EEVblog #449 - Absopulse VFC500 Variable Frequency Converter Teardown

[EEVblog]

Inside the Absopulse Variable Frequency Converter
http://www.absopulse.com/pdf/sine%20wave/fc%20-%20vfc/VFC%20500.pdf

[ProBang]

Hello.

Sometimes big factories have their own standard for mains.
I know one with 200V/100Hz.
To repair their stuff you need a converter like yours.

BTW: Diodes in parallel?
I´ve learned that´s useless. One diode sacrify itself, the other one isn´t needed.
Any variations?

 

[megahz]

It sounds like a space ship!!!!

[Len]

BTW: Diodes in parallel?
I´ve learned that´s useless. One diode sacrify itself, the other one isn´t needed.
Any variations?

To quote the video:

They'll share reasonably well. Good enough for Australia anyway.

 

[opablo]

...
BTW: Diodes in parallel?
I´ve learned that´s useless. One diode sacrify itself, the other one isn´t needed.
Any variations?

Is it possible that only one is needed in terms of current specs and the second one is there just for redundancy in case of an open circuit failure of the other one ? (to increase the MTBF ?)

[Rufus]

The output isn't a transformer driven by MOSFETs. 500W at 40Hz would need a transformer half the size of the case. I'm pretty sure it will just be an H bridge with PWM modulation producing an average sine wave of the desired frequency and amplitude followed by lots of LC filtering to get rid of the PWM.

The big Micky Mouse ear on the input isn't a varistor it is a SURGE-GUARD NTC thermistor for inrush current limiting.

Next to the jumper link connecting the input board capacitors in series is space for two more which would put them in parallel presumably for a lower voltage model with twice the output current.

The performance of IEC mains filters is significantly degraded if you don't have a solid connection between the filter case and chassis.

A quote from Schaffner :-
"One of the most common problems with grounding is the installation of filters by directly attaching them to enclosures with painted or otherwise treated surfaces, in that way interrupting the ground connection and leaving the filter almost useless."

The extension of the filter case to make the flaps around the mounting holes under the plastic bezel is intended to make that connection. If done properly and their being two of them it should be more reliable (if less obvious) than a bit of green and yellow with crimps each end. That does raise an interesting question of what the finish on the aluminium chassis and cover is. Some kind of Alochrome I would imagine. You could check it with a meter

[NiHaoMike]

One of those gadgets is better known as a "Variable Frequency Drive". Most VFDs are 3 phase output, I don't see much application for a 1 phase unit. (No, you can't effectively use one to vary the speed of a single phase motor by much, it takes a two phase unit and some rewiring of the motor to do that.) Maybe try "overclocking" a cheap desk fan with that unit (maintaining the V/Hz ratio if possible), but I think the motor would lose a lot of torque and actually start slowing down at higher frequencies as the capacitor no longer provides the correct phase shift.

BTW, the part when you turned up the frequency sounds a lot like a hybrid or electric car accelerating from a stop. Or if you add in some weird sounding squeaks and buzzes at the beginning (a sensorless drive determining the rotor position) and wind up the top end frequency about 10 times higher (think jet engine), it would sound almost exactly like Brittany Benzaia's hybrid air conditioner.

[MCRIPPPer]

that thing sounds cool. it would make an interesting piece to a song. 



that is strange that the mains input is not hard wired to the case. here int he U.S. the neutral wire is tied to ground. even "non grounded" (thins without a ground prong on the plug) devices get connected to ground through the neutral wire. im not sure about how the wiring is done in other countries. maybe the device uses the neutral to get its ground connection.

[firewalker]

Is it ok to use banana socket in the output? I would expect mains socket.

Alexander.

[Ed.Kloonk]

Is it ok to use banana socket in the output? I would expect mains socket.

Alexander.

I'd expect a banana socket since the output is so widely variable. And the load is random gear which may not even have a plug or even a lead.

With a proper socket, im(h)o, the expectation is there for the output to be fairly normal like an ordinary power outlet.

If you'd have only mains appliances to test with the thing then you would have to simply make up a socket lead.

[casper.bang]

Lovely ending with the sound demo. I'm curious though, what physical cause is there for why switching regulators emit this and surely it's not the direct switching frequency (usually betweeen 100KHz and 1Mhz?)?

[peter.mitchell]

Lovely ending with the sound demo. I'm curious though, what physical cause is there for why switching regulators emit this and surely it's not the direct switching frequency (usually betweeen 100KHz and 1Mhz?)?

Generally, it isn't the regulators, but the inductors or transformers that make the noise, the magnetic forces cause them to ever so slightly vibrate.


An interesting use of such an effect is in hobby grade brushless motor controllers, where the controller uses a winding as a speaker for notification sounds.

[Phoenix]

Let's see if I can work this out from the video...

The output DC-AC stage looks pretty easy to understand. The 4 MOSFETs are in a H-Bridge configuration, with an inductor (labelled MSI 2000) on each phase outputs (not transformers). I would assume this is followed by the AC capacitor in parallel then the common mode choke before the take off point (can't quite make out all the tracks) forming an LCL type filter.

Those extra diodes are a little odd from what I can see, but they are clearly in some way anti-parallel diodes for the MOSFETs (as MOSFETs body diodes are slow). What is odd to me is that the to-220 package diodes are 45V 16A schottky diodes, but then appear to be seriesed with the axial diodes (where 2 of those are in parallel, can't make out the model number). It would be fairly easy to trace out with the board. Can you get the model of the axial diode?

If I'm not mistaken the current transformer is located on the DC side, but the current will be pulsed anyway. I guess some low pass filtering is done at a later stage in the measurement.

As for the input stage, it's a bit hard there is a lot hiding away. The label says 2x200V 1.5A, so I guess it has a 400V DC bus (which is about right for a 1phase 264V output via PWM). SG100 is surely an NTC themistor. The unit won't have regeneration capabilities as it's got dumb rectifiers in it.

I would say there is the input EMI filter stuff then a bridge rectifier (single package) on the first heatsink plate. That means the first magnetic component would have to be an inductor on one of the DC rails - for the boost PFC stage no doubt. Why the PFC needs 2 MOSFETs I don't know, but the third component on that heatsink plate is a diode (has 2 pads).

I'll take a guess at the second stage and say that the primary side of the transformer is center tapped and the two MOSFETs are in push-pull configuration. There are clearly 2 output windings (eventually seriesed) that are rectified to produce 200VDC each. What that third magnetic component is for I am not sure, as the core is shared between windings for the positive rail of both outputs, possibly just used for inductive filtering.

Cheers
Stewart

[nitro2k01]

Lovely ending with the sound demo. I'm curious though, what physical cause is there for why switching regulators emit this and surely it's not the direct switching frequency (usually betweeen 100KHz and 1Mhz?)?

Here's a spectrum plot of the recorded audio. The "weird noise" is quite obviously aliasing effects. Here's my interpretation of what I'm seeing:
There seems to be two things at play here:
1) The PWM frequency, which seems to be 10 kHz. If you put in a frequency above that limit, you will get aliasing (reflections). The 10 kHz signal itself is the band around the 10 kHz vertical line.
2) The sample rate of the sine wave. The sine wave reaches from 40-440 Hz, and it's produced from a table of sine values, which are churned out at a higher frequency, which will also be superimposed. In this case, this frequency seems to be 8 kHz for the starting 40 Hz sine wave, meaning the sine table is 200 samples big. So, the PWM breakpoint is updated 8000 times per second to begin with, and proportionally more often as the frequency is increased. The strange noises are (repeated) spectral reflections of the sine sample frequency for the sine generator, around the 10 kHz mark  See second image. Green = the actual frequency component. Blue = aliasing.
I expect the output from the terminals to filtered and much cleaner. Dave, care to scope it/spec it?

[c4757p]

here int he U.S. the neutral wire is tied to ground. even "non grounded" (thins without a ground prong on the plug) devices get connected to ground through the neutral wire.

No. No, no, no, no, no. No. Please never wire any mains equipment. 

[Fezder]

nice video, again new equipment to see

[FLL-Freak]

I was a little disapointed not to see wha the output waveform looked like under various loads. The whole introduction was on how to make a good sin wave output and we did not get to see it. For all we know it could have been a "crustry" output.

[geekysuavo]

What could be the reason for breaking out the different functions into physically separate PIC chips? Did they happen to run at odd- or non-integer multiple system clock rates?

~ Brad.

[N2IXK]

here int he U.S. the neutral wire is tied to ground. even "non grounded" (thins without a ground prong on the plug) devices get connected to ground through the neutral wire.

No. No, no, no, no, no. No. Please never wire any mains equipment. 

The original poster is absolutely correct. The neutral (white wire) is bonded to earth ground (green or bare wire) at each building's service entrance. The secondaries of transformers and other "separately derived systems" are treated in the same way. An ungrounded (floating) circuit is only allowed in specialized applications (like the isolation transformer used on an electronics test bench). Neutral and earth are ONLY supposed to bonded at the one point, and are kept separate throughout the system, with the neutral used as the current return path, and the earth only used as a safety ground and EMI/RFI/Static drain.

The grounded neutral is NOT supposed to be connected to equipment frame ground, even on devices with a 2-wire mains plug.  At one time, the frame ground wire was permissible for use as a neutral return on some types of equipment (clothes dryers and electric ranges operating at 240V, but needing 120V for timers or indicator lights), but this is no longer permitted by current codes. Such equipment is now required to be fed with a 4-wire circuit, with 2 hot phases, a neutral, and an earth ground.

[c4757p]

Neutral is connected to ground at one point, but treating it like a ground is reckless. It's just as "live" as phase. Saying ungrounded devices get connected to ground is technically correct but practically unsafe. Ungrounded devices are ungrounded.

[emcarro]

It woulg be very interesting seing that "sine" wave. But its very dificoult for an oscilloscope that kind of waves. 
The carrier frecuency would not be too high, too much noise.

[staxquad]

we were still using 25Hz hydro electrical power in some parts of Ontario up until last year 

I guess that Absopulse wasn't made for Northern Ontario.

Three generating stations on the Upper Mattagami River – Wawaitin, Sandy Falls and Lower Sturgeon, built 100 years ago, were still generating electricity at 25Hz for the mines up until 2011.

Next month, the last four 25 Hertz (Hz) generators at Wawaitin GS will cease production. That will be the end of the 25 Hz era in Northeastern Ontario.
The electricity we use in our homes is produced as alternating current, which has a frequency of 60 cycles
per second, also known as 60 Hz. In the Timmins area in the early 1900s, 25 Hz was adopted as the
frequency standard for operating pumps and motors for local mines. The Upper Mattagami River plants
had 25 Hz changed to 60 Hz through a frequency  converter, transformer and switching equipment
located in Sudbury.

They didn't mention the 25Hz lights, besides the pumps and motors.

Sandy Falls 25Hz


Wawaitin 25Hz
This was the last operating 25Hz station in Canada.


Ontario Power is a 25Hz plant built at the base of Horseshoe Falls (Niagara) by an American company.
decommissioned in 1999

[k8tek]

volt-amperes does not equate to watts. 

[staxquad]

volt-amperes does not equate to watts.

in DC it does

in AC, there's impedance, one lags the other, so can't be multiplied

[paulvos]

volt-amperes does not equate to watts.

in DC it does

in AC, there's impedance, one lags the other, so can't be multiplied

Unless you know the power factor ( cos phi ).

[RobUSVI]

Neutral is connected to ground at one point, but treating it like a ground is reckless. It's just as "live" as phase. Saying ungrounded devices get connected to ground is technically correct but practically unsafe. Ungrounded devices are ungrounded.

Reckless and shocking when a careless electrician has reversed the white and black wire (neutral and hot) and you touch the "grounded" chassis.  (Where I live, anybody that can hold a screwdriver is an electrician... Sadly, I see hot neutrals all the time when troubleshooting.)

[staxquad]

Neutral is connected to ground at one point, but treating it like a ground is reckless. It's just as "live" as phase. Saying ungrounded devices get connected to ground is technically correct but practically unsafe. Ungrounded devices are ungrounded.

Reckless and shocking when a careless electrician has reversed the white and black wire (neutral and hot) and you touch the "grounded" chassis.  (Where I live, anybody that can hold a screwdriver is an electrician... Sadly, I see hot neutrals all the time when troubleshooting.)

which is why I plug in one of these anytime I work on the electricity in the house
identifies open ground, open neutral, open hot, hot&ground reversed, hot&neutral reversed and correct wiring
the button is to trip GFCI to test that it's functioning

[Rufus]

Neutral is connected to ground at one point, but treating it like a ground is reckless. It's just as "live" as phase. Saying ungrounded devices get connected to ground is technically correct but practically unsafe. Ungrounded devices are ungrounded.

Reckless and shocking when a careless electrician has reversed the white and black wire (neutral and hot) and you touch the "grounded" chassis

A device with neutral connected to the chassis becomes live if there is an open circuit fault in the neutral anywhere between the device and neutral/ground bonding point. It doesn't need an incompetent electrician to be dangerous.

[elgonzo]

Lovely ending with the sound demo. I'm curious though, what physical cause is there for why switching regulators emit this and surely it's not the direct switching frequency (usually betweeen 100KHz and 1Mhz?)?

Generally, it isn't the regulators, but the inductors or transformers that make the noise, the magnetic forces cause them to ever so slightly vibrate.


An interesting use of such an effect is in hobby grade brushless motor controllers, where the controller uses a winding as a speaker for notification sounds.

Another, quite famous example of such "coil whine" is the Siemens Taurus electric locomotive, also known as the musical loco.
Similar to what Dave demonstrated at the end of his video, you can hear the motor converters doing their work during the acceleration of the loco:

To achieve optimal traction on slippery rails, the power output of each motor on each axle is regulated individually to avoid wheel spin. With the right (bad) weather conditions and with a sufficiently heavy train the Taurus actually sounds like music:

[Paul Moir]

That's really neat elgonzo.  Thanks!

[Cognito]

we were still using 25Hz hydro electrical power in some parts of Ontario up until last year 

Three generating stations on the Upper Mattagami River – Wawaitin, Sandy Falls and Lower Sturgeon, built 100 years ago, were still generating electricity at 25Hz for the mines up until 2011.

What is the reason to use 25 Hz instead of 60 Hz? Could it generate a good speed for the pump in the mines?

Nice pictures =)

[elgonzo]

we were still using 25Hz hydro electrical power in some parts of Ontario up until last year 

Three generating stations on the Upper Mattagami River – Wawaitin, Sandy Falls and Lower Sturgeon, built 100 years ago, were still generating electricity at 25Hz for the mines up until 2011.

What is the reason to use 25 Hz instead of 60 Hz? Could it generate a good speed for the pump in the mines?

Nice pictures =)

Wikipedia says: https://en.wikipedia.org/wiki/Utility_frequency#25_Hz_origins

[Alana]

Another, quite famous example of such "coil whine" is the Siemens Taurus electric locomotive, also known as the musical loco.
Similar to what Dave demonstrated at the end of his video, you can hear the motor converters doing their work during the acceleration of the loco:

To achieve optimal traction on slippery rails, the power output of each motor on each axle is regulated individually to avoid wheel spin. With the right (bad) weather conditions and with a sufficiently heavy train the Taurus actually sounds like music:

I think they re-program those controllers for real music in some passenger locos during EURO 2012 football championships.

[KeithBrown]

Did any (other) sharp-eyed Canuks notice that there are actual components in the unit that are themselves made in Canada? At 28:38 we some resistors from http://www.renfrewelectric.com/index.html! Anybody spot anything else?

[elgonzo]

Another, quite famous example of such "coil whine" is the Siemens Taurus electric locomotive, also known as the musical loco.
Similar to what Dave demonstrated at the end of his video, you can hear the motor converters doing their work during the acceleration of the loco:

To achieve optimal traction on slippery rails, the power output of each motor on each axle is regulated individually to avoid wheel spin. With the right (bad) weather conditions and with a sufficiently heavy train the Taurus actually sounds like music:

I think they re-program those controllers for real music in some passenger locos during EURO 2012 football championships.

Well, there is a video on YT that presumably shows a German ICE 3 express train that plays the german national anthem while being in service mode.
If this video is indeed true, you can be certain that this is in no case part of the operating software of the train, but controlled through an external interface. Mind you, these trains do service speeds of 300 km/h with hundreds of passengers. In case such easter eggs would be discovered in the trains operating system, they would either be put out of service for safety-related investigations, or at least their allowed operational speed would be restricted severely and some serious complaints directed to the manufacturer...

[G7PSK]

25 HZ was also used by some train company's here in the uk as they said it was better for traction purposes, I have an old reed frequency meter from a power house of such a railway.
When Ferranti first started to push ac transmission he had a 96rpm 33 pole pair's alternator giving 100 amps at 1100 volts at 26.4 hz. He also had a co-ax power cable that he demonstrated the safty of by having his foreman knock a cold chisel through while live after the trips went he would simply replace the section of conductor and reset the trips. Ferranti was also a consultant on the first Niagara falls power plant in Canada.

[SeanB]

New ringtone done............

[staxquad]

we were still using 25Hz hydro electrical power in some parts of Ontario up until last year 

Three generating stations on the Upper Mattagami River – Wawaitin, Sandy Falls and Lower Sturgeon, built 100 years ago, were still generating electricity at 25Hz for the mines up until 2011.

What is the reason to use 25 Hz instead of 60 Hz? Could it generate a good speed for the pump in the mines?

Nice pictures =)

Wikipedia says: https://en.wikipedia.org/wiki/Utility_frequency#25_Hz_origins

Interesting read.

Japan is as f***ed up in frequencies as Australia is in rail gauge

In Japan, the western part of the country (Kyoto and west) uses 60 Hz and the eastern part (Tokyo and east) uses 50 Hz. This originates in the first purchases of generators from AEG in 1895, installed for Tokyo, and General Electric in 1896, installed in Osaka. The boundary between the two regions contains four back-to-back HVDC substations which convert the frequency; these are Shin Shinano, Sakuma Dam, Minami-Fukumitsu, and the Higashi-Shimizu Frequency Converter.

no use for the Absopulse whatsoever:

The 15 kV AC rail networks, used in Germany, Austria, Switzerland, Sweden and Norway, still operate at 16? Hz or 16.7 Hz

[Hypernova]

no use for the Absopulse whatsoever:

The 15 kV AC rail networks, used in Germany, Austria, Switzerland, Sweden and Norway, still operate at 16? Hz or 16.7 Hz

It's not likely that the designers at Absopulse had a locomotive in mind when they designed it.

[Ed.Kloonk]

no use for the Absopulse whatsoever:

The 15 kV AC rail networks, used in Germany, Austria, Switzerland, Sweden and Norway, still operate at 16? Hz or 16.7 Hz

It's not likely that the designers at Absopulse had a locomotive in mind when they designed it.

But if you were building a miniature model city like that one in Germany, you'll need to power the little miniature trains that should run on 16.7Hz if it's to be truly authentic.

[ConKbot]

Interesting that its AC->DC->DC->AC, instead of just AC->DC->AC,   I would have expected that The PFC rectifies and boosts to a ~380v ish voltage, feeds it to the h-bridge to drive your isolation transformer with high frequency, then your LC filters on the output. 


Anyone have any ideas on why the intermediate DC/DC step? 2250vdc isolation is doable with a double isolated transformer (special rated primary wire and tape etc...)   and then it would need a much smaller supply for internal functions and thats it.  2 separate transformers for hi-rel purposes? off the shelf transformers because it turns out they arent as custom wound as I would expect?

[tom66]

Possibly it drops the rail voltage for lower output voltages to improve efficiency (the switching transistors operate at a higher duty cycle.)

[Phoenix]

Interesting that its AC->DC->DC->AC
...
2 separate transformers for hi-rel purposes?

I think you missed an AC stage 

Rectifer+Boost PFC           ->           Push-Pull??+Rectifier                                       ->                                           Inverter
        AC-DC                                         DC-AC-DC                                                                                                     DC-AC
   1 Inductor                              1 Transformer, 1 Unknown Coupled Inductor                 2x Inductors, 1x Coupled Inductor (Common Mode Choke)

Anyone have any ideas on why the intermediate DC/DC step?

I'll take a stab - High frequency isolation requires a smaller transformer. If you were to put a transformer on the output of the H-bridge it would have to be big enough to operate at the 47Hz fundemental, but have little capacitive coupling to ensure blocking of the high frequency switching. You would also have to contend with the (less predictable?) leakage inductance, magnetising current and any transformer/magnetic nonlinearitries especially at low voltages - not sure exactly what effect that would have on the output quality, but it won't be good!

Possibly it drops the rail voltage for lower output voltages to improve efficiency (the switching transistors operate at a higher duty cycle.)

Don't think the efficiency benefit would be worth while (it's not very dependent on duty cycle - current will either flow in the FETs or antiparallel diodes), but the power quality benefit of runing at a higher duty cycle might be worthwhile. At a higher duty cycle there is greater ratio of fundemental voltage to switching harmonics. You may also drop pulses at very low duty cycle (when the FETs don't get a chance to fully switch on).

Could put a mutlimeter on to see if it's a constant 400VDC or variable? I would suggest it's a constant 400V as the PFC inductor and an output capacitor have 2x200V/1.5A each written on the side (22:19 and 26:33 in the video).

Perhaps one more reason -  the DC board looks like it's designed for mutliple uses as there are pads on the DC output that could be connected to parallel the two 200V transfomer outputs (27:04). As well as some pads to link to ground reference one of the rails.


The part that is confusing me is the purpose and arrangement of the magnetic component just before the DC output, after the high frequency isolation transformer, that has the flying torroidal component attached to it (28:32). It lookes like both the positive output rails of the transformer go through it, but share a common core (looking at the tracks)... but then some fly leads go off somewhere from the top...?

[ilanko]

Someone forget the DaveCade ... And what's PWD ?

[ilanko]

[ilanko]

Hi Dave, I hope to see a video testing common house electronic functioning under extreme frequency or low voltage.

[Yansi]

Hi
I've just seen this nice video. I was quite interested to see, what topology have they used. So I carefuly analyzed what I've seen in the video.

Mains input is rectified with a bridge, decoupled by the first 4u7 MKP cap gunked onto the first transformer - whcih is a choke actualy - from the PFC controller. Behind the choke, we can clearly see a TO247 mosfet and a diode. Then there is the filterbank.

The DC/DC converter then takes the PFC's  output of probably around 400-420 volts and produce 2x200Vdc in series on the two output caps. The main DC/DC is a standard halfbridge topology, with two secondaries, two bridge rectifiers and double output choke.

The rectangular cutouts in the board are for airflow cooling, not isolation gaps.

Now the more interesting part, the actual convertor: They've used H-bridge topology. Each half-bridge has its own output choke. Between the other ends of the two chokes, there's the second polypropylene capacitor. After that, commonmode choke was placed. The output is immediately after the commonmode choke, near ? a varistor. Seems there are no additional filtering caps. Pretty underestimated filtering. I expected somewhat better filtering there.

The most interesting part is, that the four mosfets in the H-bridge have all blocked the internal body diodes. The four TO220 two-legged diodes Dave showed in the video, are connected in series with drain electrodes of the mosfets. Then there are 8 small plastic diodes (like DO-41 ones) antiparallel with that. God why? I thought, that outblocking the mosfets internal diode is worth only for high switching frequencies, like 100+ kHz, which I think won't happen here, considering the quite bad board layout. Why have they did it? Your guess is good as mine. Maybe they have chased each miliwatt of heat and this helped a little to increase efficiency a little.

How do you think the current transducer works there? It is obviously connected on the plus side of the supply to the H bridge. Wtf? DC biasing the transformer / measuring DC current with it? No? I quite don't understand the location of the transformer. Very strange to me.  Source shunts would do much easier.

As I mentioned, the bord layout is not as good as I think it should be. Note the big curent loop, between the H-bridge stage and the electrollytic caps. Crazy. The cap connected after the output chokes is also connected via too long leads and traces. The layout should minimize stray inductances there to gain better HF suppression on the output.

The four PIC micros on the front panel are just funny touch to whole, as is the chassis grounding connection. Even if it would be connected (where?), it wouldn't respect the starpoint earthing - which is here important for reliability. It is not good to earth something somewhere to the chassis, and then elsewhere connect the chassis to the PE lead. There are to many interconnections in between, which could possibly make trouble.

I am also concerned, that the inverter's filter design is dodgy. There should be cap in between inverter supply ground and the choke output. The cap between the chokes only does not suppress the HF switching components between inverter ground and outputs. Oh yes, there's a commonmode choke there, but it is more like "how ya doing" than a properly designed filter. If there is enough capactity between inverter's ground and chassis (PE), it is something I would really be quite concerned about when testing the thing on EM compatibility. For more clearity, I redrawn the schematic of the inverter part of the circuit.

Here are some reverse-engineered pictures of the thing and its topology. Interesting to note, that one of the output post is labeled as "N" (neutral) on the front panel. Douesn't make any sense, if the whole inverter section is possibly floating, so both outputs symmetrical.

Edit: the last shot explains the labeling "PH N GND" quite well. grrrrrrrh.  They've just shorted one side of the floating output to PE with some crappy foil cap.

[NiHaoMike]

There are also asymmetrical inverter designs where one side is switched at a high frequency and the other side at the output frequency. That allows lower conduction loss (but higher switching loss) transistors to be used for one side, boosting efficiency.

[Yansi]

Think about it, twice. The topology you described is useless, because you would need the HF switching side to make sharp steps from V+ to V- side. This will introduce inherent distortion of the waweform around zero crossings and  problems with EMC, because both outputs of the inverter will jump at output frequency (50Hz or so) with square wave of 350Vpp. And you can't filter out this square wave component from the output. This asymmetrical approach is both unpractical and useless.

And I do not understand what other efficiency boost it will bring, than saving switching losses on one side. You still can use lower conduction loss transistors on both sides, regardles of the switching frequency.

The only assymetrical approach usable is to just ground one lead and have symmetrical supply of +-350Vdc. Then you will chop only one side in between these positive and negative rails. But this is not usually done, because as obvisous, you would need power switches designed for twice the voltage swing on them -> higher loss and EMI. This approach would require 1200V IGBTs and is not worthy for smaller output power, like these 500VA.

[bktemp]

Think about it, twice. The topology you described is useless, because you would need the HF switching side to make sharp steps from V+ to V- side. This will introduce inherent distortion of the waweform around zero crossings and  problems with EMC, because both outputs of the inverter will jump at output frequency (50Hz or so) with square wave of 350Vpp. And you can't filter out this square wave component from the output. This asymmetrical approach is both unpractical and useless.

You may be right with the fast edges around 50Hz when used at the output side, but this topology exist. I have seen it in some UPSs used at the low voltage side driving the primary winding of the transformer.

Another variant uses IGBTs for the low side switches and mosfets for the highside ones. Because IGBTs do not have a body diode, you can use faster external ones without the need to disable the internal ones. The IGBTs switch at ac frequency, the highside mosfets modulate the waveform. This may not be usefull for a universal ac supply, but it works well for loads like gas discharge lamps.

[Yansi]

There is no "may be right", in this one, there are only facts. The topology you described, is just evil crap, if you don't understand that enough. Any sensible engineer won't use that anywhere, because of what I have tried to explain you. With that topology, you wont save anything apart from switching loss on one side. From the circuit design point of view, it is not simplier though. It is only negligence or lack of expirience or both, or whatever evil will make you use that kind of topology. I have never seen it anywhere, because it just don't work as on the paper. There is a big difference in between the slewrate of the 50Hz side and the output of the reconstruction LC filter, there will be crap on the output around zerocrossings, when changing the polarity of the output. And then the whole output floats on that squarewave.

It is evil in every aspect, almost none advantage, except that it is probably cheaper, cause the lack of one output choke and highspeed mosfet driver. (They can drive the slow halfbridge side using a few discretes). But cost cutdowns are NOT a valid excuse for not doin' that inverter right. And I will not advise anyone to build things that aren't right. So it is nice, that such topology exists (you will for sure find much more possibilities), but be aware some of them simply aren't going to work well or good enough.

Lowvoltage side sinusoidal modulators are a whole another category, please don't mix em here or make another topic for em.

That hybrid IGBT/NMOS topology with only lowside diodes needing to be fast does not make sense. Please describe it more or check that you aren't mixing things together.

[bktemp]

That hybrid IGBT/NMOS topology with only lowside diodes needing to be fast does not make sense. Please describe it more or check that you aren't mixing things together.

Here is a 250W HID metal halide electronic ballast using this configuration:
http://www.st.com/web/en/resource/technical/document/application_note/CD00192161.pdf

[NiHaoMike]

And I do not understand what other efficiency boost it will bring, than saving switching losses on one side. You still can use lower conduction loss transistors on both sides, regardles of the switching frequency.

There's a tradeoff between conduction and switching losses. You can't optimize for both at the same time.

The inverter I actually used the design in was supplied by an internal isolated DC/DC converter, so the output neutral was just bypassed to ground with a capacitor. As it was intended for motor loads, EMI wasn't a huge concern. It's certainly much less noisy than a common modified sine inverter or a filterless VFD.