Why use 2 decoupling capacitors?

[Oracle]

I've seen posts here that talk about the low component count of the MAX7219 only needing 2 capacitors, and I've always used it with a 0.1uF cap across the power supply, so I wondered why they said it needs 2.  I finally looked at the schematic that keeps getting linked and it has a 10uF and 0.1uF cap in parallel across the power supply.

What is the purpose of doing it that way?   I understand that the 0.1uF is intended to smooth short, sharp voltage spikes and using a 10uF is intended to smooth out ripples over the longer term, but won't the 10uF also deal with the spikes if used alone?  It's not like the 0.1uF will act faster or anything. 

Thanks.

[cmiyc]

It's not like the 0.1uF will act faster or anything. 

Wrong.  It's exactly for that reason.  The smaller cap will react faster to high frequency content.

[retrolefty]

I've seen posts here that talk about the low component count of the MAX7219 only needing 2 capacitors, and I've always used it with a 0.1uF cap across the power supply, so I wondered why they said it needs 2.  I finally looked at the schematic that keeps getting linked and it has a 10uF and 0.1uF cap in parallel across the power supply.

What is the purpose of doing it that way?   I understand that the 0.1uF is intended to smooth short, sharp voltage spikes and using a 10uF is intended to smooth out ripples over the longer term, but won't the 10uF also deal with the spikes if used alone?  It's not like the 0.1uF will act faster or anything. 

Thanks.

In theory any cap bypass cap has less and less capacitance reactance (AC resistance) to ground (which is good , acts like a high pass filter to ground) as the frequency of the noise or spike increases. However in practice different capacitor's dialectic material used for a specific kind of cap can have different ESR (equivalent series resistance) values at same frequencies. So larger electrolytic caps are favored for their effectiveness at lower frequencies (for say AC power ripple frequencies) while ceramic caps are better at higher frequencies. Also where the caps are placed can have a  big effect on how well bypass filtering work. The larger caps work better by placement at where power enters the board or right at the output of a voltage regulators, where as the smaller popular .1ufd caps are usually more effective mounted right at the Vcc terminals of any ICs being protected. You will note that the arduino board design for a 328p chip has 3 .1ufd caps wired close to the chip at the Vcc, Avcc, and Aref pins, all used for bypass filtering.

That leads to what do words like required, needed, recommended and such mean when designing ones own circuits. A lot is gained from past experience. Also checking the datasheets for any ICs being used and be useful, some ICs are assuming and require such external bypass filtering, while other chips are less sensitive to the typical noise levels. I usually when designing a circuit used the minimum number of filter caps I thought I could get by with, however I also used at least one .1ufd ceramic cap mounted right at the Vcc pin of any IC used in the circuit. I typically had less problems following that rule. However there is no perfect rules that will always solve and prevent noise caused problems, following good general practices is just the starting point and further debugging may always be required in some cases. A good O-scope is very useful for problem circuits you are first building. But so far I've found simple AVR 328p circuits to be pretty trouble free using just typical bypassing rules.

Others may have better or further recommendations based on their specific experiences.

Lefty  

[winner10920]

The larger cap also has larger inductance, many times perhaps that of the smaller value cap, the inductance limits the frequency range that it will effectively filter to, the leads of the capacitor has alot of inductance, so keep non smd caps tight
the best is a .1uf smd cap right at the supply pins as well as the 10uf electrolytic

[Oracle]

In theory any cap bypass cap has less and less capacitance reactance (AC resistance) to ground (which is good , acts like a high pass filter to ground) as the frequency of the noise or spike increases.

There's a lot of information in there, thank you.  So first off, if there are several chips, I would just put the single 0.1uF near each chip's power pins as I normally do now, and not worry about the 10uF near each chip, that's something to think of as part of the power supply circuit as needed.  For a commercial switching power supply, would the electrolytic near it make a difference ?

Second it's not simply the values that are important but the chemistry, so am I right that if I were to use a 10uF ceramic SMD cap right next to the 0.1uF ceramic cap near the power pin, it would have no more benefit beyond the 0.1uF cap alone?

[winner10920]

I think it wouldmt make a difference on an ceramic smd part
But I know electrolytic is basically two long foils, sometimes a couple feet long, so they havealot of inductance,  I can't imagine the 10uf smd counterpart having that problem tho
and for a project with many ics, definetly one smaller one at each ic, and maybe a 10ud for every two or three ics spread out, especially if its a larger board, the other side from the supply would benefit the most from the 10uf cap

[retrolefty]

In theory any cap bypass cap has less and less capacitance reactance (AC resistance) to ground (which is good , acts like a high pass filter to ground) as the frequency of the noise or spike increases.

There's a lot of information in there, thank you.  So first off, if there are several chips, I would just put the single 0.1uF near each chip's power pins as I normally do now, and not worry about the 10uF near each chip, that's something to think of as part of the power supply circuit as needed.  For a commercial switching power supply, would the electrolytic near it make a difference ?

Switching voltage regulators can be tricky finicky things to build from scratch and referencing the datasheet and application notes for the switching IC is always recommended. On finished switching regulator modules one can generally assume I would think that the basic requirements for the regulator are already mounted on the module and just using .1ufd caps at the ICs in your project should be all that is required?

Second it's not simply the values that are important but the chemistry, so am I right that if I were to use a 10uF ceramic SMD cap right next to the 0.1uF ceramic cap near the power pin, it would have no more benefit beyond the 0.1uF cap alone?

No, that is too general and encompassing a statement I think. You have to state the desired frequency bandwidth of the noise you are trying to effectively bypass to ground, larger cap sizes will almost always be more effective at lower frequencies. Recall the formula for capacitance reactance to see the effective AC resistance of two different cap sizes. Lower resistance to ground is always the desired property of a good bypass cap.

Lefty

[sdturner]

Always keep in mind that there is no such thing a an ideal capacitor. In school and in SPICE a capacitor is a capacitor. In the real world there are many different kinds of capacitors and each has a different set of limitations and dissadvantages.

All capacitors have some resistance and inductance. Small ceramic caps have the least, but within any technology family higher value caps will usually have higher parasitic resistance and inductance.

Since the capacitors are really RLC tuned elements they will only behave like their rated capacitance within a certain frequency band. At other frequencies the effective capacitance will be lower. In general small capacitor values will react to higher frequencies and larger values will react to lower values.

In most general purpose circuits this is not a concern, but when you are pushing parts to their limit of high frequency, small size or high power it can be critical.

Steve T

[BillO]

There are 2 things going on here and each capacitor serves a different purpose.

The 10uF capacitor, in conjunction with the supply leads, is acting as a line-load decoupling circuit.  It keeps the power more even at the load point.  They are generally not required where load currents are fairly constant, or where the power supply is very close to the load circuit and connected via a low impedance line.

The .1uF capacitor, as already pointed out is a bypass filter that bypasses very high frequency spikes to ground.  These need to be placed as near as possible to the power leads of each IC.  They act to suppress both the spikes generated by the closes IC and and spikes that may try to enter any IC.

Decoupling usually deals with relatively lower frequency phenomenon than does the bypass capacitor, so it's response time in not as critical and neither is the capacitor.  Cheap aluminum electrolytics are usually just fine for most of our purposes.  Bypass capacitors must deal with near pico-second pulses that might be much sorter in wave length while traveling through copper than the distance between ICs.  They need to have the fastest possible response time.  While deciding on the suitability of a particular capacitor for bypass needs is not a simple thing, ceramic capacitors offer the best bang for the buck in the frequency ranges that we are usually dealing with.  Ceramic disks are probably the best if you can afford the space requirements, but hey are not very compact and stick up a bit from the board.

[MarkT]

I don't think there's any frequency dependence in the terms bypass or decoupling - you can say high-frequency decoupling, low-frequency bypass - well I'm brought up on Art of Electronics and I don't think they make that distinction.

Perhaps this is a 'dialect' thing.

Anyway my two-pennies-worth on decoupling is inductance - to be effective high-speed decoupling caps need to be on a low-inductance path from the power pins of the device in question.  If an electrolytic is wound so that the leads both go to the same end of the foil strip then there is a low(ish) inductance path to the first part of the foil (and it doesn't take much foil area to have 0.1uF capacitance).

If however the leads go to opposite ends of the foil before its wound up there is a much more tortuous path - more inductance, and high speed signals are not going to see that capacitance so well (or so fast).  So details matter.

Of course at audio frequencies the two arrangements behave the same.

[BillO]

Note the use of the term 'usually'.  My response was in context of the use.

[JChristensen]

Glad to see this thread, am currently designing a board with a MAX7219 and I'd forgotten the recommendation of 10µF + 100nF.

I recently found some 10µF MLCC types in an 0805 package that are very competitively priced as compared to the 100nF caps that I've been using. Looking at the charts in the datasheets, the 10µF caps actually have lower impedance than the 100nF caps. The datasheet for the 100nF caps doesn't have anything to say about ESR, but the ESR for the 10µF caps is 0.01? or less from 20kHz to 20MHz. Conclusion: one 10µF near the MAX7219.

[copiertalk]

A question off the original if I can intrude into the thread.

When two or more caps are in parallel are the not added? such as

C = (10 ?F) + (20 ?F)

    = 30 (?F)

[JChristensen]

A question off the original if I can intrude into the thread.

When two or more caps are in parallel are the not added? such as

C = (10 ?F) + (20 ?F)

    = 30 (?F)

Yes.

[Docedison]

In my few years at the Electronics... I don't believe I have Ever seen a board fail due to too many Bypasses,, But I've seen a bunch that failed because of too few and more than a few of mine suffered that initial condition.
Having said that I utterly fail to find any sense in deciding how few you can get away with using in a design. The parts cost much less than the possible time figuring out why it doesn't behave "exactly" the way you thought it should. MY Very simple Rule is a 100 to 330 uF electrolytic at the Vcc connection, a .1uF cap per Ic and a 10uF cap for any board and increased by one every 3 IC's.
What Hasn't been mentioned well here is the inductance of the power supply leads and more important the traces carrying power on the PCBI. It's like a whole series of little inductors... in series with both power and grounds distributed across a board.
The Very best boards I Ever worked on had 4 layers top and bottom were grounded, more as an EMI shield that anything else, there was a Vcc plane, Split as necessary for ADC's and other more sensitive or low noise devices and an internal ground plane, Granted they were early Military stuff but there were no board issues and they were high speed synthesized radio PCB's... with 100 Mhz system clocks.
My design philosophy is very simple... You Don't have to populate the pads and holes... But it is sure hard to "Stick" em on later.
Place the parts on the board or at least the footprints... Stuff them all on the first copy and then remove the ones you think... extra, and see that your power supply is clean with an O'scope... If not start putting them back until it is.


Doc

Doc