Re: DC power supply again

From: 	Harri Suomalainen[SMTP:haba-at-cc.hut.fi]
Sent: 	Thursday, August 28, 1997 6:31 AM
To: 	Tesla List
Subject: 	Re: DC power supply again

On Sun, 24 Aug 1997, Tesla List wrote:
> > > loser in a number of systems, I have decided on a completely 
> > > different approach. I am currently designing an off-line flyback 
> > > switcher to run at the 2kW+ level. What I am looking for:
> >  
> > > - efficiency in the supply of around 80 - 85%
> > > - being able to use any size cap without having to deal with primary 
> > >   charging resonances
> > > - good gap quenching. I am looking at using optical feedback from the 
> > >   gap to shut the supply down if necessary although that is 
> > >   peripheral to getting the supply up and running. Experiments 
> > >   suggest that with this type of supply, quenching is solely dependent
> > >   on output discharges anyway.
> > > - continuously variable power setting. Break rate is obviously 
> > >   dependent on gap and power setting.
> > > - Max output of 20kVDC
> > > - No smoothing cap required. The supply charges Cp raw.
> > > - Incorporation of power factor correction by using the input 
> > >   haversine to modulate the internal reference (not yet tested).

Humm.. I'll be very interested to hear how practical prototype will
perform out. I've hade the same idea for quite a long time ago already.
However, I've thought there may be some serious problems with flyback

Assume the gap fires, primary resonates and transfers only partially the
energy during 1st half cycle to the secundary. Then, the primary current
after resonance to 0 volts will start going partially to the cap (charging
it negative!) and partially through the flyback secundary and diode.
All the energy left in the cap (now negative) will have to be dissipated
somewhere. Possibility of distruction? Have you thought about this?

Therefore I had the conclusion that the topology must be something else
to be safe. Almost every topology will still be in trouble when the
partially charged cap is reverse biasing the system. I can think of no
other solution but preventing that like with a dc cap and large inductor
after that. Unfortunately that means a large, expencive, bulky inductor.
There are some other rather exotic solutions but I'm not for them either.

I'd still love the SMPS approach becouse of the same benifits you
described. Variable power & voltage power source would be quite usefull
for other purposes as well.

For optical feedback I'd also suggest using that to determine need of
eg. large magnetic fields in the gap to quench it. Rather large magnetic
fields are quite quickly generated with suitable coils driven by say a
FET from a 400V source.. I bet that kind of device would be actually
rather simple to build and it would allow tailoring quenching times just
what you like. I guess that could also be used for preventing gap firing
before the suitable moment.

> > >and hopefully switch losses of the same order. Please don't bombard 
> > > me with "how's it going". I am currently suffering financially and
> > > time to completion is indeterminate at this stage. I see this as a 

The same old problem. I usually get the same problem as well. :(

> > > mature this technology and have tested a number of designs in order 
> > > to arrive at a well behaved one that is scalable to the kW range.

Those kind of devices for kW use are mentions even in literature. Usually
they are nowadays resonant mode devices directly charging a cap. Those
tend to be rather cheap.

The best transformer solution may be something like primarys in parallel,
secundarys in series (for non flyback). Usually many small separate
transformers is better than a huge one made from E-cores. Large ones
do not dissipate heas as well. If it is made of small units anyway, I'd
certainly go for multiple transmormer system. It can be prototyped with
low output voltage and expanded later as needed.
We have phone numbers already, why would we need IP-numbers! -unknown person

Harri.Suomalainen-at-hut.fi - PGP key available by fingering haba-at-alpha.hut.fi