-=> Quoting Roy McNeill to John Tserkezis <=-

Hello Roy,

 JT>  So much for "modified square wave".

 RM> aha - so that's what "mod squarewave" is... I'm familiar
with it, but
 RM> I'd always lumped it in with the straight squarewave, and called it
 RM> variable duty cycle squarewave. I thought mod sq was the above with
 RM> partly rounded edges, to reduce rfi and ringing. Innocent me...
 
 That's what I thought, maybe only the ones I've come across were the bodgie
ones.  Wouldn't be suprised, I've worked on some real cheapies.

 RM> The choo choo inverters, btw, use method 1, but their main
 RM> purpose in life is to drive motors.
 
 JT> You mean the PWM type? The fancy nancy train controllers do yes.
 JT> The simpler ones just use a high power variable resistor in series.
 JT> (Bodgie yes, but cheap).

 RM> Odd, that. An electronic setup would be similar in price to a power
 RM> pot, wouldn't it? (although the assembly cost could be a bit higher,
 RM> I spose)

 I'm not talking decent high power pots, just simple half-turn levers that
are cheap.

 RM> terminal to give it continuous power. Before its fuse popped, a
 RM> tantalum cap protected the power supply from the 240 by shorting
 RM> with a bang. I like tantalums. (Hi Brenton!)
 
 At least you know which one blew.  I've come across the bloody things that
were shorted with no physical signs of stress.  It was a bugger to find. :-(
 
 RM> Did you read my Exploding Speed Controller story from a couple of
 RM> years ago? I troubleshot that one at the leccy's shop, with two 100
 RM> watt light bulbs in series with each incoming phase. Good thing,
 RM> too - when I bent the control board just so, all six bulbs got very
 RM> bright.

 Erg, good fun.

 RM> Regulation: The simplest way would be to depart from the rigid sine
 RM> specification, and just alter the pulse widths in a fudgy way, eg
 RM> inc the hi pulses in the first part of the quarter cycle by
 RM> multiples of 1, and in the second part by multiples of 2, etc.
 RM> There's plenty of processor time available, it can spend all the
 RM> free time in one half cycle setting up the offsets for the next
 RM> half cycle. The current program has an empty main loop, it only
 RM> does something if there's a timer interrupt, and that something
 RM> in the current program takes around 36 usecs using a 10MHz clock.
 
 Blotching up the sine wave like that defeats the purpose of having a sine
inverter in the first place.  Sure it would be cleaner than a square only,
but if you go the trouble of making a sine output inverter, you may as well
actually USE a sine wave. :-)

 JT> One PWM inverter I've seen,
 JT> fairly high power (around 2KW) used an eprom to generate the PWM
 JT> signal on the mains 340V dc side.

 RM> Good technique. With a big enough eprom, regulation could be done
 RM> by having a heap of time arrays, and selecting them by connecting bits
 RM> from an a/d converter to address lines of the eprom. Chip count is
 RM> a bit high for what I want, though. Eprom, a/d, at least two
 RM> counters, probably a latch, various bits of glue logic.
 
 In the one that I've seen, feedback was taken from the output, and fed back
to the first DC-DC converter stage.  This way they didn't have to muck around
with the eprom side at all.

 JT> If it is just a square wave, you don't really need a PIC, as it can
 JT> be done cheaper with garden variety components.

 RM> Marginally cheaper. A PIC and crystal is typically less than $15.

 I suppose it only matters on mass production.  In a one-off, you probably
would be better off with something like a pic, since you can play with other
regulation/sine generation techniques.
 
 JT> I think EA had a
 JT> simple low power (40VA) 12 inverter that used this principal.

 (stuff deleted)
 RM> on period). Being a lazy sod, I heatsunk the transistors rather
 RM> than fix the real fault...

 Tsk, tsk.  :-)

 RM> We fixed a 600VA DSE inverter recently (broken wire) that used the
 RM> "modified" squarewave. Methinks it was a tad overdesigned - there
 RM> were at least a dozen ICs on its control board.

 Whoo hoo!  Probably a real modified square wave.  The cheap ones I've seen
had no more than a few ics.

John Tserkezis, Sydney, Oz. Fidonet: 3:712/610  Internet: jt{at}suburbia.com.au

... I quit Paranoids Anonymous.  They were all plotting against me.
---