Benchtop Power Supply Options

Problem is, it will also follow the significant ripple on its collector.

Nah, that's a capacitance multiplier. As long as you don't let it saturate, it's the bomb for supply rejection.

Cheers

Phil Hobbs

That still does not remove the ripple from the caps, what the F is wrong with you? I understand Phil better now.

Jamie

Quick version: I am almost certain John is talking about the DC *on C1 and C2*, not the DC at the output of the supply.

Long version...

There will *always* be some ripple "before" the LM317 - across C1 and C2 in the diagram you posted.* If you had a transformer with a 24 V AC secondary, and you were drawing 1 A (DC) from the output of the power supply, and you put an oscilloscope probe across C2, you'd see a DC voltage varying between about 30 V and 32 V DC. The variation would be at 120 Hz. You can reduce the amount of this ripple by making C1 and C2 bigger, but you can never get it to go away completely.

(* From "Getting Started in Electronics" by Forrest Mims, I think.)

There will be *much less* ripple "after" the LM317 - across C3 - as long as you have R1 adjusted to give an output voltage less than about 27 V (in this example). If you put an oscilloscope probe across C3, you'd see a DC voltage varying between (say) 27.000 V and 27.005 V. You can reduce this ripple a little by putting another capacitor across R1 - see the LM317 data sheet. A lot of things you would run from this power supply won't care about this. Audio stuff might care a little (you might hear a 120 Hz hum in the audio), depending on the audio voltage levels you are using.

The main reason you have to care about the ripple "before" the LM317 is that it effectively sets the highest output voltage you can get from the supply. There will always be a couple of volts of drop "through" the LM317 - in other words, even if you set the ADJ pin for maximum output, the OUT pin will always be a couple of volts less than the IN pin.

If the voltage on C1 and C2 is rippling between 30 V and 32 V, and you set R1 for 27 V output or less, then there is always at least 3 V available to lose in the LM317, and the output will be stable at 27 V. If you tried to turn up R1 to get 29 V on the output, there would only be between 1 and 3 V available to lose in the LM317. 1 V isn't enough for the LM317, so part of the time, you wouldn't get the full 29 V on the output.

There is an article by Don Lancaster on how to pick the filter capacitor size on PDF page 104 of

Matt Roberds

So far I don't recall ever having seen a PTC in front of a transformer PSU.

Over the past year or so my web trawling has dredged up all manner of weird and wonderful arrangements - most often a relay to switch a resistor in/out of the primary feed, the variety is in the circuits controlling the relay.

OOPS! - typo - that should read NTC.

Sure, as long as the input ripple is small, a few tenths p-p. But the suggestion was to put the c-multiplier between the bridge+filter caps and the main linear regulator, where one might expect amps of current and volts of ripple.

The full solution involves the copper loss and the leakage inductance of the transformer, the diode characteristics, the capacitance and ESR of the filter capacitors, and maybe the impedance of the AC power line. All that is easier to Spice, or just to test, than to calculate.

Don's math is pessimistic, because it assumes that the cap is charged instantaneously and discharges for a full half cycle. A real sine wave is mostly flat on top, so the cap doesn't discharge for 8.333 msec.

Lately I've been using shunt regulators on the base string of cap multipliers. AFAICT the best is the Exar SPX431A. You can adjust that to lop off the ripple if you like, though of course it isn't as efficient as using a huge cap.

Cheers

Phil Hobbs

That's interesting, the Exar is "best" because it has the least noise to start with?

As you've said, the reference/regulator can be the noisiest part of a design. I've got an LT3080 in an instrument. (cap multiplier after it.) And I've regretted it. Much better to just make your own, (reference->opamp->transistor)

George H.

The Exar is quieter and has a lower minimum cathode current.

Cheers

Phil Hobbs

That circuit necessarily increases Vce, which has additional benefits.

Lots of circuits have good supply rejection at low frequencies and need help at high frequencies. In those cases, just RC power filtering works well. Polymer aluminum caps are great for that.

I've also done closed-loop opamp regulators with a huge RC tau on the output. That lops off the HF part of the noise.

Who suggested that then - I never even mentioned the bridge and filter caps.

Nice, I like that you've got the cap multiplier inside the feedback path.

Thanks, George H.

It's possible to get lower (SPX431A wants over 400 uA, and TLV431 wants over 50 uA) but there's some cathode-voltage-range issues, too.

Have you considered direct shunt regulation?