CIS-60 30Bd, π/2-DBPSK

this CIS OFDM 60-tone modem, probably also it belonging to the CIS HDR family, was heard on 29 September at 0600z on 14259.0 KHz on USB. It exhibits a rather intriguing waveform:

1) manipulation speed is 30Bd and channel step 44.4Hz (Pic. 1), values are reasonably approximated since the signal was not sampled at its native frequency (look at "k" parameters);

Pic. 1

2) modulation format is DBPSK with these features: 

a) a service-tone (w/out modulation) is sent each 3 symbols in all (!) the carriers

b) 24 tones (say black-carriers) have π/2-DBPSK (Pic. 2);

c) 24 tones (say blue-carriers) have the same π/2-DBPSK modulation but with the evidence of the service tone (Pc. 3);

d) 12 tones (say red-carriers) have regular DBPSK (Pic. 4).

Pic. 2 - 24 (black) carriers: π/2-DBPSK

Pic. 3 - 24 (blue) carriers: π/2-DBPSK and service-tone

Pic. 4 -  12 (red) carriers: DBPSK

Pic.5 - phase angles in black/blue and red carriers

By writing the 60 coloured carriers in a 10x6 array, it's interesting to note that they are are arranged in a sort of diagonals format (Pic. 6):

Pic. 6 - carriers arrangement

About the ACF value, it is 200 ms, that makes 6 symbols although the real measure is 202.3 (Pic. 7) ms but you must consider that the signal is not perfectly in-frequency: infact the measured baudrate is not the expected 30.0 Bd but rather 29.9 Bd. Anyway, calculating the mathematical proportion:

(symbols-per-second : 1000) = (symbols : ACF)
(29.9 : 1000) = (x : 202.3) ; x = 6.04 

you get 6.04 symbols per frame. Then, given the exact 30Bd baudrate, I think the real ACF is 200. A further evidence in this direction is that just each 6 symbols the service-tone occupies the same position (Pic. 8).

Pic. 7 - 6 symbols ACF (200 ms)

Pic. 8 - each 6 symbols causes 200 ms ACF