Double the lifetime of an inverter?

[DaveDE2]

You can build an Inverrter for any electrical or electronic equipment to last a Life Time. Super easy to do. Most just cannot afford it. Not to say they are not made because they are. Utilities, commercial, and Industrial, and Military customers have the deep pockets to do that. Consumer grade products are made cheap so they can sell, and keep reselling to you.

...and you could make it last a DOUBLE life time (think grand kids) by then cooling it by 10C...

[DaveDE2]

I agree with everything you say here with the added comment that a 10x power reduction improvement (in the fets and drive circuitry at a given switching frequency) is a tall order and the smaller devices always seem to make advances more quickly than the larger ones. Micro inverters can benefit from this sooner than large string inverters, which is great, then again they really could use that because they really bake up there on the roof. Non-micro inverters evidently are not there yet because (for example) the latest SMA inverters still have lots of electrolytics. I'm not saying we'll never get rid of the electrolytics but as of now all conventional inverters have them.

[Sunking]

You can build an Inverrter for any electrical or electronic equipment to last a Life Time. Super easy to do. Most just cannot afford it. Not to say they are not made because they are. Utilities, commercial, and Industrial, and Military customers have the deep pockets to do that. Consumer grade products are made cheap so they can sell, and keep reselling to you.

[jflorey2]

Whilst increasing the switching frequency by a factor of five decreases the needed bridge capacitance it would also increase the switching transistor power dissipation by a factor of five assuming the rise and fall times are the same

Only if the devices themselves are the same - and they are not. MOSFETs are much faster (and less lossy at high frequencies) than the bipolar devices they replaced. And gate capacitance energy (the primary loss factor in non resonant switchmode power supplies) keeps going down as new MOSFETS are released. Resonant gate drives eliminate most of the gate capacitance loss in MOSFET designs. New devices (like GANFETs) reduce capacitances by another order of magnitude, making a 1MHz design no more lossy from a switch perspective than an older 100KHz design.

And all the above make it that much easier to move away from electrolytics to film and ceramics. Which is why you are seeing fewer electrolytic caps in inverters today.

Inverter manufacturers probably care a lot more about efficiency than increasing longevity at this state of the art.

I think they care about efficiency, longevity, size and cost. Enphase would go out of business quickly if local dealers stopped carrying them because they were getting sick of frequent warranty repairs on their inverters.

[DaveDE2]

EDLC's also have a max voltage rating of just a few volts so would never be considered for bridge capacitance.

Whilst increasing the switching frequency by a factor of five decreases the needed bridge capacitance it would also increase the switching transistor power dissipation by a factor of five assuming the rise and fall times are the same, so there are a lot of trade offs.
Inverter manufacturers probably care a lot more about efficiency than increasing longevity at this state of the art.

[J.P.M.]

Just 2 cents. Thermal cycling is inevitable in a PV inverter every day. So perhaps any efforts to limit the max temps
reached each day will lower the temp delta & cycling stress. And switching converters truly stress the caps in the
power conversion section; we are still working on taming this beast. Bruce Roe

As has always been the case, and as in most things of that sort, methods of cooling equipment come down to knowing/learning what's going on, defining goals, knowing/finding how to best meet the defined goals, and meeting those goals in a safe, serviceable and cost effective way. An ongoing process as you note.

[bcroe]

Just 2 cents. Thermal cycling is inevitable in a PV inverter every day. So perhaps any efforts to limit the max temps
reached each day will lower the temp delta & cycling stress. And switching converters truly stress the caps in the
power conversion section; we are still working on taming this beast. Bruce Roe

[jflorey2]

Inverters are still full of electrolytic caps. The mfgs are trying to eliminate them but it's very difficult as no other technology has the charge density that aluminum electrolytics have, so they are still in your inverter and mine and if we can cool them they will last longer.

Well, EDLC's have far higher charge density, but have similar problems.

However, the path towards elimination is not replacement with an equal charge density, more reliable part; it's removing the need for them to begin with. Capacitors are needed in inverters because energy storage is needed, and energy storage needs are inversely proportional to frequency. In general, an inverter with a switching frequency (not inversion frequency) of 500KHz needs one-fifth the bridge capacitance of an inverter that runs at 100KHz. Once your frequencies get high enough, electrolytic caps can be replaced by film caps or even ceramics, both of which have much longer lifetimes. For example, SolarBridge (recently acquired by Sunpower) uses only ceramic and film caps in their microinverter designs.

[J.P.M.]

I
@JPM
Good points, but in my situation the the inverters have a heatsink with large straight fins approximately 1/2" apart and will probably never clog from dust and can easily be blown clear of any surface dust with my garage air-hose. Good point about dust though and I can see doing that once per year or so just for piece of mind.

I'm planning on giving the secure power supply something to do by having it supply fan power. So when the sun is up the fans are on, sun down they are off. I can't see how this could cause thermal stress.

The flow will probably never be "blocked". It's the dust buildup on the fins that reduces their effectiveness. depending on the nature of the dust in the air, the fouling layer need only be a fraction of a mmm. to inhibit performance and negate some or all of the improvement afforded by the increased film coefficients afforded by the increased velocities of forced convection.

It's a bit complicated, but using a natural convection design and then imposing or forcing more air past the fins will change things some in ways I spent a good part of an engineering career working with, through and around. In general, more cooling will be effected, but there are some considerations often overlooked, such as cleaning, and the idea of thermal shock, which is often more a function of the rate of cooling change introduced, and also when that change is initiated (usually, turn the fan on before the inverter as you seem to suggest). Blowing the fins with an air blast 1X/a while will probably be adequate as a cleaning expedient.

BTW, forced convection may also help keep critters like spiders out of the fin portion of a heat sink. They seem to like heat, or maybe the lower air velocities of nat. convection catch more bugs in the nets.

Take what you want of the above. Scrap the rest.

[DaveDE2]

Well, you didn't list any mechanisms; you listed an observation, which is that higher temperatures increase failure rates. The mechanism is what causes that increased failure rate. For example, one common temperature related failure mechanism is dopant diffusion. Dopants are intended to be stationary with respect to the semiconductor structure; if they move around, the characteristics of the device change and it eventually fails. Increased temperatures increase diffusion rates within semiconductors. There are several other failure mechanisms like this one which obey the general relationship described by the Arrhenius Equation, which describes an exponential increase in failure rates with increasing temperatures.

We have known about this relationship for decades. During that time we have been working to reduce its incidence by improving passivation, ohmic connections within semiconductors etc. In modern semiconductors - especially power electronics - they are now well controlled. That means that other failure mechanisms that depend on other factors begin to have a larger influence. Some examples include:

Electromigration. This depends on temperature as well as voltage stress and current density. Thus voltage, as well as temperature, plays a role in lifetime. A scheme that attempts to reduce operating temperatures by operating at higher voltages (a common engineering tradeoff) may therefore increase failure rates.

Die fracture. This occurs when differing thermal expansion coefficients cause microcracks in a semiconductor substrate. This is caused by _changes_ in temperature, rather than absolute temperature. Thus a scheme that attempts to minimize average temperature by cooling the device more rapidly may increase, rather than decrease, failure rates.

All of which means that following the belief that semiconductor lifetimes always decreases by a factor of two for every 10C increase in operating temperature can get you in trouble. If, for example, you aggressively cool a device and by doing so increase the range of temperatures it sees (and the rate at which it is warmed and cooled) you might cause the very problem you are seeking to avoid.

That being said, the primary failure mechanism in electrolytic capacitors _is_ very closely tied to the Arrhenius Equation. Thus keeping 1990's-era inverters cool was pretty important; they were full of electrolytics, and I saw at least two inverters that failed due to capacitor failure. As technology improves, and especially as operating frequencies increase, manufacturers have used fewer of these devices, and the link between temperature and failure rates has become weaker.

Thank you.

Everything you have argued here indicates that temperature is the primary conduit to failure, and that "we have known about this for decades". Touche!

Some of the stuff in your argument depends on thermal shock; electromigration, die fracture, etc. Running a fan across a massive aluminum heat sink on the outside of the box is hardly going to thermally shock anything. It takes a long time, on the order of 30 minutes or so for said system to stabilize. Who's trying to use a "scheme that attempts to reduce operating temperatures by operating at higher voltages (a common engineering tradeoff) may therefore increase failure rates"? I'm not. Say what?

Inverters are still full of electrolytic caps. The mfgs are trying to eliminate them but it's very difficult as no other technology has the charge density that aluminum electrolytics have, so they are still in your inverter and mine and if we can cool them they will last longer.

[jflorey2]

jflorey2
Please tell me what the other mechanisms for electronic failure are for a device in a properly designed circuit as described above.

Well, you didn't list any mechanisms; you listed an observation, which is that higher temperatures increase failure rates. The mechanism is what causes that increased failure rate. For example, one common temperature related failure mechanism is dopant diffusion. Dopants are intended to be stationary with respect to the semiconductor structure; if they move around, the characteristics of the device change and it eventually fails. Increased temperatures increase diffusion rates within semiconductors. There are several other failure mechanisms like this one which obey the general relationship described by the Arrhenius Equation, which describes an exponential increase in failure rates with increasing temperatures.

We have known about this relationship for decades. During that time we have been working to reduce its incidence by improving passivation, ohmic connections within semiconductors etc. In modern semiconductors - especially power electronics - they are now well controlled. That means that other failure mechanisms that depend on other factors begin to have a larger influence. Some examples include:

Electromigration. This depends on temperature as well as voltage stress and current density. Thus voltage, as well as temperature, plays a role in lifetime. A scheme that attempts to reduce operating temperatures by operating at higher voltages (a common engineering tradeoff) may therefore increase failure rates.

Die fracture. This occurs when differing thermal expansion coefficients cause microcracks in a semiconductor substrate. This is caused by _changes_ in temperature, rather than absolute temperature. Thus a scheme that attempts to minimize average temperature by cooling the device more rapidly may increase, rather than decrease, failure rates.

All of which means that following the belief that semiconductor lifetimes always decreases by a factor of two for every 10C increase in operating temperature can get you in trouble. If, for example, you aggressively cool a device and by doing so increase the range of temperatures it sees (and the rate at which it is warmed and cooled) you might cause the very problem you are seeking to avoid.

That being said, the primary failure mechanism in electrolytic capacitors _is_ very closely tied to the Arrhenius Equation. Thus keeping 1990's-era inverters cool was pretty important; they were full of electrolytics, and I saw at least two inverters that failed due to capacitor failure. As technology improves, and especially as operating frequencies increase, manufacturers have used fewer of these devices, and the link between temperature and failure rates has become weaker.

[ButchDeal]

Those are old photos. They have replaced much of the glue logic with silicon asic and ceramic capacitors. Dramatically reducing part counts and increasing reliability

http://www.solaredge.com/sites/default/files/solaredge-reliability-methodology.pdf

[DaveDE2]

Totally agree that cooling should be considered, and that the chances it helps probably outweighs the chances it hurts (vibration, fouling, etc). I think any suggestion that inverter life will double with the addition of a muffin fan is ridiculous.

I wouldn't rule out outdoor installations... better natural airflow could actually result in lower temps relative to inside. Also keep in mind that IGBT failures are up there with caps and fets in the possible failure modes.

I believe that the lifetime of a semiconductor or capacitor doubles for every 10C temperature decrease. This is based on all scientific data I've read including the two documents I referenced. If someone can show me other-wise then please do. This assumes that the electronic design of the inverter is done properly and all components are operating within the manufactures specified limits including max voltage, current, junction temps etc. I agree with you that if a small fan can not reduce the operating temperature of those devices by 10C then the assertion is ridiculous but so far my measurements hold promise there and I'm going to pursue it (I used a 10W fan to see an 18C decrease, but this is not average operating temp, investigation to be continued).

As far as outdoor installations, I ruled that out from the git-go here because of the risk of premature failure due to increased stress of environmental thermal cycling (solder joints), moisture ingress, etc. There may be an ideal climate where an outdoor installation would prevail but it sure isn't here in Colorado.

jflorey2

Please tell me what the other mechanisms for electronic failure are for a device in a properly designed circuit as described above.

@JPM
Good points, but in my situation the the inverters have a heatsink with large straight fins approximately 1/2" apart and will probably never clog from dust and can easily be blown clear of any surface dust with my garage air-hose. Good point about dust though and I can see doing that once per year or so just for piece of mind.

I'm planning on giving the secure power supply something to do by having it supply fan power. So when the sun is up the fans are on, sun down they are off. I can't see how this could cause thermal stress.

ButchDeal
There is no comparison between forced air cooling of my inverter fins and computer servers. I own a large restaurant that has a server and 12 other point of entry computer terminals. Each one of them is a PC which draws air into its innards and across the components, then in most cases exhausted by the power supply. Yes, they get very dirty with dust and have to be blown free of it a few times per year so that premature temperature related failures do not occur. In the case that I'm talking about here, forced air convection only occurs on the exterior of the box (heatsink) so there will be no clogging whatsoever.

I've attached a photo of the inverter fins.

[J.P.M.]

I don't think it has to do with its quality.

First failure is immediately after panel addition with disturbance AFCI error (human error)

I think within 3 months, 2nd failure with no error code. I did see the burn mark on L1 input and also found water in J Box (another human error)

Understood. Stuff happens. Thank you.

[silversaver]

Something doesn't sound quite right there. If another failure happens in 1+ yrs. or so, I'd look beyond inverter quality. SMA stuff seems pretty robust with a good reputation for quality.

I don't think it has to do with its quality.

First failure is immediately after panel addition with disturbance AFCI error (human error)

I think within 3 months, 2nd failure with no error code. I did see the burn mark on L1 input and also found water in J Box (another human error)