[rescue] phase converters, was Re: Rescued: SGI Altix 450

[Ron Wickersham]

On Thu, 28 Jul 2016, Dave McGuire wrote:

>  The issue here is that with a whole lot of big transistors operating
> in their active region, which would be needed for a 60Hz sine wave (or
> three of them), which means a much bigger heat sink than those little
> DIN-rail-mounted boxes could contain.  So they use a few banks of
> smaller transistors restricted to either cutoff or saturation, not
> active region but just switching, and create a stepped rough
> approximation of a sine wave.  This will tear a switching power supply
> apart, but with a big fat motor they're fine because the winding
> inductance of the motor smooths out those stepped waveforms.

hi Dave,

one could build an inverter power amplifier that operates in the linear
region of the output devices, but all large inverters use rail-to-rail
switching (popularly referred to as Class D) with pulse-width modulation,
so the output devices are always in full-on or full-off mode.

one way to acheieve this is with a triangular wave at the switching
frequency (say 60 KHz) fed to one input of a comparator, and a 60 Hz
sine wave to the other input of the comparator.   when the 60 Hz is
at zero, the output of the comparator is a square wave, and when the
input sine wave is at the negative peak the rectangular wave is virtually
all to the negative rail, with very narow glitches to the positive rail,
and when the sine wave is at the positive peak the rectanular wave is 
mostly at the positive rail.    if you want galvanic isolation, then
the transformers can be at the switching frequency with small magnetic
laminations (or ferrites) without saturating the core, while 60 Hz
magnetics require large magnetic laminations support enough primary
inductance to avoid saturation.

this amplifier technique (an several variations) are popular for power
amplifiers for audio in addition to inverter applications.

-ron