Is this a good “pure” sine wave inverter design?

I'm in the initial stages of designing a pure sine wave inverter. My design is as follows:

Wien Bridge Oscillator (5V at 50 Hz) -> Class D Amplifier -> Output transformer

My question is, is this a good design? Are there better designs out there? The inverter will utilize a 12V lead acid battery as its source and output 220V at 50 Hz. The Amplifier will amplifier the signal upto 12V and the output transformer will then step-up the voltage to 220V.

An alterate design that comes to mind is:

Wien Bridge Oscillator (5V at 50 Hz) -> Class D Amplifier (Amplify to 220V)

Since the battery is only 12V, by using a DC-DC Converter the DC source can be stepped upto 220V DC - which the amplifier than utilizes to amplify the sine wave signal to 220V.

How good are these designs? Will they work at all? If not, why so? And if they do, how good are they for a general purpose inverter?

• What do you mean by "pure"? – endolith Jul 29 '11 at 17:38
• I mean that it should only contain the Fundamental and no harmonics. – Saad Jul 29 '11 at 17:44
• Using a Class D amplifier will always generate harmonics which have to be filtered out. Since you have to filter anyways, why not generate the sine wave digitally to begin with? Even switching to a class A amplifier will create some amount of harmonics and noise, so the question is, quantitatively, how pure does the signal have to be? – Theran Jul 29 '11 at 18:12
• If you have a output transformer, it will filter just about everything but the fundamental from the output waveform, so the amplifier does not have to be too great. – Connor Wolf Jul 29 '11 at 23:35
• Plus, if this is a one-off project, you can just use a 220V-12V transformer backwards as your step-up device (derate the transformer a bit). These are much easier to acquire then 12V-330V DC-DC converters. – Connor Wolf Jul 29 '11 at 23:38

You say you're not familiar with digital oscillators yet, but I think it's worth looking into. Using a microcontroller with PWM output you can input the signal directly in the amplifier's format. So you won't need an A/D conversion at the amplifier's input stage.
The code for the microcontroller is very simple. Basically it's like this: you need a sine lookup table, and a timer interrupt generating your sample clock. This can be 5kHz, for instance, then you get 100 samples per cycle, which is enough for an inverter application; the transformer will work as a low-pass filter. Every 200$\mu$s you take the next value from the table and set the PWM register with this value. The output will be pulse-width modulated following the sine pattern.
This means that your class-D amplifier doesn't need the input stage with the triangle wave generator + comparator (this would otherwise be you A/D converter), so it only consists of the driver stage for driving the output MOSFETs.

PS: a center-aligned PWM signal will give you a cleaner sine approximation, but many microcontrollers only can generate edge-aligned PWM. This Freescale controller is available in an SOIC-8 package and is an example of a microcontroller which can do center-aligned PWM.

image from this document

• This sound does sound very attractive. I have a few questions - you mention that I can do PWM modulation inside the microcontroller. Following that the modulated waveform can then be converted into an analogue waveform via onboard DAC. This waveform can then be fed directly to the amplifier but the amplifier does not have to do PWM on this signal. Instead, it can directly use its Mosfets to amplify the signal. The output can then be filtered out via the transformer. Is this what you are suggesting? Am I correct in my understanding? – Saad Jul 29 '11 at 18:30
• Yes, there's no need to have an analog signal, so no DAC. You can use the microcontroller's PWM to drive the MOSFETs directly. – stevenvh Jul 29 '11 at 18:36
• That is amazing! Thank you very much for such an insightful reply! Will definitely look into this ASAP. – Saad Jul 29 '11 at 18:39

A better option would be to use a high voltage DC-DC converter to step 12V up to ~330V DC. You can do this at a much higher frequency than 50/60 Hz so the DC-DC converter can use smaller transformers. You'd either use two DC-DC converters, fully isolated, to get -330V, 0V and +330V with two MOSFETs (a half bridge), or one 330V and a full H-bridge (four MOSFETs) to generate the PWM'd sine. You'd need a good filter on the output to convert the PWM to a pure(ish) sine.

• 330V? Why 330V? The required output is 220V... as I type this I realize that the peak value of the 220 AC waveform is about 311V. Is this why the DC needs to be stepped up to 311V? I plan on using an H-bridge so I'm guessing I just step up the battery to 311V and I'm good to go? – Saad Jul 29 '11 at 18:34
• @saad I use 330V because it works out to 233V AC, approximately 230V AC, and that is the actual nominal voltage in countries rated "220V" or "240V". It might be a good idea to have the converter output nearly 350V, as that would allow for some sag in the output leads and internal wiring, giving you 250V at no load. N.B. 220/240V is rated at 230V +10% -6%, or 216V - 253V. – Thomas O Jul 29 '11 at 19:41
• @saad Be careful with 330V. A high power DC-DC will be difficult to design, requiring extensive knowledge of boost/flyback power supplies, and a good source of custom wound transformers. Also, at that kind of voltage, full isolation from the battery is essential, optical and/or galvanic. – Thomas O Jul 29 '11 at 19:43
• I see your point about 330V. I'm afraid I do not possess an extensive knowledge of DC-DC conversion. With this in mind, do you think my initial approach is better? Amplify the signal to 12V and then use a transformer to up the voltage to 220V. – Saad Jul 29 '11 at 19:47
• @saad Probably. But you will be looking at a very expensive and very heavy transformer for anything over about 100W. – Thomas O Jul 29 '11 at 20:05

It all depends on how well the Wien-bridge oscillator is designed. Distortion can be very low if the amplitude control feedback is correct. Here is a useful document on sine wave oscillators.

Another option is a digital oscillator, using software DDS techniques or PWM.

• Would a Bubba oscillator be easier to realize than Wien bridge? I am not familiar with digital oscillators yet. – Saad Jul 29 '11 at 17:18
• Thanks for the edit. Making the sine wave via digital techniques definitely seems to be the way to go! – Saad Jul 29 '11 at 18:36