Re: Recent s.s.t.c work

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: "K. C. Herrick" <kchdlh@xxxxxxx>

We've got 3 people commenting on this so I'll copy|paste from several 
postings & add comments:

1.  Terry: "> b) an offset from the zero line

You should track that one down for sure!  Could be a bad FET or 
something.  Could be just a scope probe offset too...":

The scope inputs are set for dc.  But now that I think on it further, 
perhaps that offset is merely indicative of the net dc current drain 
from the electrolytics.  Remember that that shunt is in series with them.

However...the shunt waveform in my "5" image does not now seem to 
jibe with a simulation I've just run.  I simulate an H-bridge, 
feedback-driven thru a 5-winding gate-driver transformer (1 pri, 4 
sec's).  With k specified as a perfect 1 for all windings, the 
bridge-supply current is indeed a perfect "full wave rectified" sine 
shape, as I had conjectured in a prior posting, and IGBT switching 
occurs at exact load-current zero-crossing.  But when I set the k's 
to 0.98, lo & behold the switching has become delayed so as to occur 
at a full 30% of the maximum supply-current amplitude.  30% for such 
a small decrease in k!

In my hardware, I'm using a commercial 2 mH common-mode toroidal 
choke for the 1:1 transformer in the low-level feedback loop.  Surely 
the k of that xfmr is pretty poor since the 2 windings are not 
bifilarly wound but rather occupy <1/2 of the toroid's circumference, 
each.  So I'll soon be checking the shift across that xfmr and likely 
be replacing it.

2.  Terry:  "I just tested mine like this:

<http://hot-streamer.com/temp/CurrentShunt-test.jpg>http://hot-streamer.com/temp/CurrentShunt-test.jpg":

Your shunt looks exactly like mine.  There probably is some skin 
effect but I'd not think that would be inductive, so it would merely 
cause an apparent increase in the indicated current.  That the shape 
of the wave in my "5" image is essentially the same as that in the 5A 
simulation image seems indicative to me that it is reasonably 
acurate.  (And in that 5A image, I rather wish I'd thought to change 
the scale of one of the curves so that they did not overlap.)

And it looks as if, were the switching to have occurred at the 
desired z.c. place, the "5" image would indeed be that of a f.w. 
rectified sine.

3.  Steve Conner:  "I explored a whole bunch of different ideas, but 
I kept coming back to the PLL.":  Yeah, I'm starting to think you are 
right.  A PLL is surely the "standard" way of deriving a sync'd 
signal that is identical in frequency but variable in phase (which is 
precisely what we want).  If there were a simpler way to do that it 
would already be "out there".

"It lets you access an extra tuning...":  By extra do you mean at a 
harmonic?  Are not the choices of output frequency limited by which 
ones will maintain the phase lock?

"My driver has five pots to twiddle and the other Steve's had just 
one adjustment last time I looked.":  Ouch!  I'll take 1, or even 2, any day!

"BTW- I think I might have signed up for that "Scientists Called 
Steve" thing already? I'd better check. "  If not, make yourself #645 
or whatever, pronto!  For the benefit of others, "Steve" is in 
reference to the late scientist and author Stephen J. Gould, who was 
a firm advocate for evolution; and the "Steve-o-Meter" is a listing 
of other like-minded scientific "Steves" intended as a light-hearted 
poke-in-the-eye for Intelligent Designers.

"> Interesting...what's the squiggle, I wonder?

Well, the primary current was flowing in the anti-parallel diodes 
from the switching instant up until the squiggle. So I'm guessing the 
squiggle is the recovery transient of the diodes...":

 I suppose that's very likely right.  If & when I correct that phase 
I'll have to see what has become of those squiggles.

"I'm impressed by the output from your scope camera":  Well...it's 
just the Canon A80 doing its thing, of course, aided by the mere husk 
of a used-to-be-fine Tek C53 camera.  It's nice that with the digital 
camera one can just snap away willy-nilly at no cost until one gets a 
decent image, instantly visible.

KCH