Hi.

While analyzing an OFDM signal sampled at 40Mhz, with the following Spectrum:

fskorofdm1.png

Using mainly cross-correlation, preamble of the size 3658 samples is being detected.

The preamble consist of 2 similar parts – each – 3658/2 samples.

The preamble, however, has the following spectrum:

fskorofdm2.png

Tuning the timing and the frequency yields the following:

fskorofdm3.png

Is the preamble 8FSK modulated?

If so – what would be the best approach for demodulation?

Thanks in Advance.

Hi SenSen.  David again.

Is this the same data set that we discussed before?

I don't believe that there is any FSK sequence for the preamble.

I am posting a plot from your previous data with some markers related to the explanation.

Plot of ABS(FHSS_1_sh) starting at sample 1000

This plot is the abs(FHSS_1_sh) and I have annotated several important point in the waveform.

Markers at 7717 an 1.138e4 are the centers of two identical training sequences for the OFDM burst. Looking carefully, we see that these sequences are symmetric.  This indicates that the OFDM symbol is modulated only in the real axis. Even symmetry in time means real only in frequency, or something like that.

Markers at 2281 and 5865 indicate the beginning and ending (approx) of the first training sequence centered at 4066.  Looking closely, we see again that this sequence has symmetry.

Now looking at the first three markers at 924, 1840 and 2256, we see a repetition of the same prefix of the first training sequence.

If you take the entire sequence and subtract a delayed version of the abs of itself delayed by the 3658 you mention above, you will discover the duplicate training that I mentioned, but you also find the symbol boundaries because of the cyclic prefix process.  I have not done the excruciating analysis to find the ramp-up/ramp-down of the crossover between symbols, but I think that you could derive that from the process here.

Use the following statements to see the result of my comment above.

diff_test = 3658; diff_sig = abs(FHSS_1_sh(1:size(FHSS_1_sh)-diff_test))-abs(FHSS_1_sh(diff_test:size(FHSS_1_sh)-1)); figure; plot(diff_sig)

Back to your question.  I have no belief at all that the preamble is FSK, so there is no need to try to decode it that way.

I hope that this is useful.

David

Hello David.

How are you?

Thank you for the detailed answer regarding the previous data set.

I will review it carefully tomorrow, and update.

As to the post above - this is a different data set - where the data symbols are OFDM, however, the Preamble seems to be modulated differently. 

You can see the attached samples below sampled at 40Mhz and moved to BB(the first half of the preamble - the second is similar):

half_preamble.mat

Will be glad to know your opinion.

Many thanks.

Best Wishes.

Hi SenSen,

The sequence is so short, that I cannot make much of it at all.

I performed the standard test for FSK looking for differential phase (frequency):

1) ann = angle(half_preamble)

2) ann1 = unwrap(ann)

3) ann2 = ann1(2:end)-ann1(1:end-1)

Nothing obvious there.

David

Thank you David.

Regarding the original sequence, our findings are aligned.

Using the shifted diff as well as cross and auto correlation of different segments of the burst, the same components and sizes were derived.

One section the purpose of which was not clear to me is the 5368-5865.

Regarding the second data set - further analysis reveals that LFSR stream is being transmitted using OFDM. 

Short update.

The time and frequency tuning of the signal described above, yields LFSR based preamble pattern, as can be seen below:

fig1.png

However, same approach applied on the data symbols, does not yield the required tuning, as can be seen below:

fig2.png

Wonder whether channel estimation based on the known preamble should be performed and if so – what is the appropriate method.