Suggestions for 1kW, unregulated SMPS

[Haude Daniel]

"Unregulated SMPS" probably sounds funny, but this is it: At my lab
we frequently heat up metal samples in vacuum by placing them near
a hot tungsten filament and applying a high voltage, essentially
making the sample the anode of a vacuum diode. This typically
happens at voltages of around 1kV and currents of .5A, but 1kW
total power isn't unheard of.

Now we have some homebuilt units that take care of this. They
consist of a custom-wound mains transformer with a bridge rectifier
for the HV, and the power is regulated via filament temperature.
These beasts work OK, but the transformer makes them heavy and
expensive.

Now we need a few more, and the natural path to follow would of
course be to make them switched-mode -- except that I've never done
an SMPS. I may be wrong, but I think this should nevertheless be
doable since I need no regulation at all; the first version would
just have a pot to adjust the duty cycle. So this thing would
consist of no more than a PWM controller, an IGBT H-bridge, a
transformer and a diode bridge. The output needs to be
short-circuit proof; I'd do that by adding some series inductance
to make the primary dI/dt slow enough to be safely cut off by the
overcurrent trip.

The only tricky part that I'm aware of is the design of the gate
drive circuit -- it mustn't ring and it must be quick to reduce
heat dissipation and avoid cross conduction. The transformer, of
course, is a critical part in any SMPS design, but I figure that a
poorly wound xformer can't have any ill side effects except
having too little output power due to stray inductance.

Ah yes, and an 1kW SMPS running from a single-phase 230V line needs
PFC and soft-start. Will have to read up about that.

Should I do it? Maybe it's a bit stupid, seeing that only 5 of
these things are needed at the moment and there exists a design
that works and can be assembled and tested by the techie who
originally designed it (he wouldn't touch a switcher with a 10-foot
pole). However I'd like to have a bit of fun of my own but I don't
enjoy having to pick bits of black epoxy out of my face.

And I know that a 1000V supply capable of delivering 1A is no
joking matter. But since the secondary won't consist of anything
besides the rectifier bridge and the output jack I won't have to do
much there.

BTW, I can't use the filament/sample constellation itself as a
rectifying diode. For one thing this would require some kind of
flyback design, and, more importantly, the anode gets often hotter
than the cathode.

--Daniel

 

[Paul Mathews]

"Unregulated SMPS" probably sounds funny, but this is it: At my lab
we frequently heat up metal samples in vacuum by placing them near
a hot tungsten filament and applying a high voltage, essentially
making the sample the anode of a vacuum diode. This typically
happens at voltages of around 1kV and currents of .5A, but 1kW
total power isn't unheard of.

Now we have some homebuilt units that take care of this. They
consist of a custom-wound mains transformer with a bridge rectifier
for the HV, and the power is regulated via filament temperature.
These beasts work OK, but the transformer makes them heavy and
expensive.

Now we need a few more, and the natural path to follow would of
course be to make them switched-mode -- except that I've never done
an SMPS. I may be wrong, but I think this should nevertheless be
doable since I need no regulation at all; the first version would
just have a pot to adjust the duty cycle. So this thing would
consist of no more than a PWM controller, an IGBT H-bridge, a
transformer and a diode bridge. The output needs to be
short-circuit proof; I'd do that by adding some series inductance
to make the primary dI/dt slow enough to be safely cut off by the
overcurrent trip.

The only tricky part that I'm aware of is the design of the gate
drive circuit -- it mustn't ring and it must be quick to reduce
heat dissipation and avoid cross conduction. The transformer, of
course, is a critical part in any SMPS design, but I figure that a
poorly wound xformer can't have any ill side effects except
having too little output power due to stray inductance.

Ah yes, and an 1kW SMPS running from a single-phase 230V line needs
PFC and soft-start. Will have to read up about that.

Should I do it? Maybe it's a bit stupid, seeing that only 5 of
these things are needed at the moment and there exists a design
that works and can be assembled and tested by the techie who
originally designed it (he wouldn't touch a switcher with a 10-foot
pole). However I'd like to have a bit of fun of my own but I don't
enjoy having to pick bits of black epoxy out of my face.

And I know that a 1000V supply capable of delivering 1A is no
joking matter. But since the secondary won't consist of anything
besides the rectifier bridge and the output jack I won't have to do
much there.

BTW, I can't use the filament/sample constellation itself as a
rectifying diode. For one thing this would require some kind of
flyback design, and, more importantly, the anode gets often hotter
than the cathode.

--Daniel

You'll probably get lots more detailed responses than this one, but
here are a few questions and pieces of advice:

1) Do you need isolation? If not, you could consider using a boost
converter with 1000 V ouput as a single stage PFC/PWM. There are also
PFC topologies that provide isolation. Control loop BW is always rather
poor, though.

2) Full-bridge output SMPS poses extra challenges keeping DC out of the
transformer. Be careful, or use a series capacitor right from the
start. Don't believe the app notes that say that current mode control
prevents DC imbalance (flux walking). That promise is only realized
under controlled conditions.

3) Push-pull topology requires a larger transformer, but circuit is
less complex.

4) There is much literature available on gate drivers. For a project
like yours, simply choose one of the many good integrated gate drivers.

5) The most important design element is layout. Study SMPS layout
carefully. Where di/dt is high, no inductance is too small to be
ignored. Where dv/dt is high, same is true for capacitance.

6) You will be needing some high voltage differential probes.

7) Get ready to make some smoke.

Paul Mathews

Paul Mathews

3)

 

[Mike Harrison]

"Unregulated SMPS" probably sounds funny, but this is it: At my lab
we frequently heat up metal samples in vacuum by placing them near
a hot tungsten filament and applying a high voltage, essentially
making the sample the anode of a vacuum diode. This typically
happens at voltages of around 1kV and currents of .5A, but 1kW
total power isn't unheard of.

Now we have some homebuilt units that take care of this. They
consist of a custom-wound mains transformer with a bridge rectifier
for the HV, and the power is regulated via filament temperature.
These beasts work OK, but the transformer makes them heavy and
expensive.

Now we need a few more, and the natural path to follow would of
course be to make them switched-mode -- except that I've never done
an SMPS. I may be wrong, but I think this should nevertheless be
doable since I need no regulation at all; the first version would
just have a pot to adjust the duty cycle. So this thing would
consist of no more than a PWM controller, an IGBT H-bridge, a
transformer and a diode bridge. The output needs to be
short-circuit proof; I'd do that by adding some series inductance
to make the primary dI/dt slow enough to be safely cut off by the
overcurrent trip.

The only tricky part that I'm aware of is the design of the gate
drive circuit -- it mustn't ring and it must be quick to reduce
heat dissipation and avoid cross conduction. The transformer, of
course, is a critical part in any SMPS design, but I figure that a
poorly wound xformer can't have any ill side effects except
having too little output power due to stray inductance.

Ah yes, and an 1kW SMPS running from a single-phase 230V line needs
PFC and soft-start. Will have to read up about that.

Should I do it? Maybe it's a bit stupid, seeing that only 5 of
these things are needed at the moment and there exists a design
that works and can be assembled and tested by the techie who
originally designed it (he wouldn't touch a switcher with a 10-foot
pole). However I'd like to have a bit of fun of my own but I don't
enjoy having to pick bits of black epoxy out of my face.

And I know that a 1000V supply capable of delivering 1A is no
joking matter. But since the secondary won't consist of anything
besides the rectifier bridge and the output jack I won't have to do
much there.

BTW, I can't use the filament/sample constellation itself as a
rectifying diode. For one thing this would require some kind of
flyback design, and, more importantly, the anode gets often hotter
than the cathode.

--Daniel

This may be of interest :
http://www.stevehv.4hv.org/ccps1.htm

Also, have you looked at microwave oven transformers as a cheaper alternative to a custom unit.

 

[JoeBloe]

3) Push-pull topology requires a larger transformer, but circuit is
less complex.

4) There is much literature available on gate drivers. For a project
like yours, simply choose one of the many good integrated gate drivers.

Get an audio amp, class D at the expected wattage capacity plus a
good margin. Pump it at 17 plus kHz. Get a Large "E-I" style ferrite
transformer core and bobbin. Wind a transformer with a low turns
count of SPC teflon coated wire at like 18 Gauge on top of your High
strength, high temp, mag wire secondary.

You will be making a step up so you can feed a multiple stage
voltage multiplier. This is where you will get you isolation (vacuum
impregnate with varnish US HV Xfmr tapes). Then you need to decide how
much energy you need to store to do the job, but not store so much
that failures become more catastrophic, and costs inflate to make it
do more than is needed.

Sometimes the energy in the multiplier stages is enough and extra
storage at the end isn't needed. It all depends on the desired
consumption rate.

 

[Haude Daniel]

2) Full-bridge output SMPS poses extra challenges keeping DC out of the
transformer. Be careful, or use a series capacitor right from the
start. Don't believe the app notes that say that current mode control
prevents DC imbalance (flux walking). That promise is only realized
under controlled conditions.

Yeah, I can't imagine a current trip being fast enough to limit the
inrush into what's practically a sudden dead short.

3) Push-pull topology requires a larger transformer, but circuit is
less complex.

That's why I wanto to go PP.

4) There is much literature available on gate drivers. For a project
like yours, simply choose one of the many good integrated gate drivers.

5) The most important design element is layout. Study SMPS layout
carefully. Where di/dt is high, no inductance is too small to be
ignored. Where dv/dt is high, same is true for capacitance.

Sure. I've been reading quite a lot on the subject, out of interest,
and that's what actually got me started on this idea. This is one
of the rare opportunities where it makes sense to design a custom
SMPS, and a simple one to boot.

--Daniel

 

[Tony Williams]

Sure. I've been reading quite a lot on the subject, out of
interest, and that's what actually got me started on this idea.
This is one of the rare opportunities where it makes sense to
design a custom SMPS, and a simple one to boot.

Your requirements are similar to those of a microwave oven.
Use whatever they use, and if any microwave oven does have
an SMPS it is going to be the simplest SMPS available.

In fact a search on <microwave oven smps> does seem to
produce some interesting results.