99.5% Efficient Inverter Design

[Roswell Bob]

I intend to run DC-DC converter portion open loop. I'm seeing 170v-200v from 200W to 5kW load. I think I can run a slow feedforward loop on the inverter side and do fine.

What a perfectly odd topology......
Its certainly not a half bridge circuit as the author suggests.
More like a kind of weird push pull flyback topology driving a voltage doubler through a 2:1 step up transformer.
Its a rather sneaky way to generate a high voltage though.

As he states, the mode of operation totally changes below 50% duty cycle, it becomes discontinuous, which would make the design of a stable voltage feedback network rather challenging. At very low loads it MUST operate at lower duty cycles than 50%.

The series capacitor on the output kinda resembles the half-bridge. I think you're right, voltage doubler may have been a better term.

 

[Kinetic]

Has anyone been able to get the schematics or layout from the author on this? I've written nicely a few times using different email accounts and gotten no response.

 

[Shunt Czar]

Transformerless inverters must be carefully designed.
See

 

[Seven30]

I run my LG washer right off solar panels skipping the boost section of an old modified inverter. Efficient enough without having to build anything new.

I just ran across this interesting statement. Does this mean feeding the inverter switching section directly bypassing the high voltage DC supply?

 

[Seven30]

Certainly an interesting approach, but way beyond the average hobbyist and home constructor to get working.
Many people in Forums such as this one, expect to be able to home build a high power inverter themselves, do not even own an oscilloscope.

I get many complaints that the Warpverter is far too complex and uses far too many components, and its just four hard switched low frequency square wave bridge inverters.
It would also take several of these circuits to get the individual fully isolated supplies needed to generate a very rough stepped sinusoid.

Its still interesting though, but unfortunately way beyond what any electronics novice could hope to build and get going.

Its interesting that Tektronix 7000 series scopes introduced a resonant switcher in the early 70s:
"The frequency at which the inverter runs is determined basically by the resonant frequency of a series-LC network placed in series with the primary of the _power transformer. "

 

[efficientPV]

I just ran across this interesting statement. Does this mean feeding the inverter switching section directly bypassing the high voltage DC supply?

Yes, it will work with older discrete MSW inverters (likely only with non microprocessor designs. It will work from 20V up to the inverters rating. About 45ma of 12V has to be supplied to the control side of on switch. Inverters 12V power terminals are not used. So, for impedance protected motors of fans and pumps you could probably operate at 45V till voltage rises enough to run.

 

[Warpspeed]

Resonant mode dc/dc converters have been around for quite a long time, there is nothing new about the basic concept.
What has changed are some of the newer and more clever topologies.

Designing such a beast to be 99% efficient is certainly possible, but only with a fixed or fairly fixed load. That is not usually a problem when the dc/dc converter is designed to operate WITHIN a piece of equipment. Something like a computer monitor or some kind of instrumentation like an oscilloscope where the load is forever constant. The supply is either switched off or working at its normal fixed unvarying load.

Now an inverter is a very different beast. It must run with zero load, or any load up to full rated continuous power, and maybe even a short term surge load of MULTIPLES of its full load rating for (say) induction motor starting.

This is where resonant mode supplies are problematic. Continuous mode operation with some kind of choke cannot work at zero load. Zero load means zero output current, and current through an inductor cannot be continuously flowing if its not there at all !!
Likewise trying to meet the needs of for example, a 300% short term overload surge might be rather difficult as well, with many resonant topologies.

So 99% efficiency at only one output power level only is not useful in an inverter used with mixed domestic loads.
Especially if it simply cannot run without any load at all, or any type of horrible surge overload.

People get all excited reading about 99% efficiency, not realizing that it has other crippling shortcomings making it totally unsuitable for a domestic inverter application.

 

[RCinFLA]

The scheme is similar to the original Trace inverter design but with batteries supplying the three sub-voltages directly.

This is using 9x12v = 108v, 3x12v = 36v, and 1x 12v =12v.
Original Trace inverter uses 135v, 45v, and 15v.

The H-bridges on each voltage level can produce +, -, or 0 voltage outputs.

Putting together the three voltages in series builds the pseudo sinewave, being a little choppy, with the discrete steps.

This gives 3^3 = 27 maximum possible discrete levels, although for regulation between battery voltage and AC output voltage the result is less levels being used much of the time. High battery voltage and low AC voltage gives the least steps with choppiest output waveform.

The Trace inverter used three transformers. I call them 'pappa bear' (135v), 'mama bear' (45v), and 'baby bear' (15v). Primary side of each transformer is driven by its own H-bridge from single battery (24v or 48v models). Secondary sides of transformers are just tied in series.
Since all series connected transformers' secondary sides have the same current, the 135v transformer supplies most of the power and is biggest transformer, the 15v transformer is the smallest.

Problem with this particular, direct from battery, transformer-less design is no isolation to AC output. Also, the charging of batteries in this transformer-less design cannot easily be done directly from the same path and would likely require three separate chargers.

 

[SpongeboB Sinewave]

The scheme is similar to the original Trace inverter design but with batteries supplying the three sub-voltages directly.

This is using 9x12v = 108v, 3x12v = 36v, and 1x 12v =12v.
Original Trace inverter uses 135v, 45v, and 15v.

The H-bridges on each voltage level can produce +, -, or 0 voltage outputs.

Putting together the three voltages in series builds the pseudo sinewave, being a little choppy, with the discrete steps.

This gives 3^3 = 27 maximum possible discrete levels, although for regulation between battery voltage and AC output voltage the result is less levels being used much of the time. High battery voltage and low AC voltage gives the least steps with choppiest output waveform.

The Trace inverter used three transformers. I call them 'pappa bear' (135v), 'mama bear' (45v), and 'baby bear' (15v). Primary side of each transformer is driven by its own H-bridge from single battery (24v or 48v models). Secondary sides of transformers are just tied in series.
Since all series connected transformers' secondary sides have the same current, the 135v transformer supplies most of the power and is biggest transformer, the 15v transformer is the smallest.

Problem with this particular, direct from battery, transformer-less design is no isolation to AC output. Also, the charging of batteries in this transformer-less design cannot easily be done directly from the same path and would likely require three separate chargers.

View attachment 85572View attachment 85573

Ahhh ! Yes, that picture on the left is the one I captured while at Trace around 1998.

The RMS voltage regulation was done with different waveforms from ROM.

BTW, this topology is pretty old. The oldest place I saw it was in an inverter book from the early 1960s.

Without some kind of filtering, usually involving magnetics, EMI will be an issue. Resonant would definitely help if it worked perfectly at all conditions.

You might be able to charge all three batteries from the grid, only because it is a synchronous but maybe not because you don't have the turns ratio at your disposal. Charging lithium, might not be that easy either way.

Highest frequency would be from the baby bear portion at a few hundred Hz. Also the batteries don't have to be the same size or voltage.
The problem with the high voltage part of this is that the voltage is higher --- at least 200V FETs instead of 80V or 100V which doesn't allow nice big low voltage FETs and would have more conduction and switching losses.

Designing inverters to be sold as a viable product isn't that easy or quick especially if you are going to have it UL or ETL or NRTL listing.

None of this stuff is "easy" though. We need more technical people in the US. There are definitely some very smart people on this forum that seem to understand this stuff !

boB

 

[Warpspeed]

That is EXACTLY how my Warpverter works, except I use four inverters and four transformers creating 81 steps instead of 27 steps. That improves the total harmonic distortion from about 3% to less than 1%.

Yes different waveforms from ROM can produce slightly different output voltages and provide voltage regulation. My Warpverter uses 256 different 1K lookup tables. This idea is not new, my first attempt at building an inverter like this was about forty years ago. I have built quite a few different versions since then, each one has been better and with simpler circuitry.

EMI is only a problem with fewer steps. The more steps the closer the current in ALL of the transformers becomes a pure sinusoid. The voltages are obviously rectangular waveforms that are combined in the secondaries, but all of the currents in the primaries are pure sine waves.
This is what 81 steps looks like, and my four transformers:

 

[curiouscarbon]

looks rather sinusoidal to me!

 

[Warpspeed]

This is what 27 steps looks like, it is perfectly acceptable with usually less overall distortion than the grid at my location.

The ROM produces eight bits anyway, which can easily drive a fourth inverter, which only needs to be quite small.
My inverter works from a dc input voltage of between 80v and 160v. Its nominally 100v from 30 Lithium cells.

This is how the output waveform changes to maintain a constant ac output voltage with varying dc supply.
Notice how the individual steps get progressively taller as the dc voltage increases, the steps narrow, and become less in number as the dc voltage is increased. By jumping between lookup tables right at the zero crossings, there is no waveform discontinuity.

 

[SpongeboB Sinewave]

Interesting ! What processor are you using ?

How much power is this warp-verter good for and have you ever weighed it ?

What kind of FETs ? Have you shown this inverter on Otherpower (Field Lines) forum ?
It looks like it might be semi familiar.

boB

 

[Warpspeed]

Earlier versions used several different microcontrollers over the years, but the latest version is purely a hardware design.
Its designed for 5Kw continuous, but can easily handle multiples of that for short periods.
Its easily scalable up in power with no real practical upper power limit.

I have no idea what it weighs, but its very heavy indeed.
The transformers are bolted to a 6mm thick steel plate sitting on four heavy swivel castors.
Over that slips the outer shell of a two drawer filing cabinet turned around so the back becomes the front.

The electronics are all fitted onto a single large heatsink attached to the lid, which hinges upwards on a pair of gas struts.
It all opens up for excellent access for maintenance, everything is clearly visible and easy to get to.
It uses large (slow) IGBT power modules rated for 1200v and 200 amps for the two larger inverters, and individual IGBTs for the two smaller inverters. It all makes for a very simple and compact inverter, although the overall outside enclosure is huge.

Its been running now successfully for four years without missing a beat, and its at least known about on several solar type Forums. It may be new to some members here, but in fact its been around and known about for quite while.
There are now at least fifteen of these now up and running around the world that I know of, and possibly others. All are in the 5Kw to 7.5Kw class.

At least two software guys have come up with alternative driver boards using a Nano microcontroller. These are completely functionally identical to my hardware driver board, and I have run them experimentally in my own Warpverter, and its not possible to know which board is actually in there without looking.

 

[BiduleOhm]

I love the full logic gate design ?

I'm curious about one thing: why did you choose IGBTs over FETs?

 

[Warpspeed]

Many reasons.
IGBTs are far more robust and withstand current overloads much better than mosfets.
IGBTs tend towards having a fairly constant forward voltage drop (somewhat like a diode). Power dissipation is very roughly proportional to current.
The particular IGBTs I am using are rated for 200 amps continuous, but the data sheet says they can handle a 1,000 amp 10mS half sinusoid. They are easily protected by a normal C curve circuit breaker on the ac output of the inverter. Each big white block contains a half bridge of two IGBTs bolted beneath a pair of heavy aluminium busbars.
Very simple construction, very easy to change an IGBT in minutes without a soldering iron. Very short leads to the huge low esr electrolytics.
All very simple and compact.

Mosfets act like resistors when in the on state, so power dissipation goes up proportional to current squared. This makes them much more fragile to current overload, and even a brief surge overload results in an explosive release of destructive heat. A fifty amp mosfet will die with any more current than that. A fifty amp rated IGBT can survive a higher surge, provided its for a short enough duration.
Mofets are very fast though, with far lower switching losses.
In this type of inverter, high switching speed is not required. The largest inverter switches at a leisurely 50/60Hz.

While mosfets are definitely the best choice at lower dc voltages and high frequency PWM, IGBTs become superior at much slower switching rates, especially at higher dc voltages.

Warpverters have been built with mosfets to run at 48v but as many as 40 mosfets have been used to do that by some. While 40 mosfets end up being a lot cheaper than the big IGBT power bricks, if a dozen or so mosfets ever need replacing, its not a task I would look forward to.

The big IGBT modules are about $50 US dollars each, but the convenience far outweighs that IMHO.

 

[SpongeboB Sinewave]

NIce !

But I don't understand why you are using 1200V 1/2 bridge switches ?

What is your battery voltage ? Something higher than 48V ? I couldn't quite tell...

boB

 

[Warpspeed]

Battery voltage here is 100v nominal (thirty lithium cells).

These high power IGBTs are usually rated for 600v or 1200v, in various sizes up to around 500 amps.
The particular ones I used happened to be 1200v 200 amp versions advertised on e-bay at the time.

This is a very similar type advertised on e-bay right now, 1200v, 220 amps $57 US.
https://www.ebay.com.au/itm/402912221839?hash=item5dcf709a8f:g:pNoAAOSw7A9gxvlN&frcectupt=true

 

[BiduleOhm]

Many reasons.
IGBTs are far more robust and withstand current overloads much better than mosfets.
IGBTs tend towards having a fairly constant forward voltage drop (somewhat like a diode). Power dissipation is very roughly proportional to current.
The particular IGBTs I am using are rated for 200 amps continuous, but the data sheet says they can handle a 1,000 amp 10mS half sinusoid. They are easily protected by a normal C curve circuit breaker on the ac output of the inverter. Each big white block contains a half bridge of two IGBTs bolted beneath a pair of heavy aluminium busbars.
Very simple construction, very easy to change an IGBT in minutes without a soldering iron. Very short leads to the huge low esr electrolytics.
All very simple and compact.

Mosfets act like resistors when in the on state, so power dissipation goes up proportional to current squared. This makes them much more fragile to current overload, and even a brief surge overload results in an explosive release of destructive heat. A fifty amp mosfet will die with any more current than that. A fifty amp rated IGBT can survive a higher surge, provided its for a short enough duration.
Mofets are very fast though, with far lower switching losses.
In this type of inverter, high switching speed is not required. The largest inverter switches at a leisurely 50/60Hz.

While mosfets are definitely the best choice at lower dc voltages and high frequency PWM, IGBTs become superior at much slower switching rates, especially at higher dc voltages.

Warpverters have been built with mosfets to run at 48v but as many as 40 mosfets have been used to do that by some. While 40 mosfets end up being a lot cheaper than the big IGBT power bricks, if a dozen or so mosfets ever need replacing, its not a task I would look forward to.

The big IGBT modules are about $50 US dollars each, but the convenience far outweighs that IMHO.

Ok, I see.

I was curious because you can have far lower losses with FETs than IGBTs for high currents, but I can see how the higher robustness of the IGBTs can be interesting here

 

[RCinFLA]

IGBT's are pretty rugged but they have the bipolar Vce saturation characteristic. For low voltage, high current, MOSFET's will usually provide lower loss with lower Vds drop at high current compared to IGBT's Vce voltage drop. IGBT's Qg gate charge is a little lower than MOSFET's so gate driver power consumption is a little lower for IGBT's. IGBT's have lower maximum switching speeds compared to MOSFET's.

Just about all high freq hybrid inverters use MOSFET's for battery side H-bridge to ferrite transformer and IGBT's for HV DC synchronous rectifier bridge and output HV DC to sinewave PWM H-bridge. Since the battery side MOSFET's and IGBT's synchronous rectifiers have to be synchronized, the switching speed has to be carefully selected compromise that is acceptable to synchronous rectifier IGBT's.

IGBT's win out for high DC voltage switching. No hard fixed rule, but less than 150 vdc is MOSFET territory.

Interestingly, based solely on Youtube India HF inverter repair video examples, the output PWM IGBT's seem to be dominate failure mode. Probably the result of not-so-smart connecting AC input source to inverter AC output terminals. The side-by-side, input-output AC port terminal strip placement on inverter makes this a common mistake. Battery side MOSFET toasting is usually the result of excessive overload surge on inverter when the high freq ferrite transformer saturates.