Re: [Discuss-gnuradio] question about the synchronization symbol used in ofdm example

[Martin Braun]

On 11/02/2014 04:44 PM, xianda wrote:
> Hello all:
>              Thank you in advance.I have two questions about the symbol
> used in ofdm example.
>              1.The first synchronization symbol is:
>                
> n1=00000001.414213538169860[...]

Waah lots of numbers.

If the first half wasn't identical to the second half, this wouldn't
work. Also, it's a mathemetical necessity given every 2nd sub-carrier is
zero that the halfs be equal.

However, they don't have t have the same *sign*. See this line in the
OFDM sync block:

64 gr::filter::fir_filter_ccf::sptr delay_ma(
    gr::filter::fir_filter_ccf::make(
        1, std::vector<float>(
            fft_len/2, use_even_carriers ? 1.0 : -1.0
        )
    )
);

The reason is basic FFT math (hint: phase shift in one domain equals
what in the other domain?).

Literally all you needed to do was plot the ifft() results and you would
have seen this.


>             The first half of the symbol in time domain isn't same to
> the second half of the symbol.It is a fault?
>              2.I read the document fft_vcc_fftw.cc
>               gr_complex *dst = d_fft->get_inbuf();
>               unsigned int len = (unsigned int)(floor(d_fft_size/2.0));
> // half length of complex array
>               memcpy(&dst[0], &in[len], sizeof(gr_complex)*(d_fft_size -
> len));
>               memcpy(&dst[d_fft_size - len], &in[0],
> sizeof(gr_complex)*len);
>              Before doing the ifft operation,should I first do this?

It is unclear what you mean by "this". My guess is, the answer is no.

M