Combination switch mode and regulated power supply

[darrenb]

Here's an idea that's been going around in my head for a while.  It seems like it would work, and makes sense, but I've never seen it done before.

The problem with switch mode power supplies is that you'll always get some ripple which isn't ideal if you want a really clean supply, say for a lab supply.

The problem with regulated supplies is that they waste all of the excess power that you don't want.  If the gap between the input voltage and the regulated voltage is high then there is a lot of waste and a lot of generated heat that has to be dealt with.

Why not use a switch mode power supply to supply the voltage that you want plus say 2 volts (or whatever the headroom required for the linear regulator is).  Then use a linear regulator to regulate the required volts + 2 down to the required volts?

[T3sl4co1l]

Oddly enough, it's been done.

One downside is attenuation: LDOs typically stink at high frequency PSRR, and there's not much that can be done about it.  2V is enough headroom to do a HDO (like a BJT emitter follower), but collector capacitance is quite high at low voltages, again limiting the performance.

One advantage over a passive filter would be speed; you could move that active circuit a lot faster (~kHz bandwidth?) than a pile of inductors and capacitors, for the same attenuation.

On a practical note, either way, with careful design and few stages (one or maybe two, plus some ferrite beads and such to take off the really high frequency stuff), by the time you're talking microvolts, you're worrying more about what the wires from PSU to circuit will pick up environmentally.  So it's not a big deal either way.

Tim

[darrenb]

Hmmm.  Great answer.

I'm still thinking that because you have a known next stage after the switch mode stage (the linear regulator, instead of whatever the user might connect), you could do enough passive filtering between stages to satisfy the regulators ripple rejection and still end up with reasonable speed and a clean regulated output.

There must be a reason why almost all lab supplies are linear even with their attendant weight and power wastage and I'm guessing that is because in the lab you want to start with a very clean supply and not have to factor in power supply ripples, however small.  In the lab weight and power use shouldn't be a problem but maybe this would be ideal for compact or battery driven supplies?

Anyway, it's just an idea.  I haven't tried any of this and the technical issues you've raised may well be a showstopper.

[T3sl4co1l]

Yeah, not really anything against it, one way or another.  Main thing for lab supplies is, no one cares about efficiency because testing is expensive any way you cut it.  A production test might take higher priority, but lab supplies generally don't find their way into that setting (well, fancy GPIB digital ones for ATE notwithstanding, I guess).

The main thing I'd look for is, big watts -- hundreds to thousands of watts, or high voltages that are difficult to regulate otherwise.  One downside with post-regulation on an application like this: if the supplies get out of sync, the linear regulator still has to handle -- potentially -- the full output voltage, so you can't safely run, say, 200V at 2A with just an LM338.  Short circuit transients would be a big problem, discharging the switcher's filter caps through the regulator -- just dumping it through a 30V TVS for instance isn't going to be very nice to the TVS or the regulator.  On the upside, high power applications are unlikely to demand microvolt noise levels, so a straight switching output, maybe with some better than average filtering, is likely more than adequate.

That said, I've done low power testing on 10kW equipment using a cheap POS bench supply, one of those $300 1kW jobbies.  The common mode noise is so intense, it causes measurement errors (spurious waveform triggering, spikes, peaking, and control loop oddities) in the equipment -- and this is with the case wide open, probes everywhere, looking at the raw, unfiltered inverter waveforms!

Tim

[kizzap]

This is exactly what I am working on at the moment:

Link.

Only designing it to be a 0-30V 0-3A supply, however I have done some calculations and instead of having to dispate 120W+ through a heatsink to get lower voltages, I am managing to only dissipate 12W total, and that is only while drawing the maximum voltage/current output.

-kizzap

[Richard Head]

The problem with switch mode power supplies is that you'll always get some ripple which isn't ideal if you want a really clean supply, say for a lab supply.

There is no real advantage to having a linear post regulator other than very fast transient response it can provide.
A switchmode PSU can be made VERY clean if that is required, but generally it's not required.
If a noisey SMPS is placed in a Faraday cage with bypass capacitors to chassis on inpu and output then the radiated and conducted noise is going to be way down.
Unfortunately this is not normally done with cheap PSU's due to cost but is routinely done for military PSU's due to the tighter specs.

Dick

[tszaboo]

Linear technology made AN-30 application note where Figure 46 describes this. That was in 1989.

[oldway]

http://www.hameg.com/9.18.0.html

This series is based on a classical concept with a mains transformer, high efficiency electronic pre regulators and linear post regulators. This concept yields the high power in the smallest space with the highest efficiency.

This is a switch mode pre regulator (chopper).
Actually, two cost saving options seems to be used:
- linear regulator with transformer taps switching by relays.
- SMPS without linear regulator.