23 December 2023

Swedish Defence, unid datalink protocol over MS-110A

The analysis concerns short transmissions monitored on 3824.0 KHz/USB within the 80mt Ham Band (1) thanks to SM1OTX Airspy HF+ in Sweden and OZ1AEF KiwiSDR in Denmark. According to my observations, at least on this channel (3824 KHz/USB), daytime transmissions occur at intervals of 15 minutes but not with a fixed schedule (for sample: hh05, hh20, hh35, hh50; hh10, hh25, hh40, hh55; hh08, hh23, hh38, hh53) and if there are messages to be sent. The transmissions seem less frequent at night, but I don't have enough IQ monitoring to say for sure; however I have noticed that - at night - the same frequency is sometimes occupied by CIS-12 transmissions.

I ran several Direction Finding tests using the TdoA algorithm and all the results point to an area south from Stockholm, probably the NAVCOMMCEN of the Sweden Defence (Figure 1).

Fig. 1 - Direction Finding results

In this regard, it's worth noting the use of the 3824 KHz frequency. Indeed , at that high latitudes, due to the long dark periods and the sun low on the horizon during winter, the low portion of HF must be used (lack of F layers).  The 80 meter band is certainly the best choice for medium distance contacts (eg  ships in Baltic Sea or Gulf of Bothnia) as it guarantees 100% reliability over 24 hours (Figure 2) and is suitable for NVIS (Near Vertical Incidence Skywave) mode propagation for local contacts.

Fig. 2 - Circuit Reliability for 80 mt comms between medium distances

Below the results of my analysis, aimed to understand and write down the Protocol Data Units (PDUs) of the datalink protocol and the way their fields are encoded. Please notice that the "designations" used are only mine and are introduced just for convenient reference.

The MIL-STD 110A (MS-110A) 1200bps Short Interleaver is the used HF waveform (Figure 3).

Fig. 3 - MS-110A waveform and ACF (66.6 ms)

Sometimes it may happens to see fast exchanges (bursts 1-4) before a "usual" data transfer: judging by the fading patterns in Figure 4, two nodes are involved.

Fig. 4 - fast exchanges before a data transfer

Discarding the MS-110A waveform overhead, the bitstreams exhibit a 104-bit period: data blocks, well-defined structures and repeated patterns are clearly visible in Figure 5.

Fig. 5 - a bitstream after the removal of MS-110A waveform

Blocks are delimited by long series of logical "1s" and are composed of up to 17 datagrams. Each datagram starts with a synchronization-type PDU (S_PDU) followed by data-type PDUs (D_PDU):

Fig. 6 - general structure of a MS-110A transmission

synchronization-type PDU (S_PDU)
The S_PDU consists of a common 40-bit/5-byte sync sequence (2)

[0111111001111110100010111001100010111001] 0x7E7E8B98B9

followed by a 104-bit/13-byte chunk of data which is  repeated three times. After synchronizing the bitstreams on the sync sequence and a length of 144 bits (ie: sync + ACF), the 13-byte chunk is formed of (Figures 7,8):

- a specific 8-bit field h (designated a header)
- a 40-bit/5-byte field m
- a common (!) 32-bit/4-byte sequence s
[11101010011010100010001001100010] 0xEA6A2262
- a 3 bits ("000") string
- a 21 bits field n

Fig. 7 - a synched bitstream (first 144 bits)
Fig. 8 - bit-field map of the synchronization-type PDU (S_PDU)
data-type PDU (D_PDU)
The D_PDU consists of 80-bit/10-byte field p (designated a payload) followed by a 3 bits "000" string and a final 21 bits field q (Figures 9,10): notice that the final part is the same as the one of the S_PDU.
Fig. 9 - datagrams consisting of a S_PDU followed by D_PDUs
Fig. 10 - bit-field map of the data-type PDU (D_PDU)

 The general structure of a datagram is shown in Figure 11:

Fig. 11

I spent some days studying the bitstreams and trying to understand the meaning of the fields of  both  S_PDU and D_PDU, and I came to the following conclusions.  

Header field
The argument of the 8-bit header field of the S_PDU is coded as described in CCITT V.42 paragraph, which specifies that the least significant bit of a 1-byte argument shall be transmitted first. Therefore, after its conversion into a decimal number xyz, I found that the field it indicates the number of the useful bytes of the datagram and the eventual presence of a filler in the last D_PDU of the datagram.
Given that:
- each D_PDU has a payload of 10 bytes (40 bits)
- the header field of the S_PDU is a 8-bit field
the system is designed so that each datagram may contain up to 255 bytes of data (11111111). For example (see Figure 12):

xyz = 130: 130 useful bytes, 13 complete D_PDUs, no padding
xyz = 126: 126 useful bytes, 12 complete D_PDUs + 6 useful bytes in the 13th D_PDU (the remaining space is filled with four 0 value bytes)
xyz = 118: 118 useful bytes, 11 complete D_PDUs + 8 useful bytes in the 12th D_PDU (the remaining space is filled with two 0 value bytes)

So, the max value of the header field (255) stands for: 255 useful bytes, 25 complete D_PDUs + 5 useful bytes in the 26th D_PDU (the remaining space is filled with five 0 value bytes).

Fig. 12 - meaning of the header field of S_PDU

That meaning of the header field is also confirmed by analyzing the bistreams of the bursts in Figure 4 

- burst 1 (header: 01101110) carries 110 bytes of data within a single datagram of 11 D_PDUs
- bursts 2,3,4 (header: 00001110) carry short 14-byte "messages" in 2 D_PDUs and six padding bytes

Fig. 13

(It is practically impossible to define the nature of these short 14-byte messages, one can only assume that it involves some form of negotiation between the two nodes. Likewise, it cannot be said whether these messages are due to the operator intervention or whether they are automatically generated by the protocol, as is the case with STANAG-5066) 

The argument of field q of the D_PDU  consists of the 21-bit Cyclic Redundancy Check (CRC) applied and computed on the first 83 bits (ie payload p + "000"). The same consideration can be made up regarding the S_PDU, where the argument of field n is the 21-bit CRC field and fields h+m+s + "000" form the 83 bits which shall be coded.
Indeed, thanks to the help of my friend cryptomaster, we found a (104,21) matrix generated by the polynomial: 


which is well suited to the specific case.

Fig. 14 - (104,21) matrix used for CRC

I coded a short Octave script and tested the check sub-matrix on the first 83 bits of both the types of  PDUs, results are shown in Figures 15,16: the check-matrix works like a charm! So, if we consider that fields are structured in bytes, the 000 bits act as kind of padding added to the word being coded and a (104,83) FEC coding is used (BCH? LDPC?).

Fig. 15 - check matrix applied to the S_PDUs

Fig. 16- check matrix applied to the D_PDUs

In this regard  I extracted and merge together more than 500 D_PDU payloads into a single stream then I performed some tests on it.  The quality of the cryptography can be evaluated with a statistical method or by calculating the Shannon Entropy (3) and the Index of Coincidence (IC) (4) of the stream.
The statistical test (Figure 17) determines the randomness, the number of single bits in the stream is counted, then the double bits, then the triple bits and so on to the end. The result is a graph: if the information is not systematic, the adjacent columns should be half the size of the previous ones. The test shows good encryption quality. 

Fig. 17

The measure of the Shannon Entropy can be used, in a broad sense, to detect whether data is likely to be structured or unstructured. 8 is the maximum, representing highly unstructured, 'random' data. Properly encrypted or compressed data should have an entropy of over 7.5 while a low IC generally means that the text is random, compressed or encrypted (Figure 18):

Shannon entropy: 7.948409571238646
Index of Coincidence: 0.039349499261437125

Fig. 18 - Shannon Entropy and Index of Coincidence

If my guess is correct, the string m of the S_PDU is an Initialization Vector (repeated 3 times) of a stream cipher and thus the preamble is actually a COMSEC preamble consisting of bit sync, header, Initialization Vector, frame sync and a final CRC field.

A Secure Text system?
According my analysis, the PDUs of the used datalink protocol consists of the following formats (please notice that the 3-bit field encoded with the value "0" could be specified as NOT USED):

Fig. 19

But what kind of comms we are speaking about?
- although a datagram may contain up to 255 bytes of data, according my recordings each datagram is composed of no more than 130 bytes of data: such brevity could mean that the datagrams are made up of short text messages;
- transmissions occur almost regularly every 15 minutes, as if the sender station collects  messages coming from one link and then group and forward them into another link;
- transmissions are in a "blind" way, i.e. they are not preceded by 2G/3G ALE neither by voice calls and are not followed by ACKs: this could mean transmissions addressed to "stared" receiver(s);
the above issues make me think of a Secure Text system, ie a system that allows text and data exchange up to and including the level of secret and protected by cryptographic devices (5).
A good point in favor of my guess is that Swedish Armed Forces currently use a variety of national secure text messages that are not interoperable with international systems and are distributed just through the MaRA Naval Communications Centre (NAVCOMMCEN) for maritime platforms and operations [1].

...some speculations
Since Sweden Defence use  national crypto for domestic comms (6), what about the MGZI "Kryptomodem 1401" (Kryapp 1401, Crypto Modem for serial communication) commissioned by FMV (Swedish Defense) to Sectra [2]?

Fig. 20

The Swedish Defence Materiel Administration (Swedish: Försvarets materielverk, FMV)[3] is a Swedish government agency that reports to the Ministry of Defence. The agency is responsible for the supply of materiel to the Swedish defence organisation. In this case they also provide Kryptomodem 1401 to the Swedish Navy. Notice in Figure 21 that maybe the "Stri 8000" could be identified with the HF-8000 SDR Radio System prioduced by ELBIT (ELBIT Sweden is one of the supplier of Swedish Defence).

Fig. 21



(1) The 80-meter or 3.5 MHz band is a band of radio frequencies allocated for amateur radio use, from 3.5 to 4.0 MHz in IARU Region 2, and generally 3.5 to 3.8 or 3.9 MHz in Regions 1 and 3 respectively. European common allocation is AERONAUTICAL MOBILE (OR) FIXED LAND MOBILE 

(2) It must be said that in many messages the sync sequence seems to be 48 bits starting with [00000000] 0x00

(3) In the context of information theory, Shannon's entropy is a measure of the rate at which information is produced by a source of data. It can be used, in a broad sense, to detect whether data is likely to be structured or unstructured. 8 is the maximum, representing highly unstructured, 'random' data. English language text usually falls somewhere between 3.5 and 5. Properly encrypted or compressed data should have an entropy of over 7.5

(4) 0 represents complete randomness (all characters are unique), whereas 1 represents no randomness (all characters are identical). English text generally has an IC of between 0.67 to 0.78 whereas 'Random' text is determined by the probability that each letter occurs the same number of times as another. A low IC generally means that the text is random, compressed or encrypted. 

(5) Secure text systems enable the armed forces to transmission and receipt military request, reports, orders, plans and policies and can be achieved in different ways, both regarding the format of the message and its method of transmission.

(6) A few words on Swedish cryptology
For two centuries Sweden has upheld neutrality in the sense that the country has not been member of any military alliance. Only recently did the country apply for NATO membership although it has been a public secret that Sweden since the end of WWII has had an informal, but quite strong collaboration with NATO countries, especially the Scandinavian countries including Finland with which strong historical, cultural and linguistic bonds exist. The cooperation has manifested itself in the area of signals intelligence, exchange of intercepted information and interoperability of communications equipment including crypto equipment, and more recently in a close collaboration with NSA.
The credibility of neutrality has been supported by strong defense forces, including cryptologic efforts. A few cases serves to lustrate this: In spite of its neutrality, Sweden was able to just a short time after Germany attacked Denmark and Norway in April 1940 by systematic interception to reading German military telegraph traffic, which transited Sweden, by cracking the so called Geheimschreiber (a kind of substitution cipher where the key was changing with each character and the initialization vector was altered with a few days’ interval) primarily based on the attack by a mathematical genius, Arne Beurling. 
Another feat was the successful evacuation – operation Stella Polaris - of the entire (nearly) Finnish SIGINT staff, their families and crypto analytic material and equipment after the Russo-Finnish war in 1944 ended in Finnish defeat.
Already in 1942, all cryptological and cryptanalytic effort were concentrated in FRA, Försvarets Radioanstalt (Defense Radio Establishment).
Thus, for many years the crypto efforts of Sweden's defence forces and government communications have had a high priority as a national endeavour, mainly due to the centuries long Swedish neutrality, as described above. Today, as Swedish membership of NATO is imminent pending ratification by the Turkish parlament, this collaboration of long duree ensures that the transition from neutrality to NATO mebership will be smooth. There is no doubt that the navy of Sweden in the actual geopolitical situation will play an important regional role in the Baltic Sea and the approaches to this ocean area.
The designations of Swedish crypto equipment is quite straightforward: 

MXY[Z] - Maskinkrypto; Machinegenerated crypto
X - G = Gemensamt (common for all service branches), M - Marinen (navy), L - Flygvapnet (air force)
Y - Typ Y; Type Y
[Z] - Optional designator; Z =  I, for international use, Z = U, for use abroad
The equipment is graded according to its degree of signal protection (Sw., SG, signalskyddsgrad), i.e. SG R(estricted), C(onfidential), S(secret) and T(op) S(ecret).

[1] https://docplayer.net/45022367...internet.html
[2] https://communications.sectra.com/case/encryption-solutions-for-the-swedish-defence/
[3] https://www.fmv.se/english/

17 November 2023

16Kbps/16KHz encrypted wideband FSK (WBFSK)

Unid 16Kbps/16000 wideband FSK (WBFSK) heard and sent me by my friend Martin on 31875 KHz, just a bit beyond the limit of the HF region (LVHF). Notice in Figure 1 the sharp filtering at the edge of FSK transitions.

Fig. 1

The signal has strong 196 ms ACF spykes which are due to the inserted bursts limiting the blocks in which the transfer is arranged (Figure 2).

Fig. 2

The bitstream resulting after demodulation of part of the signal has a 3136-bit length framing, as expected since the 196 ms ACF, consisting of a 191-bit known sequence (the above mentioned inserted bursts) followed by 2945-bit of unknown data; the inserted burst are probably used for (re)synchronization purposes. Despite several runs, I did not find a valid LFSR for that sequence. Also note - zooming the bitstream - further 33 bits that could not be part of the following data block: this way it could be a 192-bit sequence followed by a 32-bit CRC... but it's only a mine speculation!

Fig. 3

Voice or data? The 16 Kbps speed and the way it sounds (like a white noise, to make it difficult to intercept) make me think of a vocoder. Indeed, in many VHF radio stations the analog speech signal is converted into digital form using CVSD (Continuous Variable Slope Delta) [1] just running at 16 kbit/s and then encrypted probably using a VINSON family crypto device such as KY-57/KY-58 [2].
L3Harris RF-7800H-MP could be a possible candidate among other radios since it allows both VHF WBFSK and CVSD vocoder.

Likely, the 191-bit (or 224-bit) inserts are not a CVSD specific feature but are added by the crypto device.


[1] https://en.wikipedia.org/wiki/Continuously_variable_slope_delta_modulation
[2] https://en.wikipedia.org/wiki/VINSON

13 November 2023

a new family of wideband OFDM waveforms

A new family of wideband OFDM signals has recently appeared on-air, most likely tests of Russian origin, and which - at least according to the observations carried out so far - seems to be composed of six waveforms which for convenience I call OWF (OFDM Waveform):

OWF3: 3 KHz bandwidth, OFDM 64-tone
OWF6: 6 KHz bandwidth, OFDM 128-tone  *
OWF12: 12 KHz bandwidth, OFDM 256-tone
OWF24: 24 KHz bandwidth, OFDM 512-tone  
OWF48: 48 KHz bandwidth, OFDM 1024-tone    
OWF96: 96 KHz bandwidth, OFDM 2048-tone

 * the OWF6 waveform has not been observed "directly" but its characteristics and main parameters can be obtained from early-test transmissions [1]

Note that, unlike the western 188-110D App.D (single-tone waveforms, not OFDM), the bandwidths of 9, 15, 18, 21, 30, 36, 42 are missing.
The six waveforms share some common features:

1. initial preamble consists of an LFM segment followed by MFSK-64 37.5 Bd segment
2. MFSK segent is followed by up-chirped FMCW (Frequency Modulation Continuous Wave) sweeps
3. data transfer is arranged in 15 OFDM blocks, separated by chirps as in 2
4. OFDM blocks have all the same speed of 37.5 Baud and same channel separation of 46.8 Hz
5. in some waveforms (OWF1, OWF24, OWF48) the initial MFSK preamble is followed by the same up-chirped sweeps as in 2

It's interesing to note that OFDM parameters fit DRM-B ones.

OWF3 (3 KHz bandwidth)

Fig. 1 - 3 KHz bandwidth waveform

Fig. 2 - 3 KHz bandwidth OFDM

OWF6 (6 KHz bandwidth)

Fig. 3 - 6 KHz bandwidth waveform 

Fig. 4 - 6 KHz bandwidth OFDM

OWF12 (12 KHz bandwidth)

Fig. 5 - 12 KHz bandwidth waveform 

Fig. 6 - 12 KHz bandwidth OFDM

OWF24 (24 KHz bandwidth)

Fig. 7 - 24 KHz bandwidth waveform
Fig. 8 - 24 KHz bandwidth OFDM

OWF48 (48 KHz bandwidth)

Fig. 9 - 48 KHz bandwidth waveform 

Fig. 10 - 48 KHz bandwidth OFDM

OWF96 (96 KHz bandwidth)

Fig. 11 - 96 KHz bandwidth waveform

Fig. 12 - 96 KHz bandwidth OFDM

While the vast majority of Western countries use single tone waveforms (see the aforementioned MS-110D) Russian friends favor the development of multi-tone waveforms, as OFDM in this case: I am not able to make a comparison of performances, there is certainly a valid reason behind preferring one approach or the other.

Thanks to my friends ANgazu, Joni, Linkz and KarapuZ for sending me their recordings.

(to be continued)


[1] http://i56578-swl.blogspot.com/2016/05/cis-ofdm-64-tone-qam-16-40bd.html

11 November 2023

unid MFSK-17 125Bd

Unid (probably Russian) MFSK-17/125 125Bd burst transmission heard on 21292 kHz USB (15mt HAM band) USB at 0830 UTC. 

Fig. 1

Fig. 2

The same waveform was already noted six years ago(!) and reported in this blog:


9 November 2023

Akula 250Bd/500 FSK version

A friend of mine, whom I'm grateful, sent me this interesting and quite rare example of the Akula 250Bd/500 FSK waveform. Transmission was recorded on 9202 KHz around 0800 UTC using a Japanese SDR: as you see, the values of the FSK parameters are the half of the usual ones (500Bd/1000) 

Fig. 1 - Akula 250Bd/500

The demodulated bitstream shows the normal Akula "stuff" (Figure 2):
- reversals
- sync group (6 code words followed by 6-bit "0"s separator)
- preamble group (7 code words with two different, but varying values arranged as 4 x 1st code word + 3 x 2nd code word)
- data block
- End-Of-Message group + EOT group (which never varies and consists of the five code words 101111 100010 100010 101111 011110)

Also notice in Figure 2 the slight difference between the preamble of this sample:
3 x 100101 + 110101 + 2 x 110001
and the characteristic one obtained from the demodulation of the 500Bd/1000 waveform:
4 x 100101 + 2 x 110001
further registrations are needed before we can say that this is the characteristic preamble of the 250Bd/500 waveform.

Fig. 2 - Akula 250Bd/500 demodulated bitstream

It's worth noting in Figure 1,3 the continuous carrier in absence of messages, as already seen in other recordings [1]: probably this is due to the adopted ship-shore "paradigma". While in many of the western navies the shore stations are used to broadcast a list of available listening frequency (FABs/CARBs) (1), it could be that the Russian shore stations transmit a carrier on their known listening frequencies at scheduled times on behalf of subs which have something to comunicate to the shore station itself.  That's obviously my and my frield's guess.

Fig. 3

It's very interesting to note that a day after, and on the same frequency, a short speech was noted: "GREYDER ya DALNIE", more over the Ministry of Defense of the Russian Federation web site reports about an anti-submarine exercise in Peter the Great Bay by the Pacific Fleet just on 8th November (Figure 4) [3]. Note that "Peter the Great Bay" is located in the Sea of Japan, northwestern Pacific Ocean, in the Maritime (Primorye) region of far eastern Russia and that the Akula sample was heard using a remote KiwiSDR in Nagano, Japan. Just a coincidence?

Fig. 4 - https://function.mil.ru/news_page/country/more.htm?id=12484855@egNews


(1) CARB is the acronym for Channel Availability and Receipt Broadcast, these transmissions radiate information on the frequencies available for ship-shore traffic and to pass control and receipt messages; sometimes also indicated as FAB or Frequency Availability Broadcast. These procedures are used to automatically perform a channel-link before a message could be sent [2]. 

[1] http://i56578-swl.blogspot.com/2022/04/akula-quite-unusual-session.html
[2] http://i56578-swl.blogspot.com/search/label/FAB
[3] https://function.mil.ru/news_page/country/more.htm?id=12484855@egNews

26 October 2023

NILE/Link-22, likely QAM traffic waveforms

Two NILE/Link-22 close channels recorded this morning (26th October) on 14.656 KHz and 14.659 KHz using a remote AirSpy server located in Romania [1], traffic in the two channels flowed in alternating mode.

Fig. 1 - NILE/Link-22 transmissions

Link-22 use the TDMA (Time Division Multiple Access) waveforms, as for STANAG-4539 Annex D: modulation technique consists of phase shifting of a 00 Hz sub-carrier and speed of 2400 baud. In TDMA mode each user is allowed to transmit only within specified time intervals named as "Time Slots" so that different users transmit in differents time slots (1). According to S-4539 Annex D, a TDMA slot is the high level structure in which information will be transmitted/received and it is composed of a Preamble, a certain number of Media Code Frames and a Guard Time.


A Media Code Frame is composed of 270 symbols to be transmitted at the modulation rate of 2400 baud and using different Traffic Waveforms and modulations. Each Traffic Waveform is composed of a sequence of different Data blocks and Mini Probe (MP) blocks: the Data block contains coded information symbols and the MP block contains known training symbols to be used by the equaliser. The Mini Probes symbols, prior to scrambling, are all symbols number 0. Since the 270 symbols of the Media Code Frame, regardless the used Traffic Waveform, it's ACF has a value of 112.5 ms.
Figures 2 & 3 show that the two Link-22 channels use two different Traffic Waveforms to arrange the 112.5-ms/270-symbol length of the media code frame; also notice that they use the same time slot duration since both send 15 media code frames per time slot.

Fig. 2 

Fig. 3

Given the poor SNR of the signals I was unable to find clear constellations and therefore identify the type of the Traffic Waveforms that were used; attempts suggest the use of QAM modulations (Figure 4).

Fig. 4

Unfortunately I don't have the 2019 edition of Stanag-4539 but only the first edition dated 2005 which specified only three waveforms (QPSK & PSK8): as you may see, the framings resulting in figures 2,3 do not correspond.


(1) separation among users is performed in the time domain https://en.wikipedia.org/wiki/Time-division_multiple_access

[1] sdr:// 

24 October 2023

CIS VFT 12-FSK async channels system (P-327-12)

Very interesting and rather rare capture sent me by my friend AngazU, of a CIS VFT (Voice-Frequency Telegraphy) waveform which uses twelve  independent FSK channels multiplexed onto a single 3 KHz wide HF channel. The transmission was recorded on 4567 Khz USB , just while monitoring the close well-known "buzzer". Regardless of the nickname we would like to give to these signals, the name of the equipmnet is P-327-12, or in Russian language П-327-12  (thanks to Cryptomaster for the info).
The P-327-12 equipment provides up to 12 channels of Voice-Frequency Telegraphy with a speed of up to 100 baud in one HF channel or 6 channels in wo standard HF channels. The occupied band of each VFT channel is 160Hz. The characteristic (upper and lower) frequencies of a single channel are determined by the formulas:
f1 = 180 + 240n Hz
f2 = 300 + 240n Hz

where f1 and f2 are the lower and upper characteristic frequencies (ie the FSK shift) of the nth channel. If there is no traffic onthe  nth input channel, only the higher frequency (f2) is transmitted. The "pilot tone", as usual, is transmitted on 3300 Hz.

Fig. 1 - CIS 12-FSK spectrum

It's worth noting that the P-327 system is asynchronous(!), i.e. the transmission speed of each channel can be any in the range of 0-100 or 200 Baud. For example, in this sample channels 1 and 3 have a speed of 50 Baud while channel 6 is keyed at the speed of 100 Baud (Figs. 2,3).

Fig. 2a

Fig. 2b
Be careful not to confuse the CIS-12 (AT-3400D equipment) and CIS VFT-12 (P-327-12 equipment) waveforms: although "at first sight" on the waterfall they look like the same signal, they sound different and have a slight difference in the occupied bandwidth; furthermore, while CIS-12 consists of 12 continuous PSK channels, the T-237-12 signal consists of twelve not simultaneous FSK channels (Figure 3).
Fig. 3 -comparison of the spectra of VFT-12 and CIS-12 signals

At least in this sample, the data do not appear to be formatted according to the Russian CIS-11 and/or CIS-14 standards (Figure 4).
Fig. 4 -  CIS-11 & CIS-14 test results
P-327 system was adopted as weapons to replace the P-318M voice-frequency telegraphy equipment. Thanks to the use of modern digital components and new technical solutions, it was possible to improve the quality of telegraph communications, simplify the process of operating equipment, and reduce weight and size indicators.

Fig. 5 - P327-12 system: note the 12 "slots" of the modem (https://infopedia.su/18xc490.html?ysclid=lo41wlbhdo570983862)