Re: [Discuss-gnuradio] Fourier Transform on multiple OFDM symbols

[Ron Economos]

Here's the best I can do. This is with an OOT module that I developed for fooling around with spectrum painting, but it can be used to generate specific OFDM signals. The secret sauce is that although the amplitude of each carrier is controlled, the phase is random. Also, a super long cyclic prefix helps to separate the carriers.

https://github.com/drmpeg/gr-paint

Here's an example spectrum paint from an image.

https://twitter.com/drmpeg/status/664060519880355840

The receiver program is Gqrx. It's more useful since it can use very large FTT's (262144 in this case).

https://github.com/csete/gqrx



If you want to re-create this, I've uploaded the flow graph and test image.

http://www.w6rz.net/paint_tx.grc

http://www.w6rz.net/paint.bin

You'll also have to change two lines in gr-paint/lib/paint_bc_impl.cc

Line 51, change the FFT size to 64 so that there are less carriers in the 2 MHz bandwidth.

ofdm_fft_size = 4096;

ofdm_fft_size = 64;

Line 127, increase the amplitude since we've decreased the FFT size.

magnitude /= 10000000000.0;

magnitude /= 1000000000.0;

Ron

On 01/04/2016 01:42 PM, Aditya Virmani wrote:

Hello Sir

Yes, you are right. But, why is it so? If everything else is right, I should have been getting a similar graph without averaging it as well; isn't it?

Can it happen because of non-equal data? But, I am feeding in same bytes (Random Uniform Source: Min = Max = Seed = 255), and further also, if I employ QAM, which should give me same Pre-OFDM data, then after OFDM modulation, again, I should be getting the same symbols which should lead to such peaks in the FFT output.

My problem is that I have to analyse each sub-carrier separately, where average case won't help.

Second thing which I fail to understand here are the large peaks in the relatively empty part (unoccupied part of the OFDM symbol), and in the middle.

Can you please help me with this.