diy solar

diy solar

99.5% Efficient Inverter Design

HaldorEE How can you say this!!?

My Chineseium 500W Grid-tie worked very well until the wires wore out and the brass studs got tired and started to lean over.

20190622_183407.jpg20190622_184032.jpg
 
Filtering inductors in the output are not what I am talking about.
One could say the primary use of inductors is for filtering. Take for example a buck regulator. We would agree the inductor is an integral part of buck regulation. It acts as a current filter. But we could remove the inductor and still achieve some sort of pwm regulation. The output would have higher ripple and switching noise. We would then state "it also may not be a bad idea to have a small inductor on the output".

I looked at his BOM. The Power Board has 3 H-bridges: using 3 different voltage rating MOSFETs: 40V, 1.3 milliOhm; 75V, 1.46 milliOhm and 150V, 4.8 milliOhm. There's a Driver Board with 12 x (fet driver chip, opto isolator and dual output dc/dc).

Edit, Has the author mentioned a THD spec?
 
Last edited:
I would like to read this paper, but I cannot seem to download. Do you have this in pdf, or can you tell me how to download? When I snap on it it is too small to read.
Sorry, it comes up on both my cell (Droid) and PC (Win).
 
I would like to read this paper, but I cannot seem to download. Do you have this in pdf, or can you tell me how to download? When I snap on it it is too small to read.
On my computer, at the very bottom and right hand side there is a slider button. -_________________+ to zoom in.
 
For anyone curious, this tech has been around for a while now (1990s) and is REALLY STOUT. They have great surge/overload capabilities. They are called Multilevel Inverters and are showing prominence in MV grid tie solar applications. They have extremely low switching frequencies <1kHz and require no output filtering unlike their HF counterparts; this is a huge advantage as a large portion of inverter losses are switching losses. It appears in this specific case, the design was a 1:3:9 (ratios of voltage supplies) cascaded half bridge inverter using a NLM (nearest level modulation) scheme. This 1:3:9 ratio is referred to as asymmetrical and allows for much higher efficiency by increasing the number of levels (or steps) in the sine wave per cycle for a given amount of hardware (switching) components. In this case, this should be a 27 level inverter. The inverter would have an output THD (distortion) of around 1-2% (R load). They are very flexible and expandable. In my mind they are the best type of inverter for solar energy applications. I've been thinking of making something much like this. I'm in the coding stages right now, but I want to make mine programmable/customizable.
 
Well, they're great in MW applications, but for home applications we are talking 3 battery banks floating at dangerous (and changing) voltages so you need separate and floating SCC and PV panels for each bank...

That's not a problem on big installations where anybody who goes in knows all the dangers, but in a home it's not a good idea, at all.
 
Well, they're great in MW applications, but for home applications we are talking 3 battery banks floating at dangerous (and changing) voltages so you need separate and floating SCC and PV panels for each bank...

That's not a problem on big installations where anybody who goes in knows all the dangers, but in a home it's not a good idea, at all.
True. This definitely isn't practical for most people to make or even use.
 
For anyone curious, this tech has been around for a while now (1990s) and is REALLY STOUT. They have great surge/overload capabilities. They are called Multilevel Inverters and are showing prominence in MV grid tie solar applications. They have extremely low switching frequencies <1kHz and require no output filtering unlike their HF counterparts; this is a huge advantage as a large portion of inverter losses are switching losses. It appears in this specific case, the design was a 1:3:9 (ratios of voltage supplies) cascaded half bridge inverter using a NLM (nearest level modulation) scheme. This 1:3:9 ratio is referred to as asymmetrical and allows for much higher efficiency by increasing the number of levels (or steps) in the sine wave per cycle for a given amount of hardware (switching) components. In this case, this should be a 27 level inverter. The inverter would have an output THD (distortion) of around 1-2% (R load). They are very flexible and expandable. In my mind they are the best type of inverter for solar energy applications. I've been thinking of making something much like this. I'm in the coding stages right now, but I want to make mine programmable/customizable.
Good luck with this project.
 
If I had to guess, it works like using a cascade air bank for filling scuba tanks? You have to switch on the next voltage above until it almost matches, and then the next one, and so on. Coming down you are just doing the opposite.
 
If I had to guess, it works like using a cascade air bank for filling scuba tanks? You have to switch on the next voltage above until it almost matches, and then the next one, and so on. Coming down you are just doing the opposite.
Essentially, yes.
 
Essentially, yes.
Hi, all.

Then you would like this forum: if you don't already know it.

We are calling it the "warpverter", named to the man that introduced us to this multi step inverter on the backshed forum.
He build a version without uC, but there are already several versions with uC.

I will build one myself in the future, I will use 4 H-bridges, and 4 transformers for isolation and use with 1 battery (bank) and battery voltage you want, just adjust the turns on the transformers.
With 4 stages you will get 81 steps which will be close to perfect. The beauty of it is like you said the power output, because power will be sort of shared by 4 transformers, although only the 2 biggest transformers will carry most of it. Correlation will be more like 300W+900W+2700W+8100W
=12000W. Insane surge power capability and robust, it will handle much better loads like foans, ac split units, were sine is chopped to control power.
 
Forget to mention the downsides, it will much heavier, bigger and more expensive than other inverter topologies. That's why companies like victron, mastervolt,... Doesn't bother designing them. But this would be a nice DIY project.
 
Looks like the same kind of topology as the old Trace Engineering SW4024 and SW5548 inverter/chargers from the1990s...
Basically 3 mod-square-wave inverters in series but the 2 smaller ones using different waveforms and slightly higher frequency to approximate the sinewave as well as regulate the output based on battery (single battery) voltage.
I think the main reason this one may be more efficient, from what I have read here, is that there is NO transformers for this inverter.
It is also very difficult to measure efficiency accurately at close to 100% which I'm sure this one is close to. A bit too unpractical in my opinion but nice demonstration maybe.
 

Attachments

  • Copy of SW-Waveforms.gif
    Copy of SW-Waveforms.gif
    122.8 KB · Views: 8
The reality is that for a domestic inverter the average power level is quite low, huge power peaks are fairly infrequent. But at night there will be almost no load at all.
No load idling current, and low load efficiency are far more important than full load efficiency.

With regard to the direct switched battery bank idea, its been done many times before, and its a pretty rough stepped waveform.
There are two inherent problems, the first one is charging the individual battery banks while its being switched around "live" in the inverter.
That requires multiple battery chargers, multiple banks of solar panels, and multiple MPPT controllers, all fully electrically isolated !
It all becomes vastly complex and impractical.

The second problem is the individual battery banks will not see identical loading. Some will be on for longer periods than others, and the current varies depending where each battery bank is being switched on and off in the sine wave. You need a random sequencing to ensure equal discharge of each battery bank over time. Not difficult, but it needs to be done.
The whole thing sounds conceptually simple, but its not when you look into it a bit more deeply.

If you want pure sine waves PWM is still the best approach at relatively low power and low voltage. A transformer PWM inverter will be more robust than a high voltage dc/dc PWM inverter, because peak power will always be limited by the dc/dc converter stage.
If you build a 1Kw dc to dc converter you cannot suddenly get it to deliver a 3Kw surge. But a 60Hz transformer can do that, without blowing up, provided you have enough parallel mosfets to drive the transformer.

The main limitation with PWM is that it does not scale up very well. Its fairly easy up to about 4Kw or 5Kw, but above that it gets really difficult to get large numbers of mosfets in parallel to switch together and load share. Its been done, but its not the sort of project a novice is likely to get going first attempt. However PWM is still the best choice up to around 5Kw.

Above that, the Warpverter has a great many advantages. It uses four square wave inverters driving four transformers. The secondaries are all connected in series so the voltages add. The secondary voltages go up in a 1 : 3 : 9 : 27 sequence. By switching the inverters on and off in the correct sequence, you can generate 40 voltage steps up, zero, and forty steps down. Its direct digital to analog conversion on steroids.

With 81 steps peak to peak, the measured harmonic distortion is typically 0.85% without any filtering. Third harmonic is the largest and its down to -40 db. General industry practice is that anything less than 2% THD is considered a pure sine wave. Around here the grid is more like 5% THD.

Anyhow, the big advantage is that all this square wave switching goes on at a relatively low switching frequency, and that is far less critical of layout. It also reduces switching losses and is theoretically more efficient. In practice the efficiency is pretty much identical to a good PWM inverter, about 92% to 94% at flat out full power. Zero load idling power is about the same too. A 5Kw Warpverter will have an idling power of around 20 to 30 watts, depending mostly on transformer design.

The biggest advantage of the Warpverter is it can very easily be scaled up in power. If you wanted 10Kw, 20Kw, 50Kw, it would be possible if you could get the transformers wound. With IGBTs, single devices rated to hundreds of amps are available, and they are far more rugged than mosfets and can withstand high fault currents, so can be protected with a circuit breaker.
But the really big IGBT power blocks are very slow, and unsuitable for high frequency PWM. But they will switch 60Hz just fine and withstand massive abuse.

The big down side with the Warpverter is the requirement for four transformers and four switching bridges. Its a lot of parts (and dollars) so its not worth doing for low power. But for 5Kw and above, its a lot easier to build and get going. There are now about fifteen warpverters in the 5Kw to 7.5Kw range now running successfully in various countries around the world, mostly in Australia, and several more being constructed including a three phase version.

So up to 5Kw your best bet is still a PWM transformer inverter, and that would probably suit most people.
Above that power, the Warpverter is well worth consideration.

The Warpverter has no commercial value, its just far too expensive to build, and just not cost competitive. But for home brewing, if you are prepared to wind your own transformers out of recycled junk, its a lot of work, but you can do it a lot cheaper than a manufacturer can, a manufacturer has to use all new copper and steel, you do not.
 
Last edited:
The reality is that for a domestic inverter the average power level is quite low, huge power peaks are fairly infrequent. But at night there will be almost no load at all.
No load idling current, and low load efficiency are far more important than full load efficiency.

With regard to the direct switched battery bank idea, its been done many times before, and its a pretty rough stepped waveform.
There are two inherent problems, the first one is charging the individual battery banks while its being switched around "live" in the inverter.
That requires multiple battery chargers, multiple banks of solar panels, and multiple MPPT controllers, all fully electrically isolated !
It all becomes vastly complex and impractical.

The second problem is the individual battery banks will not see identical loading. Some will be on for longer periods than others, and the current varies depending where each battery bank is being switched on and off in the sine wave. You need a random sequencing to ensure equal discharge of each battery bank over time. Not difficult, but it needs to be done.
The whole thing sounds conceptually simple, but its not when you look into it a bit more deeply.

If you want pure sine waves PWM is still the best approach at relatively low power and low voltage. A transformer PWM inverter will be more robust than a high voltage dc/dc PWM inverter, because peak power will always be limited by the dc/dc converter stage.
If you build a 1Kw dc to dc converter you cannot suddenly get it to deliver a 3Kw surge. But a 60Hz transformer can do that, without blowing up, provided you have enough parallel mosfets to drive the transformer.

The main limitation with PWM is that it does not scale up very well. Its fairly easy up to about 4Kw or 5Kw, but above that it gets really difficult to get large numbers of mosfets in parallel to switch together and load share. Its been done, but its not the sort of project a novice is likely to get going first attempt. However PWM is still the best choice up to around 5Kw.

Above that, the Warpverter has a great many advantages. It uses four square wave inverters driving four transformers. The secondaries are all connected in series so the voltages add. The secondary voltages go up in a 1 : 3 : 9 : 27 sequence. By switching the inverters on and off in the correct sequence, you can generate 40 voltage steps up, zero, and forty steps down. Its direct digital to analog conversion on steroids.

With 81 steps peak to peak, the measured harmonic distortion is typically 0.85% without any filtering. Third harmonic is the largest and its down to -40 db. General industry practice is that anything less than 2% THD is considered a pure sine wave. Around here the grid is more like 5% THD.

Anyhow, the big advantage is that all this square wave switching goes on at a relatively low switching frequency, and that is far less critical of layout. It also reduces switching losses and is theoretically more efficient. In practice the efficiency is pretty much identical to a good PWM inverter, about 92% to 94% at flat out full power. Zero load idling power is about the same too. A 5Kw Warpverter will have an idling power of around 20 to 30 watts, depending mostly on transformer design.

The biggest advantage of the Warpverter is it can very easily be scaled up in power. If you wanted 10Kw, 20Kw, 50Kw, it would be possible if you could get the transformers wound. With IGBTs, single devices rated to hundreds of amps are available, and they are far more rugged than mosfets and can withstand high fault currents, so can be protected with a circuit breaker.
But the really big IGBT power blocks are very slow, and unsuitable for high frequency PWM. But they will switch 60Hz just fine and withstand massive abuse.

The big down side with the Warpverter is the requirement for four transformers and four switching bridges. Its a lot of parts (and dollars) so its not worth doing for low power. But for 5Kw and above, its a lot easier to build and get going. There are now about fifteen warpverters in the 5Kw to 7.5Kw range now running successfully in various countries around the world, mostly in Australia, and several more being constructed including a three phase version.

So up to 5Kw your best bet is still a PWM transformer inverter, and that would probably suit most people.
Above that power, the Warpverter is well worth consideration.

The Warpverter has no commercial value, its just far too expensive to build, and just not cost competitive. But for home brewing, if you are prepared to wind your own transformers out of recycled junk, its a lot of work, but you can do it a lot cheaper than a manufacturer can, a manufacturer has to use all new copper and steel, you do not.
Very good points.

NASA has been throwing some money around for the switched battery bank and I've been able to latch on to some of it...and yes, not very practical and a waste of your tax dollars as far as I'm concerned- but thanks for your tax dollars guys :)

The idle current on my 5kW/48v inverter is about 2amps. That's near 2.4kWh over a 24h period. My loads are about 6kWh per day. Without opening up this inverter and poking around, my guess it that driving the 4 strings of pig fets and magnetizing the iron eats up most of them. I shut it down every night before I go to bed.

Peak power does not have to be limited by the DC-DC converter and done correctly a HF design can perform just as well as a LF design. Just make the DC-DC converter bigger. I have been designing PWM inverters for as long as I can remember and I'm ancient. I started with Delco bipolar transistors that had a gain of nearly 1 way back in the old days. I was on the team that developed IXYS Igbt's in the early 80s and they changed the game. Large inverters are reasonably efficient these days with IGBTs at lower carrier rates. These things can be annoying and any iron in it's path will scream. The newer Gan or SiC parts allow higher carrier rates with good efficiency. Likewise these parts are great for DC-DC converters.

to be continued....
 
The reality is that for a domestic inverter the average power level is quite low, huge power peaks are fairly infrequent. But at night there will be almost no load at all.
No load idling current, and low load efficiency are far more important than full load efficiency.

With regard to the direct switched battery bank idea, its been done many times before, and its a pretty rough stepped waveform.
There are two inherent problems, the first one is charging the individual battery banks while its being switched around "live" in the inverter.
That requires multiple battery chargers, multiple banks of solar panels, and multiple MPPT controllers, all fully electrically isolated !
It all becomes vastly complex and impractical.

The second problem is the individual battery banks will not see identical loading. Some will be on for longer periods than others, and the current varies depending where each battery bank is being switched on and off in the sine wave. You need a random sequencing to ensure equal discharge of each battery bank over time. Not difficult, but it needs to be done.
The whole thing sounds conceptually simple, but its not when you look into it a bit more deeply.

If you want pure sine waves PWM is still the best approach at relatively low power and low voltage. A transformer PWM inverter will be more robust than a high voltage dc/dc PWM inverter, because peak power will always be limited by the dc/dc converter stage.
If you build a 1Kw dc to dc converter you cannot suddenly get it to deliver a 3Kw surge. But a 60Hz transformer can do that, without blowing up, provided you have enough parallel mosfets to drive the transformer.

The main limitation with PWM is that it does not scale up very well. Its fairly easy up to about 4Kw or 5Kw, but above that it gets really difficult to get large numbers of mosfets in parallel to switch together and load share. Its been done, but its not the sort of project a novice is likely to get going first attempt. However PWM is still the best choice up to around 5Kw.

Above that, the Warpverter has a great many advantages. It uses four square wave inverters driving four transformers. The secondaries are all connected in series so the voltages add. The secondary voltages go up in a 1 : 3 : 9 : 27 sequence. By switching the inverters on and off in the correct sequence, you can generate 40 voltage steps up, zero, and forty steps down. Its direct digital to analog conversion on steroids.

With 81 steps peak to peak, the measured harmonic distortion is typically 0.85% without any filtering. Third harmonic is the largest and its down to -40 db. General industry practice is that anything less than 2% THD is considered a pure sine wave. Around here the grid is more like 5% THD.

Anyhow, the big advantage is that all this square wave switching goes on at a relatively low switching frequency, and that is far less critical of layout. It also reduces switching losses and is theoretically more efficient. In practice the efficiency is pretty much identical to a good PWM inverter, about 92% to 94% at flat out full power. Zero load idling power is about the same too. A 5Kw Warpverter will have an idling power of around 20 to 30 watts, depending mostly on transformer design.

The biggest advantage of the Warpverter is it can very easily be scaled up in power. If you wanted 10Kw, 20Kw, 50Kw, it would be possible if you could get the transformers wound. With IGBTs, single devices rated to hundreds of amps are available, and they are far more rugged than mosfets and can withstand high fault currents, so can be protected with a circuit breaker.
But the really big IGBT power blocks are very slow, and unsuitable for high frequency PWM. But they will switch 60Hz just fine and withstand massive abuse.

The big down side with the Warpverter is the requirement for four transformers and four switching bridges. Its a lot of parts (and dollars) so its not worth doing for low power. But for 5Kw and above, its a lot easier to build and get going. There are now about fifteen warpverters in the 5Kw to 7.5Kw range now running successfully in various countries around the world, mostly in Australia, and several more being constructed including a three phase version.

So up to 5Kw your best bet is still a PWM transformer inverter, and that would probably suit most people.
Above that power, the Warpverter is well worth consideration.

The Warpverter has no commercial value, its just far too expensive to build, and just not cost competitive. But for home brewing, if you are prepared to wind your own transformers out of recycled junk, its a lot of work, but you can do it a lot cheaper than a manufacturer can, a manufacturer has to use all new copper and steel, you do not.
.... I want to qualify my statement on the carrier rates of Igbts. They are running super-audio ranges these days and are so much better than the last time I used them doing industrial inverters. Last commercial inverters I did was 20 years ago at ABB. I'm involved with specifying some 1MW inverters and Eaton is doing a nice job with PWW down in Raleigh,NC. So to limit PWM to 5kW is a little bit of an injustice.

Mosfets are fairly easy to parallel and their are some ham radio guys (WA1QIX comes to mind) are running well into the MHz range with large parallel arrays. But yes to get there I'm sure there are bins of blown up parts, and not for the novice. I've run a home-brew full-bridge phase shift transmitter in the 160m bands with I'm told amazing audio. The modulation is taken care of with the phase shift.

A mosfet is actually more rugged than an IGBT. When they heat up they limit their fault current and self-protect. They take a beating before they let go. In between Darlingtons and Igbts ABB did a lot of 230v inverters with mosfets. We actually removed the overcurrent trip circuits and put a counter on the hardware current limit. Under a short circuit the inverter would 'sing' for a second before we shut it down. The IGBT will never survive a circuit breaker. The FBSOA curve will tell you that you need to be off in a couple of microseconds. A breaker will take milliseconds. A semiconductor fuse isn't even fast enough to clear a fault before the IGBT let's go.

Good luck with your Warpverter it's interesting work and I will keep my eye on it. I'm doing the do it yourself open source inverter (DIYOSI). 48vdc in 120vac out 5kW with 200%v overload.. Modules will be series or parallel capable for 120 or 240 out with split phase capable and eventually line interactive. This will use the resonant push-pull converter in the front end. If you've got some links to these up and running inverters that would be good to see. Maybe we can compare when complete. I think you may kick my ass on efficiency tho. Thanks for the intel and good luck with the project.
 
I am retired now, and out of the loop a bit.

But you are quite right, IGBTs have made giant strides in the last decade or so.
All the high voltage grid tie inverters run small IGBTs and they work fine with 20Khz PWM.
My own Warpverter uses 1200v 200 amp IGBTs that are rated for a full half cycle fault current of 1,000 amps. That is a 1,000 amp half sinusoid over 10mS (50Hz here) which will definitely trip a 20 amp C curve circuit breaker !

In fact I have tripped the ac breaker several times over the last 3+ years and the Warpverter did not even blink.

2018-06-26_0001.jpg

The problem with mosfets is as you say, the safe operating area. At very high fault currents, the safe "on" time shrinks to the microsecond region, and protection by even the fastest fuses is just not a practical proposition.

The idling current of your 5Kw inverter 5Kw/48v and 2 amps is very high by today's standards.
It can only be transformer magnetizing current doing that. But its pretty typical of a large transformer designed to run at commercial flux densities.
If you use grain oriented silicon steel and keep the flux density below one Tesla (10,000 Gauss) you should be able to do a lot better than that.

A 5Kw/48v inverter should be more like 5 watts for all the electronics and 15 to 20 watts magnetizing current for the toroidal transformer.
That is pretty typical for home brewed transformer design over at "The Back Shed" Forum.

O/k on the Ham Radio too. I am VK3ALY , cheers.
 
Last edited:
VK3ALY de N9NEO.

Ok, the 2 amps is on the DC side. So near 100W of losses. The magnetizing current is reactive so (in theory) no losses. I think you are right - crappy iron and the thing is probably saturating as well. I'm not sure what the pwm rate is in the AIMS unit, but they are using strings of early IR parts which are notorious for gate charge inefficiencies. Some energy there too.

I am talking about direct short circuit survivability. Direct short the poles on the inverter and the IGBT current on a 200A part will skyrocket well beyond 1000amps. Best case is the two transistors in the fault share the bus voltage. Worse case is the short appears while in a zero state (Both upper or both lower transistors on for those not familiar) and one transistor turns on into the short. A lot of attention was paid and work required to get the early IGBTs to survive. Desat detection and soft turn off and so on.. Early IGBTs also suffered latching. The trick back then was to turn them off softly to prevent the latching. Similar problems with the darlingtons. When we went from six step drives to PWM we had problems surviving faults. The difficulties were because of the negative base drive we were using because the faster PWM rates. The negative base drive caused current crowding and boom. Where are you that you are using 50Hz?
 
Hi,

the 3 bank supply isn't a problem if you use resonant push-pull converter to made this 3 isolated supply from standard 48vdc lifepo4.
these stepup converter also made the inrush current protection like many inverter does.
Nowadays resonant push-pull converter show > 97% efficiency at low load and small power :

So maybe this is drawing some complexity but an inverter with 96% efficiency at low loading output is not common
 
Back
Top