I need more auxiliary supply rails...

[void_error]

Back to goofing around designing my bench supply (still) I've determined I need some extra voltages, all derived from a rectified DC rail (anything from 15 to 55V) and I'm a bit stuck with choosing the right way to generate them.

Here's what I need:

• +10V Op amp supply @ less than 10mA
• -10V Op amp supply @ less than 10mA
• +5V Digital supply @ less than 500mA
• +5V Analog supply @ a few mA
• -5V Bias supply @ less than 100mA

What's the best way to go in terms of cost/efficiency/simplicity for the worst case (55V input) scenario? Charge pumps/inverting regulators/step-down regulators...? Suggestions are very welcome

[Rerouter]

for the +- 10V, as op amps have good PSRR's i would say charge pump should suffice, but the wide rail voltage will not make things easy there, most have a band of about 15V they will work in,

Both 5V rails i would say use a generic buck regulator, using an LC filter on the analog supply, and planning out your current paths to avoid noise between the 2,

Finally for the -5V, perhaps a charge pump on the 5V digital rail, when you say bias supply i do not know how stable you need it,

[SeanB]

Wind a transformer and use multiple secondaries for each rail. Use the 5V rail as the rail to regulate with any common SMPS primary drive controller using a TL431 and optocoupler in the standard configuration, and use the other rails with regulation provided by LDO regulators. You probably will be using a flyback converter, so you will not have the luxury of using the forward current ramp as a opposite polarity low voltage supply as well, which can save a winding. You will need to have a winding per rail, but the low voltage ones will be thinner, and you will probably wind them to be 10% or so more turns to get enough voltage for the LDO to get the rail right. If you can live with a larger variation in voltage ( providing the load on the main 5V rail is pretty constant, say from 400-450mA) you can leave the LDO regulators off.

As well if you are using a mains transformer to supply the main rail ( probably via some relays) just use one of the secondaries and another bridge rectifier and capacitor for the aux supply to the SMPS. Makes design easier, plus you can choose the supply somewhat so you can use a telecoms SMPS controller that will run off say a 24VDC rail that is going to vary somewhat, but which will be good enough for it.

[void_error]

I prefer using off-the-shelf parts, so no winding my own transformer.

The mains transformer will only have 1 secondary winding.

The -5V bias supply should be fairly stable, +/-10%, a charge pump might do the job. Current draw will be more or less constant.

If it's possible I'd stick to a +5V step-down regulator with a multiple output charge pump based on some logic inverters driving small MOSFETS and some LDOs for extra filtering/stability, with LC filters where needed. By using Schottky diodes for the charge pump I might be able to go with some LM317/LM337 (low current ones where possible) instead of LDOs fed from 5V x 3 - (3*V_F) for +/-10V and 5V x 2 - (2*V_F). Jellybean all the way!

[Rerouter]

ok, for off the shelf,
5V using a generic buck regulator of your choice,
-5V via using a  similar buck in a inverting configuration,
5V to +-10V via a re-purposed MAX232 chip,

hows that for off the shelf?

[void_error]

ok, for off the shelf,
5V using a generic buck regulator of your choice,
-5V via using a  similar buck in a inverting configuration,
5V to +-10V via a re-purposed MAX232 chip,

hows that for off the shelf?

74HC04 + MBR0530 + capacitors + MOSFETs = better flexibility

-5V current draw turns out to be less than 50mA so I can also generate that with the same circuit as for the +/-10V. Might not have to even regulate it, as long as it's below -3V it should be fine and it's also cheaper than an inverting regulator.

[prasimix]

I'd like to suggest LM5574 if you are going over 50V. It works very nice and with two of them (seamlessly synchronized) you can have +/-10V and +/-5V where you can for instance from +10V using LDO derive additional +5V. For that you'll need coupled inductors. Basic idea about inductance you can get using Webench. Coilcraft have a plenty (i.e. MSD1260 series) of them and you can get few samples without hassle. It's possibly a little bit more expensive and not "jelly bean" class but you don't need to worry about allowable power, stability, transient figures etc.

[void_error]

It works very nice and with two of them (seamlessly synchronized) you can have +/-10V and +/-5V where you can for instance from +10V using LDO derive additional +5V. For that you'll need coupled inductors.

Wondering how I could sync it with the LTC3824 used as the pre-regulator... other than that it looks like a good choice. Or maybe get two LTC3824s and sync them, slightly more expensive but a lot more flexible . Oh, coupled inductors...

[prasimix]

It works very nice and with two of them (seamlessly synchronized) you can have +/-10V and +/-5V where you can for instance from +10V using LDO derive additional +5V. For that you'll need coupled inductors.

Wondering how I could sync it with the LTC3824 used as the pre-regulator... other than that it looks like a good choice. Or maybe get two LTC3824s and sync them, slightly more expensive but a lot more flexible .

Possibly as it's advertised: easily . I already tested such thing between two different family (LM5574 and LM5118) and it works great. You only have to take care about voltage level (which here is not an issue if LTC3824 can survive up to 6V and LM5574 will be the "master clock"), frequency range and frequency difference: one has to be set to lower then another but not too low (usually that is 15-25%).

Oh, coupled inductors...

Required if you'd like to have split supply (+/-10V and +/-5V). If you'd like to use LM5574 as a "bias pre-regulator" just to drop Vin from 50V+ to 5 or 10V from which you're going to derive others than "single coil" inductor is sufficient.