Author Topic: why we tend to not use electrolytics for decoupling to gnd?  (Read 12469 times)

0 Members and 1 Guest are viewing this topic.

Offline amateur_25

  • Regular Contributor
  • *
  • Posts: 106
why we tend to not use electrolytics for decoupling to gnd?
« on: January 25, 2012, 11:29:52 pm »
Hi
I've always used electrolytics for decoupling until I was told it'd wrong to do so. For example what so wrong in using electrolytics across the vcc and gnd pin of a atmega which is what I was doing at the time when my mate said not to do that. He said "they should be used in power supplies" and yet on Dave's schematic I only see one across the supply intake. why?
 

Offline IanB

  • Super Contributor
  • ***
  • Posts: 9628
  • Country: us
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #1 on: January 25, 2012, 11:51:02 pm »
If you look at advice in data sheets they may suggest tantalum capacitors first with electrolytics as an alternative, for example "1 uF tantalum or 25 uF aluminium electrolytic". There may also be a recommendation to put a 0.1 uF disc capacitor in parallel with the electrolytic.

It's not "wrong" to use an electrolytic capacitor, but it is often necessary to use a much bigger aluminium electrolytic than other choices to achieve the necessary performance, and so in a production design this might not be desired. But in a home build design, if you use a big enough electrolytic capacitor and you have room for it, it will be OK.
« Last Edit: January 25, 2012, 11:55:07 pm by IanB »
I'm not an EE--what am I doing here?
 

Online Fraser

  • Super Contributor
  • ***
  • Posts: 9164
  • Country: gb
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #2 on: January 25, 2012, 11:53:22 pm »
Electrolytics are inductive due to their foil spiral construction. Not great for decoupling at higher frequencies.

Aurora
 

Offline amspire

  • Super Contributor
  • ***
  • Posts: 3787
  • Country: au
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #3 on: January 25, 2012, 11:54:59 pm »
For lower frequencies and low current - like a 10MHz microcontroller, an electrolytic would probably work fine. So in a small board that needed an electrolytic capacitor across the power rails anyway, it could double as the micro decoupling cap if it is close to the micro.

At higher frequencies, ceramic decoupling capacitors have lower inductance and resistance then electrolytics, so they are strongly recommended. The SMD packages are preferred over axial or radial lead packages.

I have always felt that an electrolytic or tantalum capacitor is a part that degrades with age, so I prefer to use other capacitors as much as possible.

Atmel has a good discussion on the subject of capacitor choice:

http://www.atmel.com/dyn/resources/prod_documents/doc0484.pdf

Richard
« Last Edit: January 25, 2012, 11:59:25 pm by amspire »
 

Offline amateur_25

  • Regular Contributor
  • *
  • Posts: 106
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #4 on: January 26, 2012, 12:09:08 am »
Thanks you guys. Before there had been many views but no replies so I thought I had asked a stupid question.
After abit of research I actually found my answer. I have found that it's all to do with reasonant frequency which I should of known since am
training to be an electrician. It was useful to have you guys answer my question to confirm my understanding.
 

Offline IanB

  • Super Contributor
  • ***
  • Posts: 9628
  • Country: us
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #5 on: January 26, 2012, 12:16:45 am »
There can be resonance effects, but basically the answers from Aurora and Richard give the main reason, which is high frequency response. Electrolytic capacitors tend to present an increasing impedance and thus become less effective as the frequency goes up. Tantalum and ceramic capacitors work much better at high frequencies and so are preferred.
I'm not an EE--what am I doing here?
 

Offline MarkS

  • Frequent Contributor
  • **
  • Posts: 530
  • Country: us
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #6 on: January 26, 2012, 12:32:48 am »
Atmel has a good discussion on the subject of capacitor choice:

http://www.atmel.com/dyn/resources/prod_documents/doc0484.pdf

Richard

 :o

I've almost always used Tantalum as decoupling caps! Wow!
 

Offline IanB

  • Super Contributor
  • ***
  • Posts: 9628
  • Country: us
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #7 on: January 26, 2012, 01:32:17 am »
I've almost always used Tantalum as decoupling caps! Wow!

Just to give the background to this statement, here is the short summary from Atmel:

What Types to Use: Multi-layered Ceramic and Plastic Dielectrics

What Types Not to Use: Aluminum Electrolytic, Tantalum, and Anything Axial

But I guess the answer really is "it depends" ... it depends on the frequency and switching speed you are dealing with.
I'm not an EE--what am I doing here?
 

Offline siliconmix

  • Frequent Contributor
  • **
  • Posts: 466
  • Country: wales
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #8 on: January 26, 2012, 09:38:31 am »
"Electrolytics are inductive due to their foil spiral construction. Not great for decoupling at higher frequencies." auarora
how do you mean ? could explain this a little ?

 

Offline ejeffrey

  • Super Contributor
  • ***
  • Posts: 2054
  • Country: us
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #9 on: January 26, 2012, 10:36:18 am »
"Electrolytics are inductive due to their foil spiral construction. Not great for decoupling at higher frequencies." auarora
how do you mean ? could explain this a little ?

Aluminum electrolytic capacitors are basically made of two sheets of aluminium film wound in a spiral with an electrolytic solution between them.  The theory is that the spiral winding is an inductor, which causes poor high frequency response.  I find this dubious at best: the current flowing down the anode is exactly opposite the current flowing down the cathode, so the magnetic fields should cancel.  I expect the main source of inductance in an electrolytic capacitor is the leads, and the way the leads are attached to the foil.  That would mean that an electrolytic capacitor has roughly the same inductance as any through-hole capacitor.  However, that means that a 10 uF aluminum capacitor will be no better than a 0.1 uF ceramic through-hole capacitor, and a surface mount ceramic capacitor will be much better.  At high frequency, the extra capacitance doesn't do you any good, you would prefer a smaller surface mount package with lower inductance.

Aluminum capacitors are not good high frequency choices for other reasons.  They do have relatively high series _resistance_  and they have large dielectric losses due to the electrolyte.  If made to conduct a large RF current they will cause unnecessary dissipation and overheat.  That is why it is better to use electrolytic as bulk energy storage, and use one or more small surface mount capacitors for high frequency decoupling.
 

Offline siliconmix

  • Frequent Contributor
  • **
  • Posts: 466
  • Country: wales
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #10 on: January 26, 2012, 12:14:42 pm »
so if the electrolytic has inductance and capacitance.it should oscillate
 

Offline amspire

  • Super Contributor
  • ***
  • Posts: 3787
  • Country: au
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #11 on: January 26, 2012, 12:59:28 pm »
so if the electrolytic has inductance and capacitance.it should oscillate
It is a passive component, so it cannot oscillate, but there is a possibility it could resonate. However, the third factor is the ESR (Effective Series Resistance). At the resonance frequency, the ESR is usually dominating so you probably will not see any actual ringing.  However, above the resonance frequency, the capacitor starts to look inductive and once the frequency gets high enough so the inductance starts to dominate the ESR, it is not much use as a capacitor.

Richard.
 

Offline siliconmix

  • Frequent Contributor
  • **
  • Posts: 466
  • Country: wales
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #12 on: January 26, 2012, 01:52:26 pm »
sorry resonate i meant .good elplantion.
 

HLA-27b

  • Guest
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #13 on: January 26, 2012, 02:25:48 pm »
From Analog Seekrets by Leslie Green

CH6 page 81
Quote
Immediately above resonance, a capacitor is an inductor.

man do I like this book or what!

Btw. electrolytics are less inductive than what they should be because the winding direction changes in the middle, so two opposite magnetic fields cancel each other out.
 

Offline jahonen

  • Super Contributor
  • ***
  • Posts: 1046
  • Country: fi
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #14 on: January 26, 2012, 04:50:06 pm »
It should be noted that only thing that supply rails care, is that impedance should stay as low as possible for as wide frequency range as possible. It is irrelevant whether the capacitor impedance is capacitive (below SRF) or inductive (above SRF). Electrolytic capacitor is good for decoupling in sense that it has low Q-factor thus does not resonate aggressively. Another effect is that minimum impedance region is wider than for SMD ceramic. Bad thing is that the ESR tends to rise significantly when temperature drops, making decoupling essentially disappear at low temperatures.

And, it is not just the inductance of a bare capacitor that matters. Inductance which adds due to physical connection to the device to be decoupled is often the most limiting factor. Even with carefully designed multilayer board, the board level decoupling tends to become ineffective above something like 200-300 MHz. I once compared the impedance difference between 0603 SMD capacitor and strip of copper tape. Copper tape was not actually significantly better. Both were limited by the interconnect inductance.

Inductance of a capacitor depends only on the physical construction (geometry), not on dielectric. This is intuitive since inductance is a property of a magnetic field. Thus if capacitor has same form factor but different capacitance, inductance tends to be the same.

Regards,
Janne
 

Offline olsenn

  • Frequent Contributor
  • **
  • Posts: 993
Re: why we tend to not use electrolytics for decoupling to gnd?
« Reply #15 on: January 26, 2012, 05:22:13 pm »
Don't worry to much about the inductance of the capacitor and all that stuff; just remember to use a high value (100uF - 470uF) electrolytic capacitor where you bring power to your board, and use ceramic capacitors (~ 100nF) at the power pins of each IC you are using. If you have an op-amp or another IC that uses three power points (+V, GND, -V); some poeple will place one cap between the +V and ground and another from Gnd to -V; others will just place one larger capacitor between +V and -V. Experiment if you need to

If you have polyester, polystyrene, polypropylene, or teflon film capacitors, save them for passing AC signals through, since electrolytics will badly distort your high frequency waveforms or attenuate them altogether.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf