The Great Microinverter Debate

It was only a couple years ago that Enphase released the M175 onto the market. The follow-up release of the M190 really got the ball rolling, capturing a large part of the California inverter market. The M190 was the first “micro-inverter” to really succeed.

And with that success came the slings and arrows. Traditional big-iron inverter companies started pooh-poohing the upstart, stating that there was no way the product could ever be competitive. Enphase fought back, unleashing Raghu Belur on an unsuspecting market. His argument about scaling factors following the computer market model, a Moore’s Law of inverters, struck a chord.

The debate quickly grew heated as Enphase sales continued to ramp up. The competition became increasingly frustrated, launching a variety of attacks against both the product and the company. I’ve had representatives from major inverter and panel companies badmouth them right to my face, apparently oblivious of that damage this does to the market as a whole (it’s a small sandbox kids, play nice).

And in spite of looking really closely at all of this, I still can’t make up my mind about who’s got the best argument. And when I say I’ve looked really hard, the readers that know me will understand exactly what that means. So, in lieu of any strong conclusions, I’m going to spend some time -well, a lot really- simply laying out what I’ve learned so far.

Note: I’ve published a market update you might want to read. You can click it now, it will open in another tab/window.

The basics

Solar panels produce DC power, like a battery. Your house, and most everything else, runs on AC power, what comes out of a wall socket.

Systems that convert one to the other are generically known as “converters”, but we don’t use that term all that often in practice. Instead, we call AC-to-DC converters “power supplies” (or “rectifiers”, inaccurately), and the DC-to-AC versions “inverters”. There’s a whole class of devices that convert DC-to-DC in order to change the voltage, which go by a variety of names. There’s no equivalent AC-to-AC circuit, because that task is easily handled by a transformer – that’s why we use AC for everything, transformers are cheap and efficient.

You can buy an inverter at Canadian Tire that you plug into your car’s lighter socket that lets you power small devices like TV sets or radios. They’re cheap and not that bad in efficiency terms. In the case of solar, though, these devices are missing two very important features.

The first is the accuracy of the output power. Most devices will work fine if the power is “AC-like”, where “like” varies from something entirely unlike the grid to increasingly accurate approximations. The power company will not accept this, and they demand far more accurate renditions of their power. Not only does the waveform have to look like a clean sine signal, but the signal has to match the voltage, frequency and phase that the company is using. We talk about 120V power at 60Hz as the basis for the North American grid, but in practice this can vary a whole lot in the field, and the inverter needs to match it in real-time.

The other issue is specific to solar. PV systems deliver their power efficiently only when they are presented with the proper “load”. The relationship between incoming sunlight, temperature and load is complex. This is handled by a system known as a “MPPT“, a DC-to-DC converter that loads the panel as conditions vary, and converts the output voltage so it matches what the inverter wants.

Traditional PV inverters consist of three basic parts: the DC-to-DC MPPT, the DC-to-AC inverter, and control electronics that tune the operation of both. The electronics also watch for various fault conditions in the system and on the grid, and cuts power when certain red flags come up. This is known as “anti-islanding”, and it’s intended to protect the electrical workers when they come out to fix the wires in the event of a blackout – they don’t want to face your live wire, so the inverter turns off if the grid is down.

The inverter market

For most of the history of PV inverters they’ve been based on a basic circuit design known as “PWM high-frequency conversion”. We won’t get into the technology here, but suffice it to say that practically everyone – SMA, Xantrax, Fronius, everyone – produced designs based on this concept. That’s designs, as opposed to design, for a reason…

Inverters are only really efficient when they work near a specific design power. In addition, generally speaking, the higher the voltage the better. You can get higher voltage from your panels by stringing them in series like Christmas tree lights, each one adding about 30 to 40V. There’s an upper limit due to the cabling, which normally tops out at 600V, which gets you about 15 panels or so in a “string”.

Ok, so let’s make an inverter that’s tuned for 600V and 7 kW of total power, two string of 15 panels. Ok, fine, but now I ship it to a customer with a small home that can only fit 12 panels on the roof. His efficiency goes to hell.

It’s possible to design your inverter so a single chassis can have the parts inside swapped to produce models at different sizes. In most cases you don’t even have to change that much, because the controller electronics and cables and things all stay the same. Ahhh, but that’s not true for the transformer. It’s a big block of iron wrapped in very expensive copper, and its sizing is part of the fundamental efficiency criterion. At a minimum you need to change the transformer for the different models, and when you do, a bunch of other components end up changing too.

Building small number of different things is generally not a good idea. It means you have small production runs and all sorts of inventory issues, both of which drive up costs. That’s a problem with any mass produced product, but in the case of solar there’s a more subtle problem as well. We can design our line to minimize what needs to be changed for different models, like cables and electronics. In a large inverter this might represent only a fraction of the cost, so changing from the 5 kW model to the 7 kW one scales pretty smoothly – the parts you’re swapping are the real value. But when you start scaling down that falls apart – the parts in the 1 kW model might be pretty much the same price as the 1.5 version. The price is no longer scaling with size.

This isn’t trifling matter, it means that small systems will always be less cost effective than large ones.

And once the customer has selected a model, if they want to add a panel they have to buy a new inverter. Generally both the customers and installers want to get the price and efficiency maxed out. The best way to do that is put every panel you can onto a given inverter… got a 5 kW model that says it can actually handle 5.5 kW of panels? Put 5.5 on it! But now when you buy two more panels and try to make it a 6 kW system? Ka-blooie!

It’s all bad.

It would be much better if we had a single inverter model that could be used across a wide variety of installations and different power levels, from one panel to one million.

The microinverter concept

So all of this naturally gives rise to the concept of an inverter specifically designed to work with a single panel. Panels generally come at about the same power ratings, modern ones all fall between about 235 and 260 Watts for instance. If I have a small range of power settings to worry about, it becomes possible to make an inverter that works with any “system” out there. One panel? Fine, obviously. Two panels? No problem, just wire them in parallel. Five? Ten? 100 panels? Go nuts!

There’s huge advantage that this exposes. Solar panels are funny things. When you shade them, they produce less power. No surprise there. But they also increase their resistance to power flowing through them. It’s a problem because when you string those panels together in series to feed them into your inverter, a shadow on any one panel drives down the production of the entire string. If you use single-panel inverters wired in parallel, this problem is eliminated. Sure, the shaded panel will lose power, but that will have no effect on anyone else.

And since every system from one to a million panels uses the same model of inverter, your production line is building a large number of a single basic design. That’s great for driving down the prices. And since there’s one on every panel, instead of one on every string, you’re producing maybe 10 to 20 times as many of them in total. And that can really drive the prices down. It’s kind of like what happened with computers; as soon as the microprocessor came on the market, the price of all electronics fell because you could use a single chip design to punch out all sorts of different products. And thus the microinverter was born, a great example of marketing driving the terminology.

There are drawbacks.

One is a production issue. Earlier we noted that there’s a problem with the pricing as you scale down. In the case of a micro, you can imagine that even the smallest parts like the case and wiring becomes a huge percentage of the total cost. Consider just the panel wiring connectors; and they might cost $5 in a product you hope to sell for $150. Well a 10kW inverter has the same two connectors, but it might sell for $4000. So, as a percentage, these basic parts represent a tiny fraction of the cost of a string inverter, but a major portion of the cost of a micro.

Another problem is technical. Any given wire can carry a certain amount of current. Power is current times voltage. So if I raise my voltage, I can carry more power on the same cable. In the case of common household 14-gauge wire, that limit is 15A. 15A times 120V is 1,800W, enough to run the biggest hair dryer you can buy (that’s not a coincidence). Now if I used that same cable, but bumped the voltage to 600V, that gets me 9,000W, enough to run my entire house. Since inverters pump out AC deliberately matched to the local grid, they normally operate at 120 or 240V. So, generally, micros have to use heaver gauge wire to carry the same amount of power. And copper is freaking expensive these days.

And so, there’s the rub. Micros are, by any basic measure, more expensive than a conventional string inverter. You may, as proponents suggest, be able to overwhelm those inherent disadvantages through huge production runs – after all, that’s why the cost of solar panels is so low these days. But “maybe” and “reality” are two different things, and the proof is always in the pudding.

The Enphase story

The introduction of the PWM high-frequency inverter in the 1980s/90s made for highly efficient string inverters, and it also allowed for them to be scaled down as far as you might need. It was this invention that made the microinverter possible. But facing the price pressures outlined above, this didn’t happen overnight.

The first real microinverter was the Mastervolt Sunmaster 130S from 1993, but it simply couldn’t compete with conventional designs and disappeared from the market fairly quickly. A more aggressive attempt was made with the OK4 product, but after a run of some 200,000 or so, production ended for reasons that are not well recorded (rumours of huge failure rates).

And then came Enphase. Started by ex-telecoms people from California, Enphase was able to tap into the local talent pool and vast reserves of venture capital, and came out of the gate flying. Their original M175 model was released in 2008 and followed quickly by the improved M190. At the time, the M190 sold for about $190, or in the lingo of the industry, “one dollar a Watt” ($1/W). This was definitely higher than string inverters of the same era, which were around 65 cents/W, but with panels selling at over $2/W and another $1 for all the little extras, the effect of that 40 cents on the overall system price wasn’t enormous.

For smaller projects, say up to two dozen panels, the higher price of the M190s might work out to a few hundred dollars in total. This was more than made up for by the fact that the installers could stock a single inverter design for every one of their projects. Installation was way easier too; no need to design the layout and select the right inverter, you simply stuck the micro to the back of the panel, plugged it in, and moved onto the next one. And for small projects, the 240V/20A limit of the Enphase wiring meant that systems up to 4,800 W could be run off a single wire, which encompasses a huge chunk of the residential market.

So for a while Enphase had a run at the residential rooftops that left the other inverter companies gasping. Arguments from the cost side simply didn’t gain any traction with the installers, who were perfectly happy with the value proposition of the simplified design and inventory control. In spite of the best efforts on the part of the big iron companies, Enphase’s portion of that market just grew and grew.

So then it got nasty…

A little bit about reliability

I’m sorry to have to do this, but to understand the arguments that did stick we need to delve a little into the topic of reliability. And to do that, we need to start by considering the difference between the failure rate, and operational lifetime.

The tires on your car might have a tread life warranty for 100,000 km, which is a pretty good indication that the company expects them to last that long. It’s reasonable to expect that the tires might make it to 150,000 km, because if there’s one thing Goodyear doesn’t like, it’s handing out free tires. So when you notice that one tire is worn out after 125,000 km, that likely means they all need to be replaced.

Now there’s also the chance that the tire will simply fail, blow out for some random reason. But this happens rarely, maybe once every million km or more. That’s what’s known as the “Mean Time Between Failure”, or MTBF. But that’s just one tire, your car has four. According to Lusser’s Law, the chance that a system will fail is the MTBF for a part times the number of parts. So if we stick with the numbers above, you’ll have to go about 250,000 km before one of your tires fails.

If the MTBF was lower, say 200,000 for each tire, then you’d expect to lose one around 50,000 km, long before they reached their operational life. In that case, the rest of the tires would still have lots of tread, so you’d just fix the broken one and keep going. A failure doesn’t imply anything about the other tires.

Normally these numbers are expressed in terms of years, so I’ll convert. The average car is driven about 25,000 km a year – so the warranty is 4 years, the operational lifetime is 6, the MTBF is 20 years, and your chance of having a blow-out is 1 in 10 years. All good?

Finally, I want to touch on the different types of failures, because this is an important part of the microinverter story that is often overlooked. Lusser’s Law states that having more of something means more failures. So then wouldn’t an airplane with two engines have twice as many engine failures as a single engine plane? Indeed, they do. So then, why do airliners all have more than one engine? That’s because engine failure on a single engine plane means you’re that night’s leading news story. An engine failure on a twin engine plane is a lot less interesting. One of these is a critical failure, the other isn’t.

This “critical failure mode” is an important consideration in any design.

Let the games begin

So back to micros. You’re at year ten in your PV’s system’s life, and one of your inverters stops working. Ok, did it fail, or did it wear out? If it failed, who cares? After all, it’s wired in parallel, so you don’t even have to replace it if you don’t want to, everything else is still going. But if it wore out, well, that’s different, because that means you’re going to have a bunch of other ones go soon too.

So which is it? Well if the Enphase is really just a conventional inverter in a smaller box, why should it last any longer than a conventional inverter? And those only last 10 to 12 years, something the manufacturer will be happy to tell you. But here’s one difference… microinverters are attached under the panels… on the roof. Which means that when they start to go, you need to take the entire system apart to replace them.

The big iron companies smelled blood in the water.

The fight started in earnest in 2010. Every other inverter company started hammering Enphase – a string inverter was mounted in a convenient location on the wall or in your basement, so it could be easily replaced in 10 to 12 years when it was expected to fail. But how much would it cost you to replace the Enphase kit on your roof? Ten times as much? And, of course, they’re going to fail.

Enphase fought back, but not convincingly. They started listing MTBF in the hundreds of years, but have failed to make a single statement about their expected lifetime. They were somewhat more successful in pointing out that a single Enphase failure isn’t critical, but that’s where the argument about the product’s real-world lifetime came in.

Much of this became academic in 2011. The continuing massive downward price pressure in the solar market was particularly notable on the panel side, with prices dropping roughly in half between 2010 and 2012. Panels are now widely available around $1/W, which made the $1/W of the M190 harder and harder to swallow. If that wasn’t enough, traditional string inverters were also falling in price, down to about 40 cents/W today.

During the same period the average panel’s normal power started to creep upward, from around 220W when the M190 shipped, to 245W by late 2011, and 250 to 260 was common by 2012. The 190W M190 simply wasn’t well matched to newer panels. When you consider that the inverter was the limiting factor in production, buying higher-rated panels didn’t get you any more power. But if you took those same panels and connected them to a slightly larger string inverter, presto, more power out.

The M190 was in trouble.

Ragnarök

Enphase responded to all of these issues with one sweeping product upgrade, the M215. Among many changes, the M215 was best matched with panels around 245W, cost as about 60 cents/W, and came with a 25 year warranty. That should have been enough to silence the arguments against the product.

But everything got worse.

In order to hit the new prices points, Enphase had to address those fixed costs we talked about earlier. They introduced a smaller and simpler case, shortened up the connectors and made other basic changes. But the big change was to remove the cables that ran in parallel from inverter to inverter. Instead, the M215 had a single short cable that ended in a new connector, and they were connected to their neighbours with a separate “trunk cable”. In theory this made installation even easier, because you could attach the inverters anywhere you wanted, and then just pull the trunk cable along and click them together.

In reality it was a disaster. The connectors were so expensive that they completely offset any price advantage in the design – the branch cable general sells for $15 to $20 per connector. When you add that on, *poof*, there goes any major price advantage over the M190. Worse, since the branch cable came in a long spool, you had to cut it to length to wire up the branch, and then you have to close with all the loose ends with manually-wired plastic caps. In comparison, the M190 daisy-chained together, so there wasn’t any cutting at all. And as if that weren’t enough, because you could install panels either upright (“portrait”) or sideways (“landscape”), you needed two different types of cables so the connectors would be in the right place.

So much for simple inventory!

That might have been enough to sink the product right there, but it got a lot worse. Sure, the M215 came with a 25 year warranty, but there was little change to the internal workings. If the M190 was only going to make it to 15 years, its warranty, why should we expect the M215 to last any longer? In particular, its use of a particular part inside, the electrolytic capacitor, become a rallying cry for its opponents. These are generally expected to last 10 to 15 years.

So if they have components in there that just won’t last 25 years, yet they still offer a 25 year warranty, what’s the story? That’s when people started really nasty stories about the company, all apocryphal of course. If that wasn’t enough, I had people start telling me I was the bad guy for installing them – one audience member at the SMA booth in Toronto went so far to say I was “damaging the solar market”. Yeah, sure.

But most worrying of all these developments is the lack of major downward movement on the price. Enphase’s whole argument was that once production ramped up the price would start coming down to the point where you didn’t even think about the delta. The M215, the 3rd generation product, was their chance to demonstrate this in action. But the difference in pricing was minor. By any measure, the drop in price considerably less than the drop in pricing in string inverters over the same period.

Opportunity knocking?

In spite of all the comments from the big iron companies, or perhaps because of it, Enphase has convinced the market that the micro is a good idea. And so there’s a bunch of companies jumping into the ring.

Primary among the new entrants is Enecsys, a UK company that took work from Cambridge to produce a 220V/50Hz product for the local market. The tech was licensed to investors in the US, and re-launched in a 240V/60Hz version for North America. Enecsys’ main claim of superiority is the replacement of those electrolytic capacitors with film capacitors, which have a much longer expected lifetime. There are some interesting changes to the internal construction too, but I don’t understand it well enough to comment. They also use a different cabling system that is sort of a hybrid between the one on the M190 and the one on the M215… they have a single cable coming out of the inverter, but it ends in a T that you daisy-chain extension cables into. They offer a 20 year warranty, and come in at about the same price as the M215.

A more amusing entrant is SMA, king of the big-iron string inverters. They bought the technology from OK4 and have been claiming they’ll launch an improved version of it any day now – which they’ve been saying for two years now. They really don’t offer any arguments why their technology is better, other than to stress that SMA is a “real company” and they’ll “stand behind their product” – the implication being that “other companies” might not. But what’s odd is that they can’t stop bad-mouthing it at the same time they’re trying to pitch it, continually stressing that string inverters are cheaper and you’d only want micros in certain installs. With friends like these…

Beyond that there’s dozens of smaller companies all over the world trying to break into the micro space. Most of them boast one or two features they claim make them so much better than Enphase, but in most cases it always boils down to the use of film capacitors. And there’s a definite downside to those – they’re larger, more expensive, and you need more of them to get the same effect as one electrolytic cap. There’s a serious question as to whether or not any of these companies can get their prices down, because their parts count goes up even if the individual parts price doesn’t go up – and it does.

Lightning strikes SPARQ?

The reason I wrote this article is SPARQ Systems. Sparq is a startup out of Queen’s University in Kingston, right here in Ontario, so I’ve been watching them more closely than normal. I’ve seen their product a couple of times at the Toronto shows, but they haven’t been shipping anything (yet), so I never put too much effort into studying it in depth. But if there’s ever a time to launch a micro, this is it.

What makes Sparq interesting is that it works on a totally different principle from other inverters. Most models use the “pulse-width modulation” technique, or PWM. Basically they switch the DC power from the panels on and off really quickly in a particular pattern that, once it makes it through the rest of the circuits, looks close enough to a sine wave that only minor modification is needed to match the grid. The problem with this approach is that the rapid switching puts a whole lot of strain on the electronics, regardless of what sort of capacitor they use. Don’t get me wrong, using film caps helps a lot, but it just means some other part becomes the critical failure point.

Sparq’s inverter is based on a technology known as “resonant conversion” that’s been around for about a decade, but only recently started to become really popular. This is really a fairly simple change to the design, so instead of switching on and off rapidly, the switch is slowly flicked (sorry, bad metaphor) from on to off using a signal that itself looks more like a sine wave. This doesn’t really directly change the output, and the rest of the inverter circuit is pretty much the same. But what it does is dramatically reduce the stress on the main switching elements. So much so that they are expected to last decades.

Now when you’ve got a system that has a potential lifetime out that long, then it becomes important to make sure all of the components you use can make it. Sparq uses film caps, sure, but it also eliminates optocouplers, and thyristors/IGCTs. Those are items one, two and three on the list of things that don’t last very long. So they expect their inverter to last decades, like four or five of them. And just to be sure, they offer a 25 year warranty.

That’s not all. In a traditional inverter, you have three basic parts, the MPPT, inverter and controller. In the Sparq, the MPPT and inverter are the same thing, and they are largely implemented in the controller. The inverter isn’t so much hardware as software, software that outputs such a smooth signal that you don’t need to clean it up. So, yes, they have film caps, but they don’t have nearly as many, and their overall parts count is way less. This bodes well for scalability. And since the output is from a program, they can simply change the program to deliver any sort of output you need – 230, 240, 600, single-phase, three-phase, you name it. That bodes even better for scalability, because you can sell a single box around the world. According to their engineers, they can scale to price points that compete with the string inverters – of course, Enphase said that too.

So stop and consider the entire suite of arguments against Enphase – it really boils down to the operational lifetime being too short. Too short? Well, to be exact, anything shorter than the lifetime of the panels. Panels are warranted for 25 years and are expected to last 35 to 40.  So if Sparq can deliver, they have a product where the lifetime issue simply goes away – they not only outlast micros, they outlast the strings too, and maybe even the panel. That changes the entire ballgame.

Should we believe them? Well let’s put it this way, resonant conversion is already used in a number of applications where super-high reliability and long lifetimes are needed. Like on the Space Station. And when they tested their prototype unit, it scored between “space” (the highest) and “aerospace” (the second highest). I don’t know where the other products score, but it still gives me warm fuzzes. In comparison, Enphase, SMA, they’re all just string inverters in a small box. String inverters last about 10 to 12 years. That’s not warm and fuzzy.

Of course Sparq is a Canadian start-up, not a US one, so there’s a lot less VC and a lot more bootstrapping. Time will tell if they can pull it off, but as always, I’ll be rooting for the home team.

As if that weren’t enough…

Earlier I said that a conventional inverter consists of three basic parts, the MPPT, the inverter, and the controller. Much of the performance advantage of the micro concept is due to each panel having it’s own MPPT, which allows the micro to pull power from the panel independently of the rest.

So what if you put just the MPPT on the panel? You get all the same advantages as the micro, and the output – still in DC form – goes right into a conventional inverter. This is the “power optimizer” approach, pioneered by Tigo and now offered by a number of companies, notably SolarEdge.

One thing I’ve noticed is how much smaller and lighter an optimizer is compared to a micro. The Tigo is about 1/3rd the size of a M190, and considerably lighter. Much of this is due to the lack of a transformer, the heaviest component in a typical inverter. But the Tigo is also made entirely of plastic, and gives off little heat. That lack of heat implies high efficiency of the conversion process, and to a lesser extent, less stress on the components. This can be interpreted to suggest a very long lifetime, which is how Tigo puts it. But here’s the thing I’ve failed to wrap my head around…

The main complaint against the micro concept is that you have a distributed system and have to replace all those parts at some point. OK, but in the case of the optimizer, you still have all of those distributed components, even if their lifetime is longer. In addition, the optimizer still has the string inverter, which we know is not going to last. So you still have the minor failure modes of the micro approach, and the catastrophic failure mode of the string approach.

Maybe I’m over-thinking this, but it seems to me this approach is compounding the chance of failure. At best, you can ignore the optimizers on the panel, assuming very long life, and so it’s the same thing as if you didn’t have an optimizer. At worse, you’ve increased the chance of a failure. I’ve tried my best, and I still can’t come to a strong conclusion one way or the other.

The no-conclusion conclusions

So, like I said way back at the beginning, its tough to draw any firm conclusions from all of this.

The classic inverter’s argument is simple – yeah, it’ll fail, but it will be cheap and easy to replace. Everything else is unimportant.

The microinverter’s argument isn’t so simple – we’ll get you more power and greatly simplify your design and installation. And don’t worry about all that reliability stuff.

And then there’s the power optimizers – we’ll get you all the advantages of the micro with all the advantages of a string. Unless we get you all the disadvantages of a micro with all the disadvantages of a string!

Sadly, the only way we’ll know how this really shakes out is to sit back and wait. In the meantime I’ll be watching the Sparq story very closely. Cha Gheill!