negative impedance power supply

Ye olde swinging choke PSUs were a crude version of this.

søndag den 16. juli 2023 kl. 22.24.43 UTC+2 skrev John Larkin:

I seem to remember there was once a fad of making power amplifiers with negative output impedance to compensate for wire and coil resistance

Might be okay if the negative resistance occurred only at low frequencies, didn’t look inductive, and had a very limited voltage range.

If the first rule for warships is to stay afloat, for power supplies it’s to not blow anything up.

It wouldn’t help with wiring inductance, of course.

Phil Hobbs

(Who recently prototyped a 3.3 V regulator that usually runs at about 1A, but has to handle occasional 6 A transients with < 1 us edges, holding excursions to <30 mV. )

They exist and they're often used to drive DC brush motors which are intended to maintain a constant speed in the face of a varying load. The output impedance of the supply is set to negate the DC resistance of the motor windings.

The tricky bit is that if you get the negative impedance wrong, the supply will either under-correct or over-correct for the current demand. Put a load on the motor, and the motor speeds up!

I've seen it on simple non-isolated buck converters.

I can't predict the user loads. If their loads are resistive, it's easy to compensate the wire resistance and still have net positive resistance, which would be stable. If the customer load starts with a big cap, the effective loop resistance could go negative and it will run away if we over-correct the wire resistance. So we'd need to limit the -R bandwidth.

It can be made to work but might be hard to explain to users, how to progam it. Maybe more trouble than it's worth.

Not a great solution, though. If remote sensing is too expensive (wire cost) just do point-of-load regulation instead. Negative resistance is annoying if it hits frequencies where you can oscillate, and almost NEVER do people know what those are.

What's expensive about remote sensing - the connex and harness?

RL

Connector pins, circuits in the power supplies, accomodating connection errors, cable requirements.

Philips had a standard circuit for running their little DC (brushed) motors at constant rpm more or less independent of load, that did depends on setting it up to offer a negative output impedance to the motor, over a limited range of currents.

The feedback arrangement was frequency limited so that it didn't oscillate.

I set one up once, and it worked fine, but we didn't put it into production - we ended up going for a vibrating stirrer which was even more of a pest to drive, but was neater and more compact.

I did one to control the speed of a DC motor. it worked surprisingly well. It was a linear regulator as the motor was small and the energy supply plentiful.

I didn't invent this Philips did a similar thing for a portable record player, I think they used two transistors, I used an op-amp and a transistor.

I haven't, but it's one technique for powering things at the end of very long, lossy lines, think oilwell. Only one conductor plus armour, so remote sensing would need pretty complex telemetry.

As the current drawn increases, so does the supply voltage in an attempt to keep the downhole voltage within some range.

Expensive and vulnerable to shock and high temperatures, to be avoided unless you need telemetry for some other reasons.

If you are going to feed e.g. 5 V any significant load at such distances, you are going to have problems. Why not install a DC/DC converter close to the load and fed the long line with e.g. 24 V ?

It sure looks like their shorting/compensation trick is indeed negative impedance. They must measure the wire resistance and then cancel most of it, but not enough to make the system oscillate.

They almost certainly use a low-resistance shunt to sense output current. Voltage drops in the 10s of millivolts are typical, keeping dissipation down.

There are some nice isolated delta-sigma ADCs around lately, which allows the shunt to be anywhere in the circuit without common-mode hassles. Like ADUM7703. One can play cute tricks, trading off bandwidth against resolution in an FPGA.

There were transatlantic telephone cables with *tube* repeaters every so often. They use the most carefully manufactured and tested tubes in history.

The power supplies were 10s of kilovolts on each end and the amps were in series from a power supply standpoint. Probably had the equivalent of a shunt regulator. That was all described in an old BSTJ.

Don't be silly. All they need to do is keep the compensation slow enough to prevent any lag through the wiring creating an unstable feedback loop.

They can cancel all of the resistance, but they have to roll of the gain of the compensation loop at a low enough frequency to avoid oscillation.

It helps if you know the basic of frequency compensation negative feedback loops. People who claim top do electronic design should understand that.

still have to be fast enough to keep the voltage in check on a load dump

If the load opens and the measured current drops, the power supply will react by dropping its output voltage back to the no-load setpoint value.

I wouldn't expect that my power supply would ever try to take out much wire drop, maybe a volt or so.

In real life, power supplies tend to have wild and weird transient responses.