5 replies to this topic

[Spork]

Was just looking at MakerShield and noticed that they point out that they have power filtering capacitors on the 5v and 3.3v power. Why is this necessary or desirable? I had previously purchased two ProtoShields from SparkFun. I assembled the first one as instructed, but I left the second one bare except for stacking headers and the reset switch. I checked the schematic on SparkFun and the caps are just sitting between 5v and GND. Since I'm only using 3.3v on this second shield, I guess there's no point in soldering them in where they'd be filtering 5v. Should I consider putting one (they're 0.1 uF) between 3.3v and ground? Thanks in advance for any suggestions.

[Mario Vernari]

A 100nF (good) cap is always suggested on the power supply as close as possible to the chip. When you wire the power supply (either +5 or +3.3), there is always a small resistance, together with an inductance (i.e. a parasitic electromagnetic effect) on your wire. That could lead a malfunction, especially when the frequency involved is several MHz or more. The cap is used to attenuate/eliminate these problems. That's the way it should be places very close to the chip. Note that the electrolytic caps are *NOT* a good choice, unless they are glued together with polyester/polycarbonate/ceramic caps. Bear in mind that a good grounding is much more important than the caps on the power line. Cheers

[Bill E.]

I wholeheartedly agree with Mario. The 100nf (or 0.1uf as we call them here in the USA but I think that is slowly changing as we align with the rest of the civilized world!) are actually noise "decoupling" capacitors AKA filter caps. Electrolytic caps with high uf values do the job of "smoothing" out ripple on a DC source while small value caps, like the 0.1uf (100nf) to 0.01 (10nf) caps tend to shunt quick, rapid, spurious noise spikes to the opposing lead/wire thereby decoupling these spikes from your power rail. A really good rule of thumb is: a capacitor acts like a short (low resistance) to AC but an open (high resistance) to DC. The lower the value of capacitance leads to a faster response to change. EE98 here. Always be cautious of the VOLTAGE rating of a cap. Make sure you allow at least a 25% overhead on that value. This question is actually quite a good one. A good topic for a forum, Spork. If you are certain that you will never have anything tied to the 5V pin of the shield then you would not need any decoupling/filtering on that pin. But, why not mount the cap? It's small and comes with the kit anyway. Just in case. The last thing you want is to have your entire shield stack "go into the weeds" because of a motor or relay or remote ESD discharge, right? A bag of 0.1uf (104 marked on the cap body) should be a mandatory item in any designers box. Regards!

[Spork]

Thanks for the responses. If I'm reading correctly, it seems like one could or should have several of these capacitors, perhaps one close to the voltage supply pin of each chip? My first thought was that one such capacitor between + and GND should suffice for the entire project, but I guess that the spikes could arise almost anywhere and there would be no protection if a spike originated between the cap and the supply to some chip. Correct?

[Chris Seto]

You need a decoupling cap as close as possible to each chip, prototype or production, it doesn't matter. Always follow the app circuit where applicable for how to filter the chip's Vdd.

[Mario Vernari]

Spork, if you playing with circuits on a breadboard, maybe the caps have no sense, because the whole circuit is completely a mess. Often the circuit may have malfunctions, even is totally correct. The problem is the wiring, but the Chris consideration is also correct. On a breadboard the correct wiring is almost impossible: especially when you deal with relatively high frequencies (10-20MHz and over), where it must be considered the "impedance", the length of the wirings, etc. I found a problem, together with Stefan, around the SPI usage on a breadboard. My Netduino have the SPI wiring of about 15-20cm in length connecting to the breadboard. If I set the clock up to 1-2MHz there is not any problem, but at 20MHz and over the circuit is not working anymore. If you deal with signals such these, the best way is to use the proto-board, keeping the wiring as best as possible, together as the caps close to the chips. However, the very best way is the custom PCB, but I mean that there would be a prototypal stage also. Cheers