Inverse Marx generator

Exactly. To say "Charge is always conserved" is absurd. It is conserved in some situations, not in others. The context must be stated exactly.

Charge two identical caps to the same voltage, then connect them in parallel, but with polarities flipped. ALL the charge vanishes.

On the other hand, energy is always conserved.

John

I imagine one can use the word dump. Consider this setup: A charged C1 with switch S1 to the top of grounded inductor L1, and from the inductor another switch S2 to a discharged cap C2. Close switch S1 and the inductor current builds, open S1 when C1 is discharged to 0V and the L1 current has reached a peak. Simultaneously close S2 and L1's current continues to flow, charging C2, until the inductor current stops, then open S2. (One can avoid time-precise switches by using series diodes, but they do have loss.)

So C1's energy is dumped first into L1, and then into C2. If C2 is much smaller than C1, it'll end up with a much higher voltage, and the charge out of C1 will greatly exceed the charge into C2 (same current, shorter time), even though E1 = E2, exactly, at least in principle. It must meet John's definition of "do it efficiently."

I made one of these where C1's starting voltage was 1.2kV, and the inductor's peak current was about 1.5kA. I used large IGBT switch modules with Vce(sat) of 3V, or about 0.3% loss. Pulse cap esr and a litz-wire inductor were responsible for most of the losses, about 5% or so, IIRC.

Overall efficiency, Power-into-load/Power-supplied-to-system ??

Larkin vaguely started his thread with no mention of an inductor whatsoever, then added the inductor and claimed "sloshing" forever.

What-a-pile of BS... use real switches and real inductors and real capacitors.

Real switches, in Larkin's "episode", certainly have losses.

I know that even ideal switches used to dump one capacitor into another exhibit "magical" losses (not magical to me ;-)

I'll have to analyze such switching with an inductor... I'd guess there is indeed energy loss "magically" as well. ...Jim Thompson

If you are around to observe and tell about it.

B A N G !!! Energy finds its way out of the box(es)!

I saw a datasheet recently, can't remember exactly where, for a chip that sure looked like a flying-inductor switcher.

John

Well, Heaven help us, Larkin could "slosh" forever :-)

I do know that connecting a capacitor charged to Vo to a same-valued capacitor, but uncharged... even with an IDEAL switch, 50% of the energy goes bye-bye.

I'm busy now, but I'm guessing starting the ball game, even with an ideal switch, will lose some energy.

I'll analyze it later on. I _can_, you know, and I _will_ ;-) ...Jim Thompson

Even if the tank wiring is superconductive, it seems to me that the alternating current flow is going to result in emission of energy in the form of E/M radiation, no? There's going to be some amount of "radiation resistance", when the wiring (having a non-zero physical size) acts as a small loop antenna and lets out a chirp of RF.

True... but you have to include "radiation resistance" in this, as well as the usual dissipative-into-heat resistance in the wiring.

Elaborating: In the simplified condition of an ideal tank, slosh-slosh-slosh will continue forever... nice to think about ideally, crock-a-shit if you have to close a switch to get it started... I'm betting you lose right then-and-there, just like you do in the cap -> cap situation. ...Jim Thompson

Elaborating/embellishing even further: It's _extraordinarily_ distressing to me that the young bucks lurking here have _no_clue_ how to analyze such situations :-(

What ever happened to a fundamental education in electrical engineering ?:-( ...Jim Thompson

Like the 'pile' in the Oxford bell jar? :-)

The 'clapper' actually.

I said that certain posts were untrue. Which they were.

Cite?

It helps to understand ideal circuits before you consider real circuits. The ideals are the limiting cases. You CAN transfer charge between equal value caps without loss of charge, and you can more generally transfer energy between caps without loss; just use an inductor.

For those who dislike theory, Spice will slosh charge around between two caps for millions of cycles. Try it.

John

...

Some of TI's buck-boost configurations look a bit like what you describe:

kevin

The length of time that you'll deny it!

Why then does it not work with a full and empty battery?

What is the final voltage the battery pair will be at after the same such "transfer".

Also, in both cases, how do you propose to do it without inrush damage to the empty cap/battery and outflow damage to the full cap/battery?

There are losses, because the cap has terminations that are not ideal in nature. There will also be damage sites along the contact face for the cap plate-to-terminal_interface interface. :-)

You cannot do it with ideal caps because the charge current would jump to infinity.

There MUST be resistance in the circuit to limit the charge current.

I wonder if they could make a MEMS "Oxford Electric Bell" and use it as a precision clock or counter even.

Yes. Other examples are TI's TPS63001 and Linear Technology's LTC3534, and several that use external mosfets. We feature the LTC3534 in our Li-Ion cell charger discussion in AoE III chapter 5x. It'd be nice to learn about other similar ICs.

These all keep one of the two series mosfet switches on, so that one side of the inductor is always tied in place, and it doesn't fly. It just changes mode from buck to boost.

If you have veual value caps C1 charged and C2 not, and you connect them with an inductor, you get a ringing oscillation that goes forever. Charge is continually exchanged between the caps. At the half-cycle point, all the charge and all the energy have been transferred from C1 to C2. At that instant, you could disconnect the inductor, or leave it in place and the charge will slosh back into C1.

This is just a simple ringing tank circuit. With ideal components, no energy is lost, at the instant of connection or later. Real circuits, with real losses, can closely approximate this situation. Q=200 is perfectly reasonable.

John