I am designing a board which needs 5V and 3.3V rails, at about 1.5A. If I used two buck converters, I would use a lot of board area. So, is there a way to create a multi-output buck converter? I tried creating one by using two inductors (each of a different value), in a circuit simulator, but it doesn't work; the output ripple is through the roof and both rails sit at 5V or 3.3V (I kind of expected this.)
You can sortof do it, but it's probably not worth the trouble. You would have to add a diode in front of the 5V output cap, and a pass element in front of the 3.3V output cap. The switcher regulates the 5V output, and a comparator turns on the pass element whenever the 3.3V output is low. In other words, the comparator via the pass element steals current that would otherwise go to the 5V supply, but the feedback system makes sure the 5V supply gets enough current to keep it at the right voltage. Like I said, probably not worth it.
To do this in one switcher somehow will require a inductor that can deliver 3A average. It won't be much worse to split that into two inductors that each only have to deliver 1.5V.
Another possibility is to have the switcher make 5V, then linearly regulate that to 3.3V. This wastes 2.6W in heat.
You could buck to make the 5V, then use a open loop fixed duty cycle buck converter with 3.3V/5.5V = 66% duty cycle to make the 3.3V from there.
Some of the IC makers have products that can do more than one output from the same chip. For example Maxim has the MAX5066 among others. Needing both 3.3V and 5.0V is common in laptop computers so this feature was integrated long ago. It is still two regulators but at least it's only one IC so less board area and possibly less cost.
- No servant can serve two masters - for either the one he will hate, and the other he will love; or one he will hold to, and of the other he will be heedless
As others note, this is an inexact art.
In the general case you cannot combine two independent buck regulator circuits so that they share the same inductor and same switch, because they will both have differing regulation requirements.
Because, if one circuit's output is below the desired level and the other's is at or above the desired level then a choice has to be made re which requirement to satisfy - the result will be one output which is properly regulated, with the other either too high or too low. However ...
The following explain a "trick" which works in some cases and is useful when it does.
If one supply requires much lower energy than the other you can get an OK result with two windings on the same core. Having Vin >> Vmain *eg 12V to 5V) helps. I've done this with acceptable results for my application. Yours may not be well served.
eg if you required 5V at 3A and 3.3V at 1 A.
- 5V winding has N turns
- "3.3"V winding has N x 3.3 / 5 * k
Where k is "slightly more than 1."
How big "slightly is is TBD.
Connect 5V winding as usual.
Define "dotted end" = connection to 5V.
Undotted end = other end.
Connect "3v3" winding with undotted end to ground.
Dotted end via a Schottky diode to 3V3 DC.
In the part of the cycle where the switch (MOSFET etc) is off, the 5v winding will have about 5V + a diode drop across it.
The "3v3" winding will have (5dioe drop) x 3.3/5 x k = somewhat more than 3V3.
After diode 3v3 output will be 'a bit high'.
3V3 here only gets fed on off part of cycle with energy from magnetic storage. As Vin increases on % period will decrease and off % period will increase and 3V3 drive will be reasonably reliable. As Vin drops towards 3V3 the % off period will increase and 3V3 drive will suffer.
So - if 3V3 energy is < to << 5V energy and if you can post regulate and/or if vIN >> 5v then this MAY work well enough for you.