The role of capacitors in voltage regulators

Hi,

If I had this circuit:

And if I have used a 50uF capacitor instead of the 10uF capacitor what would happen?

That's figure 18 in the datasheet, where they are saying that if you use a decoupling capacitor (C2) on the feedback network you also need 2 diodes to protect against back-driving on power off.

Most circuits in the datasheet have no such C2 and thus no need for the diodes either. The purpose of C2 is to increase ripple-rejection, which may not be necessary. Clearly increasing its value will increase the ripple rejection.

C1 is a different matter.

That also means, increasing ripple rejection will lead to a slow voltage (Vout) adjustment?

F1_: That also means, increasing ripple rejection will lead to a slow voltage (Vout) adjustment?

Ah, no, that's the clever bit, the feedback loop in the regulator maintains the output voltage with respect to the adj terminal voltage - decoupling the adj terminal doesn't compromise the feedback, just gives a more stable Vadj.

The negative feedback is from the output terminal, being compared to an internal voltage reference that's 1.2V above Vadj. Making C1 too large could have feedback stability issues, but not C2.

Got it, thank you a lot for explaining this!

MarkT: That's figure 18 in the datasheet, where they are saying that if you use a decoupling capacitor (C2) on the feedback network you also need 2 diodes to protect against back-driving on power off.

Well, it's not so much back-driving on power off, as when the power is turned off, the input capacitor (not shown) is only discharged through the regulator as it seeks to maintain the output.

The danger is that if the output is suddenly shorted, the regulator output will be reverse-biased by C2's charge; D2 protects against this.

I mean back-driving the ADJ pin, the only pin sensitive to this. Both diodes are there to prevent this.

It’s the life story of data sheets. Incomplete and sometimes vague and cryptic. Would have been nice for them to add 1 tiny extra detail about C2… such as 10uF is adequate… with something extra… like xx uF <= C2 <= yy uF.

Although… naturally if the data sheet says 10 uF… and nothing about range of allowed C2 to use… then often a good idea to stick with 10 uF. But since they mention ‘bypassing adj’ … then that is expected to imply 10uF or larger. If in doubt…then could ask the manufacturer to make sure (if necessary).

@MarkT

Ah, no, that's the clever bit, the feedback loop in the regulator maintains the output voltage with respect to the adj terminal voltage - decoupling the adj terminal doesn't compromise the feedback, just gives a more stable Vadj.

I've not studied the LM338, however just looking at the schematic I would think C2 would delay the output voltage percentage appearing on R2. This would slow the recovery to a increased load. Perhaps I am missing something?

John

JohnRob: I've not studied the LM338, however just looking at the schematic I would think C2 would delay the output voltage percentage appearing on R2. This would slow the recovery to a increased load. Perhaps I am missing something?

No, precisely the opposite - C2 is there specifically to improve the transient regulation.

The voltage across R2 is intended to be a constant. The IC sees - and regulates - the voltage across R1 and C2 functions such that the whole of a change in the output voltage is seen across R1 rather than just a proportion determined by R1 and R2.