31 December 2022

a curious AT-3400D/AT-3104 (CIS-12) modem configuration

Interesting catch of an AT-3400D modem (also known as CIS-12 or MS-5) running on 14344.0 KHz/USB with the quite uncommon 9 channel configuration (3-out-of-12) as shown in Figure 1 below

Fig. 1

CIS-12, as you know, is a pseudo OFDM 12-tone (+ 1 pilot) waveform using PSK2 or PSK4 modulation at speed of 120 Baud while the modem name is AT-3004D (or its newer counterpart AT-3104). Channels 1-10 are used for data, 11 and 12 are test/service channels, therefore the "aggregate" speed is 1200 Baud. 

Fig. 2


9 December 2022

WALE & wideband traffic, probably RapidM proprietary WB-LDL & WB-RDL waveforms

My friend ANgazu and I are almost done going through nearly 30GB of recordings of broadband transmissions that took place in the portfion of 6.9-7 MHz band during October and November, monitoring was also possible thanks to SM5TAH/Mats who made available its broadband SDR receiver (Airspy HF+, SDR-IQ). As already discussed in previous posts, we think they were trial transmissions aimed at fixing the performance of modified waveforms with respect to the relative reference standards (118-110D App.D and 188-141D App.G) especially with regard to the fourth-generation ALE. Some of these tests were most likely conducted by Harris, as they more specifically concerned the 3GWB ALE (WBALE) and WHARQ broadband waveforms, although we did not find news or information about.

We also found 188-141D 4G-ALE (WALE) waveforms that use a modified/improved initial preamble that are probably traceable to trials by RapidM [1]: also in this case no direct confirmation other than a presentation given at an HFIA Industries meeting on March 2020. The traffic waveforms following those WALE handshakes are totally "new" (or at least never meet before for me) and defintely not 188-110D App.G compliant. Obviously, since these waveforms are used in the WALE trials they are developed, along with the related modem, by the same manufacturer.

bandwidths. We could detect waveforms spreading from 3 Khz to 48 Khz, their allocation within a wideband channel is negotiated during the WALE link setup stage and follows the specifications required by 188-141D (1). The examples in Figures 1a,1b show the wideband channel allocations of some of the monitored waveforms.

Fig. 1a

Fig. 1b

paradigma. The first consideration to make is that the traffic segments look like a ARQ system, but after the last data transfers the responder' ACKs are missing (Figure 2), maybe they are forward/reverse traffic links or the latest ACK is not required/mandatory by the system. Also notice that, although the wideband channel has been negotiated, the called station uses a different channel for its sendings, this channel is anyway "within" the one of the caller. Probably the responder announces it's own 16-bit sub-channel vector in the WALE confirm PDU.

Fig. 2

modulation. All the waveforms utilize eight-ary phase-shift keying (PSK8) constellation, the symbol mapping is the same used in 188-110D (Figure 3). As usual, the modulation rate varies according to the bandwidth.

Fig. 3

waveforms. According the used framing (ACF of 25.4 and 120 ms), we identified 2 families of waveforms, each consisting of 9 waveforms differing in bandwidth (3-24 and 48 KHz) and modulation rate (2400-19200 and 38400 Baud): unfortunately, the waveforms of the intermediate badwidths 30,36,42,.. KHz have not been found and therefore (hopefully at the moment) are missing.

25.4 ms framing. The frame structure consists of a "Tx Frame" consisting of a synchronization preamble followed by N "data packets" of fixed duration of 25.41 ms (except the 7200 Bd waveform), where N depends on the chosen waveform. Each data packet consists of alternating data (Unknown) and probe (Known) symbols. The frame structure is shown in Figure 4.

Fig. 4

A transmission consists of 1 or 3 Tx Frames separated by a "dead time" or a "guard interval", only the the first Tx Frame is preceeded by a TLC sequence (Figure 5).

Fig. 5

Main features of this family waveforms (so far seen) are sumarized in Table I.

Table I

The 25.4 ms periodicity of the data packets and their number N within each Tx frame can be seen in Figures 6 and 7 respectively.

Fig. 6

Fig. 7

However, I'm facing an odd problem about the length of the (known symbols) probes; let's see for example the 2400 Bd data packet: as from Figure 8, the 9.996 ms probe makes a length of 24 PSK8 symbols but after its demodulation the bitstream shows 75-bit lenght patterns, ie 25 PSK8 symbols! 

Fig. 8

The discrepancy is probably due to the SA PSK demodulator. Indeed, it must be noticed that either the preamble sequences and the probes are characterized by the lack of the sub-carrier in their harmonic spectrum (Figure 9): this feature has already been found in the modified/enhanced WALE preambles discussed in the previous post.

Fig. 9


120 ms framing. The frame structure consists of a TLC & synchronization preamble followed by fixed length 120 ms data packets, each data packet consisting of alternating data (Unknown) and probe (Known) symbols. The most interesting aspect is that the initial TLC & preamble sections consist of the same Tx Frame of the correspondent 25.4 ms family waveforms (Figure 10): it could be said that both waveforms (25.4 ms and 120 ms) present the same "front-end" to the receiving modem.

Fig. 10

The complete frame structure is shown in Figure 11.

Fig. 11

Figure 12 shows the 120ms period of some waveforms, unfortunately not all the samples have a good quality.

Fig. 12

The main features of this family waveforms (so far seen) are sumarized in Table II. It's interesting to see what came up when comparing the values of Table II with the correspondent values of the Waveform Number 7 of 188-110D App.D: as you see, most of the framings (ie the number of the PSK8 symbols) match the ones of the correspondent 188-110D waveforms. 

Table II

Although in some cases the durations are the same, the mini-probes sequence is different so there is no compatibilty between the observed waveforms and 188-110D. Figure 13 shows two frames of the 12 KHz 9600 Bd waveforms.

Fig. 13

RapidM Since these traffic waveforms are used together with the modified WALE waveforms seen in the previous post, it is logical to assume that they too are developed by the same manufacturer, ie - in our opinion - by Rapid Mobile (RapidM) [1]. A positive feedback comes from reading the technical documentation of their RM10 modem [2] "In addition, the RapidM proprietary wideband HF packet data modems known as WB-RDL is offered in the RM10 as a software option.  The WB-RDL waveforms and protocols are integrated with the WALE/4G ALE controller for link setup and dynamic forward/reverse traffic channel bandwidth negotiation. The combination of WALE/4G ALE with the WB-RDL provide data communication in challenged channel (e.g., SNR < 0 dB, high-latitude, high interference) conditions". More precisely, the documentation talks about the future WB-LDL & WB-RDL packet waveforms used with WALE: the acronyms could mean WideBand Low latency DataLink and WideBand Robust DataLink, that is the two families we have singled out (25.4 & 120 ms waveforms).

Fig. 12

We have quite a lot of assumptions and conjectures about it but at the moment we prefer to wait for "news" from RapidM on their official website (new RM12 wideband modem?) and monitor the HF bands for (hopefully) such new trials.


(1) the wideband channels are described in WALE PDUs using 16-bit “sub-channel” vectors, each bit element of a sub-channel vector refers to a sub-channel within an assigned wideband channel and describes 1.5 kHz sub-channels, range of 24 kHz, or 3 KHz sub-channels, range up to 48 KHz

[1] https://i56578-swl.blogspot.com/2022/11/an-enhanced-design-for-deep-fast-wale.html
[2] https://www.rapidm.com/product/rm10-wideband-software-defined-modem/