locked
Why skip_cp=8? RRS feed

  • Question

  • Hi,

    SoraSDK 1.8. I observe in bb\brick11\src\Channel_11a.hpp and bb\brick11\src\PHY_11a.hpp, the skip_cp value is set to 8. From the context, 8 means the number of the samples to be skipped. In Channel_11a.hpp, the LTS is 160 sample with 32 cp, and 16 cp is pre loaded in the STS self correlation. So, I guess the skip_cp value should be 32-16=16? In PHY_11a.hpp, the skip_cp in data symbol is also 8.

    I'm not sure if I misunderstand some parts of the code. Since the results of skip_cp=16 and skip_cp=8 in my offline test are the same, it is not easy for me to generate signal with strong multipath to varify the right value. 

    By the way, does anyone have an artificial/ideal dump file for offline test? or some guide to pack the data from ModBuffer11a(...) to a dump file.

    Thanks,

    Zhice



    • Edited by YANG Zhice Thursday, May 30, 2013 4:59 PM
    Thursday, May 30, 2013 4:58 PM

All replies

  • Hi Zhice,
    Just correct one fact: 12 CP is preloaded in TCCA11a.
    If I understand your explanation correctly, you argued that skip_cp in T11aLTSymbol should be 32-12=20 instead of 12, and skip_cp in T11aDataSymbol should be 16 instead of 8. Am I right?

    In OFDM theory, the guarded interval (ie. cyclic prefix, CP) is used to reduces the sensitivity to time synchronization problems. The offset (skip_cp) starting from symbol boundary will not influence FFT result. This is also the reason to your offline test. Ideally, you should use the middle 64 samples in one symbol (80 samples) as FFT input.
    -Qi
    • Proposed as answer by Qi LuoEditor Friday, May 31, 2013 10:45 AM
    Friday, May 31, 2013 10:45 AM
    Answerer
  • Hi,

    Thanks a lot for your explanation, quite helpful.

    I still have two more questions:

    1. Why 12cp is preloaded in TCCA11a? check_sync () returns false after the last short training symbol and check_sync () is called every 16 samples. Take the figure for example. If the newest pumped-in samples are "14,15,16,1", samples in the accumulator will generate a peak with peak_index=1. Am I right? Consider all the possible values of peak_index, the preloaded cp value would be 16 to 19(16+3).

    old

    2

    3

    4

    5

    6

    7

    8

    9

    10

    11

    12

    13

    14

    15

    16

    1

    new

    2. peak_index bears 0 to 3 samples. How dose sora handle these samples? I have not seen in the context.

    Zhice


    Friday, May 31, 2013 1:54 PM
  • What are you guys talking about? Cross-correlation problem?

    Danial.F

    Saturday, June 1, 2013 4:53 PM