"Don’t place too much on the video from YouTube for measurement, the compression and multiple handling in my making the video ..."
Sorry, but, you can't 'artifact' the kind of 'shifts' (lucky we are playing with elementary sine waves here and not something more complex!) I am seeing in the video. These (coding aberrations) would result in MAJOR distortions of sound (and we are talking a 'tone' here, pretty easy to detect/observe artifacts) ... try for instance looping (continually replaying) a segment that is continuous vs one of the segments that have the 'shift' and there is noticeable sound difference as well. Be sure to join ends of each segment as the sinusoid goes through zero in the same direction (otherwise MAJOR clicks are introduced; 180 degree phase changes make for a pronounced audible artifact).
What I said was accept that the video has been handled multiple times, and it might not be a 100% exemplar of what was heard for the entire time the transmissions were on the air. I did not say the shifts you are seeing are not there, nor did I claim any “artifacts” were causing them.
What I can say FOR SURE, since I made the video and have the original files, is that playing the video back using the same receiver settings as in the video and looking at the detected audio in any program at all, results in different pulse widths than when I look directly at the I/Q data. I suspect this is probably because of the AGC, but really am not all that worried about it. When I said “about 70 ms” I was not attempting to quantify exactly, as opposed to when I said 3.00 sec pulse-to-pulse (actually what I saw over hundreds of pulses was 3.000 sec +/- 2 ms, typically no more than 1 ms, and most often 0 ms, from first detected energy to first detected energy in the I/Q recording).
Also, when I look at different segments of the I/Q recordings I see further variations in pulse duration. Pulse durations measured, using the -10 dB point you quoted as the point of measurement along both the rise and fall, anything from 59 ms to over 80 ms. Over the just under hour or so I recorded the pulses where “about” 70 ms, while the period of the video, grabbed at random, happens to average out a little shorter than that. Yes, I understand what you are saying about the delayed pulses lengthening the overall pulse, but that would not account for SHORTER pulses than 64 ms.
This (the short pulses) is one of the reasons I think the pulses may have, let me say again, may have, varied in width over time. Think outside that one short recording and over larger periods of time.
(Signal process much? ... I'm thinking maybe "no" ... please don't impose your limitations on others if your ability in an area might be a bit 'short'.)
You would be incorrect, however I do not profess to be an expert in the field. A working knowledge of it as applied to such things as DSP based Doppler radar receiver processing, phase noise measurements, and “simple” things like the development and fielding of multiple phased array radar systems, yes, we do a little signal processing in those.
Nice attitude, considering I have not disagreed with you on anything other than the possibility of this coming from a single transmitter and to say that the pulses MAY not have stayed stable in width over extended periods. Even then I did not say you were wrong about multiple transmitters, I only said a single source could also result in the same affect and that I tend to lean towards one TX.
However, when talking about RF, multipath, and how it affects pulses, both direct and reflected, I have more than a little experience, call it 35+ years worth. And what I said was that while multiple transmitters could account for what you are seeing multipath and a single source also could. The same transmitted pulse from a single source can arrive at the detector with multiple propagation delays, resulting in smearing or lengthening the signal, particularly detectable on the trailing edge of the pulse. This can be indistinguishable from a separate transmitter in the short term and results in exactly the same kind of phase shifts when the “first” pulse ends and the “second” pulse is detected under it. I see this every day, well, at least every workday.
The fact that not once, in hundreds of pulses, did the leading edge smear or change from the 3.00 sec pulse-to-pulse timing suggest a single source and multipath to me. In all of those pulses, with the fading seen (including deep enough fades to make the signal undetectable), it is most likely that at least one time the path to the “shorter” distance source would have faded while the path to a “longer” distance source would have not, resulting in a change in pulse-to-pulse leading edge timing for that cycle. This should eventually be true for a multipath single source environment also, but is much more likely when dealing with multiple geographically dispersed sources.
I am not questioning, and have not questioned, what you are seeing on the pulses or the time you came up with. All I am saying, have said, is it does not take multiple sources to cause the same affect. I also am not questioning your average pulse width findings for that section of video, and yes I understand, and have from the beginning, what you mean by:
“Perform the PW measurements taking into consideration where the individual 'sinusoidal sources' are seen to overlap/interfere and your 'pulse width' starts to converge on more like a uniform 64 ms value ...
Hint: Use either a virtual 'ruler' or something actual and measure zero-zero crossings or peak-peak crossings of the 'signal' ... you will see where the apparent 'phase' changes in an otherwise contiguous series of zero-crossings or sinusoidal peaks near the beginnings and near the ends of the 'pulse' ... “
Just so you understand, I can see the same thing you are describing in the I/Q data. You can see the transition from one pulse to the “delayed” pulse under/behind it. In fact, for a few pulses you can see when the delayed pulse gets added to the early pulse, giving a nice reference point (both leading edges, since the trailing can be smeared) to measure the propagation delay difference between the two pulses.
This is an example of what I mean. A single pulse. Not during the video segment I put on YouTube, but about 12 seconds before that started and on the same frequency (and with the same 64’ish ms pulse lengths). The total pulse width is about 70.56 ms (first detected energy to last). But, you can clearly see there is more than one pulse here. Both pulses are about 64 ms long, but the second pulse starts about 5.6 ms after the first.
Picture 1 is the marker indicating the ends of the combined pulse, and the parts of each pulse. Picture 2 is the markers only on the first pulse. Picture 3 is the markers on only the late pulse. If you subtract the “t1” of Picture 2 from the “t1” of Picture 3 you get 6 ms, rounded to the nearest full ms. Actual delay is more like 5.6 ms. The fact that the late pusle is longer than the early can also be explained by multipath pulse stretching.
And the fact that this delayed pulse is not consistently delayed (it appears variably from 2 to 11 ms behind the first pulse leading edge) argues against multiple transmitters. In the case of two transmitters you would have two “shortest” paths, and so the most repeated paths. And while you might get multipath smearing from each of the two transmitters you should essentially always get the shortest path also. This should show up as a fixed delay between the two pulses for the strongest portions of each. I don’t see anything really indicating such a fixed delay. Of course, if there were many transmitters such a step might get lost, as you could see different ones at different times. But, you would think that occasionally you would see several in a row from the same pair.
On the other hand, the wide region hearing these pulses might support multiple TX sites, but then so could good propagation. Another indicator to support multiple transmitter sites would be the 5.6 ms delay in the pictures I posted above. 1 ms is roughly 300 km, so 5.6 ms would be roughly 1700 km. One single transmitter having a multipath component that travels 1700 km further than the most direct route to me is less likely than two transmitters, one of them 1700 km further (path length) from me than the other. Remember, I knew this when I first mentioned a single TX was the way I was inclined to think, and I still lean towards the single source based on the unvarying leading edge timing. My point was that there is more than one explanation for the events you are seeing on the tail of the pulse, and that alone does not mean multiple TXs, although it is good ammunition.