How to change PWM modulation to pure sinewave

We have a solar powered 3 phase inverter which produces electricity in PWM form. It can power a three phase motor without any problem but there is some control panels which can work only on single phase pure sinewave electricity.

How can we change the electricity generated so that it will work for the control panels.

Hi Subash,

By PWM form, I'm assuming it just does square wave inversion at 50Hz?

If you need a pure Sine inverter, you will need to modulate the PWM sinusoidally at a high frequency(typically 100KHz or more) and then filter it out with a L-C circuit. There are many ways of doing this(most standard power electronic textbooks provide all the details, I'd recommend checking 'Fundamentals of Power Electronics' by Erickson, plus the slides are available freely) but note that these will require high speed FETs/IGBTs with a decently powerful microcontroller to achieve proper stability.

Since the question is open ended I'm not sure what your tolerances are and what sort of control scheme you'll need.

I assume that since it is a solar powered inverter, it uses PWM to keep the total o/p power with a certain range. That is, in bright conditions, the duty cycle will be less than 50%, and under dull conditions, more than 50%. I also assume that you need to simply convert the square or rectangular wave itself to sine - that is, the PWM signal itself is not being modulated by a sine wave which needs to be extracted from it.

If that is the case, unless we know by how much the duty cycle varies between different light conditions, we cannot say for sure if it can be converted to sine wave. For example, if it is between, say, 50 +/-5%, we can probably convert it to a sine wave (with caveats). But if the duty cycle variation is more than that then we will not be able to convert it to sine wave. For one, a sine wave has a constant DC offset or mean value (usually zero). However a sine wave generated from PWM will have a net mean value that will vary based on the on/off time of the PWM wave - and most equipment will not like this.

If the duty cycle variation is within +/-5%, you will get a pseudo-sine wave that has a net mean value that will vary by that much - and hopefully any equipment connected to it will (or should) be able to cope with it. Also, once you convert it to sine wave the total power available will also be reduced since the peak value of the sine wave will be limited to that of the square wave, but the RMS value will now be 0.707 times the peak whereas for a square wave the RMS value is the same as the peak value. This will get worse (for AC) if the duty cycle is not 50%.

Hi Rohit,

There are a few things I need to add to what you wrote. This is a long one, so please bear with me

*_I assume that since it is a solar powered inverter, it uses PWM to keep the total o/p power with a certain range_*

Well, not necessarily. The power is primarily dictated by the load and the inverters job is to keep the voltage regulated to either 230V/120V depending on the region you live in. i.e., voltage must be regulated but we can't do it by varying the duty cycle on the inverter end as we will be invariably changing the output voltage as well.

To counter this, solar inverters are multi-stage converters and the first stage boosts the voltage(most likely a flyback/resonant converter) and regulates to a fixed setpoint(generally 300~400V). This first converter modulates its duty cycle depending on bright/dull conditions.

The inverter stage is *always* a 50% square wave(assuming that it's not a sine inverter).

_*However a sine wave generated from PWM will have a net mean value that will vary based on the on/off time of the PWM wave – and most equipment will not like this.*_

This is exactly why we fix the inverter side duty cycle to make the mean exactly zero. In fact, if its anything other than zero you'll see harmonics popping up(run a quick fourier if you're curious) and you'll fail test specifications needed to sell inverters on the market.

Like I mentioned before, for pure sine inverters we modulate a fast triangular wave(100Khz or more) with a slow sine wave(50/60Hz). What you'll end up with will be a 'chopped' sine wave you can filter out with smaller L-C values(because all the harmonics are essentially at 100KHz and we can filter them out with smaller components)

Whew, that was long.

Best,
Sree Harsha

Bandlimit it to around 500kHz to 1Mhz [typical], or at least to a point where it becomes a sine wave [I assume you have a scope to verify this].

To do this, any simple low pass RC filter will do [being the fastest and cheapest method].
If you have roll-off restrictions, use a Sallen-Key LPF circuit.

Both schematics for the circuits can be easily designed from online resources and can be constructed in a short amount of time.

Hope this helps.
-Justin

Hi Justin,

What you said will work, but only for low voltage, low current sine-waves(especially the R-C approach). The question was based for a power inverter, something like a backup supply which needs to bump up the output voltages to 110/230V@60/50Hz

At these power levels, we go the L-C approach to filtering as its (more or less)lossless. Plus you don't try to filter out a direct PWM with a filter as I mentioned before. Way too many harmonics to take out before we hit the

Hi Sree,

Doesn't L-C use a capacitor?

Thanks.
Justin

Hi Justin,

Apologies for the late reply.

_Doesn’t L-C use a capacitor?_

Yes it does, but the fundamental difference lies in the fact that the Inductor can allow larger currents to pass through. To illustrate, there are two ways you can make a R-C circuit filter(or smooth out the PWM) more, either increase the R or increase the C. But, if you increase R, you're essentially putting a limiter which will contribute an I*R drop and a I*I*R power loss depending on the load. The reason I said the capacitors would be huge is because the resistor has to be small to supply any significant load.

On the other hand, the L in the L-C circuit can be made reasonably big without incurring too many core/magnetization losses. Note that by this I still don't mean feeding a 50Hz square wave into a L-C is a good idea, atleast for power inverter applications. We still need to modulate the 50Hz sine-wave with a high frequency triange/ramp wave and then use the L-C filter to make the L and C small(and light) enough.

Best,
Sree Harsha