CIS-Navy & Akula transmission

My friend Mario often monitors the frequency of 9201.0 KHz/USB in search of "Akula" signals: he mainly uses a KiwiSDR receiver located in Japan (Azumino-city, Nagano) [1] and it seems that this frequency (certainly one of the many) is quite active for this kind of transmissions, as collected by my friend Dave too. Most of the time it is usually a pair of signals that repeat at irregular intervals.
A few days ago he kindly sent me an interesting and "curious" recording of a transmission in which both an FSK 50Bd/1000 signal and Akula (FSK 500Bd/1000) are used with a central frequency of 9202 KHz (Figs. 1,2) 

Fig. 1

Fig. 2

The first thing that catches the eye is the particular "shape" of the Akula signals in which the well-known initial synchronization and preamble groups are missing but the EOM + EOT groups (101111 100010 100010 101111 011110) are exactly in their place, as can be seen from the demodulated bitstream in Figure 3. Just one year ago I had already come across these (let's call them) "anomalies" [2]. "It could depend on a malfunction of the modem or on the receiver's attack time" my friend cryptomaster says.

Fig. 3 - Akula bitstream

 

The most interesting thing however is the presence of a 50Bd/1000 FSK modulation preceding an Akula burst: something I had never seen before (and not even that type of FSK modulation). After demodulating it and reshaped to a 7-bit format, in addition to the initial inversions, I noticed a final sequence composed of five identical 7-bit words "000100" which - as far as I know - is the typical EOM sequence used in the CIS-Navy waveform (also known by the nicknames T-600, BEE-36, CIS 36-50). However, compared to the latter, it lacks the initial part consisting of a sequence of 2 bit sequence 

(usually) "100001010010111110000101001101011010110101101"

followed by 70-bit Initialization Vector (ten 7-bit words) that is repeated twice (Figure 4).

Fig. 4 - FSK 50Bd/1000 bitstream

As per previous analysis of the CIS-Navy waveform [3], its payload data consists of 5-bit characters coded into 7-bit sequence with a fixed ratio of '1's vs. '0's of 4 to 3 (or vice versa, depending on polarity of reception) so I decided to check the 4:3 ratio in this demodulated bitstream: the result (97.5%) indicates a very good probability of success.

 

Fig. 5 - 4:3 ratio in FSK 50Bd/1000 bitstream

CIS-Navy waveform has been logged with different Baud rates (36, 50, 75, 100 and 150) and shifts (85, 125, 250 and 500 Hz) so, likely, that's another variation.

https://disk.yandex.com/d/E29PupqpJg3UTQ

[1] http://jf0fumkiwi.ddns.net:8073/?f=9201.00usbz9
[2] http://i56578-swl.blogspot.com/2024/05/akula-always-reserves-surprises.html
[3] http://i56578-swl.blogspot.com/2016/10/cis-navy-50bd200-fsk-t600-bee-36-cis-36.html