Thursday, April 09, 2026

The pair of mystery strong FSK signals on 40m

There's been a bit of discussion at our radio club about the origin of the pair of FSK signals on the 40m band in recent months. Here's how they look and sound:


They sounds a bit like morse code but as you listen you soon notice that they aren't. One of our members was convinced, due to the strength, that they come from Australia. I did some internet sleuthing and the conclusion seems to be that they come from the hotspots on this map:


From Google’s AI overview:

These mystery FSK signals on the 40m ham band (approx. 7.061–7.068 MHz and 7.100+ MHz) are persistent, paired, non-amateur, wide-shift (~7 kHz) FSK signals likely originating from Eastern Europe/Russia, potentially related to regional conflicts. They resemble slow, garbled CW (20 baud) but are actually digital, military-style 5N1 encrypted data. 

Key Details of 40m Mystery FSK Signals:

·         Location & Characteristics: Usually found as a two-channel, high-power system with a 7 kHz separation (e.g., pairs near 7.0615/7.0685 MHz or higher in the band).

·         What they sound like: They sound like slow, rhythmic, or "gibberish" Morse code to the ear, but common CW decoders will not read them properly.

·         Likely Origin: The signals are believed to be related to military, government, or state-level monitoring, possibly coming from the Sevastopol/Crimea region or near the Russia/Ukraine conflict area.

·         Modulation: They are 20-baud FSK (Frequency Shift Keying) with a 7,000 Hz shift, often characterized by a constant, repeating 255-bit, 17-bit frame, 15-frame superframe, designed to appear like noise or to be jamming-resistant.

·         Timing: They appear as "key-down/key-up" pairs, meaning when one frequency is transmitting, the other is silent, suggesting a 2-channel diversity system.

·         Distinction from Amateur Activity: While some users initially suspected Winlink (a, b) or heavy traffic, these signals are widely considered "intruders" and not standard amateur digital modes like FT8 or RTTY. 

These signals are a well-documented annoyance to amateur operators in Europe and North America and have been appearing in bursts, especially between 2023 and 2026. 

On Facebook: https://www.facebook.com/groups/hfunderground/posts/1377589986956733/

What is this random strong CW signals appearing in 40m band simultaneously in three frequencies like 7050 7065 7070. We believe it may be from Russia.

These are two fsk transmissions with 7 kHz shift and 20 baud. They change frequency after about 10 minute transmission with 1.5 kHz. 

It is Iranian.. encrypted message system built by the Russians. been tracking it for weeks. It is used for sleeper cells to give them instructions. And you can take that to the bank. Jack 

Also on Facebook: https://www.facebook.com/groups/22212809595/posts/10163697406274596/

For the past week+ there has been some intruder on the 40M CW band easily heard on both coasts of North America. 

It is occasionally QSY'ing anywhere from 7.04MHz to 7.07MHz and may be outside that range goo. It is often on a whole multiple of 500Hz.

Although we are most often hearing one CW frequency with our rigs set at a normal bandwidth, I can often find an "evil twin" which is key-down while the other one is key-up, so I think it's more likely to be super-wide shift FSK. Below for example is me putting receiver 1 on 7.063MHz and the other on 7.070MHz circa 0400Z on Jan 30 2026. You see when one is key up the other is key down.

The basic signal rate seems to 0.05seconds (20 Baud). Our CW-seeking brains seem to try to interpret it as CW because if we here one element at 0.05seconds we think it's a dit and another at 0.15 seconds our brains.

I do not recognize 7kHz shift 20 baud FSK as a standard encoding anywhere but someone might know better than me.

I'm gonna record some longer patches and see if I can recognize any framing.

Tim N3QE

From SOTA Reflector: https://reflector.sota.org.uk/t/some-weird-cw-found-on-40m/38289

While trying to impove my indoor loop on 40-80m (coil added in a middle makes a noticeable difference on both TX/RX!) I’ve came across those strange CW signals. Perhaps they are trying to emulate a mechanical key with no audio control (although a vertical wouldn’t sound like that) But what could be a purpose? Another weird thing is that those sigs are present on four frequencies simultaneously with S9+:

7041.50 kHz

7048.50 kHz

7063.00 kHz

7070.00 kHz

Weird CW on 40m 

Is that something known or a new thing? There is plenty of strange stuff on air now, but I haven’t heard such lovely Morse yet :slight_smile: 

Also

Cheers David, using that “wide-band FSK” hint I’ve found another Youtube sample and also some explanation, which I quote here for reference:

https://youtu.be/zRho2iFfBCE?si=d6cvedzhd65ZZoJs 

“Sevastopol; Extremely powerful two channel 20b/7000 Hz FSK paired signals re-appear from Apr 23rd on 40m band ~ 7045cf and 7065.5cf, each channel seems to be 7 kHz from mark and space, chosen for it’s immunity from being easily jammed, at times the pair will shift frequency; originating from Eastern EU / Western RU region, now Crimea with transmitters near Sevastopol. The modulation rate of 20b with a shift of 7000 Hz. -The signal frame is 17 bits, the first of which is a 0. -The superframe is 15 frames (255 bits) and is constantly repeated so it does not appear to carry any useful information Military encryption 5N1 use in related signal connection to the conflict between Ukraine and Russia is likely. Contributor ANgazu thank you for your analysis! Recorded 12 June 2025 by Tyler Stampfli, KA0KA …”

Wednesday, April 08, 2026

Tech talk on ABC Radio - in space you still need IT to fix Outlook

As Artemis II makes its lap around the moon, astronauts on board keep in touch with their control room and their loved ones using a different type of internet on NASA's Deep Space Network.

ACMA announces new rules for telco companies surrounding their mobile coverage maps ensuring clearer and more accurate results.

Peter Marks, mobile software developer and technology commentator from Access Informatics, joins Suzanne Hill with the latest tech news. Listen here: https://www.abc.net.au/listen/programs/nightlife/tech-talk-peter-marks/106539970 

Saturday, April 04, 2026

3D printed callsign banner for backpack in OpenSCAD

Recently I purchased a rather nice backpack which features an area that can hold velcro patches. I thought it would be nice to have one with my callsign on it and went hunting for one off emboydered patches. An idea occurred to me that a 3D printed banner could work. Here's the result:


Close up you can see that I've coloured in the engraved letters with a permanent marker. Some sort of paint would be better.


I'm able to write OpenSCAD code but it turns out so can Claude Code. This is what we came up with:

// Rectangular plate with embossed text

// Dimensions

plate_length = 90;   // mm

plate_width  = 20;   // mm (adjust to taste)

plate_height =  2;   // mm plate thickness

text_depth   =  0.8; // mm how proud the text stands

rim_width    =  1.5; // mm wall thickness of raised edge

rim_height   =  1.0; // mm how tall the rim stands above the plate


// Base plate

difference() {

    cube([plate_length, plate_width, plate_height]);

    // Embossed text on top surface

    translate([plate_length/2, plate_width/2, plate_height-0.6])

        linear_extrude(height = text_depth)

        text(

            "VK3TPM",

            size    = 15,

            font    = "Liberation Sans:style=Bold",

            halign  = "center",

            valign  = "center"

        );

}

Rolls of hook and loop self adhesive strips are available in stationary stores which is 20mm wide.

Wednesday, April 01, 2026

Experiment with live noise removal from an HF SSB signal

Today at the Macedon Ranges Amateur Radio Club coffee morning, Eric, VK3EJD, brought along a new gadget that he hopes will remove noise he experiences from a neighbour's solar system.


I used to have a noise reducing speaker from BHI but was never terribly happy with it. He reminded me that I'd started building a live noise reduction application for macOS using RNNoise so when I got home I did some tinkering and got it working. Here's a demo - although the noise is not terribly bad. (I switch it in and out during the video.


Here's an example of removing solar panel noise from an AM transmission:


It's promising but needs to be trained for your specific noise. In this I'm using the default model. This runs easily on an Apple M4. I wonder when we'll start to see this built in to amateur radio gear?

Tuesday, March 31, 2026

Convert HackRF Mayhem Portapack capture to a format playable with hackrf_transfer

I had a nice capture of a busy 40m band done using a HackRF Mayhem portapack and I wanted to play it back using a HackRF connected to a computer using hackrf_transfer.

Even though the native format of the hackrf IQ is 8 bit signed interleaved samples, for some reason the portapack pads these out to 16bit so conversion is needed.

A utility to do this is here.

Here's how it looks being played back:


These captures are excellent for demonstrating receivers and playing with decoders for digital modes. Note that the files can get rather large.

I play the output file back like this: hackrf_transfer -t 40mssb.iq8 -f 7100000 -x 1 -R


Sunday, March 29, 2026

FreeDV Sunday net - good rollup

Conditions were so bad last week that I didn't bother writing up the net. Virtually no NVIS. This week was much better. 


Stations on frequency included: VK5KHZ, VK5KVA, VK5AG, VK5ABE, VK3ZD, VK3MS, VK3KEZ, VK3JCO, VK3GTP, VK3FC, VK3DQ (welcome), VK2UMZ (rx only), VK2KNC, VK2AWJ, VK5RA, VK3XCI and me, VK3TPM.

Thanks everyone for coming up. The net is held on Sunday mornings from 10:30am eastern time on 7177 unless the frequency is in use.

Friday, March 27, 2026

My experiences of using AI for ham radio projects

At this week's Radio Old Timers Club lunch in Melbourne I presented a talk outlining the good success I've had recently using Claude Code for ham radio applications. Also there's a bit of an overview of how FreeDV RADEV1 uses machine learning already.


A few AV hitches and the crowd was finishing lunch but feedback was pretty good.