1. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 VHT-LTF Design for IEEE802.11ac Slide 1 Date: 2010-07-12 Authors:

2. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 Abstract The long training field (LTF) provides a means for receivers to estimate MIMO-OFDM channels in 802.11ac systems. An excellent LTF should not only minimize channel estimation errors, but also reduce overhead as much as possible. An efficient VHT-LTF is proposed to improve the accuracy of channel estimation with less overhead. Slide 2

3. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 Background In 802.11n, the number of HT-LTFs is greater than or equal to the number of space-time streams (N STS ). If 802.11ac uses the same mechanism, in order to support 8 space- time streams, the number of VHT-LTFs should be up to 8. It is obviously a resource waste, especially when the data field is short, e.g., NDP and sounding PPDU. Can we design a VHT-LTF to achieve excellent performance with less overhead? A class of advanced training sequences, CAZAC codes, can meet this requirement. Slide 3

4. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 Introduction of CAZAC codes What is CAZAC codes? –CAZAC (Constant Amplitude Zero Auto-Correlation ) codes are a class of complex-valued pseudo-random noise sequences with cyclic autocorrelation equal to zero. –Zadoff-Chu sequence is a well-known CAZAC code, which has been widely used in wireless systems. Slide 4 Zadoff-Chu sequence expression

5. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 Characteristics of CAZAC codes Characteristics –Zero Auto-Correlation: It means that a CAZAC code is always orthogonal with its cyclic shifted versions. –Constant Amplitude Major advantages –Reduce inter-symbol interference (ISI) –Avoid interferences between multiple antennas –Lower PAPR As a result, CAZAC codes are regarded as optimum training sequences for channel estimation in MIMO- OFDM systems. Slide 5

6. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 Simulation Parameters The Frequency Domain-Least Square-Joint (FD-LS-JCE) channel estimation algorithm has a low implementation complexity [4], comparable with the FD-LS algorithm. Slide 6 Scheme 1Scheme 2Scheme 3 VHT-LTF TypeHT-LTFCAZAC (Zadoff-Chu) Num. of VHT-LTFsN VHT-LTF symbolsOne symbol N VHT-LTF symbols MIMO1×1, 2×2, 4×4, 8×8 Bandwidth40MHz Channel modelModel B Channel estimation Algrithm FD-LS and MMSEFD-LS-JCE

7. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 Slide 7 Explanation of VHT-LTF Schemes The CAZAC code for space-time stream k is generated by cyclically shifting the basic CAZAC code (for Stream 0) with k/N STS code length. The CAZAC code in Scheme 3 are repeated N VHT-LTF times. Scheme 1 Directly use HT-LTF as VHT-LTF Scheme 2 Use CAZAC code in one symbol Scheme 3 Use CAZAC code in N VHT-LTF symbols n = N VHT-LTF symbols HT-LTF1HT-LTF2HT-LTFn CAZAC Only one symbol N VHT-LTF symbols CAZAC

8. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 Preformance Comparison (1x1, 2x2) MSE denotes the mean square error of channel estimation. Slide 8

9. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 Preformance Comparison (4x4, 8x8 ) Slide 9

10. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 Performance Analysis The performance of CAZAC codes is much better than that of HT- LTFs when they use the same number of symbols. –The SNR gain is 8 ~12 dB at MSE=0.01 when HT-LTFs use the FD-LS algorithm. –The SNR gain is 4 ~ 6 dB at MSE=0.01 when HT-LTFs use the MMSE algorithm. The CAZAC codes using one symbol can also achieve better performance than HT-LTFs using the FD-LS algorithm. The CAZAC code using one symbol can still outperform HT-LTF using the MMSE algorithm when –the number of space-time streams is not too large (<= 4), or –SNR is high Slide 10

11. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 Conclusion With speical code characteristics, CAZAC codes are able to achieve better performance with less training symbols, and thus can be used to –enhance the channel estimation accuracy, and/or –improve the spectral efficiency, especially when the data field is short, e.g., NDP and sounding PPDU. Due to significant performance advantages, we recommend CAZAC codes to be a candidate of VHT- LTFs, in which less backward compatibility with 11a/n needs to be concerned. Slide 11

12. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 StrawPoll #1 Do you support to reduce the number of VHT-LTFs to increase the spectral efficiency of 802.11ac and to edit the spec framework document, 11-09-0992, accordingly? –Yes: –No: –Abs: Slide 12

13. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 StrawPoll #2 Do you support to adopt some advanced training sequences (e.g., CAZAC codes) as VHT-LTFs and to edit the spec framework document, 11-09-0992, accordingly? –Yes: –No: –Abs: Slide 13

14. doc.: IEEE 802.11-10/0779r0 Submission Guixia Kang, BUPT July 2010 References [1] Specification Framework for Tgac, IEEE 802.11- 09/0992r11, May 2010 [2] IEEE 802.11n standard [3] Hongyuan Zhang, et al., 802.11ac Preamble, IEEE 802.11-10/0070r5, March 2010 [4]Time and Frequency Domain Joint Channel Estimation in Multi-carrier Multi-branch Systems, Aachen, Germany: Shaker Verlag, 2005 Slide 14