Input/Output capacitors of linear voltage regulator

[onemilimeter]

I think wrong capacitor value may cause the output of a linear voltage regulator to oscillate. Please correct me if I'm wrong.

Is there any rule of thumb for determining the input/output capacitors for linear voltage regulators (e.g. 7805, 7905, 7809, 7909, 7812, 7912, 7815, 7915, 7824, 7924)?


Say I wish to create a regulated +24V DC voltage. I choose a transformer with the following rating:
Input: 0-115V, 0-115V
Output: 0-24V, 0,24V)

If the output of the transformer is connected to a full-bridge rectifier, do you think the output of the rectifier is high enough to guarantee proper the operation of the 7824 (with approximately 2V voltage dropout)?

[Neganur]

It's not so the capacitance value than the ESR and trace/wire length (inductance) that are problematic.
You will typically place 100-330nF (mylar, tantalum or other low ESR type)  as close as possible at the input and 100nF the output of linear regulators. Linear regulators don't like capacitive loads too much, so yes the capacitor at the output should be small.
You can also have the regulator oscillate when the load current is too small (below 5mA or so).
Easiest way to avoid this is to place a (non-low current) LED with resistor at it's output.

24V AC translates to 34V peak after the full bridge rectifier and should be plenty (34V-24V-2V dropout = 8V). Remember to have proper cooling and design the decoupling capacitor according to your load (for the full bridge it's C = I * half mains period in seconds / ripple voltage). Example: 50 mV ripple and 50 Hz at 100 mA load gives C = 0.1 * 10 ms / 0.1 V = 10,000 uF.

Edit: I guess I was in a hurry when I wrote the capacitor/ripple voltage example. That calculation is for circuits without the reulator. A typical 7824  has a line regulation (the change in output voltage for a change in the input voltage) of 25 mV or less depending on the load current.

[Zero999]

It's true that it's a problem for low drop-out regulators which have a common source/emitter output stage because the capacitance on the output will change the phase shift of the feedback network.

There won't be any problem with the LM78xx, LM79xx etc because the output stage is an emitter follower with no gain so the capacitance makes no difference to the phase shift.

[onemilimeter]

I copy the following from the datasheet of MC7800 series linear regulator (http://www.onsemi.com/pub_link/Collateral/MC7800-D.PDF):

[1] Cin is required if regulator is located an appreciable distance from power supply filter.
[2] CO is not needed for stability; however, it does improve transient response. Values of less than 0.1 F could cause instability


Based on [2] above, it looks to me that instability is easier to happen with smaller capacitor, which is contradicted to Neganur's comment:

Linear regulators don't like capacitive loads too much, so yes the capacitor at the output should be small.

[ElektroQuark]

Values of less than 0.1 F could cause instability

 

[Zero999]

Based on [2] above, it looks to me that instability is easier to happen with smaller capacitor, which is contradicted to Neganur's comment:

No he's right but it only applies to low drop out regulators. He's also talking about capacitors with a low ESR, not just the capacitance value.

[alm]

The 79xx is more sensitive to instability due to (lack of) output capacitance than the 78xx. The output capacitor for the 79xx is mandatory, the output capacitor for the 78xx is optional. Not sure how stable it is with low ESR, I've never had issues, but I've never used it with extremely low ESR capacitors like ceramic either.