Deye PV microinverters lack a safety relay (EU)

[madires]

RelayGate: Deye Solar Microinverters Lack Essential Part (https://www.heise.de/news/RelayGate-Deye-Solar-Microinverters-Lack-Essential-Part-9220307.html)

TL;DR:
- they got their microinverters certified with the required safety relay
- sold tons of microinverters without the relay (more profit, 400k+ inverters sold in Germany)
- someone opened an inverter and noticed, others checked too
- big oopsie

[Marco]

Pre-bankruptcy trickery incoming, regulators need to immediately freeze their assets and warehouse stock.

PS. Deye's explanation is so much bullshit, they knew they needed redundant disconnect and doubling up on active switches is hardly an option. Relays are the only real option.

[fourtytwo42]

Interesting but not unusual, many times I have disassembled noisy (from an EMC perspective) power supplies and found the grid side EMC filter components missing and bridged out even though they carry CE labels.

[madires]

Anker also (MI60 inverters):
Balcony power plant manufacturer reacts: First inverter replacement begins (https://samagame.com/en/live/balcony-power-plant-manufacturer-reacts-first-inverter-replacement-begins/)

[Kleinstein]

AFAIK the extra relay is not required in all EU countries, though it looks like in Germany the relay is required. The pictures of the internals show a signnificant filter capacitance on the intermediate voltage level and thus they should need some extra method to turn off - just the semiconductors is not enough there. In most cases they turn off OK  and nothing happens, but I see 2 scenarios where the more software based mechanism can fail:
1)  during / after grid down a voltage transient damages semiconductors of the inverter and this way connects the charged filter capacitor (e.g. some 10-100 µF) to the grid side.
2) after disconnection a strong ESD even damages the semiconductors and this way connect the filter capacitors that still hold charge for quite some time after disconnect.
The main problem seems to be a large Chinese company that seem to be willing to fix the problem (exchance or deliver extra relay boxes) - this time is does not look like the disapear or bankrupt if get caught scheme. Not taking the regulations seriously also happend to a German car manufacturer.



Hardware with EMI filtering removed in later versions can be an issue. There could also be legit cases with the space for extra filtering included just in case or some version and they may meet the limits also with the reduced filter version.  If they produce EMI problems and have the filter not populated this is of cause suspicous and a lower cost version should not have the CE lable removed as well.

[Marco]

AFAIK the extra relay is not required in all EU countries

EN 50549-1 requires single fault tolerance for the disconnect, would it count if you used  inverting/SEPIC/Cuk for the boost circuit with an inactive boost circuit being one isolation and the inverter the other?

With a normal boost circuit I don't see how you could get away without the relays.

[switchabl]

If there is indeed a requirement to "disconnect" that strongly suggests a relay to me (but I don't have the standard).

However, as far as I can tell, EN 50549-1 is not a harmonised standard, so it's not directly relevant to CE and enforcing it would still be a matter of national implementation.

[uer166]

Usually a device that provides isolation in a normative sense is defined per some standard. I don't believe any kind of silicon can do it, and if you use a Sepic, it'll have to use caps that are rated as such (e.g. Y-caps for both forward and return).

Usually the base standard will call out a specific relay, breaker, contactor, or switch standards that are acceptable for isolation.

[f4eru]

This norm should be updated.
Mandating relays is a nightmare, they are long term unreliable stuff that really needs to be phased out.
the new version should limit itself to just a single failure resistant disabling, like everywhere in the machine directive.

I have had, on a product in qty 20k/y, a relay contact consistently failing after a few months because of outgassing of a nearby plastic was contaminating the contacts....
and a bunch of other random failures....
in a safety impacted application.....

[NiHaoMike]

If it's to disconnect in case a power semiconductor fails shorted, wouldn't a fuse do that automatically?

[f4eru]

Nope, it is for guaranteeing to stop in case of loss of mains.
it's old style.

[Kleinstein]

If it's to disconnect in case a power semiconductor fails shorted, wouldn't a fuse do that automatically?

In many cases yes, but not necessary when this happen just in the moment or after the mains connection is lost. The fuse would than need 2 failed semiconductors or a short on the output side.
The moment when the grid connection is lost has a somewhat increased risk that there are voltage spikes, e.g. a spike from the grid side may cause a GFI to open.

Under normal conditions the relay should not open or close under load. So I see no reason for the relay to wear out, even in decades of use. It is only of the relay is poorly build or choosen that it may be an issue. The same signal that detect the loss of grid connection will open the realy and stop the inverter circuit. Here the electronic part should stop before the realy actually opens. The main reason is to make sure the mains part is safely isolated from the energy stored in the intermediate voltage (e.g. some 400 V) capacitors. For fixed istallations I would have no real problem for allowing a little longer to get a reliable discharge of the capacitor(s) instead of the relay. For the plug in types this may not be fast enough.

Those small plug in PV installation are a stupid idea anyway: they produce not much energy as they often work at poor locations and the costs are relatively high for the rather low power. The only reason why they may in some cases are profitable is because of the billing system, not because of the real value they provide. Shifting money from one person to the other could be done more efficiently with less EMI and safty risks.

[Marco]

I have had, on a product in qty 20k/y, a relay contact consistently failing after a few months because of outgassing of a nearby plastic was contaminating the contacts....
and a bunch of other random failures....
in a safety impacted application.....

How are high voltage reed relays supposed to degrade? The contacts are hermetically sealed, the coil is potted, outside of failure modes it never switches current.

[Kleinstein]

This are not reed realys, but more "normal" (still seem to need some safety rating) relays.

[f4eru]

How are high voltage reed relays supposed to degrade?

Reed relays cannot cut the required spikes and power

I see no reason for the relay to wear out, even in decades of use.

You seem to have no experience with reliability of relays in series products

[Siwastaja]

Besides, relays have the same failure mode as semiconductors, fail as short circuit. So just saying "relay is good" is cargo cult engineering. The relay has to be specifically engineered and well tested to provide the safety, even in presence of worst case currents and voltages. Exactly the same applies to semiconductor solutions!

Now admittedly designing a reliable and safe semiconductor disconnect and proving it is more difficult - but not impossible. But in the end it might end up more expensive than the relay.

[Marco]

Reed relays cannot cut the required spikes and power

That's just a question of money. You can buy anything in a hermetically sealed package if you pay.

[Marco]

Now admittedly designing a reliable and safe semiconductor disconnect and proving it is more difficult - but not impossible. But in the end it might end up more expensive than the relay.

If it's to be redundant resistance is also a factor.

[f4eru]

That's just a question of money. You can buy anything in a hermetically sealed package if you pay.

Nope. Why do you think most power relays are NOT sealed ?
It is because sealing also confines the outgassing and particles emanations of the sparking contacts, and contaminates the contact insulation, mechanical bearing surfaces and coils with abrasive conductive dust and other gassy nastyness, compromising the relay on the medium term...

Sometimes, a sealed relay can fail faster than the same unsealed.

[Marco]

But how is that relevant in this particular situation? It's zero current switching under normal conditions. Apart from the temperature, it's ideal condition.

Given the no single point of failure standard and size of the market, you'd expect purpose designed relays for consumer grid attached generators instead of having to rely on stuff made for appliances.

[f4eru]

How are high voltage reed relays supposed to degrade? The contacts are hermetically sealed, the coil is potted, outside of failure modes it never switches current.

The list of possible failures is very very long.
It is a really big mistake to take relays as a simple coil'n'switch.
Relay reliability and switching dynamics are a really very complex topic, also depending a lot on the environment, especially the source and load transient response and impedances.

Even high end sealed HV relays fail, an example :
https://www.spcd.space/proceedings/2018/1st%20Day/Lessons%20Learned%20and%20In-flight%20Experiences/Presentation/05.%20Relays%20Failures.pdf

What is crazy is that even ESA could identify a root cause only for half of the observed relay failures  And they put in real means to try to get to the bottom of the rabbit hole.

outside of failure modes it never switches current.

Wrong. An AC disconnect relay switches current on every cycle (1-5 per day in a solar inverter)
The X and Y-capacitors will be charged by a high inrush current on every switch closing, typically a microsecond scale spike of 300A if the switch closes on the max AC peak.
That erodes contacts, especially fine contacts that would not be engineered for high currents.
If a DC-link cap is also charged at that moment, it becomes approx 100x worse.
The effects, lifetime, and failures can never be fully validated and guaranteed, because of the random variation in cable length and inductance on the AC network, among others like mounting angle affecting spark rising......

[Marco]

Or you just switch on zero crossing.

[madires]

AFAIK, two contactors/relays in series are required. But for low power inverters it's tolerated when the switching bridge also does the job of one safety relay.

[f4eru]

Yep. Just another new can of worms.
in short: nice try, but does not work either, because closing/opening time can vary over lifetime.

[Siwastaja]

Now admittedly designing a reliable and safe semiconductor disconnect and proving it is more difficult - but not impossible. But in the end it might end up more expensive than the relay.

If it's to be redundant resistance is also a factor.

Relays have a lot more resistance than people seem to assume, because the metal contacts can handle high temperatures and relays are large and thus have significant surface area for cooling. With a semiconductor solution, unless you want to design in a metal heatsink (no one wants, it's expensive and you need to think about mounting), you have to engineer it for less resistance than the relay solution.