engines)
No size fits all for signal processing on FPGA (RF Engines)
Employing general-purpose processors for radio DSP
Employing general-purpose processors for radio DSP
Do's and Don'ts of Architecting the Right FPGA Solution for DSP Design
Finally a Practical ''Do and Don't'' primer on architecting FPGA solutions for DSP design. With more do's than don’ts, the article is a down-to-basics look at how to avoid the pitfalls and realize device benefits
10 FPGA Design Techniques You Should Know
Regardless of whether you are using VHDL, System Verilog, or a different design capture language, there are a number of universal design techniques with which FPGA engineers should be familiar, from the very simple to the most advanced.
How to implement double-precision floating-point on FPGAs
The real results of a double-precision matrix multiply core that can easily be extended to a full DGEMM benchmark are demonstrated.
Frequently asked questions about Filters and Transforms IP cores
What is Ultra High Speed FFT/IFFT processor?
Ultra High Speed FFT/IFFT processor is a Filters Transforms IP core from Noesis Technologies listed on Semi IP Hub. It is listed with support for tsmc Silicon Proven.
How should engineers evaluate this Filters Transforms?
Engineers should review the overview, key features, supported foundries and nodes, maturity, deliverables, and provider information before shortlisting this Filters Transforms IP.
Can this semiconductor IP be compared with similar products?
Yes. Buyers can compare this product with similar semiconductor IP cores or IP families based on category, provider, process options, and structured technical specifications.