Hoymiles Solar Microinverter FAIL

[EEVblog]

The Hoymiles Microinverter has failed after only 6 months.

[EEVblog]

Hoymiles run hot apparently.
https://mcelectrical.com.au/blog/hoymiles-microinverter-review/#:~:text=Hoymiles%20seems%20to%20work%20a,day%2C%20and%20in%20higher%20temperatures.

A viewer in the comment said he's installed over 1000 of them and the failure rate is around 2%

[EEVblog]

Hoymiles contacted me and suggest it could be firmware or grid setting option.
That doesn't explain why it worked for 4 months no problem.

[floobydust]

TAKE IT APART, DISASSEMBLE NO. 5 lol
My money is on CapXon or Chong_X lurking in there

[EEVblog]

TAKE IT APART, DISASSEMBLE NO. 5 lol
My money is on CapXon or Chong_X lurking in there

I assume identical but two channels.
AiShi

[madires]

Hoymiles contacted me and suggest it could be firmware or grid setting option.
That doesn't explain why it worked for 4 months no problem.

You're holding it wrong?

[Poroit]

Wow....mounted between a Tin roof and the rear of a Solar Panel...that thing must have been stinking hot !!

I hope you can pinpoint the fault and give us some feedback.

[EEVblog]

I hope you can pinpoint the fault and give us some feedback.

They are sending me the wireless interface thingo so I can access the temp reading inside the unit.
The panel covers practically all the tin roof, so there is no direct sun on those tin panels underneath.

[tszaboo]

TAKE IT APART, DISASSEMBLE NO. 5 lol
My money is on CapXon or Chong_X lurking in there

I assume identical but two channels.
AiShi

You said you were looking for equipment to repair.
I think you found some.

[EEVblog]

You said you were looking for equipment to repair.
I think you found some.

Hoymiles are very confident it's not a hardware failure.

[max_torque]

I'm in the "putting active" elements up on the roof is really silly camp!

In the UK, with our (sensible) health and safety laws, "working at height" is now a costly buisness, and whilst micro-inverters might in some circumstances be able to generate a higher return from a given panel array, as far as i can tell, not only do they pretty much fail to even cover there own (greater) cost, the additional installation complexity and the fact that any failure is going to be a costly replacement, just in terms of labour for most people (who realistically probably don't either know anything about power electronics nor want to clamber around a high roof)

[floobydust]

Nice teardown and the parts are surprisingly- not all from the Mainland. That would soon change lol. I have had the Golden Sample and then what follows...
I would power it up and see if the small SMPS is alive, check the fuse etc. to see if it's at least got MCU power or a blinky.

Something I surprisingly found is a silicone potting compound is not necessarily "thermal". Common ones are actually an insulator and don't transfer the heat. You can see this with a thermal IR cam. Parts cooking inside it worse than if in free air. Although, the MCU should have a temp sensor - but it's far away from the power electronics so who knows how hot it really gets.

You might take some temp readings (cam shots) of the gear on the roof, under the panels to see if there is reasonable airflow. If the microinverter is close to the panels then that's more radiated heat and  really it's got a fake heatsink with those tiny fins.

[thm_w]

Something I surprisingly found is a silicone potting compound is not necessarily "thermal". Common ones are actually an insulator and don't transfer the heat. You can see this with a thermal IR cam. Parts cooking inside it worse than if in free air. Although, the MCU should have a temp sensor - but it's far away from the power electronics so who knows how hot it really gets.

You might take some temp readings (cam shots) of the gear on the roof, under the panels to see if there is reasonable airflow. If the microinverter is close to the panels then that's more radiated heat and  really it's got a fake heatsink with those tiny fins.

Yes non-thermally enhanced silicone potting compound is a thing, but I'm not sure why you think its a worse insulator than air is. Remember the enclosure has to be sealed, there is not going to be any airflow or fans running inside.

As can be seen, a typical unfilled silicone rubber has a thermal conductivity of approximately 0.2 W/m·K and even a highly filled silicone elastomer can barely exceed 0.4 W/m·K.
The thermal conductivity of air is around 0.03 watts per meter kelvin (W/m K).

https://www.intertronics.co.uk/wp-content/uploads/2016/11/TB2007-12-Thermally-Conductive-Silicones.pdf

[EEVblog]

Something I surprisingly found is a silicone potting compound is not necessarily "thermal". Common ones are actually an insulator and don't transfer the heat.

There is no way a company making rooftop microinverters where temperature is critical would not use a thermally conductive potting compound to the die cast alloy case with finned heatsinking.

[floobydust]

Something I surprisingly found is a silicone potting compound is not necessarily "thermal". Common ones are actually an insulator and don't transfer the heat.

There is no way a company making rooftop microinverters where temperature is critical would not use a thermally conductive potting compound to the die cast alloy case with finned heatsinking.

Estimate how many watts of heat is it generating? 400W x 94% eff. is that 26W for a ballpark.
I don't know this particular potting compound, but my suspicion is based on it appearing low density in the video, and on the meager 60°C ambient rating the product has.
I'd stuff a hot transistor or resistor in that soft blob and look at with a thermal imaging cam. Should run much cooler than free air, right?
What happened for me is a product using Dow Sylgard 170 (Thermal Conductivity 0.48W/m・K "moderate") had components inside it cooking even at 1cm thickness. The stuff insulates when thickness is up there.

I think most of the heat flow is semi's under the PCB, conducted to the flat case. Not sure how thick it is, to keep HV rating there as well.
Thermal design for automotive ECU's is done much better and differently than this potted blob. The semi's are on the other side of the board so the copper thermal pads have thermal glue right to the enclosure. Not this IC expoxy in the way of the heat transfer which is another blunder.
We don't know yet why it failed, if it was heat related either.

[floobydust]

Something I surprisingly found is a silicone potting compound is not necessarily "thermal". Common ones are actually an insulator and don't transfer the heat. You can see this with a thermal IR cam. Parts cooking inside it worse than if in free air. Although, the MCU should have a temp sensor - but it's far away from the power electronics so who knows how hot it really gets.

You might take some temp readings (cam shots) of the gear on the roof, under the panels to see if there is reasonable airflow. If the microinverter is close to the panels then that's more radiated heat and  really it's got a fake heatsink with those tiny fins.

Yes non-thermally enhanced silicone potting compound is a thing, but I'm not sure why you think its a worse insulator than air is. Remember the enclosure has to be sealed, there is not going to be any airflow or fans running inside.

As can be seen, a typical unfilled silicone rubber has a thermal conductivity of approximately 0.2 W/m·K and even a highly filled silicone elastomer can barely exceed 0.4 W/m·K.
The thermal conductivity of air is around 0.03 watts per meter kelvin (W/m K).

https://www.intertronics.co.uk/wp-content/uploads/2016/11/TB2007-12-Thermally-Conductive-Silicones.pdf

Thickness matters too, this potting compound on the underbelly may be way thick although it needs that for the HV insulation rating.
You are using a thermal resistance value for air- but inside a box (unpotted) convection currents and radiation are significant factors toward cooling. It's not just the air 'conducting' the heat away.

[EEVblog]

We don't know yet why it failed, if it was heat related either.

Hoymiles are confident that it hasn't failed due to heat, because it has deliberate derating above 75C and shuts off entirely at 80C.
Given that, when it worked, I was getting the max 800W out of it routinely, I presume that it wasn't at 75C internally yet.

[thm_w]

Thickness matters too, this potting compound on the underbelly may be way thick although it needs that for the HV insulation rating.
You are using a thermal resistance value for air- but inside a box (unpotted) convection currents and radiation are significant factors toward cooling. It's not just the air 'conducting' the heat away.

True, thickness would be a constant, the design will have a gap to the enclosure, its either filled with air or silicone.
Efficiency is claimed 96.5%, it was two 300W panels? Or 800W worst case, so about 28W worst case dissipation.

I don't know if you can easily compare the two without simulating it.

https://www.nvent.com/sites/default/files/acquiadam_assets/2021-02/Spec-00488.pdf
https://www.elliottelectric.com/StaticPages/ElectricalReferences/Guides/Heat_Dissipation_Electrical_Enclosures.aspx

[EEVblog]

True, thickness would be a constant, the design will have a gap to the enclosure, its either filled with air or silicone.
Efficiency is claimed 96.5%, it was two 300W panels? Or 800W worst case, so about 28W worst case dissipation.

Two 440W panels connected. Output peaks at around 826W (see my graph), but doesn't seem to flat top, so maybe there is a little bit output power margin left there.
It's certainly not de-rating, so according to Hoymiles, it shouldn't be reaching 75C internally at that point.

[floobydust]

Is anyone independently testing or reviewing these microinverters on the basics like efficiency and operating temps? They have to last.

I've been through the engineering wringer with products relying on potting and totally hate it. I design it out.
Improper mix/cure, the extra manual labour costs and HSE, having it end up on your desk dead so you are in hell doing the autopsy.
One product used the thinner, hard cure potting compound- which would get into the RF module and detune it. Or shear off SMT parts (despite being called an "electronic" potting compound) over temperature extremes.
Or the antenna coax MCMX was not connected properly. Or the bootloader needed to be reloaded, or the blown fuse etc. etc. $10,000's tossed in the garbage over stuff like this.

If Hoymiles says it's not a heat related problem, then they must have an idea what happened. Unfortunate she croaked.

[EEVblog]

Is anyone independently testing or reviewing these microinverters on the basics like efficiency and operating temps? They have to last.

There are few independent dealers and organisations that are testing them, nothing standardised though AFAIK.

I've been through the engineering wringer with products relying on potting and totally hate it. I design it out.
Improper mix/cure, the extra manual labour costs and HSE, having it end up on your desk dead so you are in hell doing the autopsy.
One product used the thinner, hard cure potting compound- which would get into the RF module and detune it. Or shear off SMT parts (despite being called an "electronic" potting compound) over temperature extremes.
Or the antenna coax MCMX was not connected properly. Or the bootloader needed to be reloaded, or the blown fuse etc. etc. $10,000's tossed in the garbage over stuff like this.

I've worked on designs that needed helium inside welded alloy casings for thermal transfer to the case where direct thermal contact wasn't easy.
Had to use helium sniffers to check for leaks in production units.

If Hoymiles says it's not a heat related problem, then they must have an idea what happened. Unfortunate she croaked.

They have suggested "it’s possible that the issue could be resolved by updating the firmware or adjusting the grid files.", whatever they mean by that.

[tszaboo]

Is anyone independently testing or reviewing these microinverters on the basics like efficiency and operating temps? They have to last.

I've been through the engineering wringer with products relying on potting and totally hate it. I design it out.
Improper mix/cure, the extra manual labour costs and HSE, having it end up on your desk dead so you are in hell doing the autopsy.
One product used the thinner, hard cure potting compound- which would get into the RF module and detune it. Or shear off SMT parts (despite being called an "electronic" potting compound) over temperature extremes.
Or the antenna coax MCMX was not connected properly. Or the bootloader needed to be reloaded, or the blown fuse etc. etc. $10,000's tossed in the garbage over stuff like this.

If Hoymiles says it's not a heat related problem, then they must have an idea what happened. Unfortunate she croaked.

I agree, potting is bad. I have similar experience. With the explosion safe stuff, sometimes it's unavoidable, because you want to keep things safe. But then the additives, the 2nd component for this stuff is parts that have every single health hazard logo on it. Our CEO loves it, basically this is asked:
"Can we just pot it?". Every. Single. Project.
And it either works, or it goes to the garbage dump. No repairs. Stuff like MEMs sensors suddenly don't work in it, or it comes off from the production line with 4G constant acceleration reported. We had a product designed by one engineer where the battery wasn't accessible after potting. Which the firmware (of course) over-discharged. Impossible to fix, every unit goes to the garbage dump. Impossible to even disassemble to be compliant with EU waste regulations.

And yes, it ends up on your desk. And the only question I can ask is: What do you want me to do about it? What???

[coppice]

Dave,

Do you have any idea what sort of environmental temperature these things get up to in the hottest part of your summer? Things in direct sunlight can reach some really high temperatures in hot countries. Those microinverters are shielded from the direct sun by the panels, but I assume they still get quite toasty.

[floobydust]

This is also what I am wondering, what are the ballpark operating temperatures under the panels for an optimizer or microinverter.
It seems to be mounted on the rail underneath a solar panel. Also near the top, I think the heat (convection currents) would be cumulative from panels below.
I don't see any spiders or snakes there, it's gotta get hot

HMS microinverters (label) rated to 65°C ambient but we hear it was not a heat problem causing this failure.
From MC Solar & Electrical review:
"... got up to 77°C internally... on a 25°C spring day"
"If you’re concerned about longevity, the Hoymiles inverter runs ridiculously hot."
Snippet of Hoymile response to the review:
"Third, the temperature rise
1. Your test result shows that the internal temperature of the 4-in-1 Hoymiles microinverter is 10 degrees higher than that of the 1-in-1 Enphase microinverter. This is true, and higher temperature does affect the product’s life;
2. But internal temperature alone can not decide the product life and long-term reliability. Actually, in the early stage of development, we conducted a comprehensive research, simulation and test on the heat dissipation of the machine and the temperature rise of the internal components, and then we decided to place the most critical component the electrolytic capacitor (whose temperature rise has the most impact on the product’s life) at the coolest area inside the microinverter. This can ensure the longevity and reliability of the product."

[eTobey]

They have suggested "it’s possible that the issue could be resolved by updating the firmware or adjusting the grid files.", whatever they mean by that.

Maybe that firmware has a bug, where it did not derate or turn off? With 4 months of working properly, only the last month was the hottest?

One could investigate, how the weather was when it failed, compared to before. Especially having no wind at all could make a big difference.

[EEVblog]

They have suggested "it’s possible that the issue could be resolved by updating the firmware or adjusting the grid files.", whatever they mean by that.

Maybe that firmware has a bug, where it did not derate or turn off? With 4 months of working properly, only the last month was the hottest?

Correct, last month of operation was the start of summer.
Could very well be a firmware bug.