Yet Another Power Supply

[PMOS]

I'm thinking of building a benchtop power supply. (I've been using an old ATX computer supply - which works, kind of, but isn't exactly optimal.)

Please point out anything insane below, if you could.

Background:

I'm an embedded developer, with a background in computer science and Physics. I'm decent at digital stuff, and decent at the underlying theory behind e.g. bandgaps, but analog implementation has always been black magic to me. Given that it's a weak point of mine, I figured I might as well tackle it straight on - but then realized that I didn't have a decent power supply. Hence, this thread.

General parameters:

1. I'm considering my time ~free. Pretty much anything I'm doing here I'll be learning from.
2. Generally speaking, I'm leaning towards lower-level not higher-level. A "use this chip that does everything" design is neat, but I won't learn as much from it, I don't think.
3. Generally speaking, I'm leaning towards analog design over digital design. A brute-force ADC + voltage follower may well be the cheapest solution, but I already have a fair idea about digital design. Not so much analog design.
4. Don't burn the house down.
5. Don't kill myself.
6. I am perfectly aware that building my own power supply is not the easiest way to get a power supply. The easiest method would be to buy one - but then I don't learn as much.

General goals:

1. 2 independent outputs
2. Each adjustable between -12VDC and 12VDC, if possible. (Actually, possibly -15 - 15, for working with op-amps.)
3. Absolute versus relative adjustments, if possible. (In other words, the classic "here's a pot, turn it CW to decrease voltage and CCW to increase voltage" isn't really what I want. I want to be able to easily adjust to e.g. +4.3V.)
4. Separate adjustable current limits, with indicators when current-limited.
5. 1A max output at a minimum.
6. Doesn't die if the output drives it the wrong way.
7. Current limiting should lean towards conservative in terms of transient response. I'd much prefer occasionally having to add a cap to the output if necessary over the output switching on suddenly and the current limiter allowing 10A through for a moment
8. Neutral should actually be neutral. No neutral-side current-limiting, in other words - those don't exactly behave well if you, for instance, have both channels connected to the same circuit.

Sound sane thus far?

Thoughts on how to go about this:

1. I'm thinking of using a standard AC-AC wall-wart transformer as the input. Something enough above 12V peak to allow for regulation headroom, but well below lines (115 here). Probably 3A? This is in the interests of 4 & 5 - means I don't have (as) high voltages to worry about, and means that (hopefully) even if something goes terribly wrong the adapter's fuse will blow. (I.e. none of my circuitry is unfused.)
2. I'm thinking of using encoder decade resistors to adjust output voltages and currents, in the interests of 3. Though, given I haven't been able to find any encoders that both have carries and can have every digit adjusted, I'm not sure how helpful this will actually be. Might be annoying to not be able to e.g. adjust from 3.9 to 4.0V directly.
3. Is there any decent way of doing current limiting that doesn't use a load sense resistor? Or is that the "best" approach?
4. Are there any "standard" architectures for this sort of thing?
5. This will be, of course, connected to a GFI.

I'm thinking "known voltage through adjustable resistor bridge gives adjustable voltage, through buffer amp gives decent-current adjustable voltage, through linear regulator set to keep a constant voltage drop across a variable resistance give current-limited adjustable voltage".

Problems:

1. I doubt I'll find an op-amp capable of 1A at a decent price, so I'll need a transistor (or MOSFET?) follower of some sort - but I'm not aware of any decent ways to do that that cope well with both negative and positive voltages.
2. Ditto, how to get a current-limiting indicator with the classic linear regulator current limiter?
3. I'm worried about potential feedback, as per usual with analog circuitry. In particular, between the two channels.
4. I'm potentially dumping a lot of current across the linear regulator unnecessarily.

Thoughts?

[jeroen79]

Look up schematics/servicemanuals for vintage HP supplies for inspiration.
They have a separate symmetrical supply for the regulators that is tied to one of the outputs.

And go for a single floating supply first.
If you want two channels, build it twice.

What do you mean by -15 to 15 Volt outputs?
Do you seamlessly want to move the output through zero?
Usually a DC supply simply goes from 0 to max and a negative voltage is obtained by treating the + output as the neutral and the - as negative, or the other way around for a positive supply.

[PMOS]

Look up schematics/servicemanuals for vintage HP supplies for inspiration.
They have a separate symmetrical supply for the regulators that is tied to one of the outputs.

I shall do so, thank you. I was not aware they were available.

And go for a single floating supply first.

Hmm. I was trying to stay away from floating sources - I've found they can wander to surprisingly high voltages relative to ground all by themselves. Or is that only a real problem for el cheapo supplies?

Though I suppose I could build a floating source and tie one side optionally to neutral.

If you want two channels, build it twice.

So there's no real savings between multiple channels of e.g. shared power rails?

What do you mean by -15 to 15 Volt outputs?
Do you seamlessly want to move the output through zero?

Ideally, yes.

Usually a DC supply simply goes from 0 to max and a negative voltage is obtained by treating the + output as the neutral and the - as negative, or the other way around for a positive supply.

...Which means you have to swap around the leads every time. And also means that you can't sweep a reference cleanly across neutral, as you note.

I mean, there are tricks you can do (set supply 1 to produce e.g. -6V relative to neutral, set supply 2 to produce ~+6V relative to supply 1's -6V rail), but they are all iffy at best.

[rstofer]

A neat little tracking supply

https://www.eevblog.com/forum/projects/is-it-me-or-this-is-wrong/

Follow the links through to the datasheet and you can see other schematics.

You should probably have 3.3V, and 5V an for logic projects d the tracking pair for op amps.

Look up 7895,7905,7812,7912 and LD1117V33C

[PMOS]

That is a neat little circuit. May wish to augment / replace R1 or R5 with a trimmer though, as it mentions.

I'm not necessarily seeing how that's particularly relevant though. Aside from general "the LT150A exists", that is.

I already have adapters that put out fixed voltages. Sure, it may be mildly handy to have only a single adjustment when working with op-amps, but that circuit is too specific for me, I think - it's tied down to one output being the negation of the other. Even if you drop pots in place of R1 & R5 you still won't be able to adjust them across 0V.

[Kleinstein]

There is not much savings in having two supplies using the same rails. The advantage of using separate units is that one can have them in series or in parallel or fully floating if needed.

The problem with a high floating voltage again ground exists with switched mode supplies made for a 2 wire outlet. A classical transformer based one has no such trouble.

It is possible to build (and buy) a supply that goes positive to negative on the same output, but this is a more complicated design if you need high power. A simple, low end version is just one polarity. For low power like the mentioned +-15 V and 1 A, one could get away with a power OPs (e.g. L165), though current limiting is difficult in this case. Doing it all the way by hand is much like building an class AB audio amplifier - so possible and a lot to learn.

So for the start, I would keep it simple with a single supply and one polarity only.