diy solar

diy solar

99.5% Efficient Inverter Design

Unfortunately the main reason many inverter designers go to high voltage dc (from battery voltage) and then PWM that, is to save the cost and size and weight of a large 50/60Hz transformer.
As the whole motive of doing it that way is for cost saving, many of these inverters lack the sophistication of a fully bidirectional high frequency inverter, and as a result they become rather electrically fragile.
Certainly its possible to do it properly, but that requires extra complexity which pushes up the cost of manufacture.

Low cost definitely sells inverters.
Anything really good becomes far too expensive to compete successfully in the marketplace.
 
60 Hz Power Transformers have at best 96% efficiency and this is with the highest quality ($$$) toroidal transformers. Lower cost power transformers are lucky to hit 90%.

Please explain how he is getting 99.5% efficiency passing AC power through a power transformer?

The largest 100MVA transformers are already breaking 99.75% efficiency.

There's a large inverter on the market with a claimed 99.52% efficiency

Two-Battery HEECS Inverter with over 99.7% Efficiency at 2.2kW Output and Measurement Accuracy Based on Loss Breakdown

Here’s mention of a DC-AC 1500W inverter measured at 99.75% efficiency
 
The largest 100MVA transformers are already breaking 99.75% efficiency.

There's a large inverter on the market with a claimed 99.52% efficiency

Two-Battery HEECS Inverter with over 99.7% Efficiency at 2.2kW Output and Measurement Accuracy Based on Loss Breakdown

Here’s mention of a DC-AC 1500W inverter measured at 99.75% efficiency

Are you sure that the gamesa-electric inverter is a transformer inverter and not a non-isolated inverter ?

A quick look did not reveal that it had one either way. At this high of efficiency, I am somewhat skeptical.

The HEECS inverter at 99.7% also looks to be a non-isolated, transformerless inverter from the drawings at the top.

Couldn't quite understand right off how the virtual-transformer works in that paper but I am also skeptical of the word "virtual" in that context.

My mind is open though. :)


boB
 
The other author was stating that Power Transformers have "at best" 96% efficiency but he probably didn’t realize the massive Transformers we had at the Nuclear Generating Station were >99% efficiency and 99.75% transformers have been in use for decades. I was a Nuclear Safety Inspector at San Onofre. I didn’t say those inverters I linked use transformers. I was just pointing out inverters that already broke past 99.5% in the title of this thread.
 

Attachments

  • Screen Shot 2022-06-29 at 9.16.34 PM.png
    Screen Shot 2022-06-29 at 9.16.34 PM.png
    183.7 KB · Views: 12
Last edited:
The other author was stating that Power Transformers have "at best" 96% efficiency but he probably didn’t realize the massive Transformers we had at the Nuclear Generating Station were >99% efficiency and 99.75% transformers have been in use for decades. I was a Nuclear Safety Inspector at San Onofre. I didn’t say those inverters I linked use transformers. I was just pointing out inverters that already broke past 99.5% in the title of this thread.

Thanks. Yes, transformers are "usually" around 95 or 96% efficient. Especially if they are designed for price and weight and for a competitive 50/60 Hz transformer inverter.

Where about in the power curve was your high efficiency Xformers peaking at ? Half-way or more towards full rated power ?

Any idea on materials used in those high efficiency transformers ? I have never done a large transformer before. Only ~ 5 kW from M6 laminated silicon steel and copper of course.

boB
 
Yes 95% to 96% is achievable, but its very difficult to do better.
Magnetizing current and copper losses rapidly reach the point of diminishing returns.

To get from 96% to 98% means losses need to fall from 4% to 2%.
Pretty difficult to HALVE everything after you have already given it your best shot.
To go from 98% to 99% means HALVING everything again. I don't think its even possible.

Kind of like saying if you can run a five minute mile, if you try just a little bit harder you should be able to do it in two and a half minutes.

Some things are very difficult to halve or double.
Even harder to quadruple, but that is what some people here claim is being done.

A guy in a bar (pub) once told me he could easily jump fifty feet in the air in his bare feet.
Maybe he can, and maybe he is the the same guy that designs inverters that are 99.5% efficient.
 
Last edited:
Hi Warp,

Made me dig up the scope shot of the 3-musketeers.
Hi tinyt,

I trust you are okay?

Warpspeed, Tony, just about to get the big mower out and its raining, oh! how very Normandy, France.

Thought you might be interested, in regards copies of my book with your chapter in it about the very elegant design of Warpinverter of yours, as I am getting some very positive feedback, especially from well-known respected Electrical Engineers. I can print it out as is on this forum if you so wish.?

I think most of you here realise that in the commercial world of business its all about little as possible cost of components and maximum of profit margins at the expense of quality and longevity of product.
However, from my point of view and what i see, this tends to lean towards shoddy winding and manufacturing of the Inverters toroid’s, inadequate power interconnects and woefully poor cooling especially with larger inverters designed for 4Kw and above.

As regards HF Types I am still watching the 'Midnite Rosie Inverter' testing with great interest, i believe the design has merit.

Tony, sadly regards the 'nanoinverter' i did not give it sufficient examination at time of publication, but i did manage to print some of 'Podia's' nanoinverter descriptions and some of his photos. I will do a separate post here in this topic about what i have on the 'nanoinverter'.
 
'NanoInverter'

"Podia’s" nanoverter,

Using his own PWM code, in a Arduino (ATMega 328) so the 8010 chip is not used.........

‘Poida’ , Peter Birtles, comments, ……. “The inverter uses a closed loop control design which takes a DC sample of the output AC voltage on the transformer secondary.

This DC voltage (Vfb) may have a lot of 100Hz ripple (if a full bridge rectifier is used, else 50Hz ripple), and unknown amounts of much higher frequency noise mostly coming from the full bridge mosfet switching of the DC supply. We need to filter Vfb to only let DC into the control loop.

In another post I showed that the EG8010 IC only samples Vfb during a very short time compared to the 20 ms period of 50Hz output. I think I wrote that Vfb is sampled at the top of one half of the 50Hz output waveform, and the sample window is of the order of 10us. For the rest of the (20milliseconds - 10us) period, Vfb is not sampled.

This means that we need to have Vfb reflect accurately what the output voltage is during that sample time. Vfb may go all over the place during times when the EG8010 is not sampling it’s value.

The fact that Vfb is sampled synchronously with the output waveform might lead us to simple Vfb filtering designs which work fine on the test bench, usually with purely resistive loads. In the real world the current and voltage can be way out of whack. This will upset the closed loop control leading to over/under voltage of the output.

So we have to low pass filter Vfb quite heavily to allow for large disturbances from various load types (switching power supplies, SCR controlled lights, motors, flouro lights etc).

I have built a custom inverter prototype using the Arduino (ATMega 328) so as to permit complete control over all aspects of the inverter function. I can use any sort of digital low pass filtering of Vfb I choose. I could do the filtering in hardware but I prefer the flexibility of software. At last it’s time to look at some results of the closed loop output voltage control.

I use a simple design: Vfb is fed into one of the ADC ports and is filtered to remove all the high frequency stuff. The 100Hz ripple is quite high, about 0.3V p/p riding on a 2.5V DC signal. I then sample this voltage about 2,200 or 8,000 times a second and apply some digital filtering to it. Then at the zero Voltage point in the beginning of each 50Hz output, I apply this filtered Vfb value to the PID control loop to calculate the needed PWM duty cycle %.”

Poidas Nano.jpg

2 sides.png
 
Hi tinyt,

I trust you are okay?

Warpspeed, Tony,

Thought you might be interested, in regards copies of my book with your chapter in it about the very elegant design of Warpinverter of yours, as I am getting some very positive feedback, especially from well-known respected Electrical Engineers. I can print it out as is on this forum if you so wish.?

Its all open source and free to copy/pass on without any restrictions at all. I can also answer any questions and provide further help or information if anyone decides to go down this path.

Its not for everyone, low voltage high frequency pwm with a big 50/60 Hz transformer is still the best approach for most people.

However, once you start paralleling up multiple mosfets for increased power, the problems magnify, layout becomes far more critical. Low voltage high frequency pwm does not scale up easily, although it can and has has been done.

For very higher power with fewer problems (blow ups) the Warpverter scales up easily to any imaginable power level. It does require more parts and its a lot of work winding the four transformers, but for anything over 5Kw its definitely recommended for a very robust and reliable inverter.
 
Copec, I am curious how you measure efficiency ?

Also, is that with the inverter cold or hot or high Vbatt ?
Low Vbatt ? Full power ? Peak power ? That kind of stuff.

Are you using shunts ?

Thanks,
boB
Those are average efficiencies from my usage patterns gathered from shunts and CTs. I keep the space around my inverters/batteries at 75F and keep the battery between 50-90% SoC (which is ~54-56vdc).

I need to test the efficiency of the inverter in pure off-grid mode. I would think the efficiency would take a hit having to control the PF itself, without delivering an average power to reach zero at the CTs.
 
Last edited:
I'll ask in this thread because there are people knowledgeable: Does anybody know of a ~380vdc to 240vac H-bridge inverter on the market that I could use for AC coupling (behind an isolation transformer)? Is there a good place online where people discuss and share inverter designs?
 
Low voltage rated mosfets win hands down when run at low voltages.
The "on" resistance is so very low, nothing else even comes close for conduction losses and speed.

But as the voltage rating of the mosfet increases, the "on" resistance rises much faster. High voltage mosfets are pretty miserable devices.

As RC above ^^ says, these days anything that runs over about 100v to 150v that also must run at reasonably high power, will likely be an IGBT.
Over the last ten years or so, IGBTs have massively improved in both speed and conduction losses.

You will find that all the high voltage grid tie inverters run IGBTs, never mosfets.
There is still a place for mosfets in an off grid inverter running at 48v, they are still by far the best choice for high frequency pwm at lower voltages.

These big IGBT power bricks are very slow, especially the older ones, and quite unsuitable for high frequency pwm, except in variable frequency drives where the pwm might be only a very few Khz or even hundreds of Hz. But they are perfect for switching very high power at 50/60Hz.

There is also another aspect of this that nobody talks about. These big devices may be rated for hundreds of amps, and the voltage drop across the IGBT might be a couple of volts, which is really too high for a low voltage inverter. But at much lower power, say ten or twenty amps, the voltage drop is actually quite low. Not as low as a low voltage mosfet, but acceptably low for an off grid inverter.

After building this monster and running it at 100v dc over four years, I would have no hesitation building another one to run at 48v using the exact same IGBTs.
With average mixed continuous domestic loads, which are mostly below 1Kw, the inverter efficiency is really quite good, even with IGBTs.

For really high power, a hundred dc amps is probably the practical maximum, say 5Kw at 48v. If you go up to 100v its only fifty amps and the losses from voltage drops around the system will be far lower. The only practical difficulty with high dc voltage becomes the large number of battery cells to maintain. But on the whole, I am very happy how my 100v system has worked out.
Same could be said about mosfets, particularly high voltage (superjunction) mosfets.
Your "ebay example" SKM195GB126D has ~50W condution loss at 50A
600v IPDQ60R010S7 has 10mOhm RDSon and 25W conduction loss at 50A
Or better yet VMO580-02F module, about 7W at 50A
Going to parallel devices and TO247 and we have options like SUG90090E-GE3, 3pcs parallel would be 3mOhm and 7.5w conduction loss.
These are about 4 usd per piece so 3 in parallel would be still cheap

And if cost wouldn't be issue we could use 1200V! CAB760M12HM3 SIC mosfet with 1.3mOhm rds-on and 3 watt conduction loss. :ROFLMAO:
 
Yes you are quite right, the technology is improving all the time. Its truly amazing what is now available.
I have been retired for a long time, and am now out of the loop as far as the very latest tech.

There are still a few minor points to consider when designing yourself a new inverter.
The latest devices are invariably more expensive, compared to Chinese grey clones of much older less efficient devices.
Using multiple smaller devices in parallel will create issues with board layout and equalizing all the switching delays and currents.

At least with a transformer inverter, there comes a point where conduction losses in the transformer windings become larger than the conduction losses possible in the switching devices. Beyond that, the point of diminishing returns kicks in.

In my own home brew inverter, I chose to massively oversize the IGBTs. These are very large, very cheap, and very slow. But have survived massive abuse over the last four and a half years without missing a beat.

It all becomes really difficult chasing "efficiency".
its a major accomplishment to halve the losses overall, going from say 6% to 3% losses might be really expensive.
That would get you from 94% efficiency up to 97% efficiency. The cost of that extra thee percent improvement may just not be worth it.

In a domestic inverter, the load profile is usually such that the inverter runs at a quite low average power compared to load peaks.
Its far more effective as far as overall power efficiency over 24 hours goes, to try to minimize idling power rather than just achieve higher efficiency when flat out.
That is where the real gains are.
 
Last edited:
It all becomes really difficult chasing "efficiency".
its a major accomplishment to halve the losses overall, going from say 6% to 3% losses might be really expensive.
That would get you from 94% efficiency up to 97% efficiency. The cost of that extra thee percent improvement may just not be worth it.

In a domestic inverter, the load profile is usually such that the inverter runs at a quite low average power compared to load peaks.
Its far more effective as far as overall power efficiency over 24 hours goes, to try to minimize idling power rather than just achieve higher efficiency when flat out.
That is where the real gains are.
you are 100% correct about load profiles. I just went along with the original topic lines..

Too many mosfets or too big mosfets also start to eat efficiency with PWM based designs. Albeit 23khz switching frequency is still pretty low and Coss (mosfet output capacitance) is not that big issue. (”low speed” designs like warpverter of course wont suffer from this and one can use ridicuously sized overkill mosfet bank on warpverter)

95% or 98% efficiency is not big deal from energy use point of view but it makes drastic difference in power losses and cooling need. Again sometimes this is not any issue and sometimes it is literally burning hot issue.
 
It mainly becomes an important issue where one inverter drives just one load, and is otherwise switched off.
Under those circumstances idling power becomes a total non event, and achieving high efficiency under full normal load then becomes really worthwhile.

This is usually when some kind of inverter or switching power supply is built inside a piece of equipment which is either running or not running.
Its then possible to achieve quite high efficiency at just the one optimum power level, and design everything to suit.

Most of the resonant mode switching topologies are like that. Great for the intended purpose, but useless in a domestic inverter that has to run over a very wide load range including no load at all.
So when I read that someone has achieved 99.5% efficiency, I think "good for them" but its not technology that is transferable into a domestic inverter !

Also right up high on my list of important characteristics would be very high surge capacity, and the ability to drive (and survive) highly reactive loads.

I would rather give up some peak efficiency and build an absolutely reliable bullet proof inverter that has a really low idling current.

And that was the design philosophy behind the original Warpverter.
Its really too expensive to build to be a commercial success, but its perfect for a very high powered home brew project that can use recycled parts, and would also be very easily repairable by the user should disaster strike.
 
Last edited:
Many people use the term "xx% efficiency", but ... to be precise it's the peak efficiency, and for most of us, it's the efficiency on the whole "band" that is important, or at least where you use it the most. Of course inverters in industrial facilities are uses in their optimal "band" (at least should be).

Meaning a 97% Peak efficiency inverter can be less efficient in all day use then a 96% peak efficient one.
 
I'll ask in this thread because there are people knowledgeable: Does anybody know of a ~380vdc to 240vac H-bridge inverter on the market that I could use for AC coupling (behind an isolation transformer)? Is there a good place online where people discuss and share inverter designs?
A little late but the most straightforward way would be to go for a UPS that can take that battery voltage. Generally they are already isolated and of course are way over engineered with every protection and filtering imaginable.

Depending on the wattage you need and if you get lucky you can find them real cheap, just gotta pick them up ? at least in the US. Like on ebay right now there's a 225kva one for $1600 in SC. Imo that is way too big and would get complicated, but like I said it's random luck and timing.

But if you really wanna go that route there's prebuilt inverter modules from China. Like this one, which takes in 360vdc and outputs 220v at 10kw. But of course you get the 'jank' factor.

There's also some IGBT spwm inverter drivers like I tried to use in my franken-inverter project.
They do work well actually, and all you really need to add is the igbts, cooling, output inductors and voltage/current sensing. Only problem with mine is that I used 1000A 1700V igbts out of a damn wind turbine, and the gates were too big to switch at the 23khz the EG8010 uses. (I notice that the newer version of these boards skip the EG8010 and now use a PIC chip. They state 16khz carrier so should be able to move slightly large igbts. Probably 600A 1200v ones easily.)

But if you use for example a decent, newer 300A half-bridge module pair with low voltage drop (Vce) you can easily do 20kw continuous with proper cooling. And if you add a transformer for galvanic isolation it will become practically bulletproof.
 
Back
Top