1 / 13

Complexity reduction for time domain H matrix feedback

Complexity reduction for time domain H matrix feedback. Date: 2010-11-09. Authors:. Time domain H matrix compressed feedback. Time domain H matrix feedback is without any objection the best method for feedback compression for 802.11ac

Download Presentation

Complexity reduction for time domain H matrix feedback

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. Complexity reduction for time domain H matrix feedback Date: 2010-11-09 Authors:

  2. Time domain H matrix compressed feedback • Time domain H matrix feedback is without any objection the best method for feedback compression for 802.11ac • Combines feedback compression and improvement of channel estimation • Time domain feedback with TiDFT presents • the best CSI feedback accuracy (6dB MSE over uncompressed, way more over legacy compressed V matrix): • the best compression gain (4x or 8x) • Arguments against this technique: • not a legacy scheme • complexity Slide 2

  3. Time domain H matrix compressed feedback: discussion around legacy • Objective: reduce the technological step between 11n (beamforming) and 11ac (MU-MIMO) • option 1 (ACcord): choose compressed V-matrix for 11n (beamforming) and push for compressed V-matrix in 11ac (beamforming and MU-MIMO) • this would lead to the non-selection of the best compression scheme for 11ac, which would potentially reduce the use of MU-MIMO (average by the bottom) • option 2: choose H-matrix feedback for 11n (beamforming) and choose time domain H matrix for 11ac (MU-MIMO and beamforming) • H matrix already favored by a majority of 11n client companies • selection of the best compression scheme for 11ac to have the best MU-MIMO Slide 3

  4. Time domain H matrix compressed feedback: discussion around complexity • time domain H feedback with TiDFT or DCT has a reasonable complexity: • transfer matrix multiplication in STA for TiDFT (only 16kB ROM storage for all matrices), simple fast DCT for DCT • FFT or iDCT in AP (FFT already implemented in chipsets) • No additional channel estimation improvement schemes needed • Compressed V matrix also have an important complexity in receiver • Note that the compressed V matrix process is so long that after SIFS, the process isn’t completed: the preamble time of the CSI feedback is needed to end the process • In this condition, there is no way that better channel estimation will be performed on top of compressed V matrix (stay with LS estimation): that means that the quality of CSI feedback will be of the worst quality (see Celeno’s results in annex) • For all these reasons, we believe time domain H matrix feedback should be added to the spec framework. • and we show in this presentation that the complexity can even be further reduced Slide 4

  5. Proposition to reduce the complexity of time domain H feedback • Instead of generating the impulse response of the complete channel (all subcarriers) • generate time domain response based on the all Np subcarriers • feedback of the CP first elements in the time domain • We propose to generate multiple partial impulse responses corresponding to multiple groups (B groups) of subcarriers • generate B time domain responses each based on a group of NFFT/B subcarriers • The feedback includes multiple fields of the B partial impulse responses, each constituted of CP/B elements (note that we have demonstrated that all the usefull energy is gathered in the CP/B first elements) • The two next slides illustrate the classical and the new proposition with B=2 groups of subcarriers. Slide 5

  6. Slide 6

  7. Two fields in CSI feedback Slide 7

  8. Conclusion • Time domain H matrix provides the best compression with the best quality of CSI feedback • especially when considering both compression and channel estimation • Improve the efficiency of MU-MIMO and reduce the overhead • Technological gap between 11n and 11ac and between beamforming and MU-MIMO is not higher with time-domain H matrix if the feedback issue is solved by taking into account 11n and 11ac jointly. • The complexity of the time domain H matrix can be even reduced by applying the process to 2 groups of subcarriers with TiDFT : • it enables to reduce by a factor of 2 the complexity of the compression scheme and the size of the storage for TiDFT matrices, both in STA and in AP • Note that this proposition can also be applied to the DCT-based time domain feedback (which provides a lower compression gain). in this condition, up to 8 groups can be defined with overlapping between subcarriers groups Slide 8

  9. Pre-motion • Do you support updating the spec framework to require that time domain H matrix feedback, be one feedback format for both VHT SU beamforming and DL MU-MIMO? • Yes • No • Abstain

  10. References [1] Hongyuan Zhang and all, 11ac explicit sounding and feedback, IEEE 802.11-10/1105r0, Sept 2010. [2] L. Cariou and M. Diallo, Time Domain CSI report for explicit feedback, IEEE 802.11-10/0586r1, May 2010. [3] K. Ishihara and al., CSI Feedback Scheme using DCT for Explicit Beamforming, IEEE802.11-10/0806r1, july 2010. [4] Nir Shapira, Yaron shany, Doubling Number of VHT-LTFs in MU-MIMO Operation, IEEE802.11-10/1113r0, Sept 2010. Slide 10

  11. Annex • 1 Illustration of the process with DCT-based time domain feedback • 2 Simulation results from Celeno showing the performance degradation using compressed-V matrix with LS estimation instead of time domain H matrix. Slide 11

  12. Slide 12

  13. Simulation Results from Celeno [4] • 8 AP antennas, 4 users, each with 3 antennas. Channel model D. 20ms aging • Compare the case of 1,2,3,4 or ∞ (perfect CSI) LTFs per stream in NDP • Results reflect impact of the quality of the feedback (independent of the channel estimation during Data transmission) • Results equivalent to comparing a channel estimation noise reduction scheme of 0dB, 3dB, 4.5dB, 6dB and ∞ (perfect CSI) Channel estimation resultsembeded in Time domaincompressed feedback • Curve in blue: LS channel estimation (equivalent to what is feasible with compressed V matrix feedback) • Curve in red: equivalent to channel estimation improvements embedded in time domain compressed CSI feedback starting from LS channel estimation Channel estimation resultswith LS channel estimation (only estimation feasiblewithcompressed V matrix) Slide 13

More Related