Driving capacitive loads with an H bridge

[markp]

Hi All,

I need to drive capacitor with a triangle wave with no DC across the
capacitor (i.e. a symmetrical bipolar drive but triangular) but it has to be
efficient, i.e some kind of energy retrieval.

Is it possible to use standard H bridge circuits to do this? Does anyone
have any links or app notes?

Thanks!
Mark.

 

[Tim Williams]

Assuming you mean triangle voltage waveform (= square current waveform), you can do this with an H bridge and an arbitrarily large inductor in series with the supply. You will actually have an arc segment of the LC oscillation, so you need a big L to make the frequency low enough that the arc looks straight.

Tim

 

[markp]

I can't really say as I'm under NDA, suffice it to say the load is mostly
capacitive (in the order of a few uF) and musn't have any DC component
across it.

Mark.

 

[markp]

what kind of frequencies, voltages?

class-d audio amplifier?

-Lasse

I can't say much for NDA reasons, but assume >40V AC (true RMS) and a few
hundred Hertz, so there are potentially a few amps flying about in the load
capacitor.

Mark

 

[markp]

The system becomes an LC lowpass filter, so there is in theory a
higher-order filter that makes a better triangle wave. The first step
would be a smaller LC ahead of the main LC, such as to tend to
increase the drive to the output LC during the cycle. Tapped
inductors, or a compensating series transformer, would be interesting
too.

A little saturation could be interesting, too, like the linearity
correction inductor in an old teevee set.

But, as usual, the problem is underspecified.

John

A little more info then, >40V AC (true RMS) across the load capacitor (a few
uF) and at a few hundred Hertz.

I'm wondering whether a single inductor with a half-bridge PWM'd might do. I
could ground one end of the cap and use +/- V DC supplies to provide the
high and low DC to the half-bridge. One problem with that might be the
voltage swings at the PWM from a noise point of view, and like you say might
require another filter prior to the load cap. Thoughts?

Mark.

 

[markp]

Talking about driving a capacitive load with HVDC is pretty easy
when
particulars are few or missing. Driving one with AC would require a
bit
more info. And you say it is in the area of a couple uF?! Sorry, NDA
or
not, you would need to explain more to get viable responses.

You could float the voltage up just enough that it goes negative only
for
a tiny fragment of the waveform. It would still be AC then,
technically.

I'm not looking for a design as such, just a possible architecture. So far
the info you've been given is a triangle waveform across the cap, bipolar so
no DC component, >40V AC(true RMS), a few hundred Hertz and a question
whether an H-bridge architecture could do it. I'm not sure how much more
information you want...

Mark

 

[markp]

A single ended load would require a single ended source - a
half-bridge.

A triangle wave is generated by a constant current, reversing, but
residual DC can only be limited by the accuracy of the modulator.

Driving purely inductive or purely capacitive loads is as efficient as
driving a short circuit, no matter what the drive method.

RL

Come again? I can create a pair of constant current sources with transistors
and resistors and alternately charge and discharge the capacitor with them.
But that's not efficient...

Mark.

 

[legg]

I'm not looking for a design as such, just a possible architecture. So far
the info you've been given is a triangle waveform across the cap, bipolar so
no DC component, >40V AC(true RMS), a few hundred Hertz and a question
whether an H-bridge architecture could do it. I'm not sure how much more
information you want...

Mark

A single ended load would require a single ended source - a
half-bridge.

A triangle wave is generated by a constant current, reversing, but
residual DC can only be limited by the accuracy of the modulator.

Driving purely inductive or purely capacitive loads is as efficient as
driving a short circuit, no matter what the drive method.

RL

 

[markp]

[snip]

I'm not looking for a design as such, just a possible architecture. So far
the info you've been given is a triangle waveform across the cap, bipolar
so
no DC component, >40V AC(true RMS), a few hundred Hertz and a question
whether an H-bridge architecture could do it. I'm not sure how much more
information you want...

Mark

"...but it has to be efficient, i.e some kind of energy retrieval."

What does that mean? Some particulars could get you some suggestions,
vagueness will lose our interest.

...Jim Thompson

Sorry Jim, what I meant was if you drive a triangle wave across a capacitor
then the capacitor will store energy when it's charged up, but when it is
discharged the energy has to be recovered back (retrieved) so it can be used
again in the next cycle. A parallel resonant LC circuit does just this by
shifting the stored energy from the capacitor to the inductor and back
again, so for ideal components no power is needed to sustain oscillation. An
H brigde would do it by temporarily storing energy in the inductor part and
dumping that energy back to the DC supply reservoir.

I actually thought that by stating it was to be efficient was somewhat
equivalent to that, which is why I put 'i.e. some kind of energy retrieval'.
But you're right, it could have been stated better.

Mark.

 

[markp]

Use a generator and flywheel; by shaping the poles of the generator,
you can get the output to be triangle-wave-like. The flywheel
is your energy-storage item, here.

A motor driving the generator/flywheel makes up the energy
losses. Because the generator is a simple winding like
a transformer, your DC zero is assured.

Oh dear! Just to be clear, by energy retrieval I meant the stored energy in
the capacitor when charged needs to be recovered back when discharging so
the cycle can repeat and process is efficient, much like a resonant LC
oscillator but with triangle waves instead of sine waves. See my post to
Jim.

Mark.

 

[markp]

---
Here's six circuits:

On the top, an "H" bridge driving a capacitor, a series resonant
circuit, and a parallel resonant circuit, and on the bottom a bipolar
half-bridge driving the same circuits.

Does it look like any of them will do what you want?

JF

Did you forget to include the link?

 

[markp]

in
message news:[email protected]...

[snip]

I'm not looking for a design as such, just a possible architecture. So
far
the info you've been given is a triangle waveform across the cap,
bipolar
so
no DC component, >40V AC(true RMS), a few hundred Hertz and a question
whether an H-bridge architecture could do it. I'm not sure how much more
information you want...

Mark


"...but it has to be efficient, i.e some kind of energy retrieval."

What does that mean? Some particulars could get you some suggestions,
vagueness will lose our interest.

...Jim Thompson

Sorry Jim, what I meant was if you drive a triangle wave across a
capacitor
then the capacitor will store energy when it's charged up, but when it is
discharged the energy has to be recovered back (retrieved) so it can be
used
again in the next cycle. A parallel resonant LC circuit does just this by
shifting the stored energy from the capacitor to the inductor and back
again, so for ideal components no power is needed to sustain oscillation.
An
H brigde would do it by temporarily storing energy in the inductor part
and
dumping that energy back to the DC supply reservoir.

But "dumping" an inductor into a capacitor gives you a sinusoid, NOT a
triangle wave.

In my sentence "A parallel resonant LC circuit does just this by shifting
the stored energy from the capacitor to the inductor and back again", the
'just this' bit refers to energy transfer, it was not specific to any
waveform. The first sentence gave an example of a waveform that charges up a
capacitor and then discharges it, as would any, including sinusoids.

You'd need some sort of idealized current splitter, or maybe one of
John "The Bloviator" Larkin's non-charge conservation to do it.


What exactly are you trying to do? Something that can really work, or
an "idealism" ?

...Jim Thompson

I do have very good reasons indeed for wanting a triangle wave and at some
point I'll need to make a real unit to test. However I cannot discuss the
actual application, so unfortunately (however much I'd like to) I can't
elaborate on it. I wish I could because I'm sure you and others may have
ideas.

Mark.

 

[markp]

in
message On Thu, 22 Jul 2010 10:02:23 +0100, "markp" <[email protected]>
wrote:


"Jim Thompson" <[email protected]>
wrote
in
message On Wed, 21 Jul 2010 17:13:23 +0100, "markp" <[email protected]>
wrote:

[snip]

I'm not looking for a design as such, just a possible architecture.
So
far
the info you've been given is a triangle waveform across the cap,
bipolar
so
no DC component, >40V AC(true RMS), a few hundred Hertz and a
question
whether an H-bridge architecture could do it. I'm not sure how much
more
information you want...

Mark


"...but it has to be efficient, i.e some kind of energy retrieval."

What does that mean? Some particulars could get you some
suggestions,
vagueness will lose our interest.

...Jim Thompson

Sorry Jim, what I meant was if you drive a triangle wave across a
capacitor
then the capacitor will store energy when it's charged up, but when it
is
discharged the energy has to be recovered back (retrieved) so it can be
used
again in the next cycle. A parallel resonant LC circuit does just this
by
shifting the stored energy from the capacitor to the inductor and back
again, so for ideal components no power is needed to sustain
oscillation.
An
H brigde would do it by temporarily storing energy in the inductor part
and
dumping that energy back to the DC supply reservoir.

But "dumping" an inductor into a capacitor gives you a sinusoid, NOT a
triangle wave.

In my sentence "A parallel resonant LC circuit does just this by shifting
the stored energy from the capacitor to the inductor and back again", the
'just this' bit refers to energy transfer, it was not specific to any
waveform. The first sentence gave an example of a waveform that charges
up a
capacitor and then discharges it, as would any, including sinusoids.


You'd need some sort of idealized current splitter, or maybe one of
John "The Bloviator" Larkin's non-charge conservation to do it.



I actually thought that by stating it was to be efficient was somewhat
equivalent to that, which is why I put 'i.e. some kind of energy
retrieval'.
But you're right, it could have been stated better.

Mark.


What exactly are you trying to do? Something that can really work, or
an "idealism" ?


...Jim Thompson

I do have very good reasons indeed for wanting a triangle wave and at
some
point I'll need to make a real unit to test. However I cannot discuss the
actual application, so unfortunately (however much I'd like to) I can't
elaborate on it. I wish I could because I'm sure you and others may have
ideas.

Mark.

Send me an NDA and then some real information.

I'm not into guessing.

I can do a Larkinesque ideal machine that will run forever, and take
no energy to work ;-) (And produce a triangle wave

...Jim Thompson

Like this...

http://analog-innovations.com/SED/All_You_Need_Is_A_Perfect_Current_Mirror.pdf

...Jim Thompson

Brilliant! Why didn't I think of that? Shouldn't cost much to make. All I
need to do is find out where I can get component F1 and a 1 picoohm
resistor. The latter part might require a cryogenic chamber and some liquid
helium.

I think you have far too much time on your hands sir

 

[Grant]

Neither is any other method of driving a short circuit.

Given a large enough and perfect enough inductor, and ideal switches,
a short or a capacitor can be driven efficiently by reversing the
inductor's polarity at the waveform peak. That would require 4
switches.

Given imperfect and realistically sized components, a half bridge can
reverse it's output inductor current in a finite time period while
supplying approximately constant current of the correct polarity, with
a modest ripple component and reasonable losses.

Wouldn't it be easier to run the H bridge as switching current
direction only, with current through the bridge controlled by a
separate switching regulator?

Saves talking about ginormous inductors, for starters. Might
even be buildable

Grant.

 

[legg]

Come again? I can create a pair of constant current sources with transistors
and resistors and alternately charge and discharge the capacitor with them.
But that's not efficient...

Mark.

Neither is any other method of driving a short circuit.

Given a large enough and perfect enough inductor, and ideal switches,
a short or a capacitor can be driven efficiently by reversing the
inductor's polarity at the waveform peak. That would require 4
switches.

Given imperfect and realistically sized components, a half bridge can
reverse it's output inductor current in a finite time period while
supplying approximately constant current of the correct polarity, with
a modest ripple component and reasonable losses.

RL

 

[Grant]

That would meet the "shape" requirement, but I still ponder what does
"...it has to be efficient, i.e. some kind of energy retrieval" mean?

Dunno, may be this is a piezoelectric thingy that wants to squish
the current back out when being relaxed at a controlled rate?

Rotate the bridge 90' so the capacitor voltage see-saws and the
charge doesn't fall out? ;^)

* You'd need some kind of loop to keep it "centered" also.

Yeah, that too, unless that thing is taken rail to rail, self
centering?

Grant.

 

[markp]

---
Nope, the circuit list:-(


Version 4
SHEET 1 960 680
WIRE -272 -224 -336 -224
WIRE -144 -224 -208 -224
WIRE 16 -224 0 -224
WIRE 112 -224 96 -224
WIRE 192 -224 176 -224
WIRE 384 -224 384 -320
WIRE 384 -224 336 -224
WIRE 400 -224 384 -224
WIRE 464 -224 464 -320
WIRE 528 -224 464 -224
WIRE -336 -176 -336 -224
WIRE -144 -176 -144 -224
WIRE 0 -176 0 -224
WIRE 192 -176 192 -224
WIRE 336 -176 336 -224
WIRE 528 -176 528 -224
WIRE -336 -48 -336 -96
WIRE -144 -48 -144 -96
WIRE -144 -48 -336 -48
WIRE 0 -48 0 -96
WIRE 0 -48 -144 -48
WIRE 192 -48 192 -96
WIRE 192 -48 0 -48
WIRE 336 -48 336 -96
WIRE 336 -48 192 -48
WIRE 528 -48 528 -96
WIRE 528 -48 336 -48
WIRE -336 16 -336 -48
WIRE -160 96 -336 96
WIRE 80 96 16 96
WIRE 192 96 160 96
WIRE 480 96 352 96
WIRE 544 96 480 96
WIRE -336 144 -336 96
WIRE 16 144 16 96
WIRE 352 144 352 96
WIRE 480 144 480 96
WIRE -160 160 -160 96
WIRE 192 160 192 96
WIRE 544 160 544 96
WIRE -336 272 -336 224
WIRE -160 272 -160 224
WIRE -160 272 -336 272
WIRE 16 272 16 224
WIRE 16 272 -160 272
WIRE 192 272 192 224
WIRE 192 272 16 272
WIRE 352 272 352 224
WIRE 352 272 192 272
WIRE 480 272 480 224
WIRE 480 272 352 272
WIRE 544 272 544 224
WIRE 544 272 480 272
WIRE -336 352 -336 272
FLAG -336 352 0
FLAG -336 16 0
SYMBOL voltage 16 128 R0
WINDOW 3 24 104 Invisible 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR Value PULSE(-35 35 0 .005 .005 0 .01)
SYMATTR InstName V1
SYMBOL ind 496 240 R180
WINDOW 0 36 80 Left 0
WINDOW 3 36 40 Left 0
SYMATTR InstName L1
SYMATTR Value .845
SYMBOL ind 64 112 R270
WINDOW 0 32 56 VTop 0
WINDOW 3 5 56 VBottom 0
SYMATTR InstName L2
SYMATTR Value .845
SYMBOL cap 176 160 R0
SYMATTR InstName C2
SYMATTR Value 3e-6
SYMBOL cap 528 160 R0
SYMATTR InstName C1
SYMATTR Value 3e-6
SYMBOL voltage 352 128 R0
WINDOW 3 24 104 Invisible 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR Value PULSE(-35 35 0 .005 .005 0 .01)
SYMATTR InstName V2
SYMBOL voltage -336 128 R0
WINDOW 3 24 104 Invisible 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR Value PULSE(-35 35 0 .005 .005 0 .01)
SYMATTR InstName V3
SYMBOL cap -176 160 R0
SYMATTR InstName C3
SYMATTR Value 3e-6
SYMBOL cap -208 -240 R90
WINDOW 0 0 32 VBottom 0
WINDOW 3 32 32 VTop 0
SYMATTR InstName C4
SYMATTR Value 3e-6
SYMBOL voltage -336 -192 R0
WINDOW 0 -42 0 Left 0
WINDOW 3 24 104 Invisible 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V4
SYMATTR Value PULSE(-35 35 0 .005 .005 0 .01)
SYMBOL voltage -144 -80 R180
WINDOW 0 -43 116 Left 0
WINDOW 3 24 104 Invisible 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V5
SYMATTR Value PULSE(-35 35 0 .005 .005 0 .01)
SYMBOL cap 112 -208 R270
WINDOW 0 32 32 VTop 0
WINDOW 3 0 32 VBottom 0
SYMATTR InstName C5
SYMATTR Value 3e-6
SYMBOL voltage 0 -192 R0
WINDOW 0 -42 -2 Left 0
WINDOW 3 24 104 Invisible 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V6
SYMATTR Value PULSE(-35 35 0 .005 .005 0 .01)
SYMBOL voltage 192 -80 R180
WINDOW 0 -44 115 Left 0
WINDOW 3 24 104 Invisible 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V7
SYMATTR Value PULSE(-35 35 0 .005 .005 0 .01)
SYMBOL cap 464 -240 R90
WINDOW 0 0 32 VBottom 0
WINDOW 3 32 32 VTop 0
SYMATTR InstName C6
SYMATTR Value 3e-6
SYMBOL voltage 336 -192 R0
WINDOW 0 -44 -3 Left 0
WINDOW 3 24 104 Invisible 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V8
SYMATTR Value PULSE(-35 35 0 .005 .005 0 .01)
SYMBOL voltage 528 -80 R180
WINDOW 0 -38 106 Left 0
WINDOW 3 24 104 Invisible 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V9
SYMATTR Value PULSE(-35 35 0 .005 .005 0 .01)
SYMBOL ind 368 -304 R270
WINDOW 0 32 56 VTop 0
WINDOW 3 5 56 VBottom 0
SYMATTR InstName L3
SYMATTR Value .845
SYMBOL ind 0 -208 R270
WINDOW 0 32 56 VTop 0
WINDOW 3 5 56 VBottom 0
SYMATTR InstName L4
SYMATTR Value .845
TEXT -320 304 Left 0 !.tran .1

JF

Thanks for that. Yes, one of these is going to solve it - my preference at
the moment is not actually an H bridge for noise reasons but rather a
resonant type oscillator like the one in the top right of the schematic, if
I can make the waveform triangular. That actually *might* be possible by
using a class D oscillator and forcing additional current in and out of the
drive windings, not sure yet. The question is how the driving sources are
implemented. For example the top right (a capacitor driven by two voltage
sources either side) would need to have a current source as well, and that
current source mustn't be resistive so that the energy used to charge the
capacitor in one direction is recovered from the capacitor and not thrown
away by resistive losses.

Mark.

 

[Fred Bartoli]

Jim Thompson a écrit :

The OP (markp) implies this is a _real_ situation. How do you meet
his requirement "...it has to be efficient, i.e. some kind of energy
retrieval" ??

...Jim Thompson

ROFL!

You'd better not write such nonsense when you intend to blast everybody
that don't fit your taste here...

I take it that you didn't think much when writing this, which anyway
shouldn't be with the posture you choose to display, but even,
understanding this doesn't take much thinking, so?

 

[markp]

---
That _is_ an H bridge; notice that the drivers are two voltage sources
180 degrees out of phase.
---


---
How much power are you willing to lose?

After all, it's not likely that you'll be able to build a "perpetuum
mobile."

As I see it you have few choices if you want to make the waveform
across the capacitor triangular.

Here are two:

1. Use a bipolar resistive current source and drive it with a square
wave.

No good, way to lossy.

2. Change the inductance of the choke with respect to time so its
increasing and decreasing reactance does the current limiting.

Now that's interesting. A saturable core reactor that responds quickly
enough could do it. Now there's I2R losses in doing that, you have to drive
current through a coil, the big question is whether this would actually not
be that lossy.

The question becomes something like this: If you had a certain current
flowing through a saturable core inductor the energy stored in it is
(I^2L)/2. So if the current remains the same, and you change its inductance
by changing the control current, what happens to the stored energy?

Either way is going to cost you power, so what efficiency would you
_like_ to see?

Something like 85% would be nice.

 

[Fred Bartoli]

Jim Thompson a écrit :

Where's the "energy retrieval"?

An H-bridge driven from an inductor doesn't "retrieve".

I see nada of substance in your comments.

Another "Bloviator" ?

...Jim Thompson

2.9uJ loss for 360uJ transfered to the cap and "retrieved" per cycle.
That's 0.8% loss.

How do you call that?


Version 4
SHEET 1 1072 680
WIRE 96 -16 -48 -16
WIRE 336 -16 176 -16
WIRE 448 -16 336 -16
WIRE 576 -16 448 -16
WIRE 336 48 336 -16
WIRE 576 48 576 -16
WIRE 992 48 912 48
WIRE 288 64 272 64
WIRE 640 64 624 64
WIRE 288 112 240 112
WIRE 672 112 624 112
WIRE -48 160 -48 -16
WIRE 912 176 912 48
WIRE 336 192 336 128
WIRE 416 192 336 192
WIRE 576 192 576 128
WIRE 576 192 480 192
WIRE 864 192 576 192
WIRE 336 240 336 192
WIRE 864 240 336 240
WIRE 336 272 336 240
WIRE 576 272 576 192
WIRE 288 288 272 288
WIRE 640 288 624 288
WIRE 64 336 64 304
WIRE 176 336 176 304
WIRE -48 432 -48 240
WIRE 64 432 64 416
WIRE 64 432 -48 432
WIRE 176 432 176 416
WIRE 176 432 64 432
WIRE 240 432 240 112
WIRE 240 432 176 432
WIRE 288 432 288 336
WIRE 288 432 240 432
WIRE 336 432 336 352
WIRE 336 432 288 432
WIRE 576 432 576 352
WIRE 576 432 336 432
WIRE 624 432 624 336
WIRE 624 432 576 432
WIRE 672 432 672 112
WIRE 672 432 624 432
WIRE 912 432 912 256
WIRE 912 432 672 432
WIRE -48 512 -48 432
FLAG -48 512 0
FLAG 64 304 A
FLAG 272 64 A
FLAG 640 288 A
FLAG 176 304 B
FLAG 272 288 B
FLAG 640 64 B
FLAG 992 48 Vcap
FLAG 448 -16 Sup
SYMBOL voltage -48 144 R0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V2
SYMATTR Value 0.022
SYMBOL sw 336 368 M180
WINDOW 0 32 15 Left 0
WINDOW 3 32 44 Left 0
SYMATTR InstName S1
SYMBOL sw 336 144 M180
WINDOW 0 32 15 Left 0
WINDOW 3 32 44 Left 0
SYMATTR InstName S2
SYMBOL voltage 64 320 R0
WINDOW 0 -53 5 Left 0
WINDOW 3 -242 110 Invisible 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V1
SYMATTR Value PULSE(1 0 0 1E-6 1E-6 .004999 .01)
SYMBOL sw 576 144 R180
WINDOW 0 32 15 Left 0
WINDOW 3 32 44 Left 0
SYMATTR InstName S3
SYMBOL sw 576 368 R180
WINDOW 0 32 15 Left 0
WINDOW 3 32 44 Left 0
SYMATTR InstName S4
SYMBOL cap 416 208 R270
WINDOW 0 32 32 VTop 0
WINDOW 3 0 32 VBottom 0
SYMATTR InstName C2
SYMATTR Value 3e-6
SYMBOL ind 80 0 R270
WINDOW 0 32 56 VTop 0
WINDOW 3 5 56 VBottom 0
SYMATTR InstName L2
SYMATTR Value 10
SYMATTR SpiceLine Ipk=12m Rser=1
SYMBOL voltage 176 320 R0
WINDOW 0 -53 5 Left 0
WINDOW 3 -242 110 Invisible 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V3
SYMATTR Value PULSE(0 1 0 1E-6 1E-6 .004999 .01)
SYMBOL e 912 160 R0
SYMATTR InstName E1
SYMATTR Value 1
TEXT -40 480 Left 0 !.model SW SW(Ron=0.1 Roff=1E6 Vt=0.5 Vh=0)
TEXT -32 456 Left 0 !.tran 0 25 24.9 1m


Same apply for RL's half bridge, which is nothing more than an unloaded
synchronous buck.


The real question is that the OP's 40V peak at 1kHz on a 3uF cap is only
10W dissipation. Is it worth the trouble?



BTW, Bloviator yourself.