which capacitor type and when to use them / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

which capacitor type and when to use them

Move Thread

LAN_403

IamSmooth

Wed Aug 27 2014, 03:16PM

Registered Member #190 Joined: Fri Feb 17 2006, 12:00AM
Location:
Posts: 1567

I'm using a 5v voltage regulator chip for one of my projects. According to the datasheets a 0.33uf and a 0.1uf capacitor are used on the input and output of the three-lead device. When does one want to use a tantalum vs an electrolytic vs a ceramic capacitor?

Right now, I'm using a tantalum on the input and a 0.1uf ceramic on the output.

Steve Conner

Wed Aug 27 2014, 03:48PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Check the regulator datasheet, it will specify what type of capacitors to use. Most regulators like the 78xx, LM317, LM1117 etc. work best with an electrolytic capacitor of some sort on the output. The low capacitance and low ESR of a ceramic can degrade transient response and stability.

Dr. Slack

Wed Aug 27 2014, 04:30PM

Registered Member #72 Joined: Thu Feb 09 2006, 08:29AM
Location: UK St. Albans
Posts: 1659

Designing the control loop round a regulator is quite tricky, and the designers have made some assumptions about the components that will be used, usually including a finite ESR range for the output capacitor, and sometimes an input/output capacitor ratio.

This generally means that 'old' regulator designs will not tolerate the very low ESR of ceramics on the output, preferring the modest ESR of a tantalum or leaded alli 'ytic, or the higher ESR of a SMD alli 'lytic.

If the data sheet was written before large size ceramics became commonplace, they may not 'fess up to the bad behaviour when one is used on the output, so if it doesn't say 'safe with ceramics', then it's best not to use output ceramics. Even newer regulator designs took a while to catch on to the stability problem, MIC5205 is one case in point that got sprinkled around on our designs as point of load regulators, with ceramic output caps. Only a few went unstable, and only after the design was being sold (of course). So just because it's OK today doesn't mean it will be OK at some other temperature, or load current, or capacitor batch. Avoid ceramics unless the data sheet says use them.

Sometimes in order to use ceramics, the regulator data sheet will specify a minimum capacitance. And therein lies another issue, which is capacitor voltage coefficient. I don't know how the manuafactuers get away with this, but a rated 2u2 ceramic can be less than 1uF in use. Read the fine print. Usually the fine print is *not* supplied with the standard data sheet, you have to contact the capacitor manufactuer and demand it, for that particular capactiance, at that voltage, in that package. Change the voltage rating or package size and the voltage coefficient can change because the dielectric changes. Don't think that X7R defines the voltage coeff, because it doesn't. It defines the tempco, which is another story, just less hidden.

If you use a tant or leaded alli, of sufficient value, you will be pretty much bomb-proof on any regulator. Use SMD alli and ceramic with more care that you really thought was necessary.

Steve Conner

Wed Aug 27 2014, 08:49PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

And you thought 3-terminal regulators were simple...

Our standard decoupling for a linear regulator is a parallel combo of a SMD tantalum electrolytic (the cheap variety, not the expensive low ESR ones) and a 0.1uF ceramic. We argue that it will end up supplying ICs that use ceramic caps for decoupling, so the use of ceramics must be OK as long as there is also an electrolytic present to damp it.

IamSmooth

Wed Aug 27 2014, 08:54PM

Registered Member #190 Joined: Fri Feb 17 2006, 12:00AM
Location:
Posts: 1567

Some nice responses. Thanks.
This is the chip:

It does not specify what type of capacitor.

So, when should one use ceramics? I use them with crystals for the clocking on my uP.

If a datasheets suggests a 0.33uf capacitor, do I use one that is listed as 0.33uf or one that actually measures 0.33uf? Most of the caps I measure, as Slack pointed out, are well under the stated value.

Dr. Slack

Thu Aug 28 2014, 06:28AM

Registered Member #72 Joined: Thu Feb 09 2006, 08:29AM
Location: UK St. Albans
Posts: 1659

It's a wide subject, and I was only warning about LDOs, that use PNP pass transistors.

*Very* old designs, especially the high dropout versions like 780x that use an NPN pass transistor, tend to have a much easier time for stability. The clue on that data sheet is that C2 is 'not needed for stability'. Let's narrow down the warning to when the data sheet requires several uFs for stability.

The main recommnedation stands that if you stick to tant or leaded alli you *will* be bomb-proof.

Of course we all use small valued ceramics at the ICs, and the keys there are small, and at the IC, after a bit of trace for a sniff of ESR. It's when the bulk decoupling of the many uFs needed for stability is ceramic that you can run into problems.

It's only higher value ceramics, say >1uF in packages that make you think 'wow, how do they do that', that suffer the really bad voltage coefficiant. Small ceramics will only show a small drop in value, and all other technologies none at all. Any measurement shortfall with those is just tolerance. If the regulator DS says use XuF, they are assuming a reasonable tolerance, so no need to be paranoid. As the 780x DS says output cap not needed, you can of course use any value.

The voltage coeff means that even if you measure your shiny 2u2 ceramic as 2uF on the meter, it might still only provide 1uF at working voltage, and 0.7uF when hot.

Steve Conner

Thu Aug 28 2014, 07:08AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Yup. In the case of the 7805 (or TL780-05 etc) I would use a 10uF aluminium electrolytic on the output and try to mount it close enough to the main power supply filter cap that no input capacitor was needed. This is completely different to what the datasheet recommends, but as Dr. Slack points out, it's not critical.

Bob Pease published some interesting research on this topic. Before a regulator actually goes unstable, it will show noise peaking at some frequency in its output voltage. The frequency and peaking gain depends on load current. You can measure this with a spectrum analyser in prototypes to get an idea of how much stability margin you have.

I recently had a problem with an Intersil switching regulator chip that squealed at 300MHz when it got hot. It was probably due to bad PCB layout, but the chip was on a module that we were buying in from someone else.

IamSmooth

Thu Aug 28 2014, 02:24PM

Registered Member #190 Joined: Fri Feb 17 2006, 12:00AM
Location:
Posts: 1567

If my input voltage to the regulator is coming from a generic AC-DC wall adapter, would the regulator need an input capacitor? I don't know if there is a capacitor on the output stage of the regulator. In that case, based on your 10uf recommedation on the regulator output, should I just use another 10uf aluminum electrolytic on the input?

Steve Conner

Thu Aug 28 2014, 03:04PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Yes, the wall adapter probably has a capacitor inside, but the wire between adapter and your circuit is long enough that you'll need another capacitor at the regulator. In this situation, I would use another 10uF aluminium electrolytic.

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob