non-sine inverters

[Opcom]

I have three inverters that may be used. none are sine wave. I would like to discuss using an LC resonant circuit to make the waves closer to a sine wave, or in any case, less harmonic, because there are servo motors and they tend to 'sing' with square waves. I am sure others have heard this in Autotune rigs where 400Hz was used. A concern is the abrupt waveforms is hard on the motors and vibrates them, and could wear the bearings faster.

The converters are transformer coupled amps driven by oscillators. It is done with push pull class B and or class C amplifier operation (not linear amplifiers but those with 180 degrees or less conduction)

So, three 60Hz units.

One makes a trapezoidal wave. Like a steep walled triangle but the peak is clipped off and wide. It is class B. It's the least efficient but smoothest.


The other makes a square wave. It is class B.

The last makes the modified sine, which has dead space between the positive and negative pulses. That is more like class C. This one is not transformer coupled but has a step-up high frequency inverter to make 150V and a bridge to switch this to make AC.

If a parallel tuned LC circuit is put on either unit, perhaps including the transformer inductance as part of it, Would this work, the energy circulate in the tank, and not cause problems with the output transistors?

If they are 1KW units, the LC circuit must store 1KW for 1/60 second and be able to constantly transfer at least half that much in the case of the square wave inverter. Not common parts but that is not a concern right this moment.

I should do an LTSpice of this, but first ask what others have done to improve inverter waveforms and not waste power doing it.

[DMOD]

I would at first try a Corcom filter:

http://www.te.com/catalog/menu/en/27685?BML=L2VuL2JyYW5kcy9jb3Jjb20vc29sdXRpb25zLmV4dGVybmFsSGVhZGVyLmh0bWw

Phil - AC0OB

[Opcom]

I will take a look. Thank you very much!

In the meantime I went offline and played with LTSpice and found:
1.) series LC filter had little effect on the waveform
2.) parallel LC tank had little effect on the waveform

3.) Pi-network did a great job making a sine wave but wow look at the voltage multiplication. It holds whether the load was 10A or 1A. The Pi net must be upping the volts. I do not quite understand since Cin and Cout are the same, and it is resonant, why this would be? (I think it is resonant anyway..)
Vin = +/-120V square wave from inverter.
Vout = 370V peak volts on load (261V RMS)

[Opcom]

I would at first try a Corcom filter:

http://www.te.com/catalog/menu/en/27685?BML=L2VuL2JyYW5kcy9jb3Jjb20vc29sdXRpb25zLmV4dGVybmFsSGVhZGVyLmh0bWw

Phil - AC0OB

If these are the EMI/EMC filters, they would seem to reduce harmonics and noise but in LTSpice they do not seem to ease the square wave very well. The values were not given but I tried some based on some notes for design of those common mode and similar types. They seem to produce transients at the rise and fall of the square waves. They are not really made for square wave use.

On the other hand - circulating current is prohibitive in my circuit, either the pi or the LC one here. There should be a middle ground. Pity about a real nice sinewave though.

[Opcom]

Looks like no easy way out on this one.

Hams are obliged to be frugal, if not downright cheap.. So not wanting to pay for a high powered sine wave inverter of 400-1000W is a fine ham radio goal. 8-)

Some improvement may be possible with a T network.


It is still possible to build a small analog sine wave inverter if needed and I would not want to discourage anyone.

An example of a home made one where efficiency is not such a big deal would be a 12V-to 115V, 400Hz, 150VA one for a Collins 618S-1 servo, which I've already done in the past using a 6.3VCT 20A transformer and a couple of TO-3 transistors for outputs. There was no trouble with the 60Hz transformer in that circuit but some are not efficient at high frequencies.

[VE3ELQ]

.
3.) Pi-network did a great job making a sine wave but wow look at the voltage multiplication. It holds whether the load was 10A or 1A. The Pi net must be upping the volts. I do not quite understand since Cin and Cout are the same, and it is resonant, why this would be? (I think it is resonant anyway..)
Vin = +/-120V square wave from inverter.
Vout = 370V peak volts on load (261V RMS)

That would take a lot of iron for an L or PI at 60hz.  Just thinking out load here what if it were 200 or 300 khz. Easily filtered with small inductance and capacitance very similar to what we do in Pulse Width Modulators. Build a PWM with a H bridge output, feed it with a  sign wave oscillator 60 hz or 400 hz or whatever you want at 100% modulation, and drive an isolation transformer or use it direct if the bridge runs at about 160V. A half bridge is simpler but will require a BIG output capacitor. Want more power, add more FETs, the 10A 400V 11n40s are about $1.25 if you buy them 10 at a time, great little FETs. Even crappy FETs will switch 300 Khz.
73s  Nigel

[DMOD]

I will take a look. Thank you very much!

In the meantime I went offline and played with LTSpice and found:
1.) series LC filter had little effect on the waveform
2.) parallel LC tank had little effect on the waveform

3.) Pi-network did a great job making a sine wave but wow look at the voltage multiplication. It holds whether the load was 10A or 1A. The Pi net must be upping the volts. I do not quite understand since Cin and Cout are the same, and it is resonant, why this would be? (I think it is resonant anyway..)
Vin = +/-120V square wave from inverter.
Vout = 370V peak volts on load (261V RMS)

Have you tried a simple LC low-pass filter with a cut-off of say 600Hz?

[W2PFY]

Since I don't know what the heck I'm talking about and looking for answers, I have also wondered about the same condition except somewhat in reverse. Years ago I went up to Toronto to a large stand by generator lecture given by a manufacturer that escapes me now. One of the subjects was the effect of non linear loading on a generator. These were large generators that would be in ships or some ground application both single and three phase. The problem with non linear loading such as motor drives is that they put spikes on the AC line as they chop up the portion of the sine wave that is used for a given power requirement. This would cause very dangerous heating in the generators windings. Its not a problem when power is supplied from the mains because the system is VERY large. The solution was quite simple. You would run single or three phase to two identical transformers, for example you would feed 240 in and step it up to 480 and then on the second transformer, step it back down to 240 volts again. This method would totally shield the generator from the harmonics cause by the non linear loads. They also pointed out that the neutral lines needed to be the same gauge as the other wires in the circuit.

Well anyhow, I often thought about how such an arrangement would wave shape a modified sine wave inverter. I think by playing with some AC capacitors on the high voltage side of the step up transformer would perhaps shave some of the rough edges off the modified sine wave and make the load happy? Any thoughts?  

[KO6YB]

I'll join the since I don't know bus. I thought about the type of filter as used after DBM "mixers", the diplexer. One side to pass the 60 hertz, and the other to pass all above AND below 60hz.

Stan.

[WBear2GCR]

Shooting out of the dairy aire here...

Inductor input filter ought not boost the output voltage, I'd guess.
Could be all wet on that...

The trick with the two xfmrs, is puzzling to me. Entirely unclear to me why the step up/step down is helpful. The idea of putting a ton of inductance in the line does make sense. I'd guess that the step up/step down ends up with less losses than a 1:1 arrangement??

In addition the mere application of a transformer at the output will make the output rather sinusoidal - in fact the "worse" the xfrmr in terms of bandwidth, the better the job. There was some mention of toroidal power iron working as backwards connected mod iron when driven with a solid state power amp, meaning the bandwidth is good.

You could "tune" the secondary of the xfmr or a choke (in a filter) with a parallel cap, as is oft done in some HV supplies.

Are you using these fixed, or mobile? If fixed, then size is not terribly important?

                            _-_-

[Opcom]

Since I don't know what the heck I'm talking about and looking for answers, I have also wondered about the same condition except somewhat in reverse. Years ago I went up to Toronto to a large stand by generator lecture given by a manufacturer that escapes me now. One of the subjects was the effect of non linear loading on a generator.

I guess the transformers ate the spikes?

some test waveforms using rotten loads with a new generator:
http://www.bunkerofdoom.com/mil/m35/gen/test1/index.html

It was the adjustable RPM being a bit low that caused my complaint, not anything wrong with the generator or its ability to supply power to nasty loads. It was educational to do the experiments.

For VFD to 200KW where I contracted a bit last year we used purpose-made inductors to isolate the drives from the mains. We used these a lot.
http://www.transcoil.com/Products/KLR-Line-Reactor.htm
They look like 3 PH transformers, but are just 3 coils.

[Opcom]

Have you tried a simple LC low-pass filter with a cut-off of say 600Hz?

Yes. The LC is OK but two inductors gives a much better (slower) transition with a square wave input.

[Opcom]

Well anyhow, I often thought about how such an arrangement would wave shape a modified sine wave inverter. I think by playing with some AC capacitors on the high voltage side of the step up transformer would perhaps shave some of the rough edges off the modified sine wave and make the load happy? Any thoughts?  

I found that peak charging currents/pulses in the capacitors connected right across the voltage can be very high during the rise and fall times of the inverter waveform. A cap with some series resistance helps but that also limits its effectiveness. With a transformer some of the current in those pulses would probably become I2R losses in the transformer.