Best approach with multiple voltage regulators?

Hello all,

I have a project where I need multiple voltages onboard - 10v, 5v, and 3.3v. The supply to the circuit will be assumed to be an unregulated wall wart with a voltage high enough for the 10v reg. Assume these are all "standard" linear 78xx regulators. Are there any rules of thumb for designing such a power circuit?

For instance, would it be conventional to supply the wall wart voltage to the 10v reg, then supply the regulated 10v to the "in" pin on the 5v reg, then the regulated 5v to the "in" pin on the 3.3v reg? Or, would you design based on current requirements at each voltage?

What about capacitance? When using a single regulator on it's own, I'd just follow the datasheet or other convention for input and output caps. But, if I have three regs tied together like this, do I really need all the caps called for with each regulator?

For instance, assume (hypothetically!) that the 10v and 5v regs both call for a 10uF electrolytic and a .1uF ceramic on the inputs and outputs. Assume I'm feeding the output from the 10v reg directly to the input on the 5v reg (as well as to the portion of the circuit that requires 10v, of course). Do I need TWO 10uF electrolytics, and TWO .1uF ceramics, right there next to each other, "between" the two regulators?

I personally would -not- "cascade" the regulators; instead, wire up the wall-wart to a "ground" rail (negative) and a "supply" rail (positive), then hook up the regulators to that.

Actually, between the positive output of the wall-wart and the positve supply rail, you should insert a fuse with a rating slightly less than the maximum current of the lowest current rating for any of the regulators can supply (thus, if you have two regulators, one that supplies up to 1.5 amps, the other 2 amps, go with a fuse rated for 1 amp) - this is just a basic safety measure.

As far as the capacitors are concerned, you can probably get away with a single electrolytic capacitor between the supply and ground rails, rated for something like 50 volts and whatever the largest capacitance (maybe plus some) is needed by any one of the regulators, but you should put individual capacitances as defined for the individual regulators on the output pins.

The input capacitance is meant to further "smooth" the input voltage to the regulators, but the output capacitance is meant to prevent ringing and other issues within each regulator itself when it is under load, and so should have its specified capacitance according to the datasheet attached.

I personally would -not- “cascade” the regulators; instead, wire up the wall-wart to a “ground” rail (negative) and a “supply” rail (positive), then hook up the regulators to that

Any reasoning or justification for this?

My thought behind “cascading” the regulators was that, then, the input voltage to the lower-voltage regulators would be lower, which means less voltage drop - I’d be able to offload some of the heat to the 10v, essentially. Though, this is just shifting a thermal problem, not really eliminating it.

In a few designs I’ve found that use this (i.e. an Arduino shield that has a chip requiring 3.3v), the configuration is as I described - the lower-voltage reg is fed from the higher-voltage reg, not from the raw Vin. Though, I don’t know if that’s related to the fact that these designs are “shields” or not. . .

Any reasoning or justification for this?

If you mean "do I have any engineering reason" - no; other than the fact that I have never seen something like this in a standard multi-rail power supply design (but that doesn't mean it can't work, I suppose).

I guess its one of those "from what I have seen, this isn't done in the industry for standard designs, so there must be a good reason for it".

If I were to guess why, it might be because each regulator could place a particular load on the regulator before it, possibly also maybe oscillation issues cascading through the system as loads are added or removed, the heat dissipation issue (really, a case of cacaded current loads) you mentioned...

Designs that I have seen always use common rails for the input to the regulation circuitry, whether it is "single component" regulators, or a discrete component regulator design.

I will note this, though - you may ask yourself "then what about a bench power supply with regulation?" - in other words, if you have such a supply, you can easily, say, run your 5V regulator off the regulated output set for say "9 volts" on the power supply, and it will work OK.

I don't have a good answer for that - I do know, though, that if you looked at the schematic for that regulated power supply, that the designer probably didn't cascade his rails...

Maybe someone else here has a better idea of what is "best practice"?

:)

other than the fact that I have never seen something like this in a standard multi-rail power supply design

Well it happens all the time. I have used it on many projects and seen it on virtually every design I have come across in the last 30 years.

Mainly the advantage is that is minimises the voltage drop across the lower voltage's regulators thus limiting the power dissipation it has to handle. Of course this means increased current in the high voltage regulators but that means you might have to only have one heat sink instead of two. Also lower voltage regulators also tend to have a lower maximum voltage input. I would place a capacitor on both sides of a liner regulator close to the regulator irrespective of how they are wired. The main use is to lower the supply rail impedance to increase the stability, that is stop them oscillating.

Well it happens all the time. I have used it on many projects and seen it on virtually every design I have come across in the last 30 years.

Do you happen to have any handy link(s) referencing an example? I mean, I understand how such a thing would be constructed, and I can visualize the design in my head, but as I noted I have never seen one; all the designs I have seen (admittedly, I haven't looked at tons of power supply designs) use common rails. I am just curious to see such a design. I'll have to do some more research.

Mainly the advantage is that is minimises the voltage drop across the lower voltage's regulators thus limiting the power dissipation it has to handle.

That certainly makes sense, and something I had thought about as I replied previously - but I had never seen it used before.

Of course this means increased current in the high voltage regulators but that means you might have to only have one heat sink instead of two. Also lower voltage regulators also tend to have a lower maximum voltage input.

Once again, true.

I would place a capacitor on both sides of a liner regulator close to the regulator irrespective of how they are wired. The main use is to lower the supply rail impedance to increase the stability, that is stop them oscillating.

Yes.

I guess the question then is, if it is useful to do this, why isn't it used everywhere? We can see the advantages of a cascaded design - where are the disadvantages? What are the advantages/disadvantages of a common rail design?

I can think of one disadvantage of a cascaded design, which, depending on the needs and such, might be a bad thing: If one of the larger regulators higher in the chain fails or shuts down (overload condition), its failure/shutdown would affect lower chained devices (this may or may not be desireable, I suppose).

Thanks for the info, though!

:)

FWIW this is the first occurrence I stumbled across:

http://www.nuelectronics.com/download/projects/Ethernet_v1_0.pdf

This is an Ethernet shield based on the ENC28J60 chip.

The ENC28J60 chip needs 3.3v. There is a 3.3v regulator on the shield. Of course, the shield has both "raw" Vin and regulated 5v available. The designer of this shield chose to feed the 3.32v reg from the 5v rail, not Vin. I've seen probably 3 or 4 other Ethernet shields with the same design (ENC28J60 chip, onboard 3.3v reg) and they're all designed this way. That's a really small sample size, but at least it shows that the "cascaded" design is in use out in the real world.

our designs use cascaded and so does korg, for some “real world” examples, also A sony vcr I recently saved from the dumpster (well … most of it heh) used a cascaded setup

Do you happen to have any handy link(s) referencing an example?

Well no because it is all commercial stuff and my employers wouldn't be too happy if I were to post their schematics over the net.

its failure/shutdown would affect lower chained devices

Yes this is an advantage. If you loose one voltage rail and continue to power a device or supply signals to it that can also damage it. Normally there is no function if one regulator is blown so why bother keeping the others up. For example the power on LED can be taken from the lowest regulator and it will instantly indicate a power fault on any regulator.

The only real disadvantage is that the supply decoupling is worse because you are feeding a regulator with the noise from another circuit so the two supply are not as isolated from the noise point of view.

The designer of this shield chose to feed the 3.32v reg from the 5v rail, not Vin.

Yes that is what I would do as well if I required 3v3 and 5V. However if I needed say 1v8 and 1v2 I would probably regulate both of them from the 3v3 line to give the regulators some head room as normally 0.6v is a bit small.

I can think of one disadvantage of a cascaded design, which, depending on the needs and such, might be a bad thing: If one of the larger regulators higher in the chain fails or shuts down (overload condition), its failure/shutdown would affect lower chained devices (this may or may not be desireable, I suppose)

While adding components can increase the failure rate you can also get a reduction in failure rate by running each component at lower temperatures.

(* jcl *)