An inverter generates horrible RF jamming comms for hundreds of meters

[Fflint]

I need some help in my attempt to fix this inverter. It is a solar olinverter, but that really is irrelevant , because the problem is with a  high voltage DC to AC inverter module.

I know it is not a design fault because I have more than one, and only one became an RF jammer. It is a so called "a hybrid "(in quotes as it's not a real hybrid) solar inverter all in one. Its contains the following modules:
- A  bypass relay that can connect AC in to AC out or disconnect them
- A  bi-directional inverter between AC (voltage range 120V to 275V) and a high voltage (430 to 480V) DC bus. It is used to create AC from DC when the grid is down and bypass is opened. It can be used to feed power to the grid when grid is on.  In both cases it synchronises it's phase to the grid.
- a rectifier/booster - this just converts the grid input to high voltage DC for battery charging.
- a 44 to 62V battery charger powered by the HV DC bus
- two solar mppt to HV DC bus controllers.

And that is it. A relatively simple device.

When it is "off" its AC generation is off, but it still rectifiers the grid, maintains the HV DC bus, is able to charge the battery and mppts can be active depending on config. In this state it is not generating any substantial RF.

However when it is switched "on" so the inverter is creating AC it makes horrible RF noise that exits through every single cable including ground as far as I can tell. It generates huge signals at 5Mhz and it's multiples as well as the entire range 24-44mhz (every few kb there is a huge spike 40dB over background).

The inverter design from what I read is very basic. There is no isolation between the HV DC and AC. I believe it just has a bunch of mosfets in parallel and a filter. That is it.

When it initially came to me it seemed to fail to start with "HV DC bus shorted" error, but then after I disassembled it and put it back together it started working . I thought maybe some piece of wire got inside or other thing that caused a short, but now I'm not so sure.

Let's say one mosfet or a diode failed in a bunch. Could it fail open (or it failed almost open and these few jolts opened it fully) and being connected with its neighbours in parallel, can it generate such noise in result?

This noise is practically invisible with an oscilloscope on the ACs signal. It is few mV maybe while the AC is 240V.

I also wonder if there was a cold solder joint in the filter could it result with such RFI?

Thanks in advance for any responses?

P.S. I saw someone going to a great effort to shield the chokes. Put them in a cast aluminium enclosure. And so on. Is there any reason to shield the chokes?

[NiHaoMike]

Check the "X" capacitors in the AC filter, a common failure mode is for them to go open.

[Fflint]

Check the "X" capacitors in the AC filter, a common failure mode is for them to go open.

Thanks. I'll do that when I have it on my bench.

Just to add. A service manual (for a very similar device- they are all clones of eachother differing only in some options, like grid feed in, being enabled or not, and component numbers/quality) shows the inverter as just this:



If it is true there is indeed nothing else there to break other than the capacitors (perhaps a coil can have a short, but that would be visible in the output AC I guess). However, MOSFETS are shown here, elsewhere the manual talks about checking IGBTs in the "Check the INV full bridge on MAIN board", gives types, also shows pictures of drivers as well as optocouplers. So it is really just a conceptual drawing - ah a joy of Chinese manuals....

Same section has nothing about the filtering. It contains various SMD resistor values (which will come handy too).

This is the block diagram of the whole device:



BTW there is a swicth mode power supply fed from the battery and AC there as well. Unfortunately there is no schematic (for this model). I've heard about SPSes having such an effect. I doubt this would be it as the SPS is active even if the inverter is not creating AC and in such state no RF noise is seen.

[Jeroen3]

If it's jamming RF, the noise is common mode, typically invisible to probing on the outputs.

Get a TinySA 
The antenna or a set of cheap amazon near fields can probably tell you more.

[Fflint]

If it's jamming RF, the noise is common mode, typically invisible to probing on the outputs.

Get a TinySA 
The antenna or a set of cheap amazon near fields can probably tell you more.

I have a tinysa. All it tells me is that the noise it all around.

However ,it turns out the inverter is not the source. I was mistaken. So the hunt continuous.

I'll have to make a directional antenna for the tinysa. I saw a small loop online, but i doubt it has a lot of directionality.

I wonder if a ferrite rod antenna could work on 28mhz or is the frequency too high.

[Jeroen3]

If you suspect an appliance, sniff around with this:
https://www.amazon.com/Magnetic-Conduction-Radiation-Circuit-Antenna/dp/B0CBMN6PY1/ref=sr_1_1_sspa

[Fflint]

If you suspect an appliance, sniff around with this:
https://www.amazon.com/Magnetic-Conduction-Radiation-Circuit-Antenna/dp/B0CBMN6PY1/ref=sr_1_1_sspa

The noise is way too strong for these, also they are far too broadband.

I built a tiny antenna like this resonant at 29.5MHz:


With this I discovered it is not just one inverter... It's all of my solar inverters...plus Poe switch and the laptop's PSU.

While I can easily filter the laptop PSU and the PoE
Switch uses łan cables as antennas (they respond to 5 turns on a ferrite toroid very well) the solar inverters is a different story.

For some reason they generate most noise even when doing absolutely nothing. There is a tiny switch mode PSU in them. I suspect the actual inverter modules are well designed.But that PSU may have been an afterthought.

Also, I tried to find out which cables emit the most RF and very weirdly it is theDC cables to the batteries. I measured -47dB with this antenna touching and tinysa (lna on). (despite not using the battery at all! Inverters are in AC pass through mode for most of winter and neither charge nor draw much current from the battery. Maximum of few amps )

Then it's the AC neutrals with its -51dB, then it's AC live out  at -52dB and AC live in at -53dB. Finally solar wires are around -57dB and ground is around -71dB  (background is -120dB).

Ive ordered quite a few of material 31 ft240 cores to test . But now I wonder if my DC battery wires will not just saturated these ferrites.  The current there cab be in the hundreds of amps (150 charging per each of 3 inverters, discharge up to 250A per inverter).

I searched "entire Internet" and I've not found how to calculate a one turn saturation current on aFT240-31 amidon core. If anyone knows let me know please .

[jbb]

You know, the noise coming out over those battery wires could be common mode noise (ie same current from each wire). You could try putting a single ferrite ring around both the positive and negative battery wires together.

[nctnico]

Maybe there is a recall on the inverters. But solar inverters are problematic for RF reception. The Tetra system used in the Netherlands by the police, fire brigade and ambulances lost about 10% of the coverage due to interference of solar inverters. Actually there is a building called 'World Forum' in The Hague where they have to switch off the solar panels in case there is an event in the city. With the panels on, the Tetra radio system doesn't work well enough. Appearantly due to optimisers / micro-inverters used on the panels.

[Fflint]

You know, the noise coming out over those battery wires could be common mode noise (ie same current from each wire). You could try putting a single ferrite ring around both the positive and negative battery wires together.

Yes, it's possible. I'll try when my cores arrive. If this really is common mode noise I'll need lots of these cores.

Also I'm thinking what else can I do about grounding. I'm already running the shortest, the straightest possible wire (it is still about 2~4m long - ideal to radiate on the band I'm getting interference on) and I read ~-70dB of noise on it (30dB less than dc wires).

Perhaps I can tune this out somehow?

I though placing ferrite cores on grounding wires is a no-no as it prevents "shorting of the rf to the ground". But I saw some commercial rfi filters that have a common mode choke for all 3 wires (live, neutral, grounding).

Anyone has any advice on this?

Maybe there is a recall on the inverters. But solar inverters are problematic for RF reception. The Tetra system used in the Netherlands by the police, fire brigade and ambulances lost about 10% of the coverage due to interference of solar inverters. Actually there is a building called 'World Forum' in The Hague where they have to switch off the solar panels in case there is an event in the city. With the panels on, the Tetra radio system doesn't work well enough. Appearantly due to optimisers / micro-inverters used on the panels.

Chinese language has no word for recall :-D

Wow, this info from the Hague is amazing. If a big institution can't buy equipment free from EMI issues what hope a normal person with their normal budget has?

[Whales]

For some reason they generate most noise even when doing absolutely nothing.

Most SMPS (including inverters) enter a "pulse skipping" mode at low loads.  This mixes lower frequencies into the emissions profile, often low enough you can hear them by putting your ear nearby (please don't do this on an exposed inverter PCB!).

Many pulse-skipping implementations introduce a wide, unstable selection of frequencies.  This makes them sound like a crowd of people hoarse-throat screaming rather than a single tone/note.

You may or may not actually be getting more radio interference emissions power at these low loads, but the changed frequency spectra of them might make them easier to notice.

[Fflint]

For some reason they generate most noise even when doing absolutely nothing.

Most SMPS (including inverters) enter a "pulse skipping" mode at low loads.  This mixes lower frequencies into the emissions profile, often low enough you can hear them by putting your ear nearby (please don't do this on an exposed inverter PCB!).

Many pulse-skipping implementations introduce a wide, unstable selection of frequencies.  This makes them sound like a crowd of people hoarse-throat screaming rather than a single tone/note.

You may or may not actually be getting more radio interference emissions power at these low loads, but the changed frequency spectra of them might make them easier to notice.

I think that may very well be happening with the main inverter module. But I think the tiny SMPS that powers the logic runs at all times when it gets any power. However , there is huge difference between it running from DC and AC.

See my short video below. Where it says AC input on I flip the breaker to supply AC to the inverter (it is set to not use the battery to power outputs). Then it enables it's AC output and the main inverter I suppose goes into this "idle" mode. Where it is synced with the AC phase and ready to take over in under half a cycle if AC in fails.

Then I switch off AC in and it switches off (you can see caps discharging as EMI drops)

https://youtube.com/shorts/Zvgo2B5zP1w?si=vVTr46ks7GX8mnY3

[Fflint]

It is sad this whole topic of EMI generated by solar inverters (and EVs) doesn't interest the very people that could resolve it.

I saw threads on various forums offering same recycled info: line filters + toroid core common mode chokes.

Adding 3 ft240 cores (materials 2*31 and 1x43) to battery leads directly below the inverter made very little difference. Maybe a dB or two. I can't really tell.

I expected the battery leads to be the main culprit because I measured the strongest rf next to battery cases. They are not (and cannot be) grounded properly.

But as I said almost no difference. This despite coming to a conclusion, this is common mode RFI. How did I decide it? I made a coil that wrapped around both battery wires and it had the same signal level as when it was next to one wire.


So then I routed all the other cables via ferrite cores as well. These cables are ground, input AC (line and neutral), and output AC. Altogether 5 wires wrapped 3 to 5 turns on a stack if 2 ft240 material 31 toroids.

This did help by about 10db.

I also tried adding a capacitor (10nF Y type) between AC lines and ground. What I found with one type of inverter  is that only when the inverter is off, if I short neutral if the output to ground with a short thick wire the RFi drops by further 10dB. But it doesn't work when the device is on...

Another type on the other hand when I bypassed the common mode choke and grounded the case directly the RFi INCREASED by 6db roughly.

I also observed an increase in RFi when rs232/rs485 cables were plugged in. I added small unknown toroids to them and they stopped doing that.

So now I have no clue. Most likely the RFi is common mode and transmitted by everything. From the case through every wire. I imagine the only way to fix it is to keep adding chokes... Anyone has a better idea?

It is not a good result to achieve 10dB of attenuation and using up 5 expensive ft240 cores. But maybe I'll get more. If I could get another 10dB that would hopefully take it below the noise levels.

[Whales]

Ideally you would put every cable (DC, AC, signals) as a single bundle going though a toroid together, rather than putting toroids in separate bundles.  If DC cables exit one side of the board and AC cables the other then you have a dipole antenna.  Alas this would require very big ferrites, lots of insulation to be safety compliant and might cause new problems when noise in shared onto signal wires.

Cable chokes should not be the first option considered, good design of the switcher circuitry is the first step.  When you are working with off-the-shelve equipment, however, your options are limited.

[Fflint]

Indeed it is an off the shelf equipment. I have extras so I can put one on a workbench and mess with it so an easy modification is not out of the question. But I think it may well be the limitation of the design used.

You mentioned a single toroid is better than many toroids. Interesting. I thought the opposite (as long as both wires that form a potential difference are there) because I can fit more turns if each wire pair  has its own toroid (or let's say half go through one and the other half through the other).

Could you elaborate on this please ? Why should one use a single toroid for all kinds of wires? If we have a differential signal the toroids do nothing so se let's focus on common mode. Let's say we have a common mode signal on both the DC wires and ac wires. Then the toroid the DC wires go through will cause the same attenuation as a toroid the AC wires go through if the number of turns is the same.

Also I could fit in all wires through a bunch of single toroids, but the DC would get no turns at all (on the account on thickness), AC could get maybe one turn. Now I still get no turns on dc, but 4-5 turns on AC. So I'm not sure it is worth to change.

But I still have one extra inverter to do.

I had 14 ft240-31 cores and 8 ft240-43 cores (for 3 inverters). If I can get to 20dB attenuation of this rfi by getting the same amount again I'll be fairly happy.

[Whales]

Let's say we have a common mode signal on both the DC wires and ac wires. Then the toroid the DC wires go through will cause the same attenuation as a toroid the AC wires go through if the number of turns is the same.

You're probably right.  I was imagining different sized toroids on each bundle of wires.

If we have a differential signal the toroids do nothing so se let's focus on common mode.

I believe there is still an impact on differential mode as it still acts as an inductor.

Clip-on ferrites performance depends strongly on user configuration so the datasheets do not directly spec impedences @ frequencies, so instead checkout the graphs for this SMD CMC https://www.lcsc.com/datasheet/lcsc_datasheet_2306021132_MetalLions-ACM1211F-102T60_C908582.pdf