Cave Radio and the Loran-C spectrum
Posted: Mon 29 May 2017 10:29
I was disappointed that Robin Gape's article (Cave Radio and the Loran-C spectrum, CREGJ 97, pp7-9) made no reference to the significant discussion of Loran-C that has already taken place in the CREG journal, with articles by myself, Mike Bedford, Rob Gill, John Hey, Antonio Muñoz, Graham Naylor, John Rabson and Chris Trayner. With such a large amount of highly-relevant reference material already available for study, one cannot be blamed for critically asking what Robin's new contribution has to offer. Robin claimed to offer "an approach to mitigate the problem" but this appeared to amount suggesting that "alternative operating frequencies should be investigated". Then, in case the reader was not able to work it out for himself, Robin suggested that the alternative frequencies should be "field tested". This advice is superfluous, ... surely?
Robin hazarded a guess that "one could also envision the use of ... active systems, but this is most unlikely to be practical for the Heyphone". I think that Robin might be unaware of the state-of-the art of noise cancellation systems, and also unaware, I suspect, of the reports of tests of such systems in the CREG journal. The Loran signal is repetitive and predictable and, as such, is particularly suited to feed-forward noise cancellation, which can be "retro-fitted" to existing radios. (That's not to say that it would not require a dedicated filter design - off-the-shelf units might not suffice). It is true, though, that applying a noise cancellation system to the Heyphone is not necessarily the best approach. But I was surprised that Robin seemed apparently unaware of the two digital architectures proposed separately by Graham Naylor and Antonio Muñoz, which are specifically designed to combat Loran interference.
Additionally, I must take issue with Robin's statement that the spectrum of the signal comprises "spectral lines at multiples of the phase code repetition frequency [of approx 7.4 Hz]". That is true only for the strictly limited consideration of a single Loran beacon, and is not representative of the received signal, which comes from multiple beacons. Even if it were, it is not a very useful or helpful thing to say! As an example, suppose I produce a set of blasts on a 1 kHz whistle, repeating twice a second. To say that the spectrum comprises spectral lines at 2 Hz intervals would just be silly! The spectrum comprises a 1 kHz comb with 2 Hz sidelobes. It is the same with Loran - the spectrum is a 1 kHz comb with broad sidelobes that represent the different GRIs (group repetition intervals) of the different chains being picked up at the receiver. I demonstrated this in a CREG journal article, where I showed how one could use Matlab to derive a demonstration signal waveform and then perform a Fourier analysis on it. It is the broad sidelobes that make conventional analogue filtering less effective.
It is, of course, good that Robin has been interested enough to raise the topic again - we need to be continually reminded of the problems we face, in order to encourage fresh work. One point that Robin made, which I think everyone ought to be agreed on, by now, is that there is no need to constrain ourselves to 87 kHz. In fact, even when the Heyphone was first built over 20 years ago, I tried to point out that backwards compatibility was a mistake!
My own view is that we might benefit from investigating to a higher frequency, around 150-250 kHz. However, if one's main concern is Loran noise then, paradoxically, the Loran region may be the best place to host a cave radio signal, because the band is likely to be free from other interference; and if one approaches the noise elimination problem properly, one is left with a rather a quiet region. This is the same argument I used in my PhD thesis to suggest the siting of a cave radio transmission on the same frequency as a long-wave broadcast carrier but in the "quadrature phase space". The Loran signal is slightly different in that it comprises a number of superimposed signals of unknown phase and inconvenient periodicity but, nevertheless, it is certainly the case that a sound knowledge of the nature of the Loran signal (a 1 kHz comb) allows us to devise a noise elimination procedure along broadly similar lines to my quadrature phase space" concept. If, however, one wants to be conservative about receiver design, then the minor point of using the lower sideband at 87 kHz instead of the upper sideband would gain around 3 dB of signal/Loran noise; dropping to 83 kHz would gain 6 dB and dropping to 80 kHz would gain 10 dB. This can be seen from a Fourier transform of the Loran pulse, looking at the magnitude of each harmonic. (My figure 3 in the article in CREGJ 61; or, at least, a printout of the data corresponding to that figure).
Thus, I remain somewhat doubtful as to whether Robin's article actually advances the study of Loran that much; especially in the light of the large amount of information already published in the CREG journal. See, for example, the list produced by running a CREGJ search for Loran or Software Radio. Another source of information is the PhD thesis of Wouter J. Pelgrum, which has been published as a book, New Potential of Low-frequency Radionavigation in the 21st Century, ISBN 978-90-811198-1-8.
As a final footnote, I notice that the NELS website (which used to list the Loran beacon specifications) no longer exists, but you can find a useful table of Loran beacons, for Western Europe, at the Wayback Machine. Note that the " intended [...] antenna at Loophead" was eventually sited at Anthorn.
Robin hazarded a guess that "one could also envision the use of ... active systems, but this is most unlikely to be practical for the Heyphone". I think that Robin might be unaware of the state-of-the art of noise cancellation systems, and also unaware, I suspect, of the reports of tests of such systems in the CREG journal. The Loran signal is repetitive and predictable and, as such, is particularly suited to feed-forward noise cancellation, which can be "retro-fitted" to existing radios. (That's not to say that it would not require a dedicated filter design - off-the-shelf units might not suffice). It is true, though, that applying a noise cancellation system to the Heyphone is not necessarily the best approach. But I was surprised that Robin seemed apparently unaware of the two digital architectures proposed separately by Graham Naylor and Antonio Muñoz, which are specifically designed to combat Loran interference.
Additionally, I must take issue with Robin's statement that the spectrum of the signal comprises "spectral lines at multiples of the phase code repetition frequency [of approx 7.4 Hz]". That is true only for the strictly limited consideration of a single Loran beacon, and is not representative of the received signal, which comes from multiple beacons. Even if it were, it is not a very useful or helpful thing to say! As an example, suppose I produce a set of blasts on a 1 kHz whistle, repeating twice a second. To say that the spectrum comprises spectral lines at 2 Hz intervals would just be silly! The spectrum comprises a 1 kHz comb with 2 Hz sidelobes. It is the same with Loran - the spectrum is a 1 kHz comb with broad sidelobes that represent the different GRIs (group repetition intervals) of the different chains being picked up at the receiver. I demonstrated this in a CREG journal article, where I showed how one could use Matlab to derive a demonstration signal waveform and then perform a Fourier analysis on it. It is the broad sidelobes that make conventional analogue filtering less effective.
It is, of course, good that Robin has been interested enough to raise the topic again - we need to be continually reminded of the problems we face, in order to encourage fresh work. One point that Robin made, which I think everyone ought to be agreed on, by now, is that there is no need to constrain ourselves to 87 kHz. In fact, even when the Heyphone was first built over 20 years ago, I tried to point out that backwards compatibility was a mistake!
My own view is that we might benefit from investigating to a higher frequency, around 150-250 kHz. However, if one's main concern is Loran noise then, paradoxically, the Loran region may be the best place to host a cave radio signal, because the band is likely to be free from other interference; and if one approaches the noise elimination problem properly, one is left with a rather a quiet region. This is the same argument I used in my PhD thesis to suggest the siting of a cave radio transmission on the same frequency as a long-wave broadcast carrier but in the "quadrature phase space". The Loran signal is slightly different in that it comprises a number of superimposed signals of unknown phase and inconvenient periodicity but, nevertheless, it is certainly the case that a sound knowledge of the nature of the Loran signal (a 1 kHz comb) allows us to devise a noise elimination procedure along broadly similar lines to my quadrature phase space" concept. If, however, one wants to be conservative about receiver design, then the minor point of using the lower sideband at 87 kHz instead of the upper sideband would gain around 3 dB of signal/Loran noise; dropping to 83 kHz would gain 6 dB and dropping to 80 kHz would gain 10 dB. This can be seen from a Fourier transform of the Loran pulse, looking at the magnitude of each harmonic. (My figure 3 in the article in CREGJ 61; or, at least, a printout of the data corresponding to that figure).
Thus, I remain somewhat doubtful as to whether Robin's article actually advances the study of Loran that much; especially in the light of the large amount of information already published in the CREG journal. See, for example, the list produced by running a CREGJ search for Loran or Software Radio. Another source of information is the PhD thesis of Wouter J. Pelgrum, which has been published as a book, New Potential of Low-frequency Radionavigation in the 21st Century, ISBN 978-90-811198-1-8.
As a final footnote, I notice that the NELS website (which used to list the Loran beacon specifications) no longer exists, but you can find a useful table of Loran beacons, for Western Europe, at the Wayback Machine. Note that the " intended [...] antenna at Loophead" was eventually sited at Anthorn.