Solar inverter causing slow internet, line spectrum with software defined radio

[BoomBrush]

So this is more of a continuation from here. Basically my neighbour has a solar inverter which is located close to her phone line and as a result, our ADSL bandwidth suffers speed loss. This has been diagnosed by Telstra in writing. My hypothesis was the PWM frequency of the inverter had something to do with not being filtered correctly and is being induced into the telecommunications line. Because our ADSL uses frequencies under 2.2MHz, any sort of interference with a similar frequency would cause a lower signal to noise ratio.

I went over to her place today and plugged my software defined radio into her phone line to check out what the frequency spectrum looks like. A software defined radio is sort of like a spectrum analyser but it's USB and plugs into the computer and you can buy them pretty cheap. I used a Hamitup 100MHz upconverter as my one normally can't go below 25mhz or something like that. I also used a voltage divider to reduce the signal going into the SDR, I used a 10k and 1k resistor and had it in a 10:1 ratio. It seemed to work fine surprisingly.


So this is the control, where i'm looking at the telecommunications line WITHOUT the inverter operational. I turned it off at the power box. There seems to be some peaks around 1.6Mhz, no idea what they are.


This is when the inverter was turned on with a frequency range of 0 - 2MHz. There is a very clear centre frequency around about 1.55MHz, with harmonics extending either side of it. As ADSL uses frequencies below 2.2MHz, I believe these harmonics on the telecommunications line to be the cause of the slower internet. If some frequencies are unable to be used for adsl bandwidth, the modem will have to negotiate a connection with the DSLAM equipment at a slower speed.


This is identical as above, but the frequency has been shifted to be between 2MHz - 4MHz. As can be seen in this example, it appears the harmonics are dissipating frequency after about 3MHz.

So in theory, anything above 3MHz should still be able to be used without problem. Because the national broadband network's FTTN uses frequencies up to 25MHz (VDSL2), this means that while 3MHz might be unusable, the rest of the spectrum should be fine. I'll try to post an update on Monday.

[awallin]

Would be interesting to see changes (if any) in the spectrum if you start to put chokes on various wires of the system:
- DC wires from panels are probably not chopped at PWM?
- input AC of inverter? could leak 'backwards' from there?
- output AC of inverter?
- AC input of ADSL-box?

something like that? https://palomar-engineers.com/rfi-kits/acdc-power-line-chokes

[ogden]

So in theory, anything above 3MHz should still be able to be used without problem. Because the national broadband network's FTTN uses frequencies up to 25MHz (VDSL2), this means that while 3MHz might be unusable, the rest of the spectrum should be fine. I'll try to post an update on Monday.

You can't say for sure. Telco line and your resistor contraption attenuates not only interference from inverter but DSL signal as well. It's not about "can you see signal in the spectrum", but about signal to noise ratio - "is your signal stonger by __ dB than noise". High speed signals needs good SNR to operate w/o errors.

Inverter most likely does not meet EMC regulations and shall be replaced with proper one. Also running phone line along power line is bad idea, I would address both - inverter and line routing issues.

[BoomBrush]

You can't say for sure

I completely agree. My neighbour however does not have internet at her property and so no DSL signal is even present on the line. When testing at my property, whenever I plugged this into the phone line it immediately caused the ADSL to disconnect and it couldn't get a connection again, probably because I am loading the line or something to do with impedances I am not 100% sure on that aspect. Maybe I need an isolation transformer. When my modem becomes disconnected, sometimes it reconnects sometimes it doesn't. It's really inconsistent which confuses me even more.

Would be interesting to see changes (if any) in the spectrum if you start to put chokes on various wires of the system:
- DC wires from panels are probably not chopped at PWM?
- input AC of inverter? could leak 'backwards' from there?
- output AC of inverter?
- AC input of ADSL-box?

something like that? https://palomar-engineers.com/rfi-kits/acdc-power-line-chokes

Extremely useful information here, thank you. I was talking with a family friend about contacting people at my university and making this my thesis project. I am going down the line of telecommunications and he suggested if I wanted to continue this field of study it could turn into a Ph.D research area. Certainly a few years off that idea though, I'm only 19 and my grades aren't that amazing. I would however have to obtain one of these inverters, that would be the tricky bit.

[Kleinstein]

The solar inverters usually use relatively low PWM frequencies. So the emissions are probably high harmonics and the working frequency corresponds to the spacing of the peaks (maybe half as some peaks may be weak).

I would guess the inverter does not meat EMI limits - this may be a bad inverter, but also a bad installation (poor grounding) and added chokes (could be as simple as large clamp on ferrites) could solve the problem. Chances are the company that installed the system is responsible for fixing it - at least in theory.