Hacker News Reader

Comment by ajarmst | Hacker News Reader

I've not read the original report (and am unlikely to), but this quotation: '“WSPR is like a bunch of tripwires or laser beams (graphic below), but they work in every direction over the horizon to the other side of the globe,” Mr. Godfrey says.' is not an accurate characterization of this (or any) use of WSPR. It implies a degree of precision and clarity of signal that is completely misleading. I'm also very concerned about Godfrey's clear comfort with taking this very weak and contingent signal and using it to ascribe thoughts, motives and behaviour to the aircrew of MH370. You don't do that without substantial corroborating evidence from other sources when you have an intact Cockpit Data Recorder, much less whatever this is. Looks a lot like a familiar pattern: starting with a theory, finding some nice noisy data that needs 'expert interpretation', and then massaging it to confirm your theory.

View on HN

Yeah I'm not an expert in this stuff but I am an extra-class radio amateur and this sets off literally all my intellectual alarm bells. I'll take one for the team and go read the report but I agree that it seems like pure unadulterated bullshit.

The ionosphere is a very violent place for an HF radio signal. These radio waves do not travel in a straight line. They don't even come out of the ionosphere with the same polarization as they go in with. I cannot fathom how you could glean anything from a WSPR signal's RSSI.

edit 1: I'm just gonna live-text my thoughts on this as I go through it into this comment. First off, the software that this person developed, GDTAAA, doesn't seem to be available for viewing. Shame because I'd love to see it, but I guess I'll have to go by the paper.

edit 2: "An algorithm is run that calculates an expected received SNR based on the transmitter power and the short path distance" W O W, okay, not even taking into account an estimated noise for the receiving station. Nice. It also doesn't seem to take into account time of day, any of the different propagation modes like gray line, etc, or the wavelength of the transmission. Big yikes.

edit 3: WSPR signals have a transmission length of about two minutes. A 777 travels about 20 miles during that amount of time, and you're trying to use the average SNR of the entire 2 minute period to estimate whether an aircraft crossed the path of the signal. Even if the WSPR signal occupied the entire 20 mile-wide swath of sky that the aircraft was in during the WSPR transmission, the aircraft can only block about 0.2% of that swath at an given moment on account of it being absolutely tiny.

I agree it seems far-fetched, but: possibly by integrating many data samples you can remove the ionosphere noise from an underlying signal.

If the ionosphere noise were merely noise, and not just distortion, this would be true. I'd suspect that even changes in solar radiation (from the rotation of the earth) would cause e.g. more ionization in some parts of the ionosphere compared to others.

I'm also an extra-class radio amateur, and I too am unconvinced by this article. The "tripwire" concept looks good in 2D, but I'm less convinced in 3D. Most HF signals propagate by bouncing the signal off the ionosphere (it's actually refraction, not reflection, but...), so the ideal path to the observer looks like shooting a signal straight at the horizon (that's the top of the ionosphere that's farthest from you and not shadowed by the earth). The angle of incidence equals the angle of reflection on the way down, to an ideal receive location where that point that the sender aimed at is where the receiver is aiming at. You can have multiple hops; the ground reflects signal back up towards the sky. (Tangential paper: https://www.physics.princeton.edu//pulsar/K1JT/HFTOA_1.pdf You can measure how many hops a signal is taking at any given time, and it changes throughout the day.)

So if we were in a "spherical cow" universe, you could see how this would sort of act like a tripwire. A radio transmitter illuminates a single point in the ionosphere, and a radio receiver looks at that point. If a plane flies along that path, the signal goes dark, and you know something crossed your tripwire.

The sticky bit is that we aren't operating on points, we're operating on a continuum. If you aim your radio beam at the horizon, maybe 10% of the power is 10 degrees above that, and 5% is 20 degrees above that, and so on. The signal is "smeared" in the real world to the extent that it's not a plane-sized beam that is on or off, rather it's a big circle that gets brighter or dimmer as planes fly through it. And the circle is much, much bigger than any plane. I would be surprised if you'd even notice. (That is assuming an ideal one-hop HF setup. I have never heard of people routinely operating WSPR with an ideal setup, usually people use random wires that have very little directionality to them.)

An analogy that you can try right now is getting a camera with a lens that has a large front element (I just tried this with a 90mm f/2.8 lens). Take a picture at the widest aperture. Then stick your big ugly finger in front of the lens, right on the front element, and take another picture. If you look at the second picture, you won't know that your finger was there. There is no obvious finger, or even a finger-shaped shadow. If you compare it to the other picture, maybe you will notice a difference, but maybe you won't. You have to cover a lot of the lens before you notice. (Added fun: consider the 2 minute exposure time of WSPR. Take a picture with a 2 minute exposure time, and walk through the frame midway. Can you see yourself in that picture?)

I have to imagine that using WSPR transmitters is a lot like this. If you take lots of samples, maybe a single tiny airplane will make a difference. The data is technically there. But I don't think the random transmitters and receivers are consistent enough across cycles to be able to get this data from a single sample. (Remember that WSPR is 2 minutes transmit and 2 minutes receive. Does your radio heat up and cool down, shifting the frequency of the local oscillator in that interval? Mine certainly does!)

As I was reading the article, I thought "neat, but I'm not fully convinced" until I read the part where the author thinks he can do this arbitrarily and detect any airplane at any time, and that it is conveniently done with proprietary software that doesn't show up anywhere on the Internet. That caused the alarm bells to go off. I'm going to need a lot more data to be convinced.

I agree, the fact of some “anomaly” on a WSPR link is not sufficient evidence for the presence of an aircraft. There are many other possible reasons for those events.

Forget the radio heating up and cooling down, my cat walking by the coax going to my antenna is going to affect my SNR more than an airplane thousands of miles away will :)

Yeah, I'm surprised the author of this paper doesn't have a side project that shows how many WSPR stations are being chewed on by a cat right now.

As an extra-class amateur radio operator, I just wanted to say that I was telling a newly minted technician friend today that extra-class was really just for bragging rights. I didn't expect to encounter a concrete demonstration so quickly. :P

Okay, I'm also skeptical. I mean passive radar is a thing but ... at least the presentation in the article appears confused. But if the approach works (whatever it actually is) it should be easy enough to validate it against known flight data for thousands of other flights.

I'm assuming you meant "extra-class wasn't really just for bragging rights" because I, too, am an amateur extra and self-taught as I am, learned quite a lot of stuff preparing for that exam that I use regularly in and out of the hobby! And so, let me take this moment to encourage anyone reading to also go for your extra license: You won't regret it! Impress the girls! New and interesting bands to operate on! Bragging rights (just kidding but not really)! Amateur Extra has it all!

There have been attempts to validate this method. This one made it into a poster presentation at a conference:

https://hamsci2021-uscranton.ipostersessions.com/?s=EA-94-A8...

Others have considered these problems and done some work. Here’s a poster:

https://hamsci2021-uscranton.ipostersessions.com/?s=EA-94-A8...

I can't agree more , reading this every bullsh*t alarm in my brain was triggering. In his report Mr Godfrey never mentions the frequencies used for the WSPR transmissions. This mode can be used on VHF/UHF frequencies where it may reflect from aircraft. But 95% of WSPR transmissions are on frequencies below 30 MHz and most likely below 20 MHz. These have long wavelengths and highly unlikely to be affected by an small (in wavelength terms) aircraft. If it could be affected by aircraft then Mr Godfrey would have to take into account the movement of every aircraft in that region at the time which he doesn't appear to be doing. As a note of explanation I'm a radio amateur and a WSPR station operator , its a brilliant tool for propagation studies but not for this.

> Mr Godfrey would have to take into account the movement of every aircraft in that region

I'd imagine there were dozens of aircraft crossing the Indian Ocean around the time MH370 disappeared. If his technique was that good he should be able to demonstrate for any day ending in Y how he can track flight paths compared to ADS-B data from the likes of flightradar24

This would be a good rebuttal, to offer up a bunch of counterexamples of transponder data with no corresponding signal in the WSPR "data" using the same method, but also ADS-B transponders are limited in range, so there will be big gaps specifically in the Indian Ocean. But if this (frankly unbelievable) method works well enough to suss out the location of MH370 in those ocean holes then imagine how well it should work in areas saturated with transmitters and receivers. But it won't, because it really sounds like a bunch of hooey.

I don't disagree with the triggering of your BS alarm, but don't write off the ability of frequencies in the 5-30MHz range to detect an aeroplane the size of a passenger jet. I used to be an engineer on the Jindalee Over-the-Horizon-Radar and it could detect things much smaller than this using these frequencies.

Points for and against WSPR detecting MH370:

- Jindalee uses huge power: a beamformed antenna transmitting many kW and a beamformed receiver array 3km long (yes, one of the techs who was an amateur did plug his kit into one of the arrays).

- Jindalee uses a very stable timebase, meaning it can detect small Doppler shifts.

+ WSPR has a lot of receivers and transmitters, meaning the target is illuminated from lots of different directions. With the right signal processing, these signals might be synthesised into a large aperture. Are WSPR receivers typically coherent (I/Q)?

+ According to one article, there are 5000+ WSPR transmitters [1]. If each of these is transmitting 5W (what's typical in your experience?), that's 25kW. A useable power for an OTHR radar.

- The WSPRS transmitters have a low duty cycle and are not active simultaneously?

+ By its nature WSPR will sometimes fluke excellent propagation conditions, meaning it will be able to do things that might otherwise require higher power. With a large number of stations giving N^2 propagation paths, what are the chances of some of them being useable?

+ An aeroplane the size of MH370 is quite a big target for an HF radar, with quite a big Dopper shift.

+ By its nature, a received signal will contain information about the channel though which it passed. Minus: chances are this information is filtered out by WSPR's receiver processing.

- Does the WSPR database actually store enough information to post process? Ideally it would need to store the sampled waveform from a coherent GPS locked receiver. It looks as if each record is only a received SNR and a drift (presumably a frequency offset) and a timestamp in minutes?

- Are WSPR receivers typically GPS locked, meaning they might have the ability to measure small Doppler shifts and be coherently combined with each other?

I'd put the possibility of WSPR being able to detect an aeroplane into the "intriguing" category if (huge if) it stored the necessary data in its database. It would be a computational tour-de-force, involving synthesising a large number of WSPR stations into a single aperture, then using that aperture to form bins in range, azimuth and Doppler/velocity, then tracking a target as it moves between bins. Sadly, my guess is that WSPR does not store the necessary data.

(Maybe "intriguing" is overstating it. More "wouldn't it be cool if...")

[1] https://www.abc.net.au/news/2021-05-05/malaysia-airlines-mh3...

Thanks for your post. I'm aware of Jindalee and the Russian OTHRs which claim to detect aircraft (and I believe the claim) but they are very different from WSPR.

My station transmits 10 watts (and that's on the high side for WSPR) into an omnidirectional wire antenna. Some WSPR stations use directional yagi antennas but that information isn't contained in the WSPR data. A WSPR transmission lasts around 1 minute 50 seconds but the frequency shift/signal strength is only calculated the once during that transmission (not sure if by averaging). A transmission has to start within a second of the start of an even minute. Some stations are GPS locked but I just click the PTT on my radio to start the transmission while looking at the time on my digital atomic receiver clock.

In answer to your questions some WSPR stations transmit every 2 minutes while others like mine only transmit once or twice an hour.

Thanks for the informative reply. I think it's a fascinating system. On paper it could act as an OTHR if the signal processing was changed.

I'm guessing each station would need to lock its carrier to GPS and transmit as often as possible (10W would probably be okay with lots of stations). Similarly receiver local oscillators would need to be GPS locked (+/- 1ns?) and the receiver be coherent. Processing would need to be based on the raw stream of samples, but with a 6Hz bandwidth one could get away with a sampling rate of about 10 samples/s (Nyquist + a reality factor), each sample being a pair of 16-bit I/Q components. The low bandwidth might limit resolution. The signal processing would be the fun bit. Can all those receivers and transmitters be synthesised into one big array? Could the processing could be done in a distributed manner? Lots of interesting problems.

Apart from being an OTH radar, such a system would be the mother of all ionospheric sounders, as the main difference between a radar/sounder/communications is which features one chooses to extract from the received signal.

Is passive radar a thing in the amateur scene?

The aircraft detection phenomenon isn't far-fetched sci-fi; here's a video showing what it looks like:

https://www.youtube.com/watch?v=1gc0vVk3XBg

Remember that we're talking about a gigantic hunk of airborne metal, with a mandatory 50 miles of separation from any other gigantic hunk of airborne metal.

With that said, the article is claiming to do this detection without the waterfall plot or I/Q data, using nothing but a single scalar data point (receiver SNR) recorded once every two minutes. That is far-fetched sci-fi. If the WSPR stations were archiving the raw I/Q data at the SDR input, I might believe something useful could be recovered from that.

I mean, sure, measuring occlusion like this is possible in VHF, UHF and beyond. And radar is a real thing that exists, and you can even do passive radar, but to my knowledge you cannot do any of this outside the line of sight of the transmitter and receiver with an object that's only 1-4 wavelengths long. I'd love to be proven wrong because that would be wild and the kind of thing that makes me fall in love with radio all over again, but this source is completely bogus.

Over-the-horizon radar uses HF to detect distant objects. See https://en.wikipedia.org/wiki/Over-the-horizon_radar

With a big antenna array for beamforming, sure, you can do that. You absolutely cannot do over the horizon radar on 10 watts with two omnidirectional antennas though.

Wait, who is doing WSPR at 10 watts? Most people are just using an unamplified IO pin off their Raspberry Pi or something; milliwatt at best.

(It makes WSPR itself all the more impressive, however!)

If you have an SSB radio that you can hook up to a PC's sound card using VOX to activate the transmitter, you can use WSJT-X or similar software to do lots of digital modes like WSPR, PSK31, FT-8 and whatever's hot these days. I think best practice for WSPR is to keep power low though, so don't go crazy :)

While there are a lot of folks using very low power on WSPR, there are also a lot on HF doing so with 100 watts, and even more in some cases.

The 'Russian woodpecker' and the US' OTH-B systems built decades ago used pulsed signals of 1 Million watts and up, with gigantic antenna arrays. [0] Comparison to WSPR is impossible (esp. given the complete lack of non hand-wavey details in the article). If the method were at all feasible, it'd very likely be commercially available).

[0] https://fas.org/nuke/guide/usa/airdef/an-fps-118.htm

> ”mandatory 50 miles of separation from any other gigantic hunk of airborne metal.”

There is certainly no 50 miles of mandatory separation between aircraft! Horizontal radar separation is 10 nautical miles at most (much less in busy airspace near airports), and vertical separation is 1000 feet.

Of course, the Southern Indian Ocean is very remote and it’s entirely possible that there were no other aircraft within 50 miles of the MH370 flight path.

> There is certainly no 50 miles of mandatory separation between aircraft!

Over the open ocean, where the MH370 was, there absolutely is:

https://en.wikipedia.org/wiki/Separation_(aeronautics)#Longi...

   > The golden rule is the 10-minute rule: no two aircraft following the same
   > route must come within 15 minutes flying time of each other.  In areas with
   > good navaid cover this reduces to 10 minutes

At the MH370's cruising speed of 636mph a 10-minute separation is 106 miles! The Boeing 777's cruising speed is way at the high end for commercial airliners. Regardless anything flying over the open ocean between continents will almost certainly be going at least half that speed.

That is only if they are at the same altitude. You seem to have missed the first sentance on that link "If any two aircraft are separated by less than the vertical separation minimum, then some form of horizontal separation must exist. "

That might be the golden rule, but it definitely isn't "the rule".

If you look on Flightradar24, over the Atlantic for example, you'll see several cases where there are multiple planes well within 15 minutes flying time of each other, on the same track, with separation of 2-3 thousand feet vertically.

One example as I write has AA9820 and UA46 following each other at 40,000ft and 37,000ft, at 15nm horizontal separation, and QR740 at 36,000ft about 30nm behind them. When airline traffic is at 100%, you'll routinely see them much closer than that.

> following the same route

There are many routes

Furthermore, even if such a rule were in place, one cannot assume it was being observed in the case of MH370. With ADS-B not functioning, no-one else knew where it was - which, of course, is exactly why it is a mystery!

Im always amazed at the ability of people to make stuff up and just state lies as facts.

I am not an expert but one who actually worked on a over-the-horizon radar system.

This entire article is predicated on the RF reflectivity of ionosphere being stable during its measurement of computerized triangulation, when in fact that the sky was in sunrise mode where the ionosphere starts to fall to a lower altitude.

  Such sunrise/sunset transitory phase is never a clean delineation of RF reflection but instead a noisy raster of bounced RF, much like an oversized blanket of aluminum chaff.

Reply on news.ycombinator.com