It's all about the Software

DXers are getting used to measuring persistent/stable frequency offsets to help identify some individual stations.

Time Difference of Arrival (TDOA) is a transmitter location technique using triangulation based on signals from different receivers (e.g. KiwiSDR). TDOA is beginning to make geographic location a realistic goal.

Could another kind of time-based analysis open another avenue with multiple signals on one receiver channel?

The particular case I'm interested in is MF broadcasting where we find many Single Frequency Networks (SFN) - good examples are UK and Spain. On a single channel it can be clear that we are hearing similar signals (the same programme) from multiple transmitters.

The audible result can be a kind of slow beat through to distinct echo(es), while the cause may be path length difference but more likely (I believe) time difference in the feeds to individual transmitters. However the raw audio is usually not enough to "resolve" how many transmitters are contributing, the time offsets, or the relative amplitudes.

From a Dummy perspective, an SDR feature that might help could be autocorrelation on a single channel I-Q stream, delivering a "spectrum" of peak amplitudes across the time axis (e.g. 0 to 1 second). I've no idea how feasible that is, or how much processing penalty would be incurred.

Please could Simon or anyone else advise?

Given that these networks exist and you can hear variable beat notes or even see them on a spectrum display, how do you distinguish a single station received via multiple paths from multiple slightly offset transmitters?

Supposing you are doing this during daylight extreme resolution FFTs might break out the various frequencies; but won't do anything about time of arrival. If the modulation is precisely the same on all the stations, a very high probability, you could correlate what is received on a frequency with a slightly delayed version of itself. This process would take place at baseband so sending it as audio to an external tool would lose little or nothing. An SSB demodulator (downconversion), perhaps very wide band, or an AM demodulator would be best to work with.

Off hand I'd request SSB with a 20 kHz audio bandwidth be provided. (I've not tried that. Maybe you can set that up in the filters editor.) Then work on that demodulated data in another application. Note that there probably is nothing to be gained by shipping the tuned frequency along with the audio because different radios have different actual step sizes in their synthesizers. To improve on this identical receivers with identical settings all locked to GPS (or equivalent) must be used.

{^_^}

jdow wrote: Thu Jan 28, 2021 8:35 pm Given that these networks exist and you can hear variable beat notes or even see them on a spectrum display, how do you distinguish a single station received via multiple paths from multiple slightly offset transmitters?

Joanne, thanks for hat, and your other suggestions. Of course there are limitations and ambiguities, not least if there are two or more emitters with identical arrival times.
At its very best, I'd hope to be able to build up some kind of picture over time, as we do with S-meter or any other measurement.
I'm merely asking how feasible it might be, this "analysis open[ing] another avenue".

I can see you building a picture of your environment. That would be an estimate of how many signals you could receive if all the others momentarily went away. Beyond that I am not sure you can build up much more data. The data going from where it is prepared to the transmitters is traveling at some unspecified rate over unspecified routes. So it is highly unlikely they are anything even remotely akin to GPS signal simultaneity. On the other hand more than a 10 ms path difference is likely to produce comb filter effects sufficient to slightly annoy listeners. 100 ms path differences would come through as echos. ("Why is that announcer in a gymnasium?")

The delays are not likely to be constant from day to day and may vary noticeably over smaller intervals.

If the station frequencies are not all GPS stable and synchronized their minute by minute frequency differences can vary with equipment temperature differences. So your data gathering will not sort nicely into frequency offset bins.

Basically a station count may be about the only thing you can be sure of acquiring and will only happen if a very high resolution FFT shows some number of stations during a time of day multipath (F2 bounce) is not likely. The maximum number acquired under those conditions would probably be a good count of stations receivable if they were alone on the frequency.

Good luck with it.

{^_^}

jdow wrote: Fri Jan 29, 2021 4:46 amI can see you building a picture of your environment. That would be an estimate of how many signals you could receive if all the others momentarily went away.

Yes, that's a better way of putting my objective, I suppose.

"The delays are not likely to be constant from day to day and may vary noticeably over smaller intervals.": I realise the first, but had not thought of the second.

"Basically a station count may be about the only thing you can be sure of acquiring and will only happen if a very high resolution FFT shows some number of stations during a time of day multipath (F2 bounce) is not likely." : I was not thinking to go so far as looking for propagation delay differences (interesting though that might be).

"If the station frequencies are not all GPS stable ... So your data gathering will not sort nicely into frequency offset bins.": Carrier offset measurement was not supposed to be part of the scope here^[1] and to be honest I'm not sure how that would affect autocorrelation.

Anyway, I'm not looking for another never-ending, repetitive and ultimately sterile debate. Thanks, you have pointed out several aspects I hadn't considered - my enthusiasm is fittingly dampened, and to the extent that feasibility seems of little importance now.

[1] I've made an earlier request for higher-resolution waterfall/spectrum to further that aim.