HF Mesh Radio

One motivation for my recent Summer Ham Project is an idea that has been kicking around in my head for a few years. I figure that low bandwidth data over HF could be used for low cost communications in parts of the world where there is no infrastructure at all. For example chat, SMS, or email type services over distances of 1000’s of km.

Now HF email exists but a solar powered HF email station (HF radio, modem, solar power system) can cost $10,000 (this is a quote from some missionary friends of mine in New Guinea). The power levels are also quite high, for example a transmitted power of 100W. It’s probably possible to do it a little cheaper (say with Ham radio gear) for a minimum or perhaps several thousand dollars. Plus 1st-world service connection fees for the server side of the HF email service.

Now the trend in radio design is to move the signal processing from analog electronics to software. This tends to reduce the cost and complexity of the analog side. Software, as we know, can be free. So it should be possible to use a low cost radio (less than $100) coupled to some DSP software (say running on a laptop) to provide low cost data transmission over HF.

The problem with HF radio channels is that they are unreliable, they fade in and out over the course of the day and the solar sunspot cycle. Having a high power level helps punch that message through, but perhaps there are other ways. Some Hams have been experimenting with modes such as QRSS, which use low data rates and DSP techniques to send signals over 1000’s of km using just milliwatts of power. PSK31 is another interesting Ham mode for “HF chat” that uses sophisticated modulation and error correction coding techniques in just 31 Hz of bandwidth.

Over the past few months I have been working with Elektra on the Mesh Potato project. This has given me some exposure to mesh Wifi networks. In mesh networks client nodes forms ad-hoc networks by relaying messages. A central access point is not required. The network protocols automatically sense when a nearby node can be reached and adapt as nodes come and go. This sounds ideal for relaying HF radio messages over unreliable HF channels.

The figure below (thanks Elektra) shows the advantages of mesh networks over regular networks with a fixed central access point. Node C may not be able to see the access point, but can easily see nodes D & B. However a mesh network can use D & B to route traffic for C.

Now Mesh Wifi nodes might be a few 100m apart and operate at 11 Mbit/sec. Mesh HF networks might be 100-1000km apart and operate at 300-9600 bits/sec. Much slower but still fine for emails, SMS, and maybe even chat. Imagine a bunch of MF mesh nodes on a Pacific Island chain. They have no other form of communication. VSAT is far far too expensive due to the monthly fees and even a $8,000 HF email station is pushing it. A packet might bounce around several of the nodes before reaching the main island, and a shared Internet gateway to the rest of the world.

HF Mesh Nodes could be built very cheaply. For example a simple, low cost radio plus a laptop or even a small embedded DSP processor. As the data rates are low the signal processing could be done on a very low end chip, even a microcontroller. Low transmit powers are adequate as the mesh automatically senses a way to get the message through even under poor radio conditions. If the HF mesh nodes are cheap there can be many nodes, enhancing the performance of the network.

The HF mesh idea is just an idea at the moment that I am brainstorming with Elektra and Kim Hawtin. The potential is basic (but very useful) communications on a global scale at very low cost and with minimal infrastructure.

15 thoughts on “HF Mesh Radio”

  1. If the network is not very powerful, perhaps the endpoints can be.

    Could we convert voice-to-text and send over the network, and play it back at the other end … a la voicemail-only network.


  2. I think PSK31 has a unique ability to perform in a mesh network due to its narrow (radio) bandwidth. Your nodes could all share a single frequency and not have to worry about interfering with each other — the gateway, for instance, could hear all nodes transmitting at the same time on the same frequency and still adequately perform. When there are only a few nodes participating, they could utalize multiple psk31 streams at a time to improve bandwidth. As long as you could devise an efficient time/spectrum negotiation algorithm you’d have something that could be really reliable.

  3. That’s a neat idea.

    Military people have been doing short messages in HF meshes for some time now, but with hideously expensive hardware and complex protocols.

    This is more like a 21st century rehash of the global telegraph system.

  4. Simon – I was thinking about something very simple like a Pixie coupled with a microcontroller for the DSP side. Let the DSP make the receiver high quality, keep the analog side simple.

  5. Ah, from Kim’s blog I’ve worked out what a Pixie is now… so one of those and a DSP chip would keep the price of each node pretty low.
    I’d be quite interested in helping out if you continue with this idea.

  6. Sounds like an interesting idea. I’ve done some work with DDS’s


    With the combination of DDS/IQ demodulator and small DSP this could be done quite cheaply.

    Frequency licensing could be a problem in the HF frequency band, the advantage of WiFi is the availability and free spectrum.

    I would be interested in helping out, either with this or the mesh potato if you would like.

  7. Hi Peter and Simon,

    Sure, help is always welcome. Re HF spectrum, yes that could be an issue. I was thinking of starting on the Ham bands to test the technical side of the idea.

    – David

  8. Dear David,

    What about JASON the low power communications software that was made
    by the same guys that build ARGO. I would like to experiment with
    it a bit, just on a station to station trial. It is a slow speed communications
    system that could be maybe modified some to do what your looking for.
    I don’t think PSK-31 would be suitable because of it’s non standard code
    set for the letters and numbers. What is needed is something that uses a
    standard ASCII code set. Just my humble opinion…
    Have fun.. take care..

    73 Dave

  9. Hi David,

    Thanks – I will check out JASON. Good point re PSK31, yes we really want something that talks ASCII (or even binary) so we can easily move packets between Mesh HF and the Internet.

    I am interested in using advanced modulation and channel coding techniques, just a low bit rates on simple microcontrollers.



  10. I came across your blog about this idea, it fits in with a concept I’ve been tossing around since 2007. I want to do HF mesh as well, but the traffic will be priimarily sensor data – meterologic sensors, but perhaps other things such an environmental monitoring (Carbon Dioxide layers).

    I’ve done a lot with wifi mesh in times past, but have set it aside lately because of the diminished participation in my local area.

    You’re working with Elektra? From Berlin? CCC? I met her back in 2007. Quite an interesing, eccentric person!

    I don’t know if PSK31 is the best mode for such a net. Though something similar might work. PSK is so narrow band that unless a radio has either god crystal control or some DSP intelligence to detect drift, you could have real problems. The MFSK modes look better in my opinion, though they do take up more bandwidth. It’s a toss-up, really, as to which is better. But recall that most of the HF narrow-band modes are made for interactive communication, and have design trade-offs for that reason. The encoding, for instance, is selected to present the most commond characters used in typographical communication, and in a proper block data compression algorithm there would be no advantage to this encoding.

    I want the data to be sent unencrypted, but cryptographically signed. This presents certain challenges, such as rotating session keys and so forth that are difficult to do on an unreliable and narrow channel. I believe that a ‘reasonable’ system can be designed, but that it would be vulnerable to a determined adversarial cryptanalyst.

    I say that I want to carry sensor data, but there would be room for other data types in my scheme. But anything more than short messages and automated QSLs would be frowned upon. No P2P, no Microsoft updates, no rag-chewing!

    I like the idea of keeping unit cost down. The easiest way to do this is to use CW. But it’s painfully slow. Power supplies are a big problem in the usage I have planned, because much of the planet doesn’t get enough sun to power even a QRP radio without a significant investment in photovoltaics. I could combine it with a wind turbine, but that’s mechanical parts that wear out and don’t survive a bad storm. I had considered a microbial fuel cell, which could work well in a swamp, but perhaps the most reliable system is that of the ‘earth battery’, or ‘telluric battery’, which utilizes that electric currents which run through the planet. These currents are very weak, but they are constant, and the equipment is simple and solid state. With a constant current charging some megacapacitors, it would be pretty simple to figure out RX and TX capabilities versus power availability.

    Anyhow, you should probably email me if you want to discuss this more.

  11. Hi, yes I work with Elektra every day on the Village Telco project, and had the pleasure of visiting her home and community in Berlin a few weeks ago.

    Thanks for your comments on narrow band versus wider band modulation schemes, and the differences between requirements for real time TTY/IM type comms and block based data.

    Re power I am hoping that very low powers can be used, with rx only most of the time. The power budget is something we must watch, for sure.

    I hope to get seriously active on this project in the second half of 2010, will contact you then.



  12. Look into http://open5066.org/wiki/ it uses a open source version of a Mill spec packet radio system (STANAG 5066).

    This also supports ALE which is a type of handshaking system that can negotiate band changes to take advantage of band openings etc.

    It’s about the most robust option I can think of for HF currently that is affordable (Open Source).

    ~Ken – vk4akp~

  13. Also look at the Raspberry Pi hardware project.



    It’s a 700 MHz ARM1176JZF-S core (ARM11 family), 256MB Ram, SD Card slot, Ethernet, dual USB, DSP. Broadcom BCM2835 (CPU, GPU, DSP, and SDRAM)

    Runs Linux.

    Combine this with Open 5066, and other packet modes. Perhaps even Q15X25 and you have a hardware solution for US$35 a unit. Or US$25 for a cut down version.

    This hardware is possibly also a nice solution for a hardware based “Codec2” Digital Voice interface. ??

    Hope to see you moving forward on some of your Amateur Radio digital area’s soon. We need new options.

    ~Ken – vk4akp~

  14. Or you could make your own low orbit satellite. Not really a satellite but a remote drone that can fly autonomously on a preprogrammed circular flight path, and make it carry a wifi mesh networking payload with just two watts power it can go the distance. With a really wide area footprint similar to a satellite when its flying at high altitude. And wifi mesh ground stations with the right parabolic antenna or yagi pointed at the general direction of the uav.

Comments are closed.