Solar Panel Interference

25 Sep 2014 - Jim Evans, G3VDB

In early august 2014, my immediate neighbour had a 16 panel PV system installed on the rear SE-facing roof. Since the system became operational, the inverter switching has caused RF noise spikes from 11MHz to around 70MHz. 24MHz (12m) is now unusable for most amateur radio purposes. [Seven years on there is no resolution to this interference, nor is there likely to be. I live with it.]

When the panels shut down at dusk, all is much quieter, until dawn, as shown in the following screenshots from a Funcube SDR. The screenshots show the start-up of the inverter at 'dawn' around 07:20 in September. Click on the image for the full size image which opens is a new Tab or Window.

This is a non-rigorous investigation of a new source of electrical noise in my neighbourhood using available uncalibrated equipment.

At the time of writing [25-09-2014], I do not have details of the equipment installed other than 16 panels visible on the ESE facing roof. We are on a housing estate of detached houses and separated by about 3.5m.

When I saw the scaffolding go up and the panel fixing frames going on the roof, I feared for my ability to hear any low level HF signals in the future. Before the installation was completed, I took a 192kHz wideband sample of each amateur band using a Funcube SDR to get an indication of the pre-installation background noise. I also have historic recordings of 50MHz from one year ago.

I was right to be concerned. I can detect wideband noise, low level carriers approximately every 2kHz and higher level carriers spaced at 19.2kHz from 11.5MHz to over 45MHz. While the carriers are less evident there is a raised noise floor from 3.5MHz and to over 85MHz. A further source of interference is detectable from 750kHz to over 70MHz with a 5-10kHz wide hash every 125kHz. I believe this is a plasma tv in the area. It is switched on from 06:50 to after midnight most days. It affects 14070 PSK badly and is even stronger on 14195kHz.

I have not raised the issue with this neighbour who, although supportive of my “wires in the garden”, is unlikely to react positively to a report that their attempts to save the planet are flawed.

All HF bands are affected to some extent and this will make DXing almost impossible during the day with background noise levels of S3 and carriers every couple of kHz at S5 to S9 [CW Bandwidth of 500Hz]. The indicated noise floor level depends on the rig used, preamp configuration and bandwidth.

On 25-09-2014 sunrise was at 06:58 with a cloudy start. After a few “false starts”, the hash started continuously at 07:17:38. See screen shot above.

On my F450D and FT1000MP I detect warbling “carriers” every 19.2kHz. On 24MHz the FT450 shows these as S6 and the FT1000MP (which has a wideband frontend above 14MHz) shows S9 with the preamp on and S6 with the preamp off. These readings may be summarised as S6[S9/S6]. With the FT450 [FT1000 PreAmp On/PreAmp Off] the background intermediate noise floor is a minimum of S3 [S7/S0] but there are other peaks about every 2kHz at S5 [S8/S3]. I believe this wideband noise overloads the initial circuits of the FT1000MP MkV because switching the pre-amp off can reduce the meter reading by more than the preamp gain.

These noise problems can also be seen with a (Funcube) SDR with HDSDR software where 192kHz can be displayed at once. With the antenna terminated in 50ohm, I set the S-meter close to S zero. The spectrum of the HDSDR software then gives (on this uncalibrated scale) a base reading of -152dB.

On connecting the antenna, at 24.9 MHz at no point is the noise below -133dB. Carrier peaks at ~19.2kHz from the SolarPV are -100dB (S6 on FT450). Intermediate SolarPV “carriers” at 1.9kHz intervals are typically -115 to -110dB. The higher carriers are clearly modulated, with multiple 50Hz sidebands.

The antenna consists of 20m and 40m horizontal dipoles across my garden at right angles on the same Coax at 8m (at roof ridge level and about 20 feet from the nearest PV panel).

I detect hash from this PV system from 1.5MHz to 70MHz or more. The highest noise level is around 42.8MHz (which may be related to the resonance of the 20m dipole on its third harmonic). At 42.8MHz the wideband noise floor is -110dB and the 19.2k separated peaks are -78.5dB on the same uncalibrated HDSDR scale. Further investigation [26-09] suggests that the noise floor at this frequency may be considerably elevated by another noise source. The base level is higher before the panels have enough light at dawn to switch on [the inverter?].

It is also possible that the FunCube has a higher front end gain or lower filter loss at this frequency; this is to be checked.