Smoothing modified sinewave Inverter

[simontay1984]

I have a Nikkai 12V to 230V 300W modified sinewave inverter. If possible, I want to modify it to output a purer sinewave (i'm not expecting perfect) similar to what the really expensive pure sinewave inverters produce.

After spending ages searching Google about how to modify an inverter like mine to smooth the output, I found this page: http://bobmay.astronomy.net/misc/drivcorr.htm

I've seen schematics for wien bridge oscillators that produce pure sinewaves with very low distortion and also sinewave generators using 555 timers.

If it's possible, i'd probably use a wien bridge oscillator design as that produces a much smoother wave compared to a 555 and connect the output to the appropriate place on the inverters PCB.

I also remember reading in a forum somewhere about whether adding a mains filter circuit (like on this page: http://www.conformity.com/artman/pub...nter_200.shtml) would improve the output.

Any advice would be appreciated.

Thanks,
Simon

[darkfenriz]

It is not possible.

Most probably your inverter is a switched-mode design and uses high frequency conversion DC7DC and bi- or tri- state output stage for AC.
Keep in mind, that the technique descibed in the link assumes using standard 50/60Hz transformer, so the whole design is very bulky and inefficient.
In contrast, expesive inverters use pulse-width-modulation output stage, similar to class-D audio amplifiers.

[Eva]

You would have to either use a huge filter or to redo the output stage of the inverter completely, and that's not easy because the new output stage has to be class D.

If you don't want to learn class D, consider buying a sinewave inverter...

[star882]

Just add some inductors in series. Close enough.

[wg_ski]

Originally posted by star882
Just add some inductors in series. Close enough.

That does not result in filtering of those "modified sine" inverters. They all have a common flaw in that they go high-impedance in the dead time between the pulses of opposite polarity. If you have inductance in the load, it stores energy during one pulse, and simply releases it back to the battery during the dead time until its field collapses - sticking to the rail! Under light inductive load the waveform ends up being even "squarer" than what you started with, not more sinusoidal.

[star882]

Originally posted by wg_ski



That does not result in filtering of those "modified sine" inverters. They all have a common flaw in that they go high-impedance in the dead time between the pulses of opposite polarity. If you have inductance in the load, it stores energy during one pulse, and simply releases it back to the battery during the dead time until its field collapses - sticking to the rail! Under light inductive load the waveform ends up being even "squarer" than what you started with, not more sinusoidal.

When the H bridge is in high impedance, the catch diodes (the ones integrated into the MOSFETs since at 60Hz, recovery time is not very important) would cause the inductor to recharge the high rail cap while powering the load. As such, the stored energy will dissipate very quickly.

If you think in the frequency domain, the inductor will be a higher impedance with higher frequencies. It will (in theory) block the most troublesome high harmonics the most and the desired fundamental the least. If it makes the wave "squarer", it would imply that it would increase the harmonics, which makes no sense.

If you need it to be closer to a sine wave, there's always the option of using Delta Sigma modulators just like in TI hybrid audio amplifiers. Have two (or three for 3 phase) modulators running from a common clock (a few kHz for motor control, above 20kHz for audio), with the two receiving sine waves of the same frequency but with one 180 degrees out of phase (or 3 with 120 degrees and 240 degrees for 3 phase).

[megajocke]

I'm pretty sure the modern ones with H-bridges do not go high impedance but shorts the output together during dead time, highly inductive loads like fluorescent lights wouldn't work correctly otherwise.

The simple primitive ones using a mains frequency transformer and push-pull primary do go high impedance of course.

[EWorkshop1708]

You folks mention class D, since this inverter in the topic is PWM. Now most Class D PWM amps use an output inductor and some capacitors to filter anything above the audio range, to make sinewaves for your speakers.

With a similar concept to class D, but for the AC inverter, what about using those High Frequency EMF filter input inductors used on the AC side of common SMPS like TV's and computer power supplies?

They are meant to pass 60Hz, while filtering out the HF interference, produced by the SMPS. Using good large AC filter capacitors, and 1 or 2 good AC filter inductors, I would hope the AC wave was at least close to a sinewave, and most likely good enough to run most equipment.

[wg_ski]

Originally posted by megajocke
I'm pretty sure the modern ones with H-bridges do not go high impedance but shorts the output together during dead time, highly inductive loads like fluorescent lights wouldn't work correctly otherwise.

The simple primitive ones using a mains frequency transformer and push-pull primary do go high impedance of course.

I've used a few of the "more modern" ones. They were no different than the old school - and fluorescent light still cause the waveform to go perfectly square. They still work ok that way - but it will raise the RMS output voltage. With enough paralleled resistive load it will settle down. But some loads may be sensitive and not like it much. Electronics are pretty forgiving as loads because their supplies are rectifiers which respond more or less to peak voltage.

The ones I've seen up to about 1000W just use a switch mode DC-DC converter to make a high DC voltage from 12V. Then they do the same waveform with a H-bridge that the old 60-Hz trafo-based Tripp-Lite units of yesteryear do. I've looked into shorting the output during the dead time, and it REALLY complicates the circuitry. You just can't make a cheap inverter that cheap with those extra switching components.

[megajocke]

You shouldn't need any extra power components, just use this switching sequence for the bridge. Plus means upper transistor on, minus means lower transistor on. One transistor in each leg is always conducting, save for a short deadtime which will be needed.

Left leg  |: - - - + :|
Right leg |: - + - - :|

Or use this, which gives a more even sharing of conduction losses:

Left leg  |: - - + + :|
Right leg |: - + + - :|

When both legs are switched in the same direction the output is shorted through the transistors/diodes.