Flight data kindly provided by the Royal New Zealand Air Force (RNZAF) has been used to check what detections and tracking of aircraft are possible using GDTAAA and WSPRnet data in the southern Indian Ocean back in March 2014, when MH370 disappeared. This is the opposite of a blind test. In this case we know exactly where the aircraft is and we look to see what WSPRnet anomalies align with the position of the aircraft.
GDTAAA output using the WSPRnet data has been shown for 22 progress indicators and 6 position indicators at various stages in the flight performed by a RNZAF Orion P-3C aircraft on 28th March 2014 as part of the AMSA Search and Rescue (SAR) after the disappearance of MH370. In total 90 data points over a period of 3 hours flying time were analysed. The detection rate was on average 31% or every 6 minutes.
This system will be used to track MH370 from departure via the point of diversion to beyond the limited range of the civilian and military radar systems and into the Indian Ocean. This data can be coupled with the completely independent Inmarsat satellite data and will give a more accurate picture of the flight path followed by MH370. This information will be passed on to Ocean Infinity, who would like to resume their underwater search for MH370. Perhaps it will be possible to conduct a further underwater search towards the end of 2022 during the southern hemisphere summer months.
The paper can be downloaded here
incredible that this actually works?!?!
@david,
Welcome to the blog!
HF radio transmissions work around the globe. Amateur radio enthusiasts have shown that this works for years. The WSPRnet proves it works on a daily basis since 2009.
Over-The-Horizon-Radar works and has been used to track aircraft over thousands of kilometres.
GDTAAA uses HF radio transmissions from the global WSPRnet network of links and detects anomalies in the network like an aircraft disturbing an invisible trip wire. The reflections and scatterings caused by an aircraft can be picked up in the global WSPRnet network just like primary radar picks up reflections from an aircraft.
Amazing, can’t wait to see the actual flight track of mh370
@All,
Geoffrey Thomas has published the following article on the GDTAAA technology using WSPRnet data:
https://www.airlineratings.com/news/mh370-tacking-to-start-using-new-technology-that-is-proven-to-work/
Geoffrey states: “The greatest mystery in aviation is much closer to being solved after the latest trials of breakthrough tracking technology”
The latest trial involved an RNZAF Orion P-3C aircraft on March 28 2014 as part of the AMSA SAR after the disappearance of MH370. Despite the fact that the Orion aircraft is much smaller than MH370 and despite the fact that the flight time was in the earlier part of the day where the WSPRnet transmissions are fewer, the aircraft could be successfully detected and tracked to and from the search area near the 7th Arc. Whenever the aircraft made a tight turn or a sharp descent or climb, the number of WSPRnet anomalies increased. At one point it was even possible to detect the aircraft at an altitude of around 600 feet.
@All,
A new article by Bernard Lagan in The Times was published today describing how MH370 could finally be located by new technology.
https://www.thetimes.co.uk/article/70133a76-22b6-11ec-a1fd-37d9030e4d65?shareToken=8efc9670edb2657a1b496a7a3bed11bd
“The tests, by the British aerospace engineer Richard Godfrey, made use of a little-known online database set up in 2009, known as the Weak Signal Propagation Reporter (WSPR), which records every interaction between aircraft in the sky and signals sent by ground-based radio transmitters.”
The mammoth task of trawling through the database to find the tracks left by MH370 — a job to be performed by specially designed software — will take two months. The clues as to what happened to MH370 may have been in the airwaves all along.
Ocean Infinity said that it was aware of the trials and was open to reaching a deal with Malaysia for a new search.
“We are always interested in resuming the search whether as a result of new information or new technology,” a spokesman said, adding that late next year or early 2023 appeared a “sensible” time frame.
Your dedication to this search is inspiring! Excellent use of science, good hunting
@William S.,
Welcome to the blog!
Many thanks for the kind words.
Bless your continuing efforts to solve the mystery of MH370.
@Russel,
Welcome to blog!
Many thanks for the kind words.
Hello
I have a fair background in working with WSPR spot data and would like to go through your analysis step-by-step to better understand your approach.
In your paper referred to in this blog post section 4.0 states, “The WSPRnet data has also been post-processed to determine the received power based on a conservative noise floor level of -150 dB.”
First, is -150 dB shorthand for a noise level in rigorous units of -150 dBm in 1 Hz bandwidth in 50 ohms? If not, what are the rigorous units you have used? If it is, I cannot see where you have translated the noise in 1 Hz to the noise in the WSPR signal bandwidth.
Second, In my experience few WSPR receiving stations achieve a noise level of -150 dBm in 1 Hz bandwidth in 50 ohms, furthermore noise is rarely constant, especially on the lower bands. A selection of real-world noise estimates contemporaneous with WSPR spotting can be found at http://wsprdaemon.org/graphs/index.html.
Third, in the Excel spreadsheet refered to in section 4, ‘Rx Power’ is calculated by subtracting the SNR from the fixed value of -150, thus, taking the first line, an SNR of -4 dB becomes a Rx Power of -146 and, taking the second line, an SNR of -11 dB becomes a Rx Power of -139 dB. That is, Rx Power goes up as SNR goes down. I can’t understand the physics behind this, I would expect that Rx Power should go down as SNR goes down.
I have paused at this point as to go further would be pointless until I understand this stage of your calculations.
@Gwyn Griffiths,
Welcome to the blog and it is an honour that you take the time to discuss my hypothesis.
All I know about WSPRnet noise levels is what I have gleaned from your excellent paper titled “Estimating LF HF band noise while acquiring WSPR spots” dated July 2019 and which I recommend to anyone who wishes to understand WSPRnet SNR data:
https://www.dropbox.com/s/mqgl0qzo1kl3cg4/Estimating%20LF%20HF%20band%20noise%20while%20acquiring%20WSPR%20spots.pdf?dl=0
I am not a radio amateur. I am an aerospace engineer with a hypothesis that the WSPRnet data can be used to detect and track aircraft.
The WSPRnet SNR values should be correlated with transmitter power and the propagation path distance. The received SNR is expected to be above a minimum detectable signal level of say -30 dB. So a respectably strong signal is possible to receive, assuming the local floor noise levels at the receiver permit it. I use Proplab Pro V3.1 which does not (yet) use local floor noise levels in computing signal strength. All I know for certain is that for any given WSPRnet link between a transmitter and receiver, the WSPR protocol was correctly deciphered and the transmitter call sign and Maidenhead Grid location as well as the receiver call sign and Maidenhead Grid location are documented along with the Tx Power, Rx SNR, Frequency, Drift and a Date Time stamp. This is proof in my view that the transmission was correctly received.
All I am doing is looking for anomalies in particular WSPRnet links. For example, during the flight of MH370 on 7th March 2014, we know from official ATC ADS-B data that at 17:16 UTC MH370 was at a position of 6.391551°N 103.325580°E, altitude of 35,000 feet and tracking on a course of 25.4°T at a ground speed of 471.9 knots. From the WSPRnet data we know that at the same time from spot id 186123049 that HB9CZF in Switzerland at JN47ch transmitted at 30 dBm and 14.097079 MHz and was received by VK1CH in Australia at QF44mp with a SNR of -20 dB and a drift of -1 Hz/min over a distance of 16,530 km.
Between 16:06 UTC and 22:06 UTC on the same day there was a total of 27 WSPRnet links between these stations. The mean SNR was -17.7 dB and the standard deviation was 3.2 dB. My first question is whether the SNR at 17:16 UTC at -20 dB is an anomaly with a deviation of 2.3 dB.
When I check the propagation path in Proplab Pro V3.1 I find that the propagation took place along the entire path in either night hours or grey zone hours. I further see that the propagation distance can be covered in 5 hops and that the interim landing point of all hops is on calm water. Finally I note that the interim landing point of the 3rd hop is exactly at the position of MH370. My second question is whether the aircraft disturbed the propagation by either some reflection or scatter. I consider the possibility of either forward or back scatter. There may be well defined directional reflections from relatively flat surfaces (for example the wing span of a Boeing 777-200ER is 427.8 m2) or a scattering from a tubular body in all kinds of directions. The WSPR footprint of an aircraft may include Doppler forward scatter, aircraft structure metallic reflection, engine exhaust particles wave disturbance and aircraft wake vortex water vaporisation wave disturbance.
The Radar Cross Section (RCS) of an aircraft is difficult to measure in the horizontal plane. The RCS is very difficult to measure if an aircraft is turning and/or climbing/descending and a ray is descending from the ionosphere or ascending toward the ionosphere, after having been reflected off the Earth’s surface (flat or rugged terrain, calm or rough ocean, flat or mountainous ice in a polar region). Calm ocean or flat ice is best.
I further check in Proplab Pro V3.1 that at this time the Maximum Usable Frequency (MUF) comfortably allows a signal at 14 MHz and that an elevation angle of 2.9° fits the propagation path hops at the required azimuth. Finally I check the signal strength at each interim landing point as calculated by Proplab Pro V3.1. The software performs a smoothing process. For every signal that reaches a specific cell, the signal strength is added to the strength of any signals that previously impacted that cell. The strength of each ray is first converted from dBu to watts before being added. The resulting strength in watts is then reconverted back to dBu (relative to a 1 ohm system).
I am writing this example up in a short paper with all the data and calculation results, which I will publish in the coming days.
Richard,
Thank you for your kind compliment on the work I have done with colleagues on WSPR noise analysis. Thank you also for the additional information on your methodology.
However, when approaching something novel, which your approach certainly is, the only way I can attain a true understanding is by taking very small steps and being sure I understand each before progressing to the next. I’m no closer to an understanding on the three very specific points in my first post.
@Gwyn Griffiths,
As I mentioned in the lengthy explanation in my previous comment, I am publishing a paper in the coming days, which should help your understanding.
I apologise that I did not remove a bunch of internal workings and out of date columns in my Excel before publishing. I have fired the member of my staff responsible (me)!
Your questions are irrelevant because unfortunately, and because of my oversight, you are trying to understand the meaningless.
Having said that, I am sorry that my attempt to respond to your initial comment left you completely disappointed.
New is not necessarily wrong. Small steps can build to giant steps for mankind.
Documenting the internal workings, the underlying equations, and assumptions is absolutely vital. I would encourage you to include them in the paper you intend to publish here in the next few days.
It would also be helpful if the revised Excel spreadsheet included the calculation formulae rather than the numeric results of the calculations to enable a full review.
My questions will always be specific; lengthy responses on other, later, aspects of your analysis are not what I’m be looking for. I’ll get to those once I understand the fundamentals of your approach.
@Gwyn Griffiths,
I agree with you “Documenting the internal workings, the underlying equations, and assumptions is absolutely vital.” I have been doing exactly that for years on all projects that I have undertaken.
I do not use Excel at all for any calculations. I use Matlab R2020b Update 3 and Proplab Pro V3.1. I only use Excel to publicly document the WSPRnet data I actually use for any particular test, as Excel is in wide spread use (personally I do not use Excel for any other purpose). I have also previously explained in papers and comments the post processing I perform on the WSPRnet data before using it.
My goal is to provide full transparency of the process that I am using, despite the fact that I have submitted a patent application on the subject. I have written several papers on the subject and plan to write several more. Please remember that I have stated in my papers and in many comments on my web site that this is a “work in progress” and can be described as a “working hypothesis”. In my view, the “full review” that you are proposing is premature and inappropriate for a working hypothesis. You initially stated that you wanted to take it “step by step”. Let me publish my next paper (which is a next “step”) and you can then let me know what you are missing.
I have previously published Matlab scripts I use but meanwhile there are hundreds. I could publish further examples, if that would help. As I explained in my previous comment to you, I use Proplab Pro V3.1 to perform propagation calculations. Although I check the calculations by hand from first principles, I am not privy to the actual algorithms used in this proprietary software. I think you are well aware of the calculation formulae and can check the calculation results in the upcoming paper, which will include all the assumptions made. I guess you are familiar with Proplab Pro which has been in use for over 30 years in the radio amateur community.
Steve Kent has gathered all the relevant formulae and published them in his recent excellent paper on the WSPRnet theory. I disagree with his conclusion but the formulae are all sound. Please be aware that Steve Kent accuses me of writing papers that are “not of interest” and described as “yet more cartoonograpy” and “sheer lunacy”, so he is not exactly a supporter of my hypothesis.
https://www.dropbox.com/s/wf6z22spssby35i/Steve%20Kent%20WSPR%20Paper.pdf?dl=0
My latest paper will explain in detail the process I follow. It is not the first paper I have published on the subject and will not be the last paper either, just part of the work in progress. I am not requesting a full review by you at this stage. I have no doubt that you will provide a critique of my next paper in due course. What you are requesting to “enable a full review” will take several months. If you are looking for a final definitive paper on the subject from me, I suggest you check back on the web site early next year.