I was thinking today about a whether it would be feasible to implement an inverter using a simple sine wave oscillator (perhaps op amp based or a Wien bridge implementation), a power amplifier stage and a step-up transformer. The more common simple circuits online use a 555 or a 4047 and as such the inverter has a square wave output as well.

However I'm sure I'm missing something here because an implementation would surely exist if it were plausible. So what are the limitations of using the the setup I mentioned earlier?

PS: I understand that buying an inverter would be more efficient and more cost effective. Just wondering if what I'm talking about would actually work at all.


3 Answers 3


To be efficient, the output amplifier has to be switching (Class D), but other than that you're not missing anything.

It's easier to do this with a microcontroller synthesizing the class-D drive signals directly (which become sine waves after filtering) than with an analog oscillator (which won't run at a crystal-controlled frequency without more parts, and requires AGC to get a stable low-distortion output level), so that's generally how it's done.

I would expect you might find some very old designs (when microcontrollers and microprocessors were relatively expensive) using just that method.

  • \$\begingroup\$ Thank you for replying. Using a microcontroller makes more sense. I spent some more time on this and it seems like amplifying the output with a class D with logic level mosfets, level shifting and then stepping up to line voltage would work out fine. However wouldn't the filter (the LPF) be too wasteful for higher power applications? \$\endgroup\$
    – Ammar
    Commented Feb 23, 2014 at 13:38
  • 1
    \$\begingroup\$ Think of an L-R filter with the R as the load. The loss in the inductor can be quite small. \$\endgroup\$ Commented Feb 23, 2014 at 14:28
  • \$\begingroup\$ If the R is the load, then wouldn't the cutoff frequency become load dependent? If I were to assume the inductor were connected directly to the step up transformer, the reflected impedence would be quite small though, which means it might work. I'm afraid I might be missing something obvious.. \$\endgroup\$
    – Ammar
    Commented Feb 23, 2014 at 16:28
  • \$\begingroup\$ Usually some C as well with this particular type of setup. theengineeringprojects.com/2012/11/… \$\endgroup\$ Commented Feb 23, 2014 at 16:37

The efficiency of a class AB power output stage is theoretically about 65% when delivering a sinewave at the maximum voltage level it can. A class AB power stage would deliver a relatively clean sinewave because it overcomes cross-over distorton. If you could live with a class B output stage (cross over distortion) you can get up to 78% power efficiency.

With class D (PWM switching), efficiency is theoretically 100% but you can easily lose up to 10% in the extra filtering and high-speed switching losses.

So if you want a 1000W output invertor you'd be wasting: -

  • Class AB - 538W
  • Class B - 282W
  • Class D - 111W

Another benefit of class D is that lowering the output amplitude to the 90% level doesn't incur hardly any further loss in efficiency compared to the linear types - the linear output stages have maximum efficiency at maximum power delivery. See this for a comparison: -

enter image description here

The graph is for a class D audio amplifier but the principle is exactly the same.


Ten years later:

It´s all about generating a PWM pulse train, including a feedback circuit. When not using a microcontroller, you have to use an Error amplifier + comparator. You also need two reference waves; line wave (sine 60 Hz or 400 Hz) and a triangle wave or sawtooth wave (carrier frequency; up to 50 kHz but usually much lower). Use an amplitude of 0,95.

Use a H-bridge (SPWM) and an output filter. Using 2 voltage levels on DC-bus, you will get a 2 level PWM but then you also need a bipolar input (=> tapped transformer). When using 3 level PWM, you can use a unipolar input (DC).

Spend extra attention to the frequency compensation.

Avoid solutions with square waves or "semi-sine". Also avoid solutions with no feedback. When using a solution with a microcontroller; it will take a lot of time.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.