12V->120V AC inverters using 'modified square waves' -- why?

[Joel Kolstad]

I was wondering if anyone could tell me why 12V DC to 120V AC (or 240V AC in
many parts of the world) inverters often output 'modified' square waves,
that is, some negative voltage, 0 volts, some positive voltage... and
repeat, rather than just a square wave? The only reason I could come up
with was so that both the RMS value of the waveform would be correct _as
well as_ the peak value (sqrt(2)*120).

Clearly the harmonic content of such an output is different than that of a
square wave, but that begs the question: What kind of loads care about
harmonics? Transformers and motors presumably will tend to low pass filter
their inputs owing to their magnetiziation inductance, so in such cases
using a modified square wave inverter rather than a sine wave inverter seems
questionable (at least at higher power levels) anyway, no? On the other
hand, if you know your load is going to be, e.g., a switching power supply
(such as a computer, TV, etc.), you could get away with a much smaller
transformer in the inverter and use a considerably higher frequency instead,
correct (while still maintaining the same RMS and peak voltage -- just move
the fundamental and all the harmonics to higher frequencies)?

Thanks,
---Joel Kolstad

 

[Martin Riddle]

I was wondering if anyone could tell me why 12V DC to 120V AC (or 240V AC in
many parts of the world) inverters often output 'modified' square waves,
that is, some negative voltage, 0 volts, some positive voltage... and
repeat, rather than just a square wave? The only reason I could come up
with was so that both the RMS value of the waveform would be correct _as
well as_ the peak value (sqrt(2)*120).

The RMS value is what is the target of the M.S. wave form.

Clearly the harmonic content of such an output is different than that of a
square wave, but that begs the question: What kind of loads care about
harmonics? Transformers and motors presumably will tend to low pass filter
their inputs owing to their magnetiziation inductance, so in such cases
using a modified square wave inverter rather than a sine wave inverter seems
questionable (at least at higher power levels) anyway, no? On the other
hand, if you know your load is going to be, e.g., a switching power supply
(such as a computer, TV, etc.), you could get away with a much smaller
transformer in the inverter and use a considerably higher frequency instead,
correct (while still maintaining the same RMS and peak voltage -- just move
the fundamental and all the harmonics to higher frequencies)?

Harmonic content is high, radio/tv equipment is much better off with a sine.
Although they can be used.
Capacitve start motors most likey will not run (or start). And some motors
require MS inverters of a 3-5kw to just start. Using a sine inverter these
requirements are not needed.
For straight resistive loads, the MS inverter is fairly efficient.

If your using a 60hz inverter (MS or sine) then the appliance is most likely
60hz too.

 

[Gabriel]

Hi,
The reason the modified sine wave is simplicity. Generating true sine wave
requires PWMing at a higher frequency, filtering the PWM output such that
ripple is reduced to an acceptable level, tracking the sinwave output such
that regulation and transient distortion is taken care off, all these
translates to complexity, hence cost. The additional switching loss and
thermal management are additional considerations.

Modified sinewave (square) has a number of effects on different appliances
frone end.

On simple rectifier-capacitor filter, like Non-PFC switching power supply,
the fast rising edge will cause a current surge larger than the normal crest
factor effect.

For appliances with a PFC frontend, probably not much of a problem if the
control loop is decent.

On a transformer, the additional harmonics will generate more Iron loss and
Copper loss. If the square wave +/- cycles are not balanced, additional net
DC current will flow into transformer primary, causing saturation if the
current is too large.

For devices that is designed for sinusoidal power, like a split phase
induction motor, the higher frequency harmonics will cause various degree of
screw-ups depends on the actual design.

For universal motor, it runs as usual.

 

[Joel Kolstad]

Thanks for the responses. One thing that was perhaps unclear in my original
posting was why one prefers a modified square wave to a 'regular'
non-modified square wave -- and it was for that question that my only guess
was so as to preserve both the correct RMS value and peak value (since you
do see designs on the web where a regular square wave is used, e.g.,
http://hobby_elec.piclist.com/e_ckt8.htm).

Intuiviely I would think the harmonic content of the MSW is a little better
than a regular square wave as well, but clearly the deleterious effects
people have pointed out are significant in both cases.

---Joel

 

[analog]

One thing that was perhaps unclear in my original posting was why
one prefers a modified square wave to a 'regular' non-modified
square wave -- and it was for that question that my only guess was
so as to preserve both the correct RMS value and peak value [...]

MSW inverters only control pulse width (4 edges per output cycle)
and therefore cannot regulate the peak of their output waveform
(which depends solely on input voltage and transformer turns ratio).
This scheme minimizes switching losses while still allowing regula-
tion of the fundamental frequency and RMS level of the ac output.
Thus, synchronous motor based clocks run on time and the lights
don't dim as the battery runs down.

Note that for inductive loads (i.e. motors) the transformer must be
shorted during the dead time between cycle halves in order to allow
circulation of reactive load current. This is most conveniently
done with a mosfet full bridge inverter driven with independent
square waves to each bridge half (where the phase shift between the
driving waveforms is the control mechanism). Note that the trans-
former primary is shorted during the dead time when both sides of
the bridge are either simultaneously high or low.

 

[Joel Kolstad]

This scheme [modified square waves] minimizes switching losses while still allowing regula-
tion of the fundamental frequency and RMS level of the ac output.
Thus, synchronous motor based clocks run on time and the lights
don't dim as the battery runs down.

Gotcha. Thanks!

Note that for inductive loads (i.e. motors) the transformer must be
shorted during the dead time between cycle halves in order to allow
circulation of reactive load current.

I was planning to parallel each of the 4 switches in the bridge with
Schottky diodes to provide this path.

This is most conveniently
done with a mosfet full bridge inverter driven with independent
square waves to each bridge half (where the phase shift between the
driving waveforms is the control mechanism). Note that the trans-
former primary is shorted during the dead time when both sides of
the bridge are either simultaneously high or low.

Yes, but presumably you still need diodes anyway to provide a current path
during the oh-so-brief-but-ever-present dead time between turning off (say)
the lower section of one leg and turning on its upper section (which,
assuming the other upper leg was already switched to power, has now
effectively shorted the transformer)?

---Joel

 

[Bill Sloman]

Thanks for the responses. One thing that was perhaps unclear in my original
posting was why one prefers a modified square wave to a 'regular'
non-modified square wave -- and it was for that question that my only guess
was so as to preserve both the correct RMS value and peak value (since you
do see designs on the web where a regular square wave is used, e.g.,
http://hobby_elec.piclist.com/e_ckt8.htm).

Intuitiviely I would think the harmonic content of the MSW is a little better
than a regular square wave as well, but clearly the deleterious effects
people have pointed out are significant in both cases.

A properly constructed modified sine wave has no third harmonic
content, much less fifth harmonic content and significantly less
seventh harmonic content than a simple 50% duty cycle square wave.

Neither the modified sine wave or the sysmmetrical square wave has any
even harmonics, and the amplitude of the higher harmonics decreases in
proportion to the harmonic number.

In short, the harmonic content of a "modified sine wave" is a lot
lower than that of a square wave.

 

[Joel Kolstad]

A properly constructed modified sine wave has no third harmonic
content, much less fifth harmonic content and significantly less
seventh harmonic content than a simple 50% duty cycle square wave.

Fair enough, but is eliminating the third harmonic a good tradeoff with
regulating the width of the modified sine wave so as to keep the output at
the correct RMS value? Between low and high battery input voltages and
minimum and maximum loads I could see the output's RMS voltage varying but,
say, 20%.

I have parts ordered to build my own inverter for my own edification...
should arrive later this week! (I say 'edification' because you can get
finished working 300W units for all of something like $20 these days...)

Thanks,
---Joel