Capacitor selection for LM3940 LDO regulator

My project requires 3.3v for operation. For the simplicity I want to power it with USB sources and this makes me to add step-down converter/regulator to get 3.3v from 5v line (let's omit input power filtering for a moment).

I've chosen LM3940 low-drop regulator because of the following:

• works with 4.5v to 5.5v input range;
• can handle up to 1A, however, power consumption of my scheme unlikely exceeds 100mA.

According to the datasheet (section 8.2.2) I should add output capacitor no less than 33uF but there is also one more requirement: the ESR value.

They give some advises in section 8.2.2.1.2 for proper capacitor selection but I'm not very familiar with ESR yet and want to get it right.

According to the plot above I should be pretty safe (let's get the very minimum) if output capacitor ESR value lies within [0.10; 0.32] Ohms range (if I assume only 0.1 Ohm and middle value between 0.1 and 1 on the plot).

Also one thing to note: my device will always work at room temperature (15-30 degrees) so now ESR oscillations should occur; with this understanding I want to use only one electrolytic output capacitor without messing up with tantalums.

So I've found some electrolytic caps (series K50-35) in local store and checked datasheet to find info about ESR. However, they only gives tan δ value and tells no about testing frequency (but it seems to be 120 Hz, as stated in Accuracy subsection). So I used this formula to calculate ESR for capacitors:

$$ESR = \frac{\tan \delta}{2 \pi f C}$$

For given f = 120 Hz I've found that only following caps are suitable to be right for LM3940:

• 1000uF / 6.3v, 10v, 16v, 25v, 35v, 50v, 63v
• 470uF / 50v, 63v, 100v
• 330uF / 63v, 100v

However, I'm somewhat confused with so large voltage/capacity values and suppose that there is some kind of error in my calculations.

Could you please say if I'm right or not in selecting proper capacitor or I've missed something? Thank you!

• Another alternative (since you seem to be at the design stage) would be to use a more modern regulator that isn't as picky about its output capacitor.
– JRE
Commented Dec 21, 2019 at 13:28
• Maybe something like the LD1117 3.3V.
– JRE
Commented Dec 21, 2019 at 13:36

2 Answers

It's important to know a bit of (pre)history here... Looking at LM3940 datasheet I see a "May 1999" date, but the chip may be older than this. This means the device was designed at a time before thin portable devices, and it will require the type of output cap that was usual back in the day, which would probably be tantalum or electrolytic. Likewise a regulator like LM317 works best with the type of caps that it was designed for, that means a high-ish value with not too low ESR.

Modern LDOs are usually designed to be stable with only ceramic caps at the output because ceramic caps have gone a long way since the 1990's and become the cheaper, smaller, thinner option. But you can usually improve transient response by adding more capacitance, if necessary.

If you want to use LM3940, then you have to respect its ESR requirements. Personally I would use a Panasonic FR 470µF 6.3V capacitor, which has an ESR of 80-130 mOhms. But my reason for choosing this particular cap is that they fit the requirements and I ordered a bag of 100 (they're not expensive), so I have them. I wouldn't recommend placing an order just for one cap though, so you have to pick one that you can get easily.

Basically, you want low ESR for good transient response, but if it is too low then a LDO that is not designed for low ESR caps will go unstable, and the electrolytic cap can form a LC resonant circuit with ceramic capacitors in parallel. This cap's ESR of 80-130 mOhms is a great compromise, it is low enough for good transient response and it produces no resonance at all with a 1µF ceramic. Also most LDOs like the high capacitance and "low but not too low" ESR.

However, I'm somewhat confused with so large voltage/capacity values and suppose that there is some kind of error in my calculations.

This is normal, most "general purpose" caps are designed for low cost, NOT for low ESR. So if you want low ESR, you have to use a very big capacitor value which will use a lot of board space and may not be practical. Also the high ESR value of general purpose caps is a feature: it makes them not resonate with the ubiquitous 100nF ceramic bypass caps even if trace inductance is high, and that's good.

The "Panasonic FR" cap I linked is specifically optimized for low ESR.

Lower ESR and high capacitance mean Polymer caps. Some have ESR as low as 6 mOhms. These are quite popular among audiophiles who mod their equipment, for all the wrong reasons. DO NOT use a polymer cap unless you understand that it has strong tendency to create resonance peaks with ceramic bypass caps and/or make regulators unstable if not used carefully.

So, for your LDO, get a "Low-impedance" electrolytic (not "general purpose"), like those used at the output of switching power supplies. If you have an old switching supply that you don't use, there's your free cap!

Personally I'd replace the LM3940 with a more modern LDL1117 which has much better characteristics, lower quiescent current, and will work fine with ceramic caps.

(note LDL1117 is different from LD1117)

• Thanks for the commenting it out. But what do you mean when you talk about ceramics? According to the datasheet there are no ceramics on output, only on input Commented Dec 21, 2019 at 17:15
• The ceramics I'm talking about are all the decoupling caps on the board that will be powered by the regulator. Commented Dec 21, 2019 at 17:28

The caps you have the data sheet for are not low impedance types so you need a large capacitance to get low ESR. Also the ESR at high frequency is what counts here since that is where the regulator will oscillate, so a capacitor with an ESR rating at 100kHz would be preferred. The ones you have are more suited to mains frequency filtering.

For example, this small SMT cap.

The larger non-low-impedance types will probably work okay, regardless, but with less confidence and they’ll be much larger.

• So, my approach to calculate ESR is right? Commented Dec 21, 2019 at 17:10
• Yes, but it's under the specified test conditions, not at a more realistic 100kHz. Commented Dec 21, 2019 at 21:57
• If manufacturer doesn't provide me dissipation factor info for other frequencies then the only way to get ESR is to use ESR-meter, right? Commented Dec 22, 2019 at 8:55