Why dual/triple rail power supplies?

[claytonedgeuk]

A question for you clever folks here....when does a power supply with multiple outputs become useful or essential?  Can you get by with a couple of separate power supplies - or does that just not cut it?

[RGB255_0_0]

Space, easier to work with op-amps (split). But there's no reason why you can't tie supplies together, it's really inelegant and maybe some people frown upon it?

[tggzzz]

One lower voltage (3.3V,5V) supply for logic. Two higher supplies, positive and negative,  for analogue circuits. Logic requires higher current than the analogue circuits.

[Shock]

Tracking is a useful feature on some power supplies with multiple outputs, you can adjust positive/negative or multiple outputs symmetrically for op amps and in situations where you need to substitute the DUT supply for troubleshooting and testing.

[bitseeker]

In addition to having tracking, so that one or more channels' settings change when the master does, a multi-channel power supply can also turn all the channels on/off at the same time, which is useful for verifying the on/off behavior of circuits that use a multi-voltage power source.

Depending on how the power supply is designed, channels may also be used in series or parallel to increase voltage or current output, respectively.

Using separate power supplies instead of a multi-channel one can work too, depending on how they're designed. However, it often costs more money and space to use separates. If you're just getting started, you might consider getting one dual-channel and one single channel instead of a triple-channel supply so that if one fails, you have the other for troubleshooting it.

[shteii01]

In the beginning of time the operational amplifiers needed two power supplies (yes, I am looking at you 741).  One power supply supplied positive voltage, second supplied negative voltage.  A +15 volt and -15 volts was the minimum required for 741, so you would get dual power supply, tie the + of Supply 1 to ground, tie the - of Supply 2 to ground, and you have two rails for the op amp.

These days dual rail op amps are less common and the need for multi rail power supplies has gone down significantly.

[Cubdriver]

As others have already pointed out, sometimes multiple voltages and polarities are needed for a single circuit.

If they can play well together, there is no reason that separate supplies cannot be wired up and used for this purpose, but having everything in a single device will typically simplify connections.  Additionally, if the power supply has protective circuitry to limit output in the event of an overload, this would likely shut down all outputs on a single supply, whereas separate supplies would be isolated and could possibly cause damage if, for instance, a control voltage goes away due to a fault but the drive voltages from the other supplies stay on and cook something that should be turned off by the missing voltage.

Another thing is that there's likely only one switch to turn on or off to power or kill power to the whole circuit, rather than multiple ones that could be forgotten when troubleshooting (either on when they should be off, or off when they should be on).

-Pat

[tggzzz]

Another thing is that [with a single box] there's likely only one switch to turn on or off to power or kill power to the whole circuit, rather than multiple ones that could be forgotten when troubleshooting (either on when they should be off, or off when they should be on).

I've always found that dubious. The speed with which the voltage decays is load-dependent, and relative voltage constraints might be violated.

FPGAs, where the supply sequencing order is defined extremely explicitly, would be the canonical example.

[SingedFingers]

I think that is valid but the other way round. I want one switch to bring up both rails rather than two switches.

I also still use dual rails in most things with op amps. If you have an output stage that needs to source or sink relative to ground it makes life a million times easier. Plus when you have discrete transistors in circuit, 30v rail to rail is good headroom.

I use 12v AC single phase wall wart input to all my projects as you can knock out two opposing 15v rails trivially. If the PSRR of the opamps is decent you can get +/-15v out of two diodes, four electrolytic caps, two resistors and two zeners. Add a BD139/140 pair for better regulation and you're done for less than $1

[b_force]

In the beginning of time the operational amplifiers needed two power supplies (yes, I am looking at you 741).  One power supply supplied positive voltage, second supplied negative voltage.  A +15 volt and -15 volts was the minimum required for 741, so you would get dual power supply, tie the + of Supply 1 to ground, tie the - of Supply 2 to ground, and you have two rails for the op amp.

These days dual rail op amps are less common and the need for multi rail power supplies has gone down significantly.

It's funny, it's all about perspective I guess. Because I would explain it the other way around.
In the beginning most audio circuits were single rail voltage (and therefore ugly big capacitors were needed to block the DC voltage).
Symmetrical voltages are much easier in audio circuits. So (semi) modern and high performing circuits work with dual supply.

Ironically, things are going back to single voltage again due to those terrible performing bridged class-d chippies.

[jpb]

On the purely practical side, a single dual supply is normally a bit smaller than two separate supplies and only requires one socket. When you're using big linear supplies and have a small work bench this can be a consideration.

Though I've often thought that it might be more useful to setup some sort of bus system on the bench with low voltage outputs at say 3.3V, 5V and 12V and have the supplies tucked away elsewhere to free up bench space. What I'd like would be a linear supply split in two - the big transformer/control part in one box and the source sense jacks together with the meters and Voltage/Current controls in a slim panel that could go on the bench. I suppose it would have drawbacks in terms of noise pickup and voltage drop in the cable though a Kelvin connection should negate that.

[james_s]

I frequently use multiple power supply voltages, either troubleshooting existing gear or prototyping my own stuff. Multiple separate supplies works fine, but a single multi-output supply is a very convenient tool to have.

[Deridex]

A good example which i had recently at work:

We created a new control unit. 1 week after ordering all the needed stuff for it, we had all exept the power supply for the ucontroller. So we just had to supply him with the needed 5V and the rest with the 12V. That way we needed either 2 power supplys or a 2channel one.

[bson]

I use single supply all the time because I always, with no exception, put power management on board, so my boards usually only need single 5V or 12V.
The benefit is once I've finished debugging, I can just plug in a USB or barrel jack to put my board into production use. No PCB respin needed.

But surely at some point you run your board without the power management, off a PSU, so you can measure the continuous (power) and transient (current) needs for the different rails?  If nothing else, to make sure they're within what you expect from the data sheets, to verify the requirements for the power management section are correct.

[james_s]

Most people don't have access to such fancy tools, and in many cases it's just simpler to prototype with external power supplies.

[bson]

I have a PCB surface current probe for that purpose.

One, or one per rail?  I design my firmware in general so as to stagger switching transients and not perform all current hungry tasks at once.  How would I verify transients are actually properly staggered and don't badly overlap across the rails without one probe per rail?

Of course, a good lab PSU with transient analysis (or use a scope) costs in the hundreds of dollars used.  A pile of current probes on the other hand...  $5000?

[bson]

Of course, I'm not disagreeing that current probes aren't superior tools... just an expensive proposition for a starting point.

[bitseeker]

I design my firmware in general so as to stagger switching transients and not perform all current hungry tasks at once.

That's pretty cool, temporally distributing bursts of activity.

[claytonedgeuk]

So if I have got this right...If I'm playing with op amps....and need the +15v and -15v then having a split is easier if they track symmetrically...but I can get by if I manually track two bridged supplies separately? 

The symmetry doesn't need to be absolutely perfect such that I'll be able to manually manage it myself by hand?

Sent from my SM-G900F using Tapatalk

[SingedFingers]

You can do it with two supplies separately. The opamps don't need perfectly  symmetric power supplies at all so it's easy to manage by hand.

Before I had a power supply back in the 1990s, I used two 9v batteries to do the job. Not perfect but better than nothing.