Subject: Re: serial port control
To: None <current-users@NetBSD.ORG>
From: Wolfgang Rupprecht <>
List: current-users
Date: 02/12/1998 10:55:06
Brett Lymn writes:
> According to Wolfgang Rupprecht:
> >Its not clear to me yet that sine wave output wins.  Square wave
> >should be more efficient since you don't have the big dead-time
> >between, the two peaks.
> Urrgh no.  Firstly, there is no dead time - the voltage continuously
> changes.  

First, let me state that I have designed high efficiency DC-DC
switching regulators for military avionics.  You know the kind where
they want a few 100 watts at 95%+ efficiency in a 3 or 4 cubic inches.
Its the sort of stuff that makes your average UPS ciricuitry look like
someone's failing high-school project.

The first thing that becomes clear when designing things for that sort
of effieciency and density is that you want to keep the current
flowing as much as possible.  You want your output transformers,
diodes, transistors to run for as close to full-time as possible.  If
you try doing something silly like put out sine waves, you will only
be charging your output capacitors for *maybe* 10% of the time.  Thats
if you are lucky.  At this ratio, you will need 10x the size
transformers, diodes and transistors.  You will end up having 10x less
density (or higher losses etc).

> Lots of things are designed with a sinusoidal voltage in
> mind, feeding then a square wave can cause problems.  

We are talking about a UPS for the express purpose of backing up a
computer which has a switching power supply.  Computer switching power
supplies have a diode bridge as the first stage of the converision.
They'ed gladly take a square-wave input.  They will of course work
with a sine wave, but in that case they will only be charging the
input capacitors at near the peaks of the sine waves.  The duty cycles
will be very short, and the currents flowing during these times will
be much higher than the average currents flowing.  All this will lead
to higher losses than necessary.

> >  If the load is your typical PC switching
> >power supply then it would really prefer to have the full ~170v peak
> >voltage applied all the time.
> No, normally these devices have input filters to try and stop hash
> getting back into the power grid.  The filter will be busily trying to
> chop off the harmonics that it has been fed.  It will heat up, whether
> it will do damage is a matter of some conjecture.

Are we taking about a waveform from the switching supply that is so
ugly it hurts to look at it on a scope?  I guess someone could hide
all sorts of sins in a cheaply designed UPS.  You are right, I
certainly wouldn't want something like *that* anywhere near my

I was talking about a 60hz UPS, and the assumption is that the person
designing the UPS isn't a total dork and is limiting the rise time to
something reasonable.  Further I assumed that they did a good job and
didn't have horrible overshoot and ringing problems.  In short I'm
assuming that they didn't use the excuse of square wave output to hide
all sort of sins.

There is absolutely no good reason to have an excessive rise time that
will hit and dissipate huge gobs of power in any RF filter stage.  How
low a fequency of an RF filter stage are you going to see anyway?
1Mhz, 100khz?  We are talking the bare minimum needed for a *60hz*
supply here.

Seeing Ken's observation about his UPS makes me cringe.  I was hoping
that the monitor just had a transformer input stage and this was
causing the obvious problems.  Your claim about the massively
excessive RF outputs that cause people to worry about the input filter
really makes me worry about what the UPS industry is trying to pawn
off as "square wave output".

I'd certainly prefer a square wave output that had a rise time of say
5-10% of the waveform.  Thats about the right effieciency / RF trade
off for me.  If thats not possible I'll certainly take a sine output.

Wolfgang Rupprecht    <>
	  Never trust a program you don't have sources for.