RadCom April 2024, Vol. 100, No. 4

Regulars 28 April 2024 Data modes used at LF Transmitters on the LF bands are mostly switched- mode class-D or class-E designs that can only transmit a constant-amplitude signal. The amplifier itself, or the drive, can be keyed to obtain on-off CW, but avoiding key clicks requires controlling the power supply input to the PA as it can’t be done on the low-level input. The first ‘computer-generated’ data mode to be used on the original 73kHz band was ‘SlowCW’, now more-usually called QRSS. The signal is keyed on and off using Morse code at a very slow rate, far lower than suited to listening. A waterfall-type spectral display is used to display, such as in the example shown in Figure 1 of a 73kHz transmission made in August 1997. This was the first-ever use of ‘SlowCW’, and was transmitted one way from G4JNT on the South coast of England to G3PLX in the Lake District, a distance of 393km, using a dot period of 40s. Subsequently, several software authors produced QRSS software, using both direct on-off keying and with a soundcard-generated tone. A variant of QRSS, called dual-frequency CW (DFCW), was also developed where the dots and dashes were each of the same length, but at two different frequencies. This required a bit of mental agility to read off the screen, but became easier after practice. QRSS and DFCW are still used today, although software to transmit them seems a bit scarce. A 10-baud hard-switched phase-shift-keying (PSK) mode, called ‘Coherent’, developed by Bill de-Carle, VE2IQ, was used for a while by a few users on 73kHz, since hard-switched PSK can be passed though switch-mode transmitters. ‘Coherent’ soon showed the power of computer-generated data modes, being appreciably better than CW received by ear. PSK modulation at 10 baud was then used in the ‘Wolf’ mode that adopted heavy forward error correction (FEC) by virtue of interleaving six convolutionally-encoded data streams, and adding an additional 100 per cent clock/synchronisation overhead. This mode enabled the first transatlantic 73kHz QSO to be made by G0MRF. The mode was developed further into a soundcard mode called ‘Africa’ but, as with any phase-shift-keyed mode, it required amplitude shaping to minimise the key clicks caused by hard-switching the phase, so was never properly suited to the common amplifier types used at LF/MF. This anti-social characteristic is what probably prevented widespread adoption of either ‘Wolf’ or ‘Africa’ by many users; QRSS was still the weak-signal mode of choice for a long time. The first multiple-frequency-shift (MFSK) mode to appear, sending one tone at a time, was ‘JASON’, written by ZL1BPU. It used 17 tone frequencies spaced by 0.25Hz, encoding four bits at a time per shift. This allowed narrow and clean signalling, since frequency-shift transmissions can be made using constant amplitude. To cope with frequency instability and tuning errors, differential tone coding was used, where the information was encoded by the change in tone from one symbol to the next. It was a neat idea, but with no FEC the mode never came into widespread use. A few trial versions of other modes with FEC were written by a few individuals, but nothing really took over until ‘weak signal propagation reporter’ (WSPR) first appeared from Joe Taylor, K1JT. A version of ‘Hellschreiber’, using sequential multiple tones to allow for non-linear transmitters, was tried for a while, and an example of that can be seen in Figure 2 . The state of things today WSPR, now incorporated into the WSJT-X suite, is used widely on 137kHz and 475kHz for propagation tests, with reports being uploaded to a central database. WSPR is a beacon mode, so not suited to making QSOs, which were still being done using QRSS or one of the other less-used modes. The FST4 QSO mode was introduced into WSJT-X with a choice of several low-rate sub- modes specifically designed for use at LF. FST4 uses MFSK with a shaped frequency shift [1] to minimise spectral bandwidth, and has both a WSPR-like beacon mode, referred to as FST4W, and a QSO mode. Both of these have a better FEC scheme, offering 2dB more sensitivity than WSPR for the same speed, but also have options of Tx/Rx periods of 2, 5, 15 and 30 minutes. The fastest has the same duration as WSPR. The 30 minute, or ‘-1800’ sub mode, has a symbol length of 11s, corresponding to a signal bandwidth of 0.09Hz, making it easily the most sensitive of all the WSJT-X modes, with a threshold signal-to-noise ratio of around -43dB referred to 2.5kHz bandwidth. There doesn’t appear to be that much activity on 137kHz using the two slowest sub modes for some reason, perhaps because the 137kHz transatlantic path (the holy grail) is not usually stable enough for this very narrow mode, or perhaps users have difficulty maintaining frequency stability. G4JNT has made it across the Atlantic on 137kHz just once using FST4W-1800, and that required a sked to be set up beforehand, and probably a fluke propagation enhancement. There is still a lot of WSPR activity, in spite of FST4W being more sensitive and cleaner with reduced sidebands and offering a choice of speeds. An alternative beacon mode, ‘Opera’ [2] , has a few firm adherents. This mode uses on-off keying with options for a variety of symbol rates, incorporates heavy FEC, and the transmission consists of just a callsign with reports uploaded to a central database. That is the state of the LF bands today; there is a lot of activity in beacon mode using WSPR/FST4W, with two-way QSOs using FST4, QRSS, DFCW and occasionally some other mode just to break the cycle. The ubiquitous FT8 has been seen – presumably because it’s ‘what everyone uses’, but is not suited to LF as its 50Hz bandwidth is a bit antisocial. JT9A (6Hz bandwidth) is used by a few stations, but the convolutional encoding means that its FEC is less powerful than that in FST4, and hard-switched MFSK tones spread a bit wider, leading to the potential for more QRM. There is, of course, normal CW operation but, as it is a human-decoded mode like voice, it does not fit the remit of this column. References [1] Data, RadCom December 2020, describes the introduction of FST4. Its encoding on microcontrollers is covered in the June 2021 column, and its use for spectral spreading tests at HF in the April 2023 edition. [2] ‘Opera’: Data, RadCom April and June 2012, with an update in April 2014. Data Andy Talbot, G4JNT andy.g4jnt@gmail.com FIGURE 1: A spectrogram of the first-ever QRSS, or ‘SlowCW’ transmission, made on the 73kHz amateur- radio band in August 1997. The dot period was 40s, the transmitted power from G4JNT was around 1mW ERP, and the signal was received at a distance of 393km by G3PLX with what appears to be about a 3dB-5dB signal-to-noise ratio. FIGURE 2: A sequential multi-tone ‘Hellschreiber’, or ‘SMT-hell’, signal. Sequential tones, although resulting in sloping characters, allows a non-linear transmitter to be used.