Linear regulator output capacitors - ceramic, electrolytic, ?

[joeyjoejoe]

Man, electrolytic caps are a pain for hobbyists. Last night I was looking around for the same footprints off Arrow, and realized I'll probably need to change the footprint for a few of them. I thought, man, it'd be great to just use ceramics - 1206 is 1206.

I just finished reading a TI report on regulator stability with ceramics, which ruined that dream a bit. ( http://www.ti.com/lit/an/snva167a/snva167a.pdf ) My circuit does indeed fall to the left of some of those graphs - very low current (16mA) requiring high ESR for stability.

Three thoughts come to mind:

1. I'm not sure if this is specifically targeting LDO's, and if a regular 2V dropout such as a 7805/7812 is equally affected by this, or less so?

2. Or, the articles implies there might be some regulators that are designed with this in mind - how would I find these? Any product series that people can recommend?

3. Finally, would a tantalum cap have enough ESR to prevent loop stability issues and let me have more of a standard footprint on PCB's for these? They are fairly pricey though...

[Kleinstein]

The classical linear regulators (e.g. LM317, 78xx) with a drop in the 1-2,5 V range are usually not very sensitive to the output cap. However even if stable, regulation performance gets better with a suitable cap.  However even there a few (e.g. 7905) need care. LM1117 or similar ae a little in between - with a dropout in the 1 V range and still not that picky about the capacitor as true LDOs.

There are some LDOs made to work with very low ESR output capacitor (e.g. ceramic).  I don't know a good way finding them. If you are lucky the manufacturer has a corresponding column in there parametric table or a corresponding subgroup. It is often the newer one that are suitable for ceramic output capacitor. Otherwise it's looking at the descriptions.

There are quite different types of tantalum caps. The more usual ones are typically OK with there ESR for the LDOs. One should still ckeck the data.
Al electrolytic ones are usually OK at room temperature and above, but can be too high in ESR below freezing. They may need an additional ceramic capacitor in parallel.
The alternative could be using a ceramic capacitor with a resistor as a second element.

[wraper]

78xx work fine with ceramic capacitors. For LDO it depends. Some work fine with ceramic capacitors. Some require that capacitor ESR should be above certain resistance. The right answer is to read datasheet of particular IC.

[wraper]

LM1117 or similar ae a little in between - with a dropout in the 1 V range and still not that picky about the capacitor as true LDOs.

They are not in between. Depending on brand of 1117 these could be among the most susceptible. (Edit: particularly LM1117 by TI is something in between). And it has nothing to do with voltage drop as such but how output stage is made and compensation to counter that. Traditional Vregs have emitter follower, hence they are not susceptible by default. There are LDOs with very small voltage drop which are stable with ceramic capacitor.

[joeyjoejoe]

Thanks guys!

So I dug this up https://www.st.com/resource/en/datasheet/l78m.pdf, and indeed section 7 (page 24) says an output cap isn't even needed, but it's recommended, and in fact, something with a low ESR is recommended! And, unlike TI's application note which says the caps can be put further from the output, this one says it should be as close as possible to the output of the regulator.

Good news for my hobbyist inventory needs

[Wolfgang]

Hi,

a smart way to avoid stability problems is

- avoid extra low dropout voltages. If headroom becomes too low, stability issues become a real pest and regulation becomes very bad.
- simulate before you build
- better too much ESR on the output cap than too little.
- after building measure with a low frequency VNA like a Keysight E5061-3L5 or a Omicron Bode 100 (Review here on EEVBlog).

some info here:
https://electronicprojectsforfun.wordpress.com/power-supply-impedance-measurements-using-the-bode100-lf-vna/

[Zero999]

Yes, some low drop-out regulators have issues with low ESR capacitors. One way round the issue of low ESR capacitors is to add a resistor in series with the capacitor. A ferrite bead may also work, as it will look like a resistor, at the frequencies where it will oscillate at, but it may need more than one turn to be effective.

[wraper]

A ferrite bead may also work, as it will look like a resistor, at the frequencies where it will oscillate at, but it may need more than one turn to be effective.

You certainly shall not do this. It won't help at frequencies at which LDO usually would oscillate (up to few tens of kHz) at all and will interfere with main function of output capacitor. If you use SMD ferrite bead it would work only due to it's relatively large resistance.

[Zero999]

A ferrite bead may also work, as it will look like a resistor, at the frequencies where it will oscillate at, but it may need more than one turn to be effective.

You certainly shall not do this. It won't help at frequencies at which LDO usually would oscillate (up to few tens of kHz) at all and will interfere with main function of output capacitor. If you use SMD ferrite bead it would work only due to it's relatively large resistance.

Yes I think you're right. I'm sure I've seen ferrite beads in series with decoupling capacitors before though. I wonder why?

[David Hess]

1. I'm not sure if this is specifically targeting LDO's, and if a regular 2V dropout such as a 7805/7812 is equally affected by this, or less so?

It is more of a problem with low dropout regulators because of the higher output impedance of their common emitter/source output transistor.

Positive standard dropout regulators with their common collector/drain output transistor like the 7805 series are almost immune to this problem but still perform better with the higher capacitance and ESR of a small electrolytic capacitor.  This is *not* the case for most negative regulators!  They usually *must* have a suitable output capacitor because with an NPN output transistor, they use the same output configuration as a low dropout regulator and a 1 microfarad solid tantalum or 10 microfarad aluminum electrolytic is typically recommended.

2. Or, the articles implies there might be some regulators that are designed with this in mind - how would I find these? Any product series that people can recommend?

In most cases, the datasheet will say.

3. Finally, would a tantalum cap have enough ESR to prevent loop stability issues and let me have more of a standard footprint on PCB's for these? They are fairly pricey though...

Solid tantalum capacitors are ideal.

Another alternative is to use a ceramic or film capacitor with a low value resistor in series to produce the needed ESR.

- after building measure with a low frequency VNA like a Keysight E5061-3L5 or a Omicron Bode 100 (Review here on EEVBlog).

Transient response testing with a function generator and oscilloscope can also be used in place of a network analyser.  Output impedance varies significantly with load so testing should be done over the expected range of load current.  Low output currents are usually the worse case.

[mariush]

In some *1117 regulator datasheets, you'll find that they say that the ESR of the capacitor must be within 0.1 ohm to 1 ohm. I guess higher ESR would not be as much problem as too low ESR.

In some older datasheets, they specifically say "tantalum" because few such capacitors had ESR below 0.1 ohm. Otherwise, I suspect they don't mention it because they just assumed you'd use a tantalum (and the example schematic would often mention a tantalum capacitor part) or regular electrolytic capacitor, and back then fewer series of capacitors had such low ESR at low capacitance values.

I'm aware of at least one situation where a 1117 regulator was unstable with a ceramic capacitor on the output but I don't remember exactly who was .. it may have been a board bunnie from bunniestudios had mentioned in a blog years ago.  Anyway, here's a eevblog forum thread with a 1117 related problem: https://www.eevblog.com/forum/beginners/ldo-regulator-oscillation/

Ah ... just looked at the bottom of that thread, it wasn't bunnie, it was the guys from mightyohm : https://mightyohm.com/blog/2013/02/simple-fix-for-bus-pirate-power-supply-oscillations/
That board uses a Micrel MIC5025 part.