Disadvantages of Using half-wave instead of full-wave in Pure Sinewave Inverter?

[KerimF]

Every pure sinewave inverter, I knew or heard of (12Vdc or 24Vdc to 220Vac), that uses laminated iron transformer (input: 1 coil, output: 1 coil) drives its transformer with a 4-MOSFET bridge.

Lately, I am designing a pure sinewave inverter but by using 2-MOSFET push-pull driver (lower side only) with a half-wave transformer (input: 2 coils, output: 1 coi).

I used to believe that simplifying a board likely creates some disadvantages.

So, I hope some experienced engineers here could help me know in advance certain expected disadvantages in using the push-pull configuration, in this case, instead of a bridge.

Thank you.
Kerim

[PCB.Wiz]

What do you mean by 2 mosfet push pull ?
If you mean 2 N MOSFETS in half wave drive, yes, that saves 2 MOSFETS but
* you now need 2 clamp circuits
* worse, is the current flows in half the transformer copper at a time.
 
It’s the same reason half wave rectifiers ‘waste’ transformer resource, and bridge rectifiers are preferred.

[coromonadalix]

and you think you can re-invent the wheel ?

the push pull is the best, if you dont use that,  it work in a flyback fashion, if the push pull is well made you have zero cross over defects

your single side mosfets seems  more like an class A amplifier, it will heat more for nothing

and pls  give some schematics to show your idea

[Ian.M]

Also, the end of the primary not currently driven must swing above the supply rail so you need MOSFETs with double the Vds rating.

[Zero999]

I've not seen any which use a large laminated iron transformer. The ones I've seen use an isolated DC:DC converter to a get DC of the peak voltage of the mains, followed by a MOSFET H-bridge and filter.

[KerimF]

What do you mean by 2 mosfet push pull ?
If you mean 2 N MOSFETS in half wave drive, yes, that saves 2 MOSFETS but
* you now need 2 clamp circuits
* worse, is the current flows in half the transformer copper at a time.
 
It’s the same reason half wave rectifiers ‘waste’ transformer resource, and bridge rectifiers are preferred.

Good remark,
If I understood you well, the price is more copper at low-voltage side in order to get the same resistance of the 1 coil transformer (driven by a bridge).
But let us also recall that the resistance added by a bridge is 2*Ron instead of 1*Ron in case of half wave drive.

[KerimF]

and you think you can re-invent the wheel ?

the push pull is the best, if you dont use that,  it work in a flyback fashion, if the push pull is well made you have zero cross over defects

your single side mosfets seems  more like an class A amplifier, it will heat more for nothing

and pls  give some schematics to show your idea

The driver is simple. The MCU (actually an ATmega8) generates two anti-phase SPWM (sinewave PWM) at relatively high frequency (actually 15,625Hz).
The N-MOSFET driver is a conventional one that uses 3 BJT transistors. The dead time between the two drivers is 1us.

[KerimF]

Also, the end of the primary not currently driven must swing above the supply rail so you need MOSFETs with double the Vds rating.

Another good remark.
Fortunately, the new power MOSFETs have ratings (Vds and Id) suitable for power inverters if supplied by 12Vdc or 24Vdc.

[coromonadalix]

why dont you use already know smps parts, pwm ic's  etc ...

[KerimF]

I've not seen any which use a large laminated iron transformer. The ones I've seen use an isolated DC:DC converter to a get DC of the peak voltage of the mains, followed by a MOSFET H-bridge and filter.

You don't surprise me because you refer to the most efficient topology which could be used for a pure sine wave inverter.

In where I live, I can't design and produce a product which needs ferrite transformers because I can't get the types I may need in quantities.
So, I had to use the available laminated iron cores if I like to add such inverters to my collection.
I used to believe that having/doing something (not very good) is much better than having/doing nothing

For instance, the small leakage inductance of an iron transformer does the filtering if a relatively small capacitor (about 1uF, high AC voltage) is added at its output.

[KerimF]

why dont you use already know smps parts, pwm ic's  etc ...

The answer is simple. They are not available, or their prices are relatively high, in my local market. (Unfortunately, after year 2011, ordering electronic components from abroad by ordinary people like me became out of question... my golden years are things of the past after the world... saved the people among whom I was born and live).

On the other side, the price of the available ATmega8 is not high (perhaps because it is old and replaced by new ones). Also, the price of BJTs is good.
And to gain my daily bread, my designed products have to compete similar ones that are imported from abroad (lately by the privileged rich families only) or made locally.

 

[PCB.Wiz]

Good remark,
If I understood you well, the price is more copper at low-voltage side in order to get the same resistance of the 1 coil transformer (driven by a bridge).

Basically, yes. Each winding has the same peak current as a bridge, but half the time.
So you have wasted space in the transformer and can get less loss, by using a larger copper gauge, single winding in the same physical space.

But let us also recall that the resistance added by a bridge is 2*Ron instead of 1*Ron in case of half wave drive.

In 2024 that's immaterial, the days when low milliohms FET were hard to come by, are long gone.

[KerimF]

Good remark,
If I understood you well, the price is more copper at low-voltage side in order to get the same resistance of the 1 coil transformer (driven by a bridge).

Basically, yes. Each winding has the same peak current as a bridge, but half the time.
So you have wasted space in the transformer and can get less loss, by using a larger copper gauge, single winding in the same physical space.

I agree with you. And this is indeed a disadvantage which is the price for using a much simpler circuit.
 

But let us also recall that the resistance added by a bridge is 2*Ron instead of 1*Ron in case of half wave drive.

In 2024 that's immaterial, the days when low milliohms FET were hard to come by, are long gone.

Knowing that the primary resistance (in a power inverter) is around a few mR, the low values of Ron (also around a few mR) are not negligible even if every set is formed by 4 N-MOSFET in parallel.

[coromonadalix]

and you do know, i hope 

using any mcu for generating a sinal will not be pure as you need,  it will be called pseudo sinusoidal, you'll need good filtering, if you want to acheive a somewhat good sinus wave out of it

you could have some equipment compatibility problems

IE  i have  some laptop chargers who wont refuses to start even with pseudo signal inverters,  with pure sinewave they start  etc ....  some tv's  dont work too (they use ac-dc adapters)

while i salute your idea,   you may face other challenges you dont even know
i can understand  some material could not be available ...  etc ...  but its not a reason to play with safety, protection  etc .. 

[Kim Christensen]

Every pure sinewave inverter, I knew or heard of (12Vdc or 24Vdc to 220Vac), that uses laminated iron transformer (input: 1 coil, output: 1 coil) drives its transformer with a 4-MOSFET bridge.

The driver is simple. The MCU (actually an ATmega8) generates two anti-phase SPWM (sinewave PWM) at relatively high frequency (actually 15,625Hz).
The N-MOSFET driver is a conventional one that uses 3 BJT transistors. The dead time between the two drivers is 1us.

Are you attempting to drive a standard laminated iron power transformer at 15Khz instead of 50Hz or 60Hz?
How much power (Watts @ 220Vac) will this inverter be outputting?

[KerimF]

using any mcu for generating a sinal will not be pure as you need,  it will be called pseudo sinusoidal, you'll need good filtering, if you want to acheive a somewhat good sinus wave out of it

You are right in case the code is not well written. For instance, I had to write my codes, since early 80's, in assembly language only because I couldn't get, for free or else, high language compilers as of C for example.

I managed to output the SPWM signals at pins OC1A and OC1B of ATmega8 (using fast PWM of timer 1). The rate is 15,625 Hz. The number of samples per cycle (50 Hz) is 312.
The amazing thing is that the small leakage inductance of the iron power transformer does the filtering with a relatively small high voltage capacitor (around 1 uF). I am not sure how good (speaking THD) the sinewave, I got, is. But it looks on the scope screen it looks a sine wave without ripple.

Naturally, to maintain 220V at the output, the MCU reads the output voltage in every cycle (50 Hz) to adjust the gain of the PWM for the next cycle. This voltage regulation allows a soft start at boot (from 0V to 220V, in 800ms). The MCU has other functions related to safety in case of faults.

you could have some equipment compatibility problems
IE  i have  some laptop chargers who wont refuses to start even with pseudo signal inverters,  with pure sinewave they start  etc ....  some tv's  dont work too (they use ac-dc adapters)

In the last 12 years, I produced and sold about 5 thousand 'adaptive square wave' inverters (regulated by varying the duty cycle). Lately, it became clear that every device (made for 50Hz sinewave) whose power supply uses a series capacitor to limit the internal supply current, is damaged, sooner or later, by a square wave inverter. Summer is at the door and electric fans will be used soon. Most of these fans are controlled by electronic boards supplied internally by a series capacitor (no more by a small transformer). So, I decided to also produce low-cost sinewave inverters, at least for these fans.

I admit that with the limited available components around me I can't do miracles by making such inverters also with high efficiency. Losses in the iron transformer whose available laminated core is also of the lowest possible grade! (Bs=1.05 Tesla, Br=0.7 Tesla and Hc=100 A/m) is much higher than of ferrite's.

while i salute your idea,   you may face other challenges you dont even know

In the initial tests, it became clear that the transformer acts as a current source when overloaded. In this case, the output cannot reach the sine flat top and starts to look like a triangular wave instead. Fortunately this wave doesn't damage the series capacitor supplies, as the fast edges of a square wave do.

[KerimF]

Are you attempting to drive a standard laminated iron power transformer at 15Khz instead of 50Hz or 60Hz?

I am somehow surprised that most members here didn't hear yet how pure sine wave inverters can also use a laminated iron power transformer. I heard of such inverters in many other EE forums, since more than a decade ago.

Unfortunately, I also didn't believe first that this can be done, mainly without external filtering. But since my private business has slowed down lately, I had to re-think carefully about it. And I realized that the small leakage inductance of the iron transformer is good enough to acts as LPF for the sinewave PWM high frequency (modulated to produce 50Hz sinewave if filtered) by adding a small high voltage capacitor.     

How much power (Watts @ 220Vac) will this inverter be outputting?

The planned powers are 500W and 1000W on 12Vdc, 1000W and 2000W on 24Vdc.

[coromonadalix]

well you do sound like a specialist  with your written text tone     who frankly did not need to come here i think .......

[KerimF]

well you do sound like a specialist  with your written text tone     who frankly did not need to come here i think .......

Good remark.
But I never believed that I could reach perfection in knowledge (unless it is about the spiritual realm) no matter how long I may live.
So, in the material world, I see myself always in the learning stage. After all, I am not that smart. It took me about 10 years to believe that a conventional iron power transformer can be used practically in a pure sinewave inverter (though with lower efficiency).

By the way, if I didn't come asking for possible new knowledge, I may have missed, and I simply presented what I did lately, I would look also as a bad guy This happened to me when I started a thread (in the RF forum) about my simple reliable topology to demodulate DSB-SC signal which combines the two known ones, Costas Loop and Squaring Loop. It has no I-Q signals (as in Costas Loop) and has no selective filter at 2*fc (as in the Squaring Loop). Lately I called it 'Harmonic Loop' since its PLL locks to the second harmonic of the suppressed carrier frequency. I did it as an MS thesis at the American University of Beirut (AUB) in spring 1979. But I didn't submit it (as a document) because I had to return home early for financial reasons. But I took advantage of it in my short-range RF links (between home and workplace, 3 Km) for many years in the 80’s.

[Kim Christensen]

I am somehow surprised that most members here didn't hear yet how pure sine wave inverters can also use a laminated iron power transformer. I heard of such inverters in many other EE forums, since more than a decade ago.

Because there are much better ways of doing it. (I know you can't do it the standard way due to lack of parts). You're going to have large eddy current losses in the transformer which will cause it to heat up. This will reduce efficiency and require extra cooling of the transformer and/or require it's power output to be derated. After Googling about others who have done the same as you're proposing, they run into exactly this problem. The transformer overheats and requires extra cooling.

[Zero999]

Why not put the filter before the transformer?

Only half of the power in the PWM waveform is delivered to the load. Half of it is blocked by the filter. There's no point in the transformer passing the unnecessary harmonics and generating the associated losses. It's better to filter them out first.

[Xena E]

The rate is 15,625 Hz. The number of samples per cycle (50 Hz) is 312.

Interesting.
(To be pedantic the number of samples per second is 312.5 BTW).

This post really jumps out at one who designs this sort of thing for a living.

Now if I were to use standard power line frequency iron, which even if it were a low frequency design, I wouldn't  I'd say that you could perhaps expect the most efficient frequency to transfer power at would be in the range of 500Hz to 2kHz.

Any fancy synthesis of a 50Hz wave form would need to be filtered before hitting the primary of your power transformer.

Also has anyone noted that 15,625Hz is the horizontal scan frequency of the 625 line analog television standard?

Very Faringdon-esque O/P

Still, I'll continue watching, I may learn something from the other  honorable posters.

Regards Xena.

 

[KerimF]

Why not put the filter before the transformer?

Only half of the power in the PWM waveform is delivered to the load. Half of it is blocked by the filter. There's no point in the transformer passing the unnecessary harmonics and generating the associated losses. It's better to filter them out first.

For instance, Zero999 (reply #4) reminded us of the most efficient topology to build a pure sinewave:
"The ones I've seen use an isolated DC:DC converter to a get DC of the peak voltage of the mains, followed by a MOSFET H-bridge and filter."
I guess you know that in this case the DC:DC converter and the output filter have to be made on ferrite cores which I can't get in quantities.
Should I forget producing pure sine wave inverters just because the new world allows me to use the lowest grade of laminated iron only (besides many other limitations)? On the other side, almost all engineers around the world (I hope) can get, in one way or another, whatever they may need in their designs.

I even chose the half wave drive instead of the full wave one to reduce even more the cost and labor.

Now, returning to your suggestion, how the filter could be done without using a ferrite choke? I hope you get the idea.

[KerimF]

The rate is 15,625 Hz. The number of samples per cycle (50 Hz) is 312.

Interesting.
(To be pedantic the number of samples per second is 312.5 BTW).

Your calculation is right. But, on my side, the calculation is:
64 us (15,625 Hz) * 312 = 19968 us, that is 50.08 Hz

Now if I were to use standard power line frequency iron, which even if it were a low frequency design, I wouldn't  I'd say that you could perhaps expect the most efficient frequency to transfer power at would be in the range of 500Hz to 2kHz.

The main reason of generating a pure sign wave 'at any efficiency' is that, where I live, consumers are more concerned for the price and reliability (a product that doesn't need to be repaired one a while). And as long the efficiency is not too bad, they don't complain.

Any fancy synthesis of a 50Hz wave form would need to be filtered before hitting the primary of your power transformer.

I think I have to repeat what I replied to some others.
If I can get ferrite cores suitable for my designs, I simply build my new inverters as the one which Zero999 (reply #4) was referring to.
 

Also has anyone noted that 15,625Hz is the horizontal scan frequency of the 625 line analog television standard?

It is just a mere coincidence. The MCU crystal is 8 Mhz. The timer 1 prescaler is 1 (its clock = 0.125 us). The PWM register is 9 bit (512). Therefore, the PWM period is 64us (512*0.125) and its frequency is 15,625 Hz.

[KerimF]

I am somehow surprised that most members here didn't hear yet how pure sine wave inverters can also use a laminated iron power transformer. I heard of such inverters in many other EE forums, since more than a decade ago.

Because there are much better ways of doing it. (I know you can't do it the standard way due to lack of parts). You're going to have large eddy current losses in the transformer which will cause it to heat up. This will reduce efficiency and require extra cooling of the transformer and/or require it's power output to be derated. After Googling about others who have done the same as you're proposing, they run into exactly this problem. The transformer overheats and requires extra cooling.

I agree with all you said.
About the extra dissipation of the transformer, this may be a problem if the inverter is loaded at its maximum power for a long time. Actually, it is not the case for most local consumers, if not all, who have to preserve their battery charge as long as possible.
Soon after year 2011, we lost the mains network completely. With time, mains electricity has been restored gradually. And now (after 12 years), we get the mains voltage (220V, 50Hz) 6 hours per day at best.

[Zero999]

Now, returning to your suggestion, how the filter could be done without using a ferrite choke? I hope you get the idea.

By using a lower frequency.

Putting the filter before the transformer will not change how the filtering is done. The only difference is, the transformer doesn't have to pass the high frequency components of the signal, which will dramatically reduce the losses.

[KerimF]

Now, returning to your suggestion, how the filter could be done without using a ferrite choke? I hope you get the idea.

By using a lower frequency.

Putting the filter before the transformer will not change how the filtering is done. The only difference is, the transformer doesn't have to pass the high frequency components of the signal, which will dramatically reduce the losses.

Thank you for your feedback.

Just to be sure I understand to what you are referring as 'transformer's losses', a 500W transformer under test, for example, draws about 1 A DC from the 12V acid battery when in standby (no load). Do you refer to this kind of loss?

As you know, the extra losses in a loaded transformer are in the wirings (copper), not in the core. At high frequencies, the fundamental 15625 Hz and its harmonics, the resistance of wires increases due to the skin effect.
On the one hand, this increase helps the leakage inductance in filtering at the transformer's output that has a capacitor of a relatively small value.
On the other hand, this also increases the dissipation of the copper.

Unfortunately, since we get lately a couple of hours of electricity only in daylight and the same at night, it is not easy for me to do various tests and measurements anytime I like (I am fortunate for ordering a fluke 200MHz digital scope before year 2011 because this the only tool I have now).
And on my sole desk of work, I have to also test and adjust the various assembled boards of what we produce already.
Also, my acid battery (made locally, supposed to be 100 AH) is somehow old (its internal resistance is around 50mR). And the length of the thick wires (10 mm2) from the battery to the desk had to be around 3m (6m all). This relatively long distance is not a problem at low frequencies and loads. But to supply currents pulsed at 16 Khz, they act as thin wires instead (besides adding an inductive reactance).

My real problem is that, lately, I am almost broken while I have to design new products good enough to compete similar ones in the local market. Yes, there are already pure sine wave inverters in the local market that uses iron transformers which are imported from India, China or made locally. But all of them use transformers, having 1 primary coil, instead of two coils in mine. So, I suppose that their efficiency will be comparable to mine since they also use a SPWM high frequency which is filtered solely by the power transformer to also produce a clean sinewave at 50 Hz.

[Zero999]

Now, returning to your suggestion, how the filter could be done without using a ferrite choke? I hope you get the idea.

By using a lower frequency.

Putting the filter before the transformer will not change how the filtering is done. The only difference is, the transformer doesn't have to pass the high frequency components of the signal, which will dramatically reduce the losses.

Thank you for your feedback.

Just to be sure I understand to what you are referring as 'transformer's losses', a 500W transformer under test, for example, draws about 1 A DC from the 12V acid battery when in standby (no load). Do you refer to this kind of loss?

As you know, the extra losses in a loaded transformer are in the wirings (copper), not in the core. At high frequencies, the fundamental 15625 Hz and its harmonics, the resistance of wires increases due to the skin effect.
On the one hand, this increase helps the leakage inductance in filtering at the transformer's output that has a capacitor of a relatively small value.
On the other hand, this also increases the dissipation of the copper.

Unfortunately, since we get lately a couple of hours of electricity only in daylight and the same at night, it is not easy for me to do various tests and measurements anytime I like (I am fortunate for ordering a fluke 200MHz digital scope before year 2011 because this the only tool I have now).
And on my sole desk of work, I have to also test and adjust the various assembled boards of what we produce already.
Also, my acid battery (made locally, supposed to be 100 AH) is somehow old (its internal resistance is around 50mR). And the length of the thick wires (10 mm2) from the battery to the desk had to be around 3m (6m all). This relatively long distance is not a problem at low frequencies and loads. But to supply currents pulsed at 16 Khz, they act as thin wires instead (besides adding an inductive reactance).

My real problem is that, lately, I am almost broken while I have to design new products good enough to compete similar ones in the local market. Yes, there are already pure sine wave inverters in the local market that uses iron transformers which are imported from India, China or made locally. But all of them use transformers, having 1 primary coil, instead of two coils in mine. So, I suppose that their efficiency will be comparable to mine since they also use a SPWM high frequency which is filtered solely by the power transformer to also produce a clean sinewave at 50 Hz.

There are other losses at higher frequencies, than the skin effect.

Losses due to eddy currents and hysteresis also increase, which will also increase the standby power consumption.

I'm not sure if you understood my point about harmonic power. A bi-polar 325V PWM waveform has an RMS voltage of 325V, whilst the demodulated 230V 50Hz waveform only has an RMS voltage of 230V. Putting this excess voltage though the transformer results in additional losses.

Ferrite cores are used at higher frequencies because they have lower hysteresis and eddy losses, but they also aren't as magnetic so it's impractical to use them for large, mains frequency transformers and chokes.

[coromonadalix]

so far

Lots of talk,    nothing on paper  .....

[KerimF]

so far

Lots of talk,    nothing on paper  .....

Isn't it also the trend in the great speeches around the world?

[Kim Christensen]

This may be of interest to you since they briefly discuss using a low frequency transformer:

https://www.ti.com/lit/an/sprabw0d/sprabw0d.pdf

Some key points from the article:

• For a given input voltage, the voltage stress on the transistors is double in case of the push-pull topology than
Half Bridge and Full Bridge configuration.
• The center tapped primary in the case of the push-pull converter limits the operation for a higher VA rating for
the same core size when compared to the Half Bridge and Full Bridge converter.
• To prevent flux walking in the DC-DC stage, the current in both the halves need to be sensed and the duty
cycle needs to be corrected accordingly

[KerimF]

I'm not sure if you understood my point about harmonic power. A bi-polar 325V PWM waveform has an RMS voltage of 325V, whilst the demodulated 230V 50Hz waveform only has an RMS voltage of 230V. Putting this excess voltage though the transformer results in additional losses.

I hope you agree with me that, in standby mode (no load), the core losses predominate. So, if the current, supplied by the 12 V battery in this mode, is 1 A DC, we can deduce that the power, lost by the core, is 12 W.

I am afraid that the core loss doesn't increase when the transformer is loaded. Only the coils losses increase.
So, while in square wave inverters most of the wire section is used by the current (the skin effect works at the waveform edges only, 2 times in a cycle), in the sine wave inverters the effective section is smaller. This leads us to use thicker wires in order to reduce the extra power dissipation due to the skin effect.

I can't give more detail till I will have the chance to make more measurements related to powers.

I just know now that the standby power is not too high (though it can be lowered when necessary) and the output voltage is a sinewave, clean from ripple (after adding a relatively small bi-polar capacitance). For instance, without the small output capacitor, very high pulses could be seen on the scope (this time, I protected my 1000V/10MegaOhm probe with a 100K resistor in series because a few years ago I damaged one during a similar experiment).

[KerimF]

This may be of interest to you since they briefly discuss using a low frequency transformer:

https://www.ti.com/lit/an/sprabw0d/sprabw0d.pdf

Some key points from the article:

• For a given input voltage, the voltage stress on the transistors is double in case of the push-pull topology than
Half Bridge and Full Bridge configuration.
• The center tapped primary in the case of the push-pull converter limits the operation for a higher VA rating for
the same core size when compared to the Half Bridge and Full Bridge converter.
• To prevent flux walking in the DC-DC stage, the current in both the halves need to be sensed and the duty
cycle needs to be corrected accordingly

Thank you for providing the key points because I have no more the privilege to access any site related to Texas Instruments (though they were kind to send me many of their data and application books in the 70's).

Let us see each point:
[1]
"For a given input voltage, the voltage stress on the transistors is double in case of the push-pull topology than Half Bridge and Full Bridge configuration."

This is not a problem now because...
The typical 55V DS voltage is good for 12V battery.
The typical 100V DS voltage is good for 24V battery.

[2]
"The center tapped primary in the case of the push-pull converter limits the operation for a higher VA rating for the same core size when compared to the Half Bridge and Full Bridge converter."

This is true. In fact, this is the key price of using the simple low-cost topology of push-pull. A bigger transformer is needed for the same VA.

[3]
"To prevent flux walking in the DC-DC stage, the current in both the halves need to be sensed and the duty cycle needs to be corrected accordingly."

I solved this by writing a well precise code (in assembly, as I used to do for 4 decades). And I had the chance to build two discrete MOSFET drivers (3 BJTs each) which work together in harmony very well (1us dead time) while they are driven by the two MCU pins (two anti-phase PWM).

[Xena E]

@ KerimF

I'm a little lost now, what I'd like to know is if  this power transformer design you are considering is one that is readily available, or if you have built a prototype from materials that you have available?

If you are trying to use a ready made power transformer 'in reverse' then you are likely to find that copper losses are greater than if a bespoke unit is designed.

Core losses, as in magnetising current at no load can be largely tuned out, so their impact is largely null, though a starting point of only 12 Watts core loss in a 500W (?) Iron component is quite incredible.

The form and order of the windings are also important so as to provide best coupling. You use the term leakage inductance as if it is a good thing, but in this type of project particularly it will have a negative impact on the regulation not to mention efficiency.

You say that efficiency is of no concern provided that it's not 'too' bad, a sine wave inverter based on iron cored inductors working at 50Hz may not exceed a maximum of 60-70% in reality, possibly somewhat less. The suggestion you made of supplying a load of 1kW will create a great deal of heat...

In the conjunction that you synthesised a 50 Hz sine wave from a much higher frequency it would have to be filtered with suitable low loss core materials, and high frequency techniques, not heavy iron.

What I'm also unable to understand is that an inverter that is ready made can be imported into your country for sale but not the components to make a comparable product. Cost cannot be a consideration as the value of the laminations for a transformer core, and the cost of the extra copper required to create a component of high enough inductance for low frequency working, would be far in excess of the cost of suitable ferrite parts. This is the major reason these topologies have dominated.

Please let us help you by including in your next post the details of the proposed project, it's circuit diagram, and the design of any bespoke components.

Best regards,
Xena.

[KerimF]

I'm a little lost now, what I'd like to know is if  this power transformer design you are considering is one that is readily available, or if you have built a prototype from materials that you have available?

Yes, I built a prototype using the same iron laminations which I use for the square wave inverters.

If you are trying to use a ready made power transformer 'in reverse' then you are likely to find that copper losses are greater than if a bespoke unit is designed.

I design everything in a new product (hardware and software) which is made for sale or for my personal need. But I also avoid re-inventing the wheels, mainly if the needed device is cost-effective, like the ready-made SMPS units (but, if necessary, I extend their voltage rating up to 400Vrms by adding an external circuit). 

Core losses, as in magnetising current at no load can be largely tuned out, so their impact is largely null, though a starting point of only 12 Watts core loss in a 500W (?) Iron component is quite incredible.

Me too, I wasn't expecting this. But this is what I got.
 

The form and order of the windings are also important so as to provide best coupling. You use the term leakage inductance as if it is a good thing, but in this type of project particularly it will have a negative impact on the regulation not to mention efficiency.

Today, I tested a well-coupled transformer made for 1000W/12V square wave inverter. Even its tiny leakage inductance was enough to filter the 16Khz PWM with 1uF at the transformer output.
Let us calculate approximately the minimum value of such inductance (Lmin) to do the job.
In an LC low pass filter, if w*L > 20/(w*C), the attenuation (Vout/Vin) is less than 0.13%.
Lmin > 20/(w^2*C)
where
w = 2*pi*F = 2*pi*15625 ~= 98175 rad/s
w^2 ~= 9,638 rad/us
Therefore
Lmin > 20/(w^2*C)
Lmin > 20/(9,638*1uF)
Lmin > 2mH

The leakage inductance seen at the secondary (L_lk, at the high voltage side) could be approximated (assuming Ls_lk ~= Lp_lk*N^2):
L_lk = 2 * Ls*(1-K^2)
Ls = Ls_lk/(1-K^2)/2
where
K is the coupling coefficient
Ls is the secondary inductance measured while the primary is open,

If K=0.9999 we get:
Ls = 0.002/(1-0.9999^2)/2 ~= 5 H
The value 5H is comparable to actual transformer's inductance, seen at its high voltage side.
 

You say that efficiency is of no concern provided that it's not 'too' bad, a sine wave inverter based on iron cored inductors working at 50Hz may not exceed a maximum of 60-70% in reality, possibly somewhat less. The suggestion you made of supplying a load of 1kW will create a great deal of heat...

You may be right.
But, on the one hand, this depends on how much copper will be used.
On the second hand, producing 1000 W or 2000 W inverters doesn't mean that the user will discharge his battery by a permanent 1000W or 2000W load. Actually, the user's average load will be around 350W/700W in this case. The 1000 W or the 2000 W inverter is made to supply the inrush current of certain appliances (like a refrigerator and a water pump).

In the conjunction that you synthesised a 50 Hz sine wave from a much higher frequency it would have to be filtered with suitable low loss core materials, and high frequency techniques, not heavy iron.

If it happens that the same transformer does the two required tasks, increasing the voltage and filtering the 50Hz carrier, why do we need looking for something else? 

What I'm also unable to understand is that an inverter that is ready made can be imported into your country for sale but not the components to make a comparable product. Cost cannot be a consideration as the value of the laminations for a transformer core, and the cost of the extra copper required to create a component of high enough inductance for low frequency working, would be far in excess of the cost of suitable ferrite parts. This is the major reason these topologies have dominated.

You are right to be confused from where you live.
After year 2011, the local high class only are allowed legitimately to import products from abroad (and behind the scenes, those who are related to, you know, also don't have to pay the taxes which are imposed on the ordinary citizens in case they are allowed to trade).
On the other side, the prices of the imported raw materials (from the small electronic elements to copper and iron) became relatively very high lately due to political sanctions which have been increased exponentially since year 1980.
In brief, after the world saved the people among whom I was born and live, there is no more a local medium class, only high and low classes only (after all, this is the case of all peoples saved by the world since year 2011).   

Please let us help you by including in your next post the details of the proposed project, it's circuit diagram, and the design of any bespoke components.

Now, most of my work on this project is on draft papers (I mean not finalized). When I will draw its complete circuit/schematic on Kicad (a very old version, since I have also no privilege to download the newer releases) I will upload its Kicad files and its image for those who can't open them.

Kerim

[coromonadalix]

ahhhhhhhhhhh   still talk talk talk,  nothing more               pfffff  ill leave this thread

After 2 pages of  threads, nothing good will be added, it's time to leave

nothing practical, purely theorical  etc ...  like XenaE wrote

[KerimF]

ahhhhhhhhhhh   still talk talk talk,  nothing more               pfffff  ill leave this thread
After 2 pages of  threads, nothing good will be added, it's time to leave
nothing practical, purely theorical  etc ...  like XenaE wrote

Sorry that you didn’t find around here the answers to the questions in your mind.
On my side, I asked, and I had the chance to hear many interesting answers and hints.

After all, not in vain, it was said 'Ask, and it shall be given you'.
But, in real life, 'asking' is much harder for most people to do than 'giving'. I even met and knew some people who find very hard for them to ask and to give as well. But there are also some others who don’t mind asking whenever in real need and are pleased to answer/give whenever they are asked and they have the means to do it.

[Zero999]

I'm not sure if you understood my point about harmonic power. A bi-polar 325V PWM waveform has an RMS voltage of 325V, whilst the demodulated 230V 50Hz waveform only has an RMS voltage of 230V. Putting this excess voltage though the transformer results in additional losses.

I hope you agree with me that, in standby mode (no load), the core losses predominate. So, if the current, supplied by the 12 V battery in this mode, is 1 A DC, we can deduce that the power, lost by the core, is 12 W.

I am afraid that the core loss doesn't increase when the transformer is loaded. Only the coils losses increase.
So, while in square wave inverters most of the wire section is used by the current (the skin effect works at the waveform edges only, 2 times in a cycle), in the sine wave inverters the effective section is smaller. This leads us to use thicker wires in order to reduce the extra power dissipation due to the skin effect.

I can't give more detail till I will have the chance to make more measurements related to powers.

I just know now that the standby power is not too high (though it can be lowered when necessary) and the output voltage is a sinewave, clean from ripple (after adding a relatively small bi-polar capacitance). For instance, without the small output capacitor, very high pulses could be seen on the scope (this time, I protected my 1000V/10MegaOhm probe with a 100K resistor in series because a few years ago I damaged one during a similar experiment).

It's true core losses don't increase when loaded. As you've said, the copper losses will be higher and much higher, given the high frequency content. It's still a silly idea to pass the high frequency content though the transformer. You need to use a much lower frequency. Note how motor inverters use much lower frequencies than this.

I don't see the problem with getting hold of ferrite cores. You must be able to get hold of some old computer power supplies. I wouldn't worry about reusing ferrite cores, just don't do the same with capacitors.

[KerimF]

It's true core losses don't increase when loaded. As you've said, the copper losses will be higher and much higher, given the high frequency content. It's still a silly idea to pass the high frequency content though the transformer. You need to use a much lower frequency. Note how motor inverters use much lower frequencies than this.

I guess you agree with me that this extra heat will exist no matter if the transformer is driven by an H-bridge or push-pull.
So, I wonder how this type of inverters (using H-bridge) were sold locally since many years without hearing any complain about it (from users in the least).
But I still need to do my own studies and tests to have numerical results of the various dissipated powers (even approximately, since I don't have advanced tools and I have one hour daily to work on this, due to lack of electricity). And since all powers, useful and losses, are proportional to the transformer power, my next tests will be on a transformer of 100W only. I will let you know what I will get; for example, the plot of efficiency versus output power.

I don't see the problem with getting hold of ferrite cores. You must be able to get hold of some old computer power supplies. I wouldn't worry about reusing ferrite cores, just don't do the same with capacitors.

Yes, I have a few ones. But their number is good for hobby projects, not for production.
Lately, I had the idea to try asking a local retailer (there were tens of retailers before year 2011, now there are just a few ones only) if it is possible for him to order for me a certain quantity of cores/bobbins with a reasonable price. Perhaps this will work.
Meanwhile it is better that I keep working on the project I started with.

[KerimF]

So far. we (or I in the least) know from experiment that when the iron transformer is open (no load) a small capacitor (as 1uF) at its output shorts the PWM high frequencies (16Khz) while leaving the average of its pulses (50Hz).

In other words, we have here a low pass filter. The problem is that we can think of 3 forms of LPF. They are RC, LC and RLC filters. We have:
L_lk = the total leakage inductance seen at the secondary side.
R_wire = the total effective resistance seen at the secondary side.
Although R_wire is usually small at 50Hz (R_lo), it could become high at 16 Khz (R_hi) at which the effective depth of the wire is around 0.5mm (for copper).

The impedance of 1 uF at 16 Khz is:
Zc = 1/wC ~= 10 Ohms

And we found earlier that we need only 2mH (Z_lk = 20*Zc = 2000 Ohms) as a leakage inductance to get a clean since wave with 1uF.
The question is now:
Is the value of R_hi comparable to reactance Z_lk? Or it is much lower or higher than Z_lk?

In case L_lk is the dominant element in this LPF, it is better to use many thin wires of copper (D=0.5 mm for example) in parallel for the coil's windings. This will decrease both the voltage drop and the coils dissipation when the transformer is loaded. But there is here a trade-off because by using enameled thin wires instead of thicker ones the effective area of the core window will also decrease.

To be continued... if you don't mind.