# Capacitor selection with multiple regulators

I'm working on a circuit that needs three voltage regulators: +24VDC, +12VDC, and +5VDC. The power input is +28VDC from a bench power supply. The +24VDC regulator is a 78M24. It is supplied from +28VDC and the load is off-board and typically 120mA. The +12VDC regulator is a RECOM R-7812-0.5 switching regulator in a 3-terminal package (pin compatible with a linear 7812). This is also supplied from +28VDC input, and the load on this regulator is typically 125mA. This includes an off-board load and the load of the +5VDC regulator, which is a 78L05 whose typical load is 20mA.

I have specified capacitors according to the 3 regulator data sheets as shown in the circuit. C5, C6, C7, and C8 are ceramic. C9 is aluminum electrolytic, for higher ESR, per the data sheet. My question is, do I need any additional bulk or reserve capacitors at the input, outputs, or in between the +12VDC and +5VDC regulators? Many thanks for your help!

I have several follow on questions. The first answer mentions that a low ESR cap should be on the output of each regulator. Does that include the switching regulator (U8)? That data sheet specifically excludes low ESR on the input side, but says nothing about the output. Also, wouldn't changing from a 100nF to a 1uF on the output reduce the high frequencies that the cap bypasses? With regard to the input of the regulator, my understanding is the ceramic caps on the inputs are required for stability of the linear regulators (U9 and U10). Would bulk caps on the +28VDC net also suffice to provide bulk input capacitance to U9? Or does a bulk input cap for U9 need to be on the +12V net? Finally, since both U8 and U10 are supplying power to loads off-board, do they need bulk caps on the output? Many thanks again.

-
 If you are not concerned about the EMC and EN certifications, then that switch-mode regulator is an over-kill due to its price. I don't know, maybe it is also an over-kill in case of EMC and EN concerns. – abdullah kahraman Apr 21 '12 at 15:30

There should be a low ESR cap immediately on the output of each regulator. Perhaps 100 nF as you show is the minimum, but I'd put more there unless it was specifically disallowed in the datasheet. If they're supposed to work with 100 nF, then 1 µF ceramic sounds good.

As for the input, you can't have too much capacitance on the input of a regulator. Put what you can get in 0805 immediately on the input. That should be more than the skimpy values you are trying to squeak by with.

-
 Thanks for your answer. I have several follow on questions which I have added to the original post. – Mike Apr 21 '12 at 14:53

I am going to try to answer the questions the original poster added later on:

If you look at the datasheet of the switcher RECOM R-7812-0.5, you will see that without a capacitor at the output, there is an output ripple of 40mVp-p. If you add a 100uF capacitor at the output, then the ripple drops by a 5mVp-p. Of course, since the datasheet didn't mention, we do not know how was the load when they are testing for these parameters. The output current ripple increases with the load current. This 100uF capacitor mentioned in the datasheet can be a low ESR electrolytic to have a low ripple, or if you do not care about the ripple, just use a normal electrolytic capacitor.

A bypass capacitor is a backup power supply with relatively low lead inductance, and this allows it to source high $\dfrac{dI}{dt}$. Read more about it by Googling it. Here, the output caps you mention serve as decoupling caps, meaning that they will short the noise at some range of frequency to ground, and yes, 100nF will decouple higher frequencies than 1uF, but I think the difference is not that big in practice. Put a 100nF and a 1uF in parallel if you have the budget.

You need to put a bulk cap on the U9's input rather than on the U8's input if you think you need it. However, for 100mA ( as that is the maximum current as specified by 78L05), I do not think that you need bulk caps for U9.

I would not put the bulk caps on the board, however I would put bulk caps on the board that will have its source from this board, for U8 and U10.

I hope I didn't say anything wrong, if I did, please correct me or edit the answer.

-
 "100nF will decouple lower frequencies than 1uF" - isn't this backwards? All other things being equal, 100nF will have a higher resonant frequency and thus decouple higher frequencies. – Mike Apr 25 '12 at 1:37 @Mike Oops, sorry, you are right :) – abdullah kahraman Apr 25 '12 at 7:42