Buck converter switchable output voltage

[HwAoRrDk]

I'm contemplating what it will take to create a circuit with a DC-DC switching buck converter that has a switchable output voltage. My basic idea is that per output voltage required there will be a separate pair of feedback voltage divider resistors (or maybe only one per on one side of the divider), and that there will be some means to switch in the differing feedback divider according to the desired output voltage - maybe a physical switch, an analogue multiplexer IC, not sure yet.

One thing that occurred to me is that it's probably not a good idea to have the feedback network switched 'live', right? That is, while the buck converter is online and switching. Given the nature of a selection switch (whether physical or electronic) is typically break-before-make, I would imagine that interruption will momentarily cause havoc with the converter's regulation, which may cause unregulated voltage levels on the output, which is a Bad Thing™.

Any suggestions on how to mitigate this?

I thought perhaps there may be some way to momentarily disable the switching regulator (using its enable pin) during switching. In terms of end usage, turning off the output during a change of output voltage is entirely acceptable. I just can't think of a way to do it right now.

[HwAoRrDk]

I've just had a thought after looking at some buck converter datasheets.

Buck converters typically have some kind of output short-circuit or under-voltage protection, where if the feedback pin goes below a certain threshold, the converter is shut off. During a break-before-make switching of the top half of the feedback voltage divider, the bottom resistor will still be in-circuit, which will pull the feedback pin to ground, making it 0V and thus below the protection threshold, causing a shut-down.

Does that sound sensible? Should I take advantage of that?

[Zero999]

It should work. You need to provide more information.

Input and output voltage range? Current?

[dobsonr741]

Assuming 3 output voltages supported, you can have a fix divider for the lowest voltage and parallel the GND connected resistor with one of the two resistors for the higher voltages. Simple BJT or FET can be used to switch the resistor to GND. You’ll have a transient regardless, hopefully better than breaking the feedback loop completely.

If the transient is a concern you need to power down. Then the remaining concern is to discharge to output capacitor. In any case, you’d want a regulator that has specific on/off transient response specified on the data sheet.

[HwAoRrDk]

Input and output voltage range? Current?

Input voltage will be 12V. I'm aiming for 3, maybe 4, output voltages - 5V, 7.5V, 9V, maybe 3.3V. Maximum output load should be no more than 3A.

If the transient is a concern you need to power down. Then the remaining concern is to discharge to output capacitor. In any case, you’d want a regulator that has specific on/off transient response specified on the data sheet.

Oh, yes, the output caps. Hmm, I hadn't thought about that. Would be a problem especially if the output voltage is switched while there is no load connected. I guess I could somehow have a discharge MOSFET switching the output to ground through a low-value current-limiting resistor during the 'off' period while switching voltages. Or maybe just have a load resistor permanently in-circuit on the output - but it would have to be of low enough value to discharge quickly, but not too low to waste too much power.

[dobsonr741]

The output concerns would go away if you have an LDO post-regulator. Analog has a very nice ref design - I would have built one if the parts are not unobtainium at the moment: https://www.analog.com/en/technical-articles/high-performance-portable-dc-bench-power-supply.html

[langwadt]

The output concerns would go away if you have an LDO post-regulator. Analog has a very nice ref design - I would have built one if the parts are not unobtainium at the moment: https://www.analog.com/en/technical-articles/high-performance-portable-dc-bench-power-supply.html

how would an LDO fix the output when switching to a lower voltage? and LDO can't sink current, a sync buck can though most controllers won't

[dobsonr741]

The LDO stage has a very low output buffer cap - 1uF. The 10K load resistor discharges it in 10ms. The LDO obviously handles positive transitions. The ref design meant to be adjusted by a pot - it’s a bench supply. Other benefit is the low ripple, while efficient and no humongous heat sinks.

[T3sl4co1l]

Switched from what, a switch switch? Analog switch(es)?

If digital, just use a DAC -- bias the FB node with a resistor from a DAC, or any other current.  Notice this does not appreciably change the gain from output to FB, so the loop stability and everything is unchanged.  It's just offset by a fixed voltage depending on the resistor values and DAC setting, so settles to a different setpoint.

Note if using a PWMDAC, use two series resistors with the filter cap inbetween.  Or more RCs or an active filter or whatever but anyway with the resistor to FB because it's the current you're after, ultimately.

For switching in general, it's just changing between setpoints.  Things to watch out for include loop gain / compensation (depends on reg type, some have internal R+C compensation back to the FB pin so the impedance matters), transient / settling (if there's any RCs in the feedback network), CM noise or other interference (if the switch has long leads or external connection, since FB is a sensitive control node), uhh chatter/bounce, etc.  Bounce of course just manifests as multiple switch closures in rapid fire, so isn't any different from one edge, but I suppose if you had something rate sensitive that shouldn't be switched too often, that could become a concern.

Tim

[HwAoRrDk]

Switched from what, a switch switch? Analog switch(es)?

My initial idea was to use a dual-/single-throw multi-pole (e.g. 4P) rotary switch, or if the concept requires electronic control, something like a 4051/4052 analogue mux.

Basically, what I want is something where the user can select an output voltage (to power an external load, connected via barrel jack or something similar) and for that selected voltage to persist across power cycles. Ideally I'd also want an 'off' setting as well, and maybe a pass-through 12V setting too. For the latter I was thinking perhaps one switch selection would have some circuitry that disables the buck regulator and turns on a P-MOSFET that sends the 12V straight through.

The idea of a physical switch stays in my mind because once you get into electronic/digital control, the need for persistence of setting means you need some way to retain the choice, which probably means MCUs, EEPROMs, etc, etc. and this is something I want to keep as simple as possible.

Bounce of course just manifests as multiple switch closures in rapid fire, so isn't any different from one edge, but I suppose if you had something rate sensitive that shouldn't be switched too often, that could become a concern.

Ah, yes, I had forgotten about switch bounce too. I am beginning to think just straight-up physically switching the feedback network really isn't going to work well. Or maybe one of those things that "works", but does not produce pleasant conditions for the connected load nor the buck converter itself.