Capacity of decoupling capacitors

Hi,

I just successfully built an Arduino on a breadboard (http://arduino.cc/en/main/standalone) and have a question about the two capacitors before and after the voltage regulator:

I know there are several different use cases for capacitors; these are obviously used for stabilizing the voltage/current. The guide explicitly states a capacity of 10 µF.

Can I use capacitors with any arbitrary capacity higher than 10 µF at this place in the circuit? Are there any drawbacks apart from those being more expensive and taking up more space?

As an extreme example: Would there be any disadvantage of using 1 F capacitors other than being huge, expensive and probably totally unnecessary? Is "the bigger, the better" a valid rule of thumb in this scenario?

// EDIT: I just found out this is called "decoupling", so those are decoupling capacitors. I read: the higher the capacity, the slower the "reaction time" - is that true?

These capacitors are used to stabilise the regulator itself, such it will not start oscillating, because it is a very high gain regulator.

The capacitors should be placed as close to the regulator chip, as physical possible. This is because the wiring between the regulator and the capacitor will work as a radio frequency coil, meaning it can be unstable again.

Empiric values are 1µF on both input and ouput, though there are several different political correct suggestions around. That should keep you on safe grounds.

1F could mean the regulator burns out on power being applied as many amps will flow for significant time to charge the cap through the regulator.

Decoupling caps just have to provide enough charge store for short lived events, the regulator feedback amp takes over below a 100kHz or so and extra capacitance has little utility.

I am pretty sure, that a value of 1F = 1 farad = 1000mF = 1,000,000µF is a typing error from cpw83. He obviously didnt mean that, as he is otherwise referring to a 10µF cap in his question.

Anders53: I am pretty sure, that a value of 1F = 1 farad = 1000mF = 1,000,000µF is a typing error from cpw83. He obviously didn't mean that, as he is otherwise referring to a 10µF cap in his question.

He absolutely did mean it - he was making a point of extrapolation. :grinning:

And MarkT has answered it precisely.

In practical terms, larger capacitors tend to have significant effective series inductance and therefore fail to "bypass" high speed transients effectively. That is why small ceramic 0.1µF capacitors are required in addition, and to be placed even closer than the 10 µF (or in the Arduino UNO, two 47 µF are specified).

@OP This may be of interest to you: http://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf

You should always look at the data sheet for that specific manafacaturer because the decoupling capacitor requirements are diffrent for all manafacturers. They quote the minimum value.

A 1F cap is a silly value because it poses more problems than it solves. In genral it depends on the current and how choppy it is but above say 450uF the law of diminishing returns is just about over.

When the 78 series of regulators was first introduced in the 1970s there were some problems with them oscillating at radio frequency. The manufacturers recommended using capacitors of about 0.33 uF on the input and 0.1 uF on the output to prevent this oscillation. The design has hardly changed and those capacitors are still needed.

There is no need and no advantage in using bigger capacitors. In fact bigger capacitors may be less effective at RF decoupling and need a low value ceramic capacitor in parallel. A small 10 uF capacitor with short leads should be OK but I wouldn't use anything bigger. A 1 F supercapacitor will serve no purpose other than to give short term power backup to your circuit.

Russell

Let me ask you a question. You are asking if 1F capacitor is "better" than 10 µF, right?

Let's imagine you are making up a recipe where the cookbook says to add a teaspoon of salt to make the meal taste better. Would it taste even better if you added 100000 teaspoons of salt? I think not.

Capacitors on regulators are there for TWO reasons. 1/ The recommended types and values on the datasheet are there for stability of the regulator, to stop it from oscillating. Used close to the regulator. 2/ Added input and/or output capacitance can be beneficial for loads with transient currents. e.g. a 100-1000uF Low ESR cap on a motor. Or a 100n cap on the supply pins of a digital IC. Leo..

[quote author=Nick Gammon date=1428310161 link=msg=2174217] Let's imagine you are making up a recipe where the cookbook says to add a teaspoon of salt to make the meal taste better. Would it taste even better if you added 100000 teaspoons of salt? I think not. [/quote] Not a very appropriate comparison. ::)

While it is true that salt is bad for your blood pressure and should in fact not be added in cooking at all, and instead provided in a shaker on the table, just three teaspoons of salt instead of one will taste bad, whereas using a 33 µF capacitor instead of a 10 µF will not be a problem at all. ;)

Hi,

thanks for all the answers!

Yes, I really meant 1 Farad as in those two-beercan-sized caps - not that I plan on using those; as I said it’s just an extreme example.

Kind of, but it was more like: “Would it do do any harm?”.

And that’s basically the core of my question - is that a fitting comparison? If I add 100000 teaspoons of salt instead of one, not only will it have a disgusting taste, but you will also die of salt poisoning by eating even a fraction of it.

My analogy would be more like having a recipe (for something spoiling very fast) for one person and just multiplying every ingredient to make it enough for 100000 people, while there’s still just me eating it. Aside from the enormous waste of money, space and work it doesn’t taste bad, it’s not toxic and I’ll only take my one portion and had enough.

MarkT:
1F could mean the regulator burns out on power being applied as many amps will flow for
significant time to charge the cap through the regulator.

So let me see if I got this right after some googling:

The higher the capacity and the lower the resistance to load the capacitor, the more current will flow. A higher resistance will lead to a reduced current and a longer loading time.

So if I “overdose” the capacity in combination with a low resistance to load, at a certain (high) capacity with a certain (low) resistance the occuring loading current will overload and/or fry something - right?

Paul__B:
In practical terms, larger capacitors tend to have significant effective series inductance and therefore fail to “bypass” high speed transients effectively.
[…]
whereas using a 33 µF capacitor instead of a 10 µF will not be a problem at all.

So to sum this up:

There is no real mathematical procedure to determine the required capacity of decoupling caps for a certain component. Being based on empirical values, there is a debatable range of minimal/recommended capacities depending on several different factors; a couple of µFs more might or might not have a positive effect (other than providing a bigger power backup), but won’t do any harm. However at some point you’ll run into very high loading currents and a lowered effectivity in eliminating transients.

Right?

Two or three times the recommended capacity will be just fine. Even five times - a 47 µF.

Larger capacitors have a lowered efficiency in eliminating transients because of their physical size which results in intrinsic inductance due to the very length of internal paths.

And extremely - and it would be that - large capacitors begin to behave like batteries which therefore must have special charging provisions as well as substantial internal inductances.

Here is a paper from Analog Devices and another from Texas Instruments that will answer your question of "mathematical procedure to determine the required capacity of decoupling caps"

http://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf http://www.ti.com/lit/an/sloa069/sloa069.pdf

And this IBM presentation goes into it pretty well with test results http://www.pcbcarolina.com/images/01_pcb_power_decoupling_myths_debunked.pdf

[quote author=Nick Gammon link=msg=2174217 date=1428310161] Let's imagine you are making up a recipe where the cookbook says to add a teaspoon of salt to make the meal taste better. Would it taste even better if you added 100000 teaspoons of salt? I think not. [/quote]

I guess that depends on how much you like salt.

One day, my wife made a dinner for me and my mother-in-law; I forget what the special occasion was - but it was something special, because the main course was a large (5-7 lb) prime rib roast. She roasted it in the oven using the "salt crust" method.

Now - anybody who has looked into this knows that such a cooking method leaves the meat - especially around the edges - fairly salty. That said, my mother-in-law, rest her weary soul (not really - she didn't really like me at all), well - she was an aficionado of sodium.

To that end, when she was served, she grabbed the salt shaker and liberally applied more salt. Not a little, I'm talking shaking the shaker over that poor piece of prime rib for a good two minutes. I'm pretty sure the shaker was almost empty by the time she was done.

Then she ate it. With a smile. All of it. After dinner, tactful woman she was, she remarked to my wife that it could have used more salt.

Then there was the time I went to dinner at my mother-in-laws house for crunchy fried chicken. Guess why it was so crunchy...

Amazingly enough, she didn't pass away from a heart attack or high blood pressure, though I am sure that the excess salt and caffeine (from copius amounts of coffee - I don't think I ever saw her drink anything else in the 20 years I knew her) in her diet didn't help her COPD...

cpw83:
There is no real mathematical procedure to determine the required capacity of decoupling caps for a certain component.

Right?

Wrong. If you know enough about the circuits you can calculate optimum values. You can’t google them.

Russell.

Paul__B: Not a very appropriate comparison. ::)

While it is true that salt is bad for your blood pressure and should in fact not be added in cooking at all, and instead provided in a shaker on the table, just three teaspoons of salt instead of one will taste bad, whereas using a 33 µF capacitor instead of a 10 µF will not be a problem at all. ;)

Well I was talking about 100000 times as much, not 3 times as much.

I think the short answer is that a 1F capacitor would, initially, represent a short circuit on the voltage regulator, it would not dampen down high frequencies.

While it is true that salt is bad for your blood pressure and should in fact not be added in cooking at all, and instead provided in a shaker on the table

That’s not entirely true. Some foods need to be cooked with salt in them, adding it with a salt shaker at the table doesn’t give the same result.

michinyon: That's not entirely true. Some foods need to be cooked with salt in them, adding it with a salt shaker at the table doesn't give the same result.

I just knew someone would pipe up and say that! :grinning:

michinyon:
That’s not entirely true. Some foods need to be cooked with salt in them, adding it with a salt shaker at the table doesn’t give the same result.

How do you put salt in them foods ?
As I see it you are pouring salt over them foods, not in them foods.