Voltage Regulator Input Capacitor: ESR Rating

[gregallenwarner]

I'm looking at some datasheets for some simple linear regulators (7805 and 7905 types) and looking at what they recommend for input caps.

I'm looking at this wording in particular: "A 0.33?F or larger tantalum, mylar, or other capacitor having low internal impedance at high frequencies should be chosen."

The datasheet makes no mention whatsoever of aluminum electrolytic, however, as I've been scoping out capacitors, I find that, for a given price, I can get a solid aluminum polymer cap with a MUCH lower ESR rating than I can in tantalum. For example, one solid aluminum cap I found has an ESR rating of tens of milliohms, while a similarly priced tantalum has an ESR of over an ohm, two orders of magnitude higher.

I'm just wondering, how much ESR is considered low in this regard? Is there some rule or formula used?

Also, and this is unrelated to the question above, but I'm considering solid aluminum's because I like them, and I want something that can run in high heat. Don't know if this info matters, but thought I'd include it.

[mariush]

Linear regulators like 7805 or 7905 don't care that much about the ESR of a capacitor. 

You have to read those few lines of that in the datasheet as something like this:

The linear regulator needs some capacitance at the input, and that capacitor should have some low impedance. The minimum capacitance recommended is 0.33 uF.  If you want to stick to the minimum, some capacitor types that would fit both criterias (small capacitance and low esr) would be tantalum or mylar/film.

Unless the datasheet says a maximum capacitance, you're not forced to stick to that listed value.

You don't HAVE to use 0.33 uF, that's the minimum they recommend.  You can use 1uF, you can use 10uF, you can use 100uF in this case as long as the capacitor also has a reasonable esr/impedance... the datasheet tells you the minimums just in case you're hired by a cheapskate company that wants to buy the absolute cheapest part to save pennies.

There's very few electrolytic capacitors made below 1uF, and even then such electrolytic capacitors would have a high ESR so since nothing stops you from increasing capacitance, you would normally go something bigger. For example, I would go with a 33-100uF electrolytic, whatever I have at hand, and it's such a large range simply because they usually take the same volume on board (5x11 mm or something like that)

The higher the capacitance, the larger the volume of the capacitor and as consequence the ESR becomes lower.  So a very small 5x11 mm 10uF electrolytic may have an esr of 1-2 ohm, while a  47-100uF 8x10mm may be around 0.5 ohm ESR. With these linear regulators, the ESR is really not that important.

Do note... if you use an electrolytic you should also put a small 0.1uF ceramic capacitor as close as possible to the input pins of the linear regulator.

Another note... there are some linear regulators that are sensitive to the ESR of the output capacitors, so don't just assume that what works for 7805 will work for others. 

For example, a LM2931 regulator ( http://www.ti.com/lit/ds/symlink/lm2931-n.pdf see page 12 and figure 34 on page 10) likes having a capacitor at the output with an esr between about 0.1 ohm and 1 ohm, so using very low esr ceramic or polymer capacitors may not be a good idea, same for cheap small capacitance electrolytic capacitors that may have an esr higher than 1 ohm.

[gregallenwarner]

Ok, so if I understand, the larger the input capacitor, the less critical of an issue the cap's ESR becomes? So if I throw a 10uF cap on there with, say, 1.8 ohms rated ESR, I shouldn't have any troubles?

Also,

Do note... if you use an electrolytic you should also put a small 0.1uF ceramic capacitor as close as possible to the input pins of the linear regulator.

Could you elaborate on that? What's the reasoning why? Is it the same if I choose one of the solid aluminum types as well?

[janoc]

I believe that the ESR affects only the ability of the capacitor to dampen/eliminate possible oscillations of the linear regulator. That's why the 100n cap is recommended in parallel with the large electrolytic. It doesn't contribute essentially anything to the capacitance, but has a tiny ESR compared to the electrolytic one, so if the regulator decides to oscillate, it will quickly stop it (think of it like a dead short for anything high frequency).

On the other hand, be careful with LDO type regulators - those *need* some minimal ESR otherwise they will start to oscillate. So actually using an otherwise crappy electrolytic cap is an advantage there - if you used ceramic caps, the regulator could be unstable.

[mariush]

Ceramic capacitors have very low ESR by their nature, by design.... their capacitance varies with the voltage, for example a 1 uF 16v ceramic capacitor  may be 1uF with 1.5-3v but only 0.5uF with 12v, or something to that effect. There's also other issues, like risk of cracking if the ceramic capacitor is too big.

Tantalum capacitors usually have low esr but these days they're not used that much because a small percentage tends to die shorting out or bursting in flames.  Tantalum is also a conflict material (from war zones) and the price fluctuates a lot.

Electrolytic capacitors are available from 1uF and up.. of course there are exceptions under 1uF but in general electrolytics are for bulk capacitance. The ESR of such capacitors varies with the volume of the capacitor can and the physical construction (how the foil inside is etched and so on). Therefore, it's somewhat logical to expect the ESR to decrease as the volume of the capacitor increases.But again, it's not always the case ... for example a large 330uF 450v rated capacitor may have a 1.5 ohm esr, while a 330uF 25v small capacitor may have an esr that's under 1 ohm... so lesson is volume isn't everything.

Ok, so if I understand, the larger the input capacitor, the less critical of an issue the cap's ESR becomes? So if I throw a 10uF cap on there with, say, 1.8 ohms rated ESR, I shouldn't have any troubles?


What I'm saying is that this particular type of linear regulators (7800 and 7900 family) doesn't really care about esr of input capacitor. As long as it's low, it won't be a problem.  20 years ago, 0.5-1 ohm was low esr, now i'd probably expect 0.2-0.3 ohm to be "low esr".  Polymer capacitors and high end electrolytic capacitors go under 0.1 ohm esr. some polymers are even 0.05 ohm.
I wouldn't use a 10uF capacitor but that's simply because it's unlikely I'd have one in stock ... like I said, I'd use 33uF or 47uF or 100uF ... typical values.


As for that 100nF  (0.1uF ceramic), they're called decoupling capacitors and you'll see in circuits that most integrated chips will have one or several such ceramic capacitors very close to the pins. They're meant to help when the electrolytic capacitor is "too slow" to react.

I'm not sure how easy it would be for you to understand but here's a series of videos that teaches this subject:

This is another good video that explains the need for decoupling capacitors in a different way (more visual) :

[gregallenwarner]

Thanks. That 2nd video was especially illuminating.

I actually found some reasonably priced tantalums. As long as I make sure and order ones that are rated for a voltage well over my supply voltage, do I need to worry about them bursting into flames? Are there any protective measures I need to take in my design to help protect them from damage?

[T3sl4co1l]

Electrical properties have no relevance here, so long as you select something above the minimum.  The remaining selection criteria should be cost, size and reliability: get the cheapest type you can, with the reliability you need (if any).

If you don't need reliability, 10-100uF aluminum electrolytic will do fine.  If you do, film or ceramic might not be too bad.

Tantalum are more dubious to use and much more expensive than necessary in an application like this.  The main reason you would want to actively avoid tantalum is when large peak inrush current or dV/dt is available (power supplies, hot plugging, frequent transients / faults); they are perfectly safe to use in current limited situations, but this generally precludes anything upstream of a current limited regulator.

Tim

[Seekonk]

We bought a module for a project that was 120V AC.  We later decided to locate the module where line voltage wasn't available.  The transformer and diodes were removed so it could operate on 24V DC.  We asked the manufacturer to make a version for us because this would be purchased in number.  It had a 7815 or other internally.  When the first production units arrived they were singing at about 1MHz.  Someone at the factory decided that the power supply electrolytic wasn't needed.  If you ever wondered what would happen without one, now you know.

[Dago]

I think that datasheet might have been written ages ago (like in the 70s or 80s) when it might have been true that tantalums had lower ESR but it does not really hold true nowadays.

"Low ESR" back then means "humongous ESR" nowadays

[macboy]

I think that datasheet might have been written ages ago (like in the 70s or 80s) when it might have been true that tantalums had lower ESR but it does not really hold true nowadays.

"Low ESR" back then means "humongous ESR" nowadays

Quite true.

I know you are talking about the input side, but be careful of the output side. Modern low-ESR there is a poor choice for old 78xx regulators and disastrous for LDO regulators.

[David Hess]

We bought a module for a project that was 120V AC.  We later decided to locate the module where line voltage wasn't available.  The transformer and diodes were removed so it could operate on 24V DC.  We asked the manufacturer to make a version for us because this would be purchased in number.  It had a 7815 or other internally.  When the first production units arrived they were singing at about 1MHz.  Someone at the factory decided that the power supply electrolytic wasn't needed.  If you ever wondered what would happen without one, now you know.

My first experience with 7805 regulators had a similar result.  I hooked one up to my variable DC bench supply and the output of the bench supply went up.  Adding a 0.1 microfarad disc ceramic capacitor across the 7805 input solved it neatly.

[T3sl4co1l]

I think that datasheet might have been written ages ago (like in the 70s or 80s) when it might have been true that tantalums had lower ESR but it does not really hold true nowadays.

"Low ESR" back then means "humongous ESR" nowadays

Quite true.

I know you are talking about the input side, but be careful of the output side. Modern low-ESR there is a poor choice for old 78xx regulators and disastrous for LDO regulators.

Modern LDOs are available that tolerate low (or high) ESR, as the case may be.  But be sure to read the documentation carefully; like I said, if they're not talking about it, don't use it!

Tim

[gregallenwarner]

My first experience with 7805 regulators had a similar result.  I hooked one up to my variable DC bench supply and the output of the bench supply went up.  Adding a 0.1 microfarad disc ceramic capacitor across the 7805 input solved it neatly.

I'm seeing a lot of designs with a 0.1 uF across the output of a 7805, with a 10 to 100 uF bulk cap across its input. So are you saying an additional 0.1 uF across the input in addition to the bulk cap?

[David Hess]

My first experience with 7805 regulators had a similar result.  I hooked one up to my variable DC bench supply and the output of the bench supply went up.  Adding a 0.1 microfarad disc ceramic capacitor across the 7805 input solved it neatly.

I'm seeing a lot of designs with a 0.1 uF across the output of a 7805, with a 10 to 100 uF bulk cap across its input. So are you saying an additional 0.1 uF across the input in addition to the bulk cap?

The old application notes for the 7805 type regulators only recommended using a 0.1 uF decoupling capacitor close to the 7805 input if the bulk input capacitor was located some distance away which effectively was the situation when I had this problem.  If the bulk input capacitor is close, then the 0.1 uF input decoupling capacitor is not needed.

[gregallenwarner]

Thanks! That clears up a lot of confusion for me!

[janoc]

Thanks. That 2nd video was especially illuminating.

I actually found some reasonably priced tantalums. As long as I make sure and order ones that are rated for a voltage well over my supply voltage, do I need to worry about them bursting into flames? Are there any protective measures I need to take in my design to help protect them from damage?

Don't bother with tantalums, they are an expensive and potentially dangerous overkill for a linear regulator. They have also a nasty habit of of exploding and starting a fire if you solder them in backwards by accident or there is a large voltage spike (trust me, you will do it at least once at some point ...).

Electrolytic caps will at worst release the magic smoke (and their guts) with a bang if you put them in backwards but will not start a fire. And are often an order of magnitude cheaper.

[David Hess]

Thanks. That 2nd video was especially illuminating.

I actually found some reasonably priced tantalums. As long as I make sure and order ones that are rated for a voltage well over my supply voltage, do I need to worry about them bursting into flames? Are there any protective measures I need to take in my design to help protect them from damage?

Don't bother with tantalums, they are an expensive and potentially dangerous overkill for a linear regulator. They have also a nasty habit of of exploding and starting a fire if you solder them in backwards by accident or there is a large voltage spike (trust me, you will do it at least once at some point ...).

Electrolytic caps will at worst release the magic smoke (and their guts) with a bang if you put them in backwards but will not start a fire. And are often an order of magnitude cheaper.

I actually like solid tantalum capacitors but as you point out, they have some destructive failure modes which need to be taken into account and they are more expensive then an equivalent aluminum electrolytic and doubly so if you voltage derate them for higher reliability.

Their major advantage as I see it are consistent performance over temperature and time.

[gregallenwarner]

I've got some LM1117 3.3 volt fixed regulators that I sometimes use for CPLD's, and the datasheet recommends 2 10uF caps, one on each side, and the datasheet specifically recommends tantalum caps.

If I ever use one of these regulators, typically what I'll do is have my main unregulated DC source go into a 7805 first, then send the regulated 5 volts over to the 1117. Since the tantalum caps are behind the 7805 and not exposed directly to the unregulated DC, is there any danger of voltage spikes ever causing failure in these caps?

Is there an alternative capacitor chemistry that's suitable for the LM1117, even though the datasheet says specifically to use tantalums?

[David Hess]

I've got some LM1117 3.3 volt fixed regulators that I sometimes use for CPLD's, and the datasheet recommends 2 10uF caps, one on each side, and the datasheet specifically recommends tantalum caps.

The LM1117 has more stringent decoupling requirements compared to an LM317 or 7805 because it is a low dropout regulator.  A pair of good 47uF aluminum electrolytic capacitors would certainly work for it.

If I ever use one of these regulators, typically what I'll do is have my main unregulated DC source go into a 7805 first, then send the regulated 5 volts over to the 1117. Since the tantalum caps are behind the 7805 and not exposed directly to the unregulated DC, is there any danger of voltage spikes ever causing failure in these caps?

Voltage surges would not be the problem but solid tantalum capacitors have other failure modes.  Excessive surge current (dI/dT) can also cause failure.  The general solution for this is to voltage derate the solid tantalum capacitor.  In other words, do not use a 6.3 volt rated solid tantalum capacitor on a regulated 5 volt supply.  Use a higher voltage one.

Some solid tantalum capacitors are not suppose to have the surge current problem but I still voltage derate them.

Is there an alternative capacitor chemistry that's suitable for the LM1117, even though the datasheet says specifically to use tantalums?

A low impedance aluminum electrolytic capacitor (but not a polymer aluminum electrolytic!) of from 2 to 4 times the value of the recommended solid tantalum capacitor should work fine.  They will have about equal ESR.

If you want to evaluate the effectiveness of your decoupling capacitors with the LM1117, nuke the site from orbit run a load transient response test and compare the results with different capacitors.  It is the only way to be sure.

[T3sl4co1l]

A good hi-rel alternative to tantalum is ceramic + resistor.  Who said ESR needs to be internal?

Aluminum electrolytic is generally not recommended for controlled-ESR applications, because they have a huge tempco (ESR shoots way up at low temperatures, besides aging concerns).

Tim

[janoc]

A good hi-rel alternative to tantalum is ceramic + resistor.  Who said ESR needs to be internal?

Aluminum electrolytic is generally not recommended for controlled-ESR applications, because they have a huge tempco (ESR shoots way up at low temperatures, besides aging concerns).

Tim

Right, but that is still an enormous overkill of a silly linear reg.

[David Hess]

A good hi-rel alternative to tantalum is ceramic + resistor.  Who said ESR needs to be internal?

Nobody.

I have run across this recommendation before in rare application notes.

Aluminum electrolytic is generally not recommended for controlled-ESR applications, because they have a huge tempco (ESR shoots way up at low temperatures, besides aging concerns).

The electrolyte is a liquid so freezing is a problem.  I have worked on designs were this was definitely an issue which led us to completely avoid aluminum electrolytic capacitors from the start.

The temperature coefficient of high dielectric constant ceramic capacitors at low temperatures can also be a problem.  The low dielectric constant ceramic capacitors can be expensive if large values are needed.

[DVDRW]

I have this LM5008 switchmode regulator that needs 100V 1uF ceramic input capacitor.
Can I use four 50v 1uF in parallel + series instead?
Im asking because four 50v caps is ~8x cheaper than one 100v cap.

[T3sl4co1l]

You can always use more...

It's not clear if they require 1uF actual, or if they anticipate that the value will be about 30% of nominal at that voltage, i.e., they only require 0.3uF.

Are 50V 1uF caps really more than four times cheaper?

e.g. cheapest or next cheapest on Digikey at qty 1000,
http://www.digikey.com/product-detail/en/CL31B225KBHNNNE/1276-1291-1-ND/3889377 2.2uF 50V $0.051 (use 2 in series)
http://www.digikey.com/product-detail/en/CL31B105KCHNNNE/1276-1838-1-ND/3889924 1uF 100V, $0.11

Unless you're making a million of this thing, I don't see how that price difference is compelling, and besides, the one takes more than twice the layout area and requires one additional assembly step (and if you're making that many, you'll need to factor in assembly costs and production time, too).

Tim

[DVDRW]

I need ~10 or something (prototype) so Digikey, Mouser is out for me.
My input voltage is max 90Vin and 5V out.
Ebay etc is only option.
1uF 50pcs 100V 13,55$
or
1uF 50pcs 50V 0.88$
So about 15x cheaper