power supply discharge

A DC-coupled audio amplifier chip might work as a fully-controllable bi-directional power supply if your current and voltage requirements were fairly modest. They have the advantage of being relatively cheap, well-protected and very fast (by power supply standards). Some of them have the tab at input earth voltage, so they don't require isolation from the heat sink.

Be easy enough to sink current when the output voltage exceeds the set point by more than, say, 0.1V.

But there has to be a limit - connect the PS to your fully charged car battery, and set the PS to 10V, and you're not going to see a 10V output any time soon.

Sylvia.

Right, the load could be a battery. The user could set the output voltage high with some current limit to charge the battery (or some giant capacitor), and then set the voltage low.

What's complicating my life is that the regulator is a half-bridge switcher that, in that case, becomes a boost converter, pumping backwards into my bulk power supply, which could then blow up. Or if the control loop cranks the PWM duty cycle down to zero in a futile attempt to reduce the output voltage, it soon shorts the battery.

Or some yahoo could connect the battery backwards.

This is actually a nice multidimensional dilemma. I'll be using the DRV8962 quad half-bridge, which also constrains things.

As usual with data sheets, it isn't entirely clear.

Unfortunately, it has to be a switching regulator.

Years ago there were power supplies with a linear regulator fed from a thyristor chopper, the voltage to the linear stage being held just above the required output. This gave all the benefits of a linear power supply without the dissipation.

Perhaps you could use the switcher to power the audio amplifier?

...or were you thinking of using the switcher in reverse to feed excess power back into the supply? (That's where dynamotors and rotary converters scored.)

How about a nice diode in series, inside the FB loop, with the pulldown on the load side?

(You thought of that already.) ;)

Cheers

Phil Hobbs

Yes. I plan to offer a version of the product as a stepper/torque motor driver, in which case two half-bridges become one full bridge, and each half-bridge then has to push current in both directons. The current measurement has to be bidirectional too.

I think I can use a polyfuse to handle the backwards yahoo case.

The klutzy answer to the reverse-pump issue is to measure the prime 48 volts power supply voltage and, if it gets pumped above 55 or something, shut down the TI quad switcher for one second.

A depletion fet somewhere could discharge an external cap if I shut the switcher down.

I was just talking to a guy who says that big capacitor bank discharging is a common requirement. Sometimes they use a resistor on a stick.

Switching audio amplifiers have been around for decades now. I've never used any of them - discrete MOSFETs were cheaper than integrated parts when I last looked, but that was a while ago.

An even more extreme example of two PS connected together with different set points shows that no general solution exists, even in theory.

So it's down to requirements and specifications.

The reversed polarity battery case is I think usually handled with a diode and fuse. The controller can then email a manager pointing out that someone needs to be fired.

Sylvia.

Twiddling the adjustment knob on a bench supply doesn't represent a dramatic change - and most adjustible supplies don't load their output terminals with a lot of capacitance.

DC coupled programable supplies, or bipolar programmable supplies are made to drive loads in the first and third quadrants.

There are issues in the second and fourth quadrants, where the supply is expected to absorb power.

An amplifier driving a pure reactance experiences the same losses as driving a dead short.

RL

and this is a surprise because . . . . ?

RL

My ancient Farnell bench supply has a voltage adjustment pot and moving coil voltmeters. The 'up' speed is much quicker than the 'down' speed.

We always appreciate your valuable insights.

Yes, a channel-channel short is possible, especially when a pair of half-bridges drive a bidirectional motor coil.

I am making those up as we go along, but I'd like to make the product as good as I reasonably can.

The TI quad half-bridge has substrate diodes to ground, so a series polyfuse may handle the reverse yahoo connection. I'll try that. I'll need a gigantic power supply.

I suppose I should buy the worst series inductor that will work, to limit the surge current.

I'll post some schematic scribbles as it goes along.

Don't bother trying to be sarcastic - your language skills aren't up to it.

You usually end up with a better result if you work out what you are trying to do before you start designing it - or in your case, slinging it together. It's called system engineering.

When dim newbies post here, we often have to ask them what they are trying to do before we can work out how they might solve the problem they think they are seeing.

By "worst" do you mean cheapest or smallest?

LTSpice simulations tend to include the kind of information that gets left out of your pencil sketches.

I've measured a few, and got output terminal capacitance of a few hundred to maybe 2000 uF.

People here might measure some random power supplies. I leave them off and connect to a 50 ohm sinewave-output function generator and find the -3 dB point. One could use a square wave and scope the slopes too. Keeping the amplitude low will avoid turning semi junctions on.

A square wave source driving a cap illustrates C, ESR, and ESL on a scope. C-meters don't usually separate the components so trend to lie, especially with big electrolytics.

I don't understand that. An audio amp driving a 1 pF cap or a 1K henry inductor would surely cause less amp losses than a short.

Good bench power supplies should have minimal built in output capacitance; it compromises current limit performance. Anyone wanting supply decoupling caps on their test load is going to have them already at the load side anyhow.

I am a fan of the HP / Harrison Labs supplies of 1960s that have no output caps and are stable.

Some specialist power supplies for large scale battery testing can absorb power and return it to the AC supply.