Hello Stefan,

Concurring with Alan and rambling from there:

Of the many (at best rough) methods of predicting or correcting VLF phase 
(and/or estimating path parameters) at any given time of day, one of the 
simplest and most accurate (I should say least-inaccurate!) is:
Using an average of the prior three days' phase for the time of day in 
question, for the same TX-RX path.

Not nearly as accurate but still valuable is using an average of the prior 
three days' phase for the time of day in question, for a path including the 
same TX and a second RX between the TX and your RX, even if the second RX is 
1000 km to one side or the other of the direct path.  

The above methods generally do not work well for paths shorter than 1,000 km 
because of the number of significant hops (modes, i.e. modal phase interference 
nulls) that can be involved.
Standard LWPC would suggest that path lengths greater than 2,000 km are much 
less complicated from a modal interference standpoint (phase much more stable 
and estimation of path parameters more feasible), but real data (such as 
Ferguson Ionospheric Profiles 530) shows that path lengths greater than 2,000 
km are just substantially more stable but not dramatically more stable. 

Extrapolating from phase (or height) at one frequency to phase (or height) at 
another frequency, especially if either of the frequencies is 6 kHz or below, 
is one of the more difficult approaches for many reasons. For any desired 
accuracy the complexity of extrapolating across frequencies is particularly 
high.  

So modeling, predicting and/or correcting based on paths longer than 1,000 km, 
daytime only (as Alan mentioned), same frequency, using a prior-three-day 
average of phase for the chosen time of day, from the same receiver, would tend 
to be productive. Alternatively predicting/correcting using a prior-three-day 
average from a second receiver between your receiver and the TX would be 
useful, but much less so than with your own receiver.

Improving on the above by using a network of receivers and transmitters is an 
excellent idea, especially if machine learning is employed and the (relatively 
immature) Dst-index and SID-data models are included; but the complexity that 
Alan mentioned increases substantially at the accuracy levels served by a TX-RX 
network, so the TX-RX network approach could be a long-term effort.

One possible exception to the notion that a TX-RX network approach could be a 
long-term effort might be: the possibility that one of the newer 
machine-learning approaches could by chance be well suited to 
modeling/prediction/correction using data from a VLF TX-RX network. Practical 
(low computational burden) machine learning tools have proven effective with 
systems of sensitivity and degrees of freedom similar to those of VLF 
propagation. It's hard to imagine that recent generations of machine learning 
algorithms wouldn’t provide some significant benefit given VLF TX-RX network 
data to work with, so I have high hopes for many reasons that an RX-TX network 
will eventually be very effective for modeling/prediction/correction (of 
height, gradient, phase, and other parameters).

- Jim AA5BW
 

         
   

-----Original Message-----
From: [email protected] 
[mailto:[email protected]] On Behalf Of Alan Melia
Sent: Tuesday, November 15, 2016 4:54 PM
To: [email protected]
Subject: Re: VLF: EbNaut transmissions on lower frequencies?, pre-tests: 6.47kHz

Hi Stefan, if I can ramble a bit on this topic.......I think the situation is 
probably more complex than you are envisaging. I see what your objective 
is.....you effectively want to characterise the height of the Earth ionospher 
waveguide over the whole path. However you talk of QSB. Now QSB can only happen 
on a given route if thereare two paths of different length between the tx and 
rx. Thus for example take a 10000km path and think about rays rather than 
waveguide modes. The main mode will be about 5 ionospheric hops, but QSB will 
occur if there are also signals at the rx which travel by
6 hops. Their phase will be different and will also change differently as the 
apparent reflection height changes along the path. Another complication is that 
one or both paths may not be a strict great circle due to tilting of the 
ionospheric layer.

It may be possible using several source signals to extract parameters of the 
path. A UK radio amateur astronomer has done this for very short range paths 
using simultaneous data from Anthorn and Skelton to central England .....but 
not in real time, and only in daytime when the modelling is easier. Another 
problem could be that the apparent reflection heights calculated from 23.4kHz 
data may be different to those in action for frequencies below 8kHz.

It is an interesting challenge. In daytime you have the added complication of a 
narrower  "waveguide" It may be possible to model a 1000km path (one hop), and 
test it on that.

Alan
G3NYK


----- Original Message -----
From: "DK7FC" <[email protected]>
To: <[email protected]>
Sent: Tuesday, November 15, 2016 8:22 PM
Subject: Re: VLF: EbNaut transmissions on lower frequencies?, pre-tests: 
6.47kHz


> Hi Paul,
>
> Many thanks for the decodes from these streams. What is the most distant 
> location from where a stream is available?
>
> Interestingly, the phase of my signal is less stable in Bielefeld than on 
> your side. OK, it can be expected but now we saw it.
> So maybe this is an advantage of more distant locations? I remember the 
> VLF grabber of TF3HZ, who received my kite transmission on 8.97 kHz. Maybe 
> he will join in again and even provides a stream? The distance would be 
> 2440 km.
> Probably he is reading this, i would expect :-)
>
> The (unknown) phase stability over a certain path and time is the 
> question. And its reproducibility.
> It looks like we started something here, a thing that could become the 
> trick how to get world wide VLF communication by amateur radio stations. 
> You see i become a bit crazy now ;-)
> Edgar J. Twining is the man i'm thinking about! I know he has the skills, 
> motivation and RX sites to be the man for the unbelivable project. Maybe 
> he is reading this as well.
>
> Now a next question/idea comes up to me: When observing the phase of 
> available signals (DHO38, GBZ, JXN, ALPHAs...), can we conclude to the 
> phase on 6.47 kHz? I mean daily!
> Assume we would try to transfer a message to Australia by using the 
> currently tested technique. Then there will be 'good' and 'not so good' 
> days. Not only different QRN levels, also phase changes may become a 
> problem. If someone, let's say in Tasmania, would run a wideband RDF 
> spectrogram (yes Markus, i know. I will do it soon! Now the time has 
> come!) and can monitor the phase of DHO38 on 23.4 kHz, is this of any use 
> for 6.47 kHz or 8.27 kHz??
> At least it would be most useful to observe the VLF spectrum from there, 
> trying to find some correlations. We need an expert there!
>
>
>
> The linux commands and your vlf utils are a future project for me. It will 
> be a big project i assume. But maybe its better to let me be the one on 
> the TX side.
>
> Am 14.11.2016 20:09, schrieb Paul Nicholson:
>>   11th Nov      3.9   -62.5 -141.2 8K19A   60   4   'TEST'
>>  12th Nov      4.9   -59.7 -134.3 8K19A   60   4   'TEST'
>>  11th + 12th   6.6   -58.1 -137.5 8K19A   60   4   'TEST'
>>  13th Nov      8.5   -55.6 -138.5 8K19A   60   4   'TEST'
>>  11th to 13th  8.6   -55.2 -137.8 8K19A   60   4   'TEST'
>>  14th          3.5   -61.3 -136.8 8K19A   60   4   'TEST'
>>  11th to 14th  9.9   -54.3 -137.6 8K19A   60   4   'TEST'
> Thanks for the table.
> What's the result of 15 th and 11th to 15 th on your side?
>
> Its better not to mix up the day and night transmissions now. The night 
> transmissions are a separate experiment. But if their phase is stable as 
> well, one could correct the phase offset and use them for a decode. So we 
> can simply avoid the time where QSB and the Terminator passes the path. To 
> be checked...
>
> 73, Stefan
>
>
>