Driving capacitive loads with an H bridge

But, not interesting enough. There are sync circuits for CRT deflection that use a permanent magnet to so alter an inductor and get a sawtooth, but it's not symmetric (and DC zero output was a requirement).

Best solution I see is to resonate (make your C part of an LC tank) at the fundamental, and drive (transformer-coupled, low impedance winding in series with C) according to feedback to 'fill in' the sinewave on the inductor with the difference signal to get a triangle on the C.

I'm still fond of the figured-pole generator solution, but it's not something you could get off the shelf.

far

bipolar

transistors

them.

Same way you spelled it... "loss".

Build me one ;-) ...Jim Thompson

far

bipolar

transistors

them.

That does seem to be happening around here lately.

Maybe you could saw the capacitor open and put some Coulomb Glue on the plates. :-)

Ed

The four switches here

ftp://jjlarkin.lmi.net/Triangle_Cap.JPG

make an h-bridge. The a-a and b-b switch pairs are alternately turned on. So the power source can be connected to the load (the capacitor here) in one direction or the other. Or the two bottom switches could be turned on to short the load, or all four turned off to open it. The switches are usually transistors of some sort, and fully integrated h-bridge chips are common.

H-bridges are commonly used to drive motors and speakers, using pulse-width modulation to control how much drive goes into the load. The power source would ususlly be a voltage, not a current like in my circuit. They allow you to, theoretically, make a 100% efficient amplifier.

This particular circuit can drive a triangle wave into the capacitive load with, theoretically, zero power required after startup. In practise, it would be far more efficient than a linear amplifier.

Hmmm, you can also turn on all four switches in an h-bridge.

John

"John Larkin" wrote in = message news: snipped-for-privacy@4ax.com...

Hard to do for more than a few milivolts, though. For large signals, = you have to use back-to-back MOSFETs and *lots* of floating gate drives.

Incidentially, I recently made a synchronous rectifier with BJTs. PP, = not bridge. It passes up to +/-200mV in either direction.

Tim

--=20 Deep Friar: a very philosophical monk. Website:

John "The Bloviator" Larkin show his work??? Dream on.

Current driven? You could. Maybe to re-zero the drift in the capacitor voltage which is (supposed) to stay "centered".

Of course, from John "The Bloviator" Larkin, you're NEVER going to see a real-world circuit. He only argues with Slowman... someone on his same mental-competency level... ZERO ;-) ...Jim Thompson

to use back-to-back MOSFETs and *lots* of floating gate drives.

bridge. It passes up to +/-200mV in either direction.

This is a cool chip, been around forever.

I used it on a CAMAC module to drive stepper motors in microstep mode, for tuning the JLabs cavities. My customer used it to drive a 48-volt,

John

Which just goes to show what a handy word "theoretical" is. You could do a similar theoretical analysis on a conventional linear driver and demonstrate miserable efficiency and lots of power dissipated. After all, switching regulators are more efficient than linears, in theory and in the real world.

You can run Fred's nice Spice model and see for yourself. In fact, I believe you already have.

Four switches do have 16 states. There are all sorts of possibilities... pumping up my inductor at startup, for example.

John

But then one may consider delivering a constant current into / from a cap as not being efficient, since not a tuned circuit? Can you really have this both ways?

Anyway, we don't know if the 'capacitor' is the load, or something there to swamp it, in order to define and make a known waveform?

Yes.

Yeah, that's to pop the main fuse and protect the load from EMP :)

Grant.

to use back-to-back MOSFETs and *lots* of floating gate drives.

bridge. It passes up to +/-200mV in either direction.

PP, whazzat? Mind blank.

Grant.

Not instantaneously. The inductor is still there.

What we have going on here is mixing in the idea of RESONANT charging with an essentially constant current drive.

Calling it energy conserving is a _big_ stretch.

Maybe _someone_ will draw up a _real_ circuit that works?

...Jim Thompson

Sure. The sim works with no lossy parts at all, just one L and one C and the switches. An actual implementation could have only minor losses.

This *is* a resonant circuit. The switches just time-warp it in big jumps so that we only use the segments of the sine wave that appeal to us. Just skip over the parts you don't like.

John

For a triangle wave, 40V RMS = 138.56V peak-to-peak

I take that as 200Hz

I take that as 2A

That works out to a capacitor value of 36uF

At resonance, you'd need a 17.6mH inductor.

To get relatively straight "triangle" slopes, you'd need an inductor much greater than that value... maybe John Fields' 1 Henry.

Am I in the ball-park?

Or do you need to restate the problem? ...Jim Thompson

It's amazing what a tuned circuit can do, I remember driving big inductive loop (think doorway size) as tuned circuit with a bullet proof RS485 diff. driver chip and a seeing a whopping great signal in the loop, tuned with roughly binary value sequence caps with an 8way dip switch.

Grant.

Push-pull. Instead of driving both ends of the load, as H bridge, or = just one end, as half bridge, you drive a tapped winding from each side. = You can only pull down on the winding from either end, so transformer = action fills in the rest: a pull on one end looks like a push from the = other.

Tim

--=20 Deep Friar: a very philosophical monk. Website:

end, as half bridge, you drive a tapped winding from each side. You can only pull down on the winding from either end, so transformer action fills in the rest: a pull on one end looks like a push from the other.

Yeah, know what push-pull is :)

But PP recently for me is polypropylene film caps.

Rarely see push-pull transformer these days, more common to see half or full bridge drive, to utilise copper all the time instead of half the time?

Though centre-tap output windings are common, to save on expensive diodes, so copper utilisation perhaps a poor argument?

Grant.

It's a lot of fuss to save a few watts. For 2uF, 138.5Vpp and 200Hz I calculate you need a roughly 110mA square-wave drive current.

A resonant inductor would be 0.32H, which is a high inductance, and to insure a linear ramp, rather than a sine wave, you'd need a much higher inductance than that (did you say how much nonlinearity you can tolerate?). What's more, the H-bridge involved must be made from floating bidirectional switches. Ouch. So it appears any full-cycle energy-storage idea is going to be very painful.

OTOH, a +/-110mA class-D chopper current-source drive with +/-70V compliance would be relatively easy, using components created for the high-power audio market. Class D also uses energy storage you know.

But you said you'd not like chopper noise, so what's so bad about less than 8 watts of dissipation in a simple linear circuit?

Agreed!

Look back at...

Message-ID:

From the OP's loose description (he says a few amps, which I took as