RFSM-8000

from http://www.marsregionone.org/Temp/rfsm-8000.pdf
"The RFSM-8000 (Radio Frequency Software Modem) is a software by RFSM-IDE Group, it operates under standards MIL 188-110A / MIL 188-110B App. C, (also under a modified version) providing a maximum speed of 8000bps (standard mode), or 6670bps (non-standard "narrow" mode). Signal bandwidth: 0.3-3.3 kHz (standard mode) and 0.3-2.7 kHz (non-standard mode).

Some technical descriptions: 
- real-time signal spectroscope/waterfall view, and IQ-diagram;
- operates under standards MIL-STD 188-110A / MIL-STD 188-110B App. C, (also under a modified version);
- maximum speed: 8000 bps (standard mode), or 6670 bps (non-standard "narrow" mode);
- signal bandwidth: 0.3-3.3 kHz (standard mode) and 0.3-2.7 kHz (non-standard mode).
- uses adaptive correction;
- file transfers are accomplished utilizing ARQ;
- uses SSE2 optimization (if available on CPU);
- transmitter control over COM-port (DTR or RTS line, or CI-V interface);
- allow correction of sound card discretization error;
- allow simple remote control and file-based IPC (for automatic connecting and file transfer)"

I did some tests to check the MS188-110 compatibility and their differences, below are shown the results

- RFSM-8000 modem settings: no DataMasking, standard mode, synchronous
- text sent: "Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua."

- decoder used: k500 in both MS188-110A and MS188-110B/App.C synchronous mode

75, 300 bps
2400Bd PSK-8 preamble, headers MS188-110 not compatible
- 75bps: special very robust mode (32-Walsh-based), works under very poor conditions ( -8 dB ), data PSK-8, ACF ~13.6ms (32 symbols)
-300bps: special robust mode - 300bps (16-Walsh-based), works under poor conditions ( -3 dB ), data PSK-8, ACF 13.6ms

About the robust modes the spectrogram shows a sort of inserts with irregular repetition interval, probably due to Walsh modulation, which makes the corrispondent ACF spikes. The 32 tribit symbols, i.e. 96 bits, are clearly visible in the "period" highlighted by BEE (pic. 2b). Standard MS188-110 doesn't exhibit such ACF for the 75bps waveform (pic. 2). 
75bps and 300bps are a sort of proprietary waveforms and need more investigations.

RFSM-modem -> k500
75 long  -> 4800 uncoded (not recognized)
300 long -> 2400 voice (not recognized)

pic. 1

pic. 2

pic. 2b - 96 bits ACF in RFSM-8000 75bps as seen by BEE

600, 1200, 2400 bps

2400Bd PSK-8 preamble

- 600bps: data PSK-8,  ACF = MS188-110A (pic. 3)
- 1200bps: data PSK-8,  ACF = MS188-110A
- 2400bps: data PSK-8,  ACF = MS188-110A
source text is not clearly returned (7 bit shift ?) 

RFSM-modem -> k500 

600 long -> 600 long
1200 long -> 1200 long
2400 long -> 2400 long 

pic. 3 - RFSM-8000 2400bps ACF

3200 bps

2400Bd PSK-8 preamble, headers not MS188-110A compatible

data PSK-8,  ACF = MS188-110A

RFSM-modem -> k500
3200 long -> 75 long (not recognized)

4800, 6400, 8000 bps (MS188-110B/App.C High-Speed WF)

(4800bps only uncoded in MS188-110A, coded in MS188-110B/App.C )
2400Bd PSK-8 preamble

- 4800bps: data PSK-8,  ACF ~119.5ms or 287 symbols frame (pic. 4)
- 6400bps: data QAM-16, ACF ~119.5ms (pic. 5)
- 8000bps: data QAM-32, ACF ~119.5ms (pic. 5)

RFSM-modem -> k500
4800 long -> 4800 long
6400 long -> 6400 long
8000 long -> 8000 long
 

pic. 4 - RFSM-8000 4800bps frame

pic. 5 - RFSM-8000 6400bps, 8000bps

As pointed and tested by AngazU, more likely RFSM chops the info into 59 byte groups and adds its own 6 bytes to build 65 chars chunks for a total of 520 bits frame. The example in pic. 6 is a null file sent by RFSM-8000 at 1200bps: he kindly gave me this file so I could replicate his investigation and get the same results:

pic. 6 - RFSM-8000 data framing (do not mix up with protocol framing)

As you can see in pic. 7, plain MS188-110A doesn't exhibit such beavior since each "period" is just 8 bit length.

pic. 7 -  8 bits period of MA188-110A

an RFSM network on 9050.0 KHz (almost surely Bulgarian Diplo)

I followed these signals on 9050.0 KHz USB mainly in the morning and during the last days, from 0700 to 1100 UTC. Looking at UDXF logs and talking with some friends of mine, the network using 9050.0 Khz is probably Bulgarian or  Romanian Diplo (since some ALE calls such as RETEA, FARAONRETEA, SC1DSRETEA), the QRX is visible on 9055 kHz but sometimes only one frequency may be used. Is not sure if also a similar signal heard on 9040.0 KHz belongs to the same net. My friend Kristian from Germany suggested to be Bulgarian Diplo net and kindly pointed a Monitoring Times issue (May 2011) talking about this mode and Bulgarian Diplo activity:

Transmissions on 9050.0 seem to be scheduled at 0830, 0900 and 0930 UTC Mon-Fri, almost all the sessions start with a ~ 15sec tone at 9051.8 KHz (sent about two minutes before) and they last the time necessary for interchanging/sending the messages, usually few minutes.
From previous recordings they used plain MIL-STD 188-110A and now they switched to RFSM (Radio Frequency Software Modem) with Data Masking (link protect) feature, an MS188-110 ST based software-modem by "RFSM-IDE Group". 
Carrier frequency (1800Hz), symbol speed (2400Bd) and modulation (PSK-8)  are the same as that standard (pic. 1) and may be misunderstood and wrongly identified as MS188-110 4800bps/uncoded (the RFSM 75bps/L) and 2400bps/voice (the RFSM 300bps/L), as it happens with both k500 and sorcerer; the 13ms ACF is a clue in favor of RFSM (pic. 2). The structure of the signal is quite complicated, the PSK-8 preamble is followed by the data transfer in PSK-8, QAM-32 or QAM-16 according to needed data-rate. Just as aside note, a sort of 2-ISB signal + carrier has been seen on 9050.0 as central frequency: signal is very week and is difficult to say it's nature.

The RFSM-2400/8000 is a software by "RFSM-IDE Group", it operates under standards MIL 188-110A / MIL 188-110B App. C, (also under a modified version) providing a maximum speed of 8000bps (standard mode), or 6670bps (non-standard "narrow" mode). Signal bandwidth: 0.3-3.3 kHz (standard mode) and 0.3-2.7 kHz (non-standard mode).

http://www.marsregionone.org/Temp/rfsm-8000.pdf 

pic. 1

pic. 2