Weak-Signal Modes & FT8 — Vol 2: The Rest of the Family, and the Argument

2.1 Each mode is a purpose

FT8 gets the attention, but the WSJT family is a toolkit, and each tool was shaped for a specific channel. That is the question this whole sub-project keeps asking — what was this designed for, and what did it give up? — and here the answers are unusually crisp.

2.1.1 WSPR — the mode that is not a QSO mode

WSPR (Weak Signal Propagation Reporter, and everyone says “whisper”) is the odd one out, because you cannot have a conversation on it at all. That is not a shortcoming; it is the design.

A WSPR transmission is 110.6 seconds long inside a two-minute slot, carries a payload of just callsign, grid square, and transmit power, occupies about 6 Hz, and decodes down to −31 dB in a 2500 Hz reference bandwidth. Every receiving station that decodes you uploads the spot to a central database at wsprnet.org, along with the time, the frequency, and the signal-to-noise ratio.

The result is a planet-scale, continuously updated map of what the ionosphere is actually doing, built by amateurs, for free. Run a WSPR beacon at a couple of hundred milliwatts overnight and in the morning you will have a map showing which continents heard you, when the path opened, when it closed, and how strong you were at each hour. It is genuinely the best propagation instrument an amateur has ever had, and the fact that a 20 mW beacon on 30 metres routinely gets reported across an ocean is one of the more startling facts in the hobby.

I think WSPR is the most quietly valuable thing in this dive, and it dodges the entire argument at the end of this volume — nobody claims a beacon is a conversation, so nobody is disappointed.

2.1.2 MSK144 — talking off meteors

There are always meteors entering the atmosphere. Each one leaves a column of ionised air that will reflect VHF signals — for somewhere between a tenth of a second and a couple of seconds. Meteor scatter is the art of getting a message through in that window.

⟨correction⟩ Ward’s article names the mode “MFSK441.” The mode was FSK441 — Joe Taylor, QST, December 2001 — four-tone FSK at 147 characters per second, designed to blurt a message into a ping. And FSK441 is now obsolete: it was superseded in 2017 by MSK144, which runs about 4.5 times faster, adds forward error correction, a CRC, and frame averaging that combines multiple short pings into one decodable message. MSK144 is dramatically more sensitive and works off pings so short that FSK441 never saw them. FSK441 has been removed from the WSJT-X mode list entirely. Two errors in six words — the wrong name for a mode that no longer exists — which is exactly the sort of thing that happens to a nine-year-old article in a fast field, and exactly why this ledger exists.

(And note the name: MSK, minimum-shift keying — the same tail-eating modulation that ended The Shift-Keying Family, the one that is an FSK and an offset-QPSK simultaneously. Its constant envelope and compact spectrum are exactly what you want when you have 200 milliseconds of ionised meteor trail to work with.)

2.1.3 Q65 and JT65 — the Moon, and everything that scatters

JT65 was the moonbounce mode and still has a following there. Q65, introduced in WSJT-X 2.4.0 in 2021, is its successor and a considerable improvement: 65-tone FSK built on the QRA64 work of 2016, with a strong low-density parity-check code. It is aimed squarely at the ugliest propagation there is — tropospheric scatter, rain scatter, ionospheric scatter, trans-equatorial propagation, and EME — all of which share the property of being not merely weak but fast-fading and Doppler-smeared, which defeats modes that assume the signal sits still.

The EME numbers still stop me. The Moon is a poor reflector at a quarter-million miles; the round trip costs on the order of 250 dB. Before these modes it took very large arrays and a great deal of patience. Now a single long Yagi and 100 watts on 2 m will get you contacts. That is not incremental — it is an entire category of operating handed to people who could never have attempted it.

2.1.4 FT4 and FST4 — the two directions from FT8

FT4 (2019) goes faster: 7.5-second slots, about −17.5 dB, built specifically for contesting, where the exchange is short and the rate is everything. Trade sensitivity for speed, again.

FST4 and FST4W go the other way entirely, to the far end of the ladder — optional sequence lengths up to 30 minutes, reaching thresholds as low as −45 dB. Half an hour to transmit a beacon message. They are for LF and MF, the 2200 m and 630 m bands, where the noise is appalling, antennas are hopeless, and signals are almost theoretical. Forty-five decibels below the noise. That is a signal roughly one thirty-thousandth the power of the noise it is buried in.

2.1.5 JS8Call — getting the sentences back

The most interesting response to FT8’s constraints did not come from Princeton. Jordan Sherer, KN4CRD, asked in February 2018 whether long-form messages could ride on FT8’s engine, published a design in March, and released the first version in July 2018. It was originally called FT8Call and was renamed JS8Call — JS8 being his own variant of the 8-FSK modulation, plus a directed-calling protocol.

JS8Call keeps FT8’s extraordinary sensitivity and throws away the fixed message set. You can type sentences. It is slow — you are still moving a handful of bits per second — but you can hold a real conversation, relay through other stations, leave messages for stations that are not on, and run it as a genuine low-power text network when nothing else will pass traffic. For emergency communications and for people who wanted the sensitivity without the ritual, it is the answer, and it deserves far more attention than it gets.

2.2 The argument

Now the fight, and I am going to try to land it fairly, because I have been on both sides of it.

The case against FT8 is not stupid and it is not merely nostalgia:

• The operator is barely present. The software calls CQ, answers, sends the report, acknowledges, logs, and moves to the next station. In auto-sequence mode a station can work for hours with nobody in the chair. Two computers exchange four-character grid squares while the humans do something else.
• The exchange contains nothing. A callsign, a grid, a number. You learn nothing about the person, their station, their weather, their name. There is no conversation to be had — the protocol has no room for one.
• It has emptied the other modes. This is the complaint I hear most and the one that has some real force. Tune 20 m on a weekday evening: the FT8 watering hole at 14.074 is a solid wall of signals, and there are stretches of the CW and SSB portions where you can hear the band noise. Activity moved, and some of it did not come back.
• It doesn’t need much skill. Ninety percent of the work — the hard part, the decoding — is done by an LDPC decoder written by a Nobel laureate. You did not do it.

The case for FT8 is also not stupid:

• It gave the hobby back to people with bad stations. This is the argument that matters most to me. If you live in an apartment, or under a homeowners’ association that forbids antennas, or in a valley, or with a neighbourhood full of switching-supply hash, FT8 is the difference between working the world and working nobody. A wire in an attic and 25 watts will get you DXCC on FT8. It will not on SSB.
• It rescued the bottom of the solar cycle. During the deep minimum of 2018–2020 the higher bands were, by any conventional measure, dead. FT8 kept them usable — 10 metres produced contacts when SSB heard nothing at all, because a mode working 30 dB further down finds openings that do not exist for anything else.
• It is a genuine engineering triumph. Pulling a message out of noise 125 times stronger than the signal is not a trick or a cheat. It is superb work, freely given away, by someone who could have patented it.
• And the “not real operating” complaint is very old. The brass pounders said it about the bug (Semi-Automatic Bugs). The bug operators said it about the electronic keyer (Paddles & Electronic Keyers). Everybody said it about the teletype — the machine does the sending, where is the skill in that? — and this project has already documented that argument twice. Every single automation step in the history of this hobby has been met with the accusation that it is not really operating. That does not make the FT8 complaint wrong. It does mean it is not new, and that the people making it are standing exactly where their grandfathers stood.

Where I actually land. I am a CW operator. My favourite object in this entire collection is a chrome art-deco keyer from the 1970s, and the thing I like best about radio is a well-sent fist at the other end and a conversation with somebody in it. FT8 gives me none of that, and on the evenings when the CW portion is thin I do resent it a little.

And I run it anyway, because it is honest about what it is. It is an instrument, not a conversation — closer in spirit to WSPR than to a QSO. When I want to know whether 10 metres is open to Africa right now, FT8 tells me in ninety seconds, definitively, and no amount of calling CQ on SSB will. When I want to talk to someone, I reach for the key.

Both of those are radio. The mode is a tool, and the question that matters is the one this sub-project has asked from the beginning: what was it designed for, and what did it give up? FT8 gave up the conversation, deliberately, and bought 30 decibels with the proceeds. Whether that was a good trade depends entirely on what you came to the hobby for — and a hobby big enough to hold both a camelback telegraph key and an LDPC decoder is doing fine.

---

2.2.1 Sources (Vol 2)

• H. Ward Silver, N0AX, “Wireless Modes — Part 2,” Nuts & Volts, May 2017 — the WSJT modes, “moonbounce with modest equipment,” meteor trails as reflectors, and communicating “at signal strengths far below the noise.” 🔴 Names the meteor mode “MFSK441”; see the correction above. https://www.nutsvolts.com/magazine/article/wireless-modes-part-2
• WSJT-X User Guide (K1JT et al.) — WSPR at −31 dB in 2500 Hz with two-minute sequences; FST4/FST4W with sequences up to 30 minutes reaching −45 dB; MSK144 and the 5/10/15/30 s sequences; Q65 for “tropospheric scatter, rain scatter, ionospheric scatter, TEP, and EME.” https://wsjt.sourceforge.io/wsjtx-doc/wsjtx-main-2.6.1.html
• 🔴 FSK441 → MSK144: FSK441 introduced in J. Taylor K1JT, “WSJT: New Software for VHF Meteor-Scatter Communication,” QST, December 2001 (4-tone FSK, 147 char/s). MSK144 developed 2017, ~4.53× the baud rate, with FEC, CRC and frame averaging for pings longer than 100 ms; FSK441 is now obsolete and no longer listed in WSJT-X. https://k5nd.net/2020/10/msk144-vs-fsk441-meteor-scatter-modes-my-scattered-compilation-of-data-points/
• Q65: introduced in WSJT-X 2.4.0 (2021), 65-tone FSK building on QRA64 (2016). ARRL, “WSJT-X 2.4.0 Introduces New Digital Protocol Q65.” http://www.arrl.org/news/wsjt-x-2-4-0-introduces-new-digital-protocol-q65
• FT4: 7.5 s T/R sequences, introduced for contesting in WSJT-X 2.1.0 (2019).
• JS8Call: Jordan Sherer KN4CRD — design published 4 March 2018, first release to the development group 6 July 2018; originally FT8Call, renamed JS8Call (JS8 modulation + directed calling). https://js8call.com/
• WSPR spots and propagation mapping: https://wsprnet.org
• ⚠ The claim that FT8 activity has drawn operators away from CW and SSB is widely reported and matches my own listening, but the quantitative picture depends heavily on which spotting network you count and on the solar cycle. ⟨stated as an observation, not a measurement⟩
• Cross-links: Semi-Automatic Bugs and Paddles & Electronic Keyers (the same “that isn’t real operating” argument, two generations earlier); Teletype Machines (and a third time); Building a Wireless-Modes Bench (running WSJT-X and a WSPR beacon yourself).