Challenge : a buck converter providing split rail

[dannyayers]

I was just watching a Great Scott! video*, got reminded of frustration I had recently when wanting to use op amps alongside a microcontroller. A few years ago I'd have started with a split-rail supply with linear regulators suitable for the op amps and used a 7805 or whatever to drop to logic chip levels. But nowadays the mcu takes centre-stage (and there is a convenient power source over USB).

I wanted to power an ESP32 plus an ADC plus a bunch of analog gain/filter stages, audio frequency from USB or worst case a 9v wall wart. I have accumulated a drawer full of little buck/boost converters, but no combination of what I had gave an nice solution. Ended up boosting to around 28v, then using a resistor divider for 0v. Inelegant and adding a complication. I've had stability issues in the past getting a 0v this way, case grounding and suchlike can also be an issue.

I've only skimmed the catalogues but the only suitable modules available seemed to have a ridiculous dollar price compared to single-rail designs.

So here's the challenge : using standard inexpensive components, design a circuit that would take say 5-12v as input (ideally also from 3.3v if viable without too much extra effort)  and give +/-12v ... +/-15v output. Not much current needed, say 100mA (though more would be nice to have). Requires good stability & reasonably low noise (random or switching) in the audio range. High efficiency not a priority, but obviously nice to have.
 
Maybe there's a good solution I'm not aware of, but nothing sprang out. The irritation factor is such that I'll no doubt have a go myself (even though I've little experience in this corner), but I thought it'd be be good to put here as a fun design challenge, see how folks might approach it. Is far from niche - data acquisition, signal generation, lots of audio/music stuff. While audio freq analog processing can be done with single-rail op amps etc, it usually comes at a cost somewhere else in the specs (or price).

Cheers,
Danny.

http://hyperdata.it


*

[Benta]

I don't watch YT videos about electronics. Too much ignorance there.

But provided that your positive supply rail has to supply significantly more current than the negative rail (this is usually the case), piggy-backing an additional winding on a buck converter inductor (like a transformer secondary) is certainly a possibility.
Regulation is not very good, though.

Better would be a flyback or push-pull configuration, perhaps even a Royer converter; all with centre-tapped secondaries.

[langwadt]

you can't make 12V from 5V with a buck, you'll need something like sepic or boost
something like:

https://static5.arrow.com/pdfs/2011/12/6/2/18/29/399/ltc_/manual/1.25mhz_lt1961_dec02_mag_fig.7.jpg

[mariush]

For home use as a hobbyist, it would most likely be much much cheaper to just buy a transformer with two 12v ac secondary windings.
Then simply use a bridge rectifier, a couple of capacitors and optionally a LDO on each winding to set the output voltage.

Otherwise, negative voltages can be easily produced with jellybean components like 33063 / 34063 , or if the current amount is not big (ex 20-50 mA), look for charge pump chips which can double the input voltage or invert it.

[artag]

I've got some of these in my box of ebay goodies : https://www.ebay.co.uk/itm/252996434711
Haven't tried using them yet, but I'm curious as to why there is a single winding on the inductor and 3 diodes. Unfortunately they sanded the chip markings off. 120mV noise.

also these : https://www.ebay.co.uk/itm/254604745277
That's called DD1718PA and seems to have an XL6007 on it. The datasheet shows a more complex transformer for dual output so they're doing something cunning : note the + and - rails aren't quite equal in voltage and also have different current capabilities. I would think it's configured as a boost converter but has some parasitic arrangement to create the -ve output.

And of course the quality ones, often found cheaply : https://www.ebay.co.uk/itm/181883184138 15mV noise.

[Zero999]

A buck-boost can give a negative voltage, from positive, so if you're not bothered about the positive rail being regulated, just use it to get a negative rail.

If you want a higher voltage, regulated positive rail, use a boost converter.

It should also be possible to generate an unregulated split rail power supply, by using a buck converter which can both sink, as well as source current and take the reference from a potential divider, on the input.

[langwadt]

A buck-boost can give a negative voltage, from positive, so if you're not bothered about the positive rail being regulated, just use it to get a negative rail.

If you want a higher voltage, regulated positive rail, use a boost converter.

It should also be possible to generate an unregulated split rail power supply, by using a buck converter which can both sink, as well as source current and take the reference from a potential divider, on the input.

similar to this, https://www.beyondlogic.org/wp-content/uploads/2019/08/ETA3000-Typical-Application-Circuit.jpg

[PaulAm]

If I need +/-15 in a small system for an opamp, I just use one of the 5v -> +/-15V modules.

You can find them all over, here's a cheapy from ebay:

https://www.ebay.com/itm/193794301546

Or, you can find something like this from Digikey

AM1D-0515DH30Z for $3.28 USD

These things are too cheap to bother mucking around with.

[Etesla]

I've had this problem before. +1 vote for charge pumps to get your negative rail. +1 for a boost converter to get all the way up to 15 ish volts from 3.3 or 5.



For opamps and other analog things requiring dual rails I think the actual voltage is not very important (it doesn't have to be 12 or 15V, 13.57 would probably be cool), and balancing it is also not very important (you can have +15V for Vcc and -13V for Vee or something like that). There are specs that also tell you how well the opamp can handle fluxuations on it's supply pins (Supply voltage rejection ratio) but it's usually good enough that you don't really have to worry about it.

My latest solution was to use a boost converter to take 3V up to 15V (something like this thing: https://www.digikey.com/en/products/detail/diodes-incorporated/AP3012KTR-G1/4470846) which is about maybe a dollar in components for a single quantity prototype.

You can clean up that signal using this guy: https://youtu.be/wopmEyZKnYo

Then you can use a charge pump IC or just find a way to switch your HV rail like figure 6 here: https://www.edn.com/capacitive-voltage-conversion-aka-the-charge-pump/ to get your negative rail

You can also clean up the negative rail with the same circuit from dave's video but using a PNP transistor instead.

[asdf336]

You usually don’t need a negative supply. 

If you have 5V make 2.5V with an opamp buffer and use it as a virtual ground.  Note you’ll probably need a bidirectional capable buffer like an opamp (a couple bidirectional regulators also exist).   

Buck converters can also be tricked into making negative voltage.  Tie the buck output to GND and take your negative bus from what would have been the buck converter GND.  Caveat: must be a synchronous buck. 

[langwadt]

You usually don’t need a negative supply. 

If you have 5V make 2.5V with an opamp buffer and use it as a virtual ground.  Note you’ll probably need a bidirectional capable buffer like an opamp (a couple bidirectional regulators also exist).   

Buck converters can also be tricked into making negative voltage.  Tie the buck output to GND and take your negative bus from what would have been the buck converter GND.  Caveat: must be a synchronous buck.

doesn't have to be synchronous

page 18

https://www.ti.com/lit/ds/symlink/lm2576.pdf?ts=1639388752981&ref_url=https%253A%252F%252Fwww.google.com%252F

[Scottie]

You may consider a push-pull converter like the SN6501. Then filter and post regulate the outputs for clean analog supply rails.

There are bipolar examples in the datasheet too.

[Amper]

Well thats a very ignorant statement.

[ajb]

Swapping the single inductor in a standard converter with a coupled inductor is another option.  See figure 3 here: https://www.ti.com/lit/an/slyt380/slyt380.pdf.  Since the output from the second winding is isolated you can do all sorts of things.  Main downside is the limited off-the-shelf options in coupled inductors, but there are decent SMD parts such that in terms of board footprint it basically only adds an additional diode and couple of output caps.  Of course the auxiliary output isn't regulated, but you can do a linear post-regulator if that matters to you.

But for a one-off, it's hard to beat the little encapsulated modules. I'll happily reach for those instead of taking the time to spec, design, and lay out my own converter.