Разгадка сего опуса Эдгара. Сие была не Чайка.
Good evening,
Advice on noise source being solar system invertor,
"the inverters could well be operating on frequencies around 100kHz and
would spread over a significant spectrum. A typical DC/AC inverter
generally uses a MOSFET bridge chopping (square waves) at a high
frequency in the kHz region, then passes the signal through a low pass
filter to recover the 50Hz."
"the signal levels from your printout are very low, around a microvolt,
and whilst the inverter may well pass the EMI standards this does not
mean they don’t radiate anything."
Regards, Edgar
Moonah, Tasmania.
====================
Some more thoughts: Radio over the air could not be licensed in this frequency band. So it must be either PLC over mains wires, or some short-range inductive devices.
Fortunately, high-power inductive chargers for e-vehicles have not caught on much here, mostly due to their efficiency disadvantage. This is not an issue for low-power chargers (e.g. QI for smartphones). As far as I have seen, today's inductive kitchen stoves are no longer using primitive freerunning oscillators, but microprocessor-based and crystal-derived stepped frequency sources around 40 to 70 kHz.
Edgar sent a link to a 19.2 s long wav file, containing a recording of the 102 to 118 kHz band. Attached is a spectrogram with 62.5 mHz per bin, 5 ms scroll, and 30 dB amplitude range. At the top, we see a weak AM trace at 117 kHz (presumably an IM2 between two local mediumwave transmitters), which can serve as a frequency reference.
Again we find four independently drifting sequences of dashes. I have to say I am still not sure whether they are PLC devices conveying information, or an inductive charger or heater transporting energy. From the observations, it looks like we can find arguments for either:
1. As noted before, the MFSK frequencies are unevenly spaced, adhering accurately to a switchable divider scheme 100 MHz / (2*n+1)
=> While simplifying the transmitter, it would make demodulation at the receiver more difficult, requiring a Goertzel bank instead of an FFT.
2. There are four traces of 20 ms dashes, with partly overlapping frequency ranges. Overcrossings and symbol collisions do happen frequently.
=> Such a PLC system would inevitably experience self-QRM. Unwrapping channels would have to rely on long-enough orthogonal spread codes.
3. Center frequencies drift around slowly, and independently for each trace.
=> Makes no sense for digital communication, but might be needed for an inductive system to follow the coil resonance.
4. Often we find repeating four-symbol subsequences, but the subsequence changes after a while.
=> This looks too simple for a spread-spectrum PRN-sequence. It might however be a measurement technique to track the resonance.
5. Traces do have slightly different amplitude levels.
=> Obviously there are several physically separate sources.
6. Near the beginning, three symbol collisions occurred between the lower traces, all near antiphase (light grey).
=> Common frequency source, or just coincidence?
7. Timing appears to be precisely clocked by 50 Hz. The lower two traces switch ~ 6 ms before the others.
=> PLC's on different three-phase lines L1, L2, L3 with 6.67 ms lag between them?
A longer observation, including a 50 Hz reference from the grid, could perhaps decide whether the symbol timing actually follows small variations of the Australian mains frequency.
8. Ramps are visible between dashes in the spectrogram.
=> Frequency keying seems to be phase-continuous and not abrupt, perhaps meant to control spectral width.
9. Several devices seem to be on at all times.
=> This seems plausible for something like remotely read electricity meters, but less so for chargers or cookers.
I think the next step will be for Edgar to try to physically locate the signal source. Is it strongly present on the mains wires? Can you locate it by walking around with a portable receiver and ferrite loopstick?
73, Markus
