Modified Sine-Wave Inverter Enhanced

Aug 1, 2006 12:00 PM
By James H. Hahn, Associate Professor Emeritus, University of Missouri-Rolla Engineering Education Center, St. Louis

Proposed Improvement

Consider now a further modification — the addition of another level. The waveform is shown in Fig. 5. Again using the fact that the waveform has both half-wave and quarter-wave symmetry, we carry out the integration over the period 0 to π/2, with the result that:

This result has four variables, of which all could theoretically be varied to achieve minimum distortion. However, one particularly efficient approach is to choose a very simple set of values for A and B — namely B = 2×A — and then optimize the values of a and b for minimum distortion. This approach requires only two positive and two negative power-supply voltages, all of which can be generated from a single transformer in the high-frequency oscillator. (Other values of A versus B may be useful, but were not investigated because the simple relationship of B = 2×A had very good results as discussed later.)

With this restriction, evaluation of the Fourier coefficients indicates that the minimum distortion that can be achieved is about 6.5% (-24 dB), and occurs at β = 0.42π and α = 0.248π. The third harmonic is only about 0.17% (-55 dB) of the fundamental, suggesting that minimal low-pass filtering would greatly reduce the fifth and higher-order harmonics and produce a relatively clean sine wave. The third harmonic can be eliminated entirely, with β = 0.42π and α = 0.246698π, at the expense of slightly higher THD.

[Note: The Fourier analysis was carried out through the ninth harmonic for all three types of waveforms considered in this article. Harmonics above the ninth are not negligible, but any filtering applied to reduce the third through ninth harmonics will be even more effective on those above the ninth. Therefore, the higher-order harmonics are ignored in this analysis.]

Implementation

As demonstrated here, the modified-sine-wave inverter can be modified further to produce a much closer approximation to a sine wave, at a relatively small increase in manufacturing costs, simply by incorporating another level into the waveform. The design still uses switching technology in the power stage, assuring high efficiency. A patent application has been submitted for the approach described in this article.

The switching stage could be implemented with a combination of bridge and half-bridge components commonly used in power switching applications. To produce the proposed multiple-level waveform, several implementations are possible. In general, they all involve connecting the output lead to a specific voltage level with switches such as power MOSFETs capable of handling substantial current. Consider the block diagram shown in Fig. 6 where the voltages A and B correspond to the voltage levels defined previously.

Appropriate digital logic and timing circuits will be used to activate each switch at the correct time to achieve the α and β pulse widths. A table can be developed to indicate which switches must be closed for each section of the output waveform. Note that Switch #3 in Fig. 6 will need to be a bidirectional switch, since it must switch the output lead V_OUT to ground regardless of any voltage present in the load. All other switches can be unidirectional.

Unlike conventional PWM-inverter designs, which switch at high frequencies, the proposed inverter design switches at just three times the line frequency. As a consequence, the proposed inverter design will reduce switching losses from that of the PWM-controlled inverter and will save power regardless of the output power level.

For further details on implementation, contact the author at jhahn@umr.edu.