HFDVL modem: OFDM 73-tone (13 pilot tones) QPSK

(I-56578, ANgazu)

Total carriers: 73 Pilot carriers: 13 Data symbols per OFDM symbol: 60 Native sample rate (sps): 9600 FFT length (samples): 256

Modulation: QPSK Bit rate (bps): 3600 Bandwidth (Hz): 2737.5 Samples per OFDM symbol: 320

OFDM parameters, body segment

The most interesting feature is the way the 13 pilot tones are formed, and why - at a first look - we found only 1 pilot tone (the higher one).

We found that all the even tones have QPSK modulation (at least into the seen signals) :

 and the top tone (the 73th) is clearly a pilot one:

  

We found that in the odd tones there is a new element in absolute constellation that generates a PSK8 pattern in diff constellation. We guess this element is a scattered pilot and, for some reason we do not know, it shifts diff constellation of 45 degrees, giving the false PSK8:

Skipping symbols, we could see the 'no modulated pilot' feature for these tones. Then, according to our analysis, the 13 pilots coule be formed in the following way:

- 1 fixed pilot tone (73th);

- 12 scattered pilots which are chosen in turn from 3 different sets (each set contains 12 odd tones);

(...and we are quite positive that we have exactly identified all the tones of each set).

The pilot tones configuration changes at every OFDM symbol: we do not know the reason (perhaps to contrast propagation conditions?) and the algorithm that determines the initial configuration; possibly the BPSK preamble has these data inside, but it's just a guess. Anyway, as said above, it looks like that all the odd tones are pilots-candidate (we could say “virtual pilots”) and only 12 of them are used - in turn - to form the real pilots, while the remaining 24 + the 36 even tones are used to transport user data (60 tones).

Here is an example of a possible pilot tones scheme that assumes the configuration A as the initial configuration:

configuration A (73th tone + 12 tones of the set-A)
configuration B (73th tone + 12 tones of the set-B)
configuration C (73th tone + 12 tones of the set-C)
...and so on, restarting from configuration A 

The trailing segment of the signal just exhibits all the odd (pilots candidate) tones, some using BPSK modulation.

Two words about the speed. Official documentation say:

Total carriers 73, Pilot carriers 13, Data symbols per OFDM symbol 60, Bit rate (bps) 3600 (assuming Bit rate to be the “system” speed “on the air”). Since the 30 baud PSK-4 modulation, the system grants 30 * 2 * 60 = 3600 bps data rate.

The HFDVL modem (waveform is developed by the groups "CeTIC" from the Universidad de Las Palmas de Gran Canaria and "GAPS" from Universidad Politécnia de Madrid, other than Aeropuertos Españoles y Navegación Aérea) has been heard mostly on USB 14350.0 and 14828.5 transmitting from Spain: we think that they are conducting tests that involve transmitters in two diferent sites which exchanges data. In one of our 14828.8 KHz recordings is quite clearly visible the different quality of the signals:

Emailing the HFDVL modem manufacturer, we got a confirm about what we were thinking: "[...]Yes, there are two modems transmitting and at this moment there are not other frequencies involved in the test."