1 / 26

Sub- Nyquist Sampling Continuous to Finite Module Orthogonal Matching Pursuit Block

Sub- Nyquist Sampling Continuous to Finite Module Orthogonal Matching Pursuit Block. Final Presentation – Part A Performed by : Yoni Smolin. Supervisors: Inna Rivkin & Moshe Mishali Winter 2009 – Spring 2010. Sub- Nyquist Digita l System. On every clock cycle, the system solves How ?

ariane
Télécharger la présentation

Sub- Nyquist Sampling Continuous to Finite Module Orthogonal Matching Pursuit Block

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Sub-Nyquist SamplingContinuous to Finite ModuleOrthogonalMatching Pursuit Block FinalPresentation – Part A Performed by : Yoni Smolin • Supervisors: Inna Rivkin & Moshe Mishali • Winter 2009 – Spring 2010

  2. Sub-Nyquist Digital System • On every clock cycle, the system solves • How ? • Expand - computes y from input samples. • CTF - locates non zero elements of z. • DSP - computes their values. 4 Expand X3 CTF DSP Sampled data 12 Support A zsupp zsupp y Sampling matrix Asupp y sparse z

  3. CTF CTF Frame construction Support recovery • The frame matrix - a basis for recovery: Support frame matrix y70H y1H y2H … Q y1 y2 y70

  4. CTF CTF Frame construction Support recovery • Support recovery – applying OMP for Q: Support frame matrix A UQ Q support

  5. OMP – Algorithm (SBR2) Matching Solution Approximation Residual Update

  6. OMP – Adaptation for Hardware Matching

  7. OMP – Adaptation for Hardware Matching Modified Gram Schmidt Residual Update

  8. OMP – Atomic Operations 12 12 12 12 12 1 OMP iteration 101 projections 12 1 M.G.S. iteration update residual 12

  9. Implementation Considerations • CTF must fit in a single Stratix III FPGA. • Number representation: Fixed-point, Q1.16. • Fully parameterized design: N, L, m. • Guiding principle: Achieve best throughput. Implying: • A pipelined datapath. • Parallel execution of vector-matrix multiplications.

  10. CTF Block CTF Frame OMP Merge Amemory

  11. Block Diagram - OMP

  12. Block Diagram - Datapath keep max support stop ? SUB MAMU

  13. Block Diagram - Datapath keep max support stop ? SUB MAMU 18bit 21bit

  14. Matrix Multiplier (MAMU) Delay Multiplication - 2 cycles Projection - 5 cycles Hardware consumption DSP blocks – 64% 1 2 … 12 1 . . . 36 36 + 1 36 inputs inputs outputs outputs 12 A single column

  15. Inverse Square root • Implementation alternatives: • Remark: Input number must be > .

  16. FSM Controller

  17. OMP iteration - Matching Pipelined calculation = 101 cycles keep max support stop ? SUB MAMU

  18. OMP iteration – Gram Schmidt ( i-1 projections ) + normalization keep max support stop ? SUB MAMU

  19. OMP iteration – Update Residual Parllel calculation: stopping cond. & update R keep max support stop ? SUB MAMU

  20. Overflow • In theory, overflow can occur only for: → Solution: Divide R by 2 & recalculate: • Error detection flags: Overflow: MAMU, , SUB, . R R/2 some vector some vector

  21. Synthesis Report Frame OMP * Without debug environment

  22. Logical Simulation – Test Bench • Input text files are generated by Matlab. • Output text files are evaluated in Excel. Rmemory OMP Q.txt supp.txt Amemory A.txt

  23. Logical Simulation – Example Simulation output: • Support: →→→ Iteration: 1st 2nd 3rd 4th 5th→ f 9 64 38 93 z

  24. Logical Simulation – Results FM FM • Problem : recovery< 100%. • Solutions: • Regular mode: CTF can be reinitiated until successful recovery. • Iterative mode: Support merging can help overcome mistaken elements.

  25. Challenges • Swap normal ↔ iterations modes on the fly. • A must be normalized before system startup. • Q’scalculation: • Should be wide to avoid overflow. → transfer only 18 relevant bits to OMP. • Can be accelerated by upsamplingy. • Wide mux (5,184 bits) can pose a latency problem. • Stopping condition varies with input signal’s energy.

  26. Gannt

More Related