[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: help: HV SMPS coil driver

> Dreaming of a compact TC power supply and RF base feed driver...
> ...so I'm in the process of building a small, modular, directly mains fed
> 500W SMPS power supply to drive a TC secondary, 100..300kHz adjustable.

I've been playing around with those things. Actually, I've got one prototype
on my desk even at the time. However, I'm heading for much more power.
The basic idea is still the same.
> The ferrite xfrm is about ready, should give 3..5kV, but I've now started

Sounds quite high, depending on the secundary. With my 1'x1' secundary I found
surge impedance Z0=sqrt(Lsec/Csec) is quite low (assume 20kohm).
When I feed it with 1:10 transformer it would look like 200ohm load for the
low-voltage side. With 200V sine drive at the exact resonance it would draw
1A current ie. 200W power.

Check things like this before running to 3-5kV drive. I found 1,4kV peak 
(transformer) secundary voltage resulted in 1kW power draw with the actual
prototype. With high turns ratio it would always have blown up. Start with
lower turns and add more as the prototype works well. At leas provide
taps for getting lower output voltage if needed.

> wondering about insulation (a bit late maybe... ;). I've used LD-PE
> plastic bags and transparency sheets as insulation between primary and
> secondary, and also between each secondary layer (turns, one layer PE,
> one wire led across, one layer of transp sheet, next turns). Plus,
> poly-urethane varnish. 

LDPE is not a good choise. You usually design the transformer for 75-100 C
temperature and that's where LDPE might melt. Use something like mylar
instead. It will not melt with the normal operation temeratures.

> I designed the xfmr for push-pull, but with my lack of experience and
> those symmetry/saturation problems I've for now dropped that idea.

Current limiting will take care of those problems. (If it is acting fast
> I'm using a CA3825 PWM chip (with dead-time expansion), but couldn't find
> any FET buffer chip so looks like I can use only two FETs safely (?).

3825 should provide enough drive current for normal FET driving. I had no
problems running at up to around 1MHz. (A different topology though but
the same drive requirements.)
> Sooo... what is the "best" way to hook up that xfrm? 
> What would put minimal strain on the switch MOSFETs? 
> (They are 500V 8A, mains is 220V 50Hz.)

If you choose push-pull it will also define the stress. Do not forger enough
snubbers of voltage spikes will kill the fets.

Do not forget current limiting. I had a lot of problems with coil kick-back
which resulted in transformer saturation and blown fets if the current-limiting
did not act fast enough. Quick limiting circuit on the other hand tends to
suffer from noise problems.
> Other thing, I was wondring with voltage feedback from the secondary...

No idea in that one. You know the transformer secundary voltage depends
just on your transformer and drive.

> Do you use a series cap on the fet to cancel out xfmr reactance? That
> sounds wrong somehow - but how do you do it otherwise?

Yes in bridge topologies, no in push-pull.

> (By the way I've never built a SMPS before but've tried to gather as much
> knowledge as possible for starters. If you happen to know any good web
> site on SMPS, I'd be very much interested.)

Start with low powers and use variac. This will save you a lot of fets!

Sorry, no detailed info on my prototype available yet. Probably some day when
it works well. The basic design is based on resonant topologies. It is
inheritantly short-circuit proof even with no current limiting. My current
is with the automatic syncronization to the coil resonant frequency.

Sounds simple but is not with high-Q coils. Coil current is easy to sense from
primary side and is good for syncronization reference. However, unless you are
already within say 1-2kHz of the resonant point it will draw almost no current.
The primary side resonant circuit also makes things a bit more complicated.
Once I get that working it will be worth publishing the idea and design for all
of you out there.