3
\$\begingroup\$

SOMEWHERE, SOMEWHEN, I clearly remember reading a very strong recommendation that one put a 1 uF tantalum capacitor between the output pin and ground on a linear 3-terminal fixed voltage regulator, to prevent it from oscillating.

I also remember, back in the late 1970s, seeing a malfunctioning homebrew power supply, that used such a regulator and a 2N3055 pass transistor. This was a VERY common circuit topology at that time. Adding the capacitor cured the problem instantly.

Just now, I went digging through old datasheets on the Web, and found that TI says that those parts DON'T need those caps, but they MIGHT help with transient response.

Am I completely nuts? Or does anyone else remember that seeing that recommendation, and remember where they saw it?

\$\endgroup\$
3
  • \$\begingroup\$ You haven't said which parts you mean when you say "those parts". \$\endgroup\$
    – The Photon
    Nov 28, 2020 at 6:11
  • \$\begingroup\$ The oscillation issue is solved by many modern linear regulators. Ergo, you don't always need a capacitor anymore. \$\endgroup\$
    – Kyle B
    Nov 28, 2020 at 6:18
  • \$\begingroup\$ I've also seen some very strong recommendations to not use tantalum caps in anything ever, because they tend to misbehave in unexpected ways and be hard to diagnose as such. So now we have that conflict. Use them to make "X" work, don't use them because they misbehave. The moral? Always be skeptical of absolute statements. \$\endgroup\$
    – AaronD
    Nov 28, 2020 at 20:21

3 Answers 3

13
\$\begingroup\$

It depends entirely which linear regulator you are using. The datasheet for a good one should tell you specifically what output capacitor is required or if that regulator will be stable without one. In some cases there might be upper and/or lower limits on the equivalent series resistance (ESR) of the capacitor used, and this will also be spelled out in the datasheet.

But there is no general rule --- you must read the datasheet of whatever specific regulator you are using to find out if that type requires a capacitor, and what capacitor parameters are required.

\$\endgroup\$
13
\$\begingroup\$

A 3-terminal "old school" regulator like the LM317 or 78xx (7805, 7812, etc.) uses an emitter follower stage as an output. This leads to a high dropout voltage (up to 2V was/is common), but the open-loop output impedance of the final stage of amplification is low, because -- emitter follower.

True low dropout regulators* became popular in the mid 1990's. These chips used open-collector PNP or open-drain PMOS outputs. These chips tended to have very high output impedances in open loop. With this high open-loop impedance, if it was used with a circuit that has high impedance at some frequency then the loop gain could get very high in places that it didn't with the older emitter-follower style. As a consequence, the regulator compensation that worked for an emitter-follower stage doesn't work with an open-collector stage, unless that stage is driving a healthy capacitance to keep the loop gain small at high frequencies.

During that time, pretty much all of the "new" LDO 3-terminal regulators needed that 1uF tantalum you're thinking of. 1uF to be enough capacitance, and tantalum because the ESR needed to be as low as possible for the time. Since this was absolutely unnecessary for the 78xx regulators, and absolutely necessary for the newfangled LDO's, there were a lot of people banging the drum about that 1uF cap. Since it was furthermore not harmful to an emitter-follower style of regulator to have the cap, there was no need for a counter-stream of argument against it. So it became common knowledge.

Later, IC designers figured out how to make true low dropout regulators that didn't need the cap. Because cap ESR became a selling point, and high-value ceramics came out, we also all figured out that for some regulators the cap ESR could be too low.

So now, basically, the requirements for an output cap on your 3-terminal regulator are all over the map. Some need a cap, some don't, some need ultra-low ESR, some will oscillate with it. This is why @The Photon mentions in his answer that you should just read the nice datasheet and do what it says.

* For a while the term "LDO" was applied to any linear regulator, which is a severe misnomer -- some time around 2000-2010 you'd see 7805 regulators with a 2V dropout called "LDO". Ugh.

\$\endgroup\$
4
  • \$\begingroup\$ One precision : "These chips tended to have very high output impedances" in open loop \$\endgroup\$
    – andre314
    Nov 28, 2020 at 8:54
  • 2
    \$\begingroup\$ FWIW I still see lots of people using "LDO" to mean "linear regulator". It's only becoming less common because old-school non-low-dropout chips are becoming less common, not because people started to understand the difference. \$\endgroup\$
    – The Photon
    Nov 28, 2020 at 16:19
  • \$\begingroup\$ @andre314: Thanks! I've corrected that, and expanded a bit on why it makes a difference. \$\endgroup\$
    – TimWescott
    Nov 28, 2020 at 19:51
  • \$\begingroup\$ @ThePhoton Probably right. Like op-amps, you've got to check the data sheet carefully. \$\endgroup\$
    – TimWescott
    Nov 28, 2020 at 19:53
0
\$\begingroup\$

Depends on the regulator. For instance, if we're looking at National oldies, the LM320 required a minimum of 10 uF, while the LM 340 had the spec you remember. The LM317 adjustable will work with no cap, but a 1 uF tantalum is recommended.

The Photon's answer is really the best.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.