Inverse Marx generator

[Tim Williams]

Let's say you had a really high voltage, low current source. Not very useful. Think nuclear battery, or lightning.

So let's say you use it to charge a stack of caps. Then you rewire the caps in parallel. It's like an inverse Marx generator, or a synchronous C-W multiplier. How would you do it?

Tim

 

[George Jefferson]

Let's say you had a really high voltage, low current source. Not very
useful. Think nuclear battery, or lightning.

So let's say you use it to charge a stack of caps. Then you rewire the
caps in parallel. It's like an inverse Marx generator, or a synchronous
C-W multiplier. How would you do it?

What would be the point?

You still end up with the same power? Why would you go through all the
trouble when you can get 1000x more power from your outlet?

Say you have a 1M van de graff, you get maybe 1uA from it... thats just 1W.
No matter how you configure it you'll never get any more power from it.

Your caps are in series which means the voltage across them is 1/n of the
total voltage. When you wire them up in parallel your new total voltage is
1/n.

e.g., suppose you have 1M voltage and 1000 caps in series. Your new voltage
would then be 1000V and your current would be 1000 times the original.

i.e., your just making a transformer. Any time you have a predetermined
amount of energy you can only ideally transform it into a different
"form"(since energy can neither be created nor destroyed).

 

[Tim Williams]

How would you charge caps, without enough current?

By having current. See above.

Tim

 

[Tim Williams]

That's just a charge pump working in the down direction. The switches
are the problem.

Yup. So we need to make a flying capacitor switcher with >10kV isolation.

I don't think your favorite 400V optoisolators are up to the task

Tim

 

[Graeme Zimmer]

A reverse Van Der Graph Generator ?

A constantly rotating Variable Capacitor with comutator contacts to apply
the high voltage at Min Cap and then bleed off the converted charge at Max
capacitance?

Q=CV

Perhaps a cascade of them in series?

................ Zim

 

[Tim Williams]

It's an exercise in finding utility, then. How about exciting the
plate of a vacuum tube? Then, you can feed a 'suitable' frequency
into the grid and get whopping AC power output from a
transformer on the cathode circuit.

But they don't make tubes that run 500kV, at least that are meant to do any AC. Certainly no NOS radio tubes. And low efficiency, and low bandwidth, and blah, blah.

Tubes in the 10kV range might not be too horrible for doing the series-parallel conversion (sync rect with tubes?), but driving the heaters is a pain. Even big stacks of FETs are appealing at that rate. And even then, you might as well go with 500 or 1000V caps, and more stages, so you get a single transistor per stage without cascoding.

So the question remains, how does one change HV into useful V? Flying capacitors, sure, but can you tackle the isolation?

Tim

 

[Grant]

But they don't make tubes that run 500kV, at least that are meant to do any AC. Certainly no NOS radio tubes. And low efficiency, and low bandwidth, and blah, blah.

Tubes in the 10kV range might not be too horrible for doing the series-parallel conversion (sync rect with tubes?), but driving the heaters is a pain. Even big stacks of FETs are appealing at that rate. And even then, you might as well go with 500 or 1000V caps, and more stages, so you get a single transistor per stage without cascoding.

So the question remains, how does one change HV into useful V? Flying capacitors, sure, but can you tackle the isolation?

Carefully aimed pointed wires? ) Gently drifting the ion flow,


Step-down Van der Graaf upthread sounds good?

Grant.

 

[Tim Williams]

OK, so use a klystron (those DO scale to 500 kV) but your AC
output will be at a possibly inconvenient frequency.
'low efficiency' is not as big an issue as you might think; there
are 23MW klystrons, at that scale (or even at 0.5 kW in a microwave
oven) the efficiency beats out all solid state competitors.

Hmm, at that voltage, heater power is inconsequential, and if it runs at a lazy 500MHz let's say, there are plenty of schottkies around that could convert it, maybe not at great efficiency, but it's something. They don't have 500kV klystrons at AES though. It's one possible solution, if one were to make a startup, and cook their own tubes, and "harness" lightning.

Tim

 

[petrus bitbyter]

"Tim Williams" <[email protected]> schreef in bericht
Let's say you had a really high voltage, low current source. Not very
useful. Think nuclear battery, or lightning.

So let's say you use it to charge a stack of caps. Then you rewire the caps
in parallel. It's like an inverse Marx generator, or a synchronous C-W
multiplier. How would you do it?

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

By far most of the high voltage sources I know about you need a lot more
energy to operate them then you ever can retrieve. Todays nuclear batteries
as used in some spacecraft are low voltage. Only lightning consists of very
high voltages and currents but they are very, very short so the total energy
of a lightning strike is some Ah only. So harvesting this energy is likely
to use more then it yields.

petrus bitbyter

 

[Fred Abse]

A constantly rotating Variable Capacitor with comutator contacts to apply
the high voltage at Min Cap and then bleed off the converted charge at Max
capacitance?

Q=CV

But E=QV, hence, if charge is to be conserved, the system will have to
give up energy somewhere as the voltage is reduced. In practice, the
rotating mechanism would absorb that energy.

Capacitors switched from series to parallel will conserve both charge and
energy.

 

[Fred Abse]

Then it follows that E = C * V^2

which of course it doesn't.

Quite right, put it down to haste.

E=QV/2

It still implies that reducing V by increasing C involves a loss of energy.

 

[Tim Williams]

It still implies that reducing V by increasing C involves a loss of
energy.

So a flying capacitor converter is always 50% efficient?

Cap charging (in terms of conserved charge) is an irreversible process. But just like irreversible thermodynamic processes, if you make the steps small enough, it starts looking reversible.

So the point is, you reduce the cap ripple, so the delta Q is small, and increase the frequency, so the number of transfers is high = more output current.

A solid state switched circuit may not have this advantage. For example, say you drove a wad of 1.5kV MOSFETs with photovoltaic gate drivers (through fiber optics). You will lose a lot through Rds(on) and swirching time (in the ms), so it will have to run slowly, in the 100Hz range. Rds(on), of course, dominates the loss component, but charge is still conserved, so long as the on-period is several RC time constants.

Tim

 

[StickThatInYourPipeAndSmokeIt]

Tim

snipped retardedly formatted text.

72 characters max line length, idiot. Your news client doesn't do
that? Then it is not a news client. Outhouse Unexpressive is what it
is, and what you are for using it... like an idiot... without setting it
up correctly, Mr. Useitasitcomesoutoftheboxtotalretard.

 

[Tim Williams]

You can have two caps, C1 charged and C2 not, and transfer all the
charge from C1 to C2, without loss. In fact, you can slosh the charge
between them, back and forth, forever. Just don't use resistors.

What if you want equal charges on both?

Tim

 

[Fred Abse]

So a flying capacitor converter is always 50% efficient?

My original comment was in reply to a suggestion of using a rotating variable
capacitor, a sort of "reverse electrophorus", where C varies. That will
inevitably lose energy, similar to the electrophorus gaining energy from
the mechanical separation of the plates.

Switching capacitors from series to parallel does not change each individual C,
hence, neglecting switching losses, both charge and energy remain the
same. I assume that is what you mean by a flying capacitor converter.

 

[Fred Abse]

You can have two caps, C1 charged and C2 not, and transfer all the charge
from C1 to C2, without loss. In fact, you can slosh the charge between
them, back and forth, forever. Just don't use resistors.

I like your use of the word "slosh".

 

[Tim Williams]

Easy.

Without resistors? Prove it ;-)

Note: no dangling currents. Inductors carry charge, too, so that wouldn't conserve it very well.

Tim

 

[Tim Williams]

Switching capacitors from series to parallel does not change each individual C,
hence, neglecting switching losses, both charge and energy remain the
same. I assume that is what you mean by a flying capacitor converter.

What I meant was charging a cap with another cap. Without tricky quasi-resonant or inverter circuits, you inevitably have to do this (under the presumption that a constant DC output is desired), and this inevitably leads to loss. But the 50% loss only occurs to the *change* in energy, so if you make this change an arbitrarily small fraction of the supply voltage, efficiency can be quite high. Hence why things like MAX232 can be ~95% efficient despite pumping caps into caps.

When you mentioned a mechanical method, I envisioned a stack of capacitors charged by the high-voltage source, then a rotor to ferry a bit of charge at a time from the stack to a reservoir. The rotor would be equipped with capacitors, and contacts would be present on each side, so that the rotor is charged by the HV stack on one side, then connected in parallel on the other side to deliver its charge. The rotor has to be big enough, and insulating, so that arc-over doesn't occur at either end of the rotor. The capacitors on the rotor have to be big enough to deliver a useful amount of charge, enough times per second, to meet the design current and efficiency specs.

Of course, no standing capacitor chain need be provided; the rotor can simply mesh with series-connected contacts, providing all the capacitance itself. Likewise, two or more rotors could be used, in make-before-break mode, to eliminate the DC link capacitor. YMMV; a standing cap bank would be wise for lightning collection, but unnecessary for experiments (in either direction, step-up or step-down). Three rotors in make-before-break would be quite suitable for supplying a conventional (inductor based) converter, transforming, say, 1-10kV into 1.5V or 12V or 160V, etc.

OTOH, when a capacitance is changed, work is performed. An electrophorus works by applying force to seperate charges, increasing the voltage. If capacitance falls linearly, voltage rises linearly, but energy rises as voltage squared, so the energy rises linearly. The difference comes from the work input, which by hand, feels negligible against a 100g electrophorus. This might be confusing to perpetual motion types, who are fond of electric or magnetic devices with forces so weak, they seem to move of their own accord.

Tim

 

[Tim Williams]

Connect an inductor across C1 until you've bled it down to half its
charge. Now connect that inductor to C2 and charge it up to the same
charge as C1 has. Now disconnect the inductor. If you keep the L
shorted, you can save the residual energy for reuse later.

^ ^ ^ ^
Ha, so charge wasn't conserved after all. See? ;-)

Tim

 

[Fred Abse]

When you mentioned a mechanical method, I envisioned a stack of
capacitors charged by the high-voltage source, then a rotor to ferry a
bit of charge at a time from the stack to a reservoir.

The original suggestion came from Graeme Zimmer. He said "A constantly
rotating Variable Capacitor with commutator contacts to apply the high
voltage at Min Cap and then bleed off the converted charge at Max
capacitance?"

Not the same thing at all.