Caps per pin?

Go To Last Post
18 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Not really sure if this is AVR, General Electronics or Off Topic...

 

There have been more posts than I can count here on AVR Freaks that recommend caps for and close to each power pin.

 

So why do the Arduinos generally ingnore this?

 

Not counting the caps associated with the power regulator, we see:

 

  • Arduino Uno - 1 cap for ATmega328P and 1 for ATmega16U2

    (https://www.arduino.cc/en/upload...)

  • Arduino Pro Mini - 1 cap for the three power pins

    (https://www.arduino.cc/en/upload...

  • Arduino Micro - 2 caps for 4 power pins on the ATmega32U4

    (https://www.arduino.cc/en/upload...)

  • Arduino mega2560 - 3 caps for the ATmega2560 and 1 for ATmega16U2

    (https://www.arduino.cc/en/upload...)

 

The following do follow the cap per power pin guideline:

  • Arduino Gemma
  • Arduino Zero

 

Thoughts anyone?

David (aka frog_jr)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Saving a buck?

Didn't know better?

 

For a simple, general purpose, inexpensive PCB with a short distance from the power supply to the uC, and no external hardware generating noise, it will likely work without all of the caps.

 

But, as the distance between the power supply and the uC increases, and one begins adding lots of external circuitry, which both puts noise on the power rails, (and other traces), and draws its own current in small spikes, and perhaps over a range of voltages and temperatures the reliability will begin to fall.

 

For a reliable system it is just easier to follow the "rules" and install the by-pass caps. 

 

An intermittent PCB failure (bug, error, reset, etc.), from a poorly by-passed uC can be a very tricky problem to recognize and verify as the culprit, (albeit easy to fix).

 

Note when you read the data sheets for some chips like ADC's, DAC's, op-amps, RF chips, etc. some of them devote an entire section of the data sheet to by-passing recommendations.

 

For most PCB's I lay out I include all of the by-pass caps.  That said, there have certainly been a time or two when I knowingly "cheated" and skipped a cap or two.

In that case the schematic will have a statement noting that Cxx wasn't installed, as a reminder in case I have a problem down the road, or share the schematic

 

JC

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

If there is just a few mm's of track between the separate VCC pins and separate GND pins, a single capacitor of 100nF is sufficient IF there is not a lot of RFI around. Like Doc explained. For the TQFP package and smaller that can be easily achieved.

For a professional (read: industrial) design, I would definitely go for a capacitor on each Vcc pin. Cost impact is minimal. Like Doc explained smiley

 

Also note that the Arduino isn't designed for industrial use.

 

Sidenote:

A few years ago (time flies ...) I discussed a design with the folks from elecfreaks.com: we were considering to design an Anaduino: an Arduino variant with special attention paid to decoupling, specifically for the analog part. In the end the conclusion was that the market for such an optimized design was small. So we decided not to invest time and money in that design.

 

cheers

 

Nard

A GIF is worth a thousend words   They are called Rosa, Sylvia, Tessa and Tina, You can find them https://www.linuxmint.com/

Dragon broken ? http://aplomb.nl/TechStuff/Dragon/Dragon.html for how-to-fix tips

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I just checked the Adafruit pro trinkets and to my surprise they only have a 10uf bulk cap, no 100nf decoupling cap. Maybe the 10uf is good enough and close enough to not need the decoupler. In my day job I definitely aim for 1 decoupling cap per power pin, with bulk caps dotted around the chip in question. Not always possible due to space but you do the best you can...

It is better to use multiple same value decoupling caps instead of multiple values due to something called antiresonance. The varying Rs and Ls of the connections will give a broad frequency coverage to reduce a wide range of the transients we want to remove. High speed signals (hundreds of MHz) eg data busses are much more complex and you start to need to think about transmission line theory, and looking for monotonic signals.

Murdo.

There are already a million monkeys in front of a million keyboards, and the internet is nothing like Shakespeare!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

frog_jr wrote:

There have been more posts than I can count here on AVR Freaks that recommend caps for and close to each power pin.

So why do the Arduinos generally ingnore this?

 

It depends what you are chasing.

The average MCU is very tolerant of Supply decoupling, because they work over a wide supply range anyway.

 

Where it matters more, is say you use the Vcc as a ADC reference - now a few 10's of mV of noise, are significant.

 

MCUs with on-chip regulators, have stability requirements that need caps above a certain size.

 

Some circuits show 100nf + 4.7uF, but that is to allow for Electrolytic Caps.

If you use MLC SMDs then a 4.7uF is likely to be fine.

 

If the PCB is 4+ layers, then the PCB itself already has good HF decoupling.

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Lets add another dimension to this: history. When these recommendations generally evolved, 15-20 years ago, the predominant bypass caps were leaded ceramics. They did not perform as well as the modern SMT ceramics. A larger leaded ceramic, say 1uf, has a self resonant frequency in the general range of a few 10s of MHz (1uf series resonant with lead inductance) and do not bypass very well above 25-40MHz. So, the historic recommendation was to add a .01uf in parallel with the larger one, Its self resonant frequency is up in the 100MHz range (plus or minus a fair amount). This made a very effective bypass for the very narrow current spikes that are generated by many thousands of gates switching at the same time.

 

Now, lets come to the current time with good (and inexpensive) 1uf to 10uf (and more) ceramic SMT caps. These have such a high self resonant frequency that a smaller parallel cap really is not necessary. In a real sense, improved technology has made the historic rule of thumb not very useful. 

 

Personally, I use 1uf 0805 SMT caps like they are candy. This has proven to be very effective.

 

Jim

Jim Wagner Oregon Research Electronics, Consulting Div. Tangent, OR, USA http://www.orelectronics.net

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I scatter .1uF 6030 size caps everywhere: Every power pin or anything that looks remotely like a power pin. If there's any analog work to be done, I separate the analog supply from the noisy digital supply with a few microhenrys.

 

I've seen the addition of a single .1uF cap settle down whole systems full of glitches and unpronounceable problems. Consider them jellybeans, except that jellybeans cost more.

 

I've seen guys say the capacitor is a critical part of the chip's power supply. Given that our chips work at a wide range of voltages, I'd say the capacitor might be THE critical part of the chip's power supply.

The largest known prime number: 282589933-1

It's easy to stop breaking the 10th commandment! Break the 8th instead. 

Last Edited: Thu. Jul 21, 2016 - 03:31 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I use a ferrite bead to isolate analog Vcc from digital Vcc. Same principle but small and relatively cheap.

 

Jim

Jim Wagner Oregon Research Electronics, Consulting Div. Tangent, OR, USA http://www.orelectronics.net

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I just haven't found any guidance on how to pick the bead, but next time I put together a digikey order, I'll get a few hundred tiny ones.

The largest known prime number: 282589933-1

It's easy to stop breaking the 10th commandment! Break the 8th instead. 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Jim, we used FB's on the AVCC line as well, but I don't remember which one we used, it was an 0603 size bead.   What part # bead do you use.

Like Torby I'll pick some up with my next order. 

 

Jim

 

 

Click Link: Get Free Stock: Retire early! PM for strategy

share.robinhood.com/jamesc3274
get $5 free gold/silver https://www.onegold.com/join/713...

 

 

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I use a hard to find Toko 0805 with an unusually large high-frequency resistance. There are three important parameters (beyond size and cost). They are: (1) DC resistance, (2) High frequency resistance, and (3) Lower corner frequency. To a first order approximation, they look like a resistor followed by a parallel combination of an inductor and resistor. At low frequencies, you see the series R (the DC resistance). At frequencies where the reactance of the inductor is larger than its parallel resistor, you see that parallel resistor (the high frequency resistance). The lower corner frequency is where the inductive reactance equals the DC resistance. 

 

I try to get the lower corner frequency well below the period of the likely signal of concern (usually the MCU clock). The lower, the better. In this case, better = more effective filter.

 

You also need to be a bit careful of the DC resistance. It generates a DC drop depending on the average current (if it is well bypassed on the load side). There can also be power dissipation issues with this resistance. There MAY be a maximum current rating so be careful of that; it is probably dissipation limited but I would not count on that.

 

Jim

Jim Wagner Oregon Research Electronics, Consulting Div. Tangent, OR, USA http://www.orelectronics.net

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I find lots at digikey for 3 or 4 cents each qty 100.

The largest known prime number: 282589933-1

It's easy to stop breaking the 10th commandment! Break the 8th instead. 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ka7ehk wrote:
I use a hard to find Toko 0805 with an unusually large high-frequency resistance.
Do you have a part number or spec on those?

David (aka frog_jr)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Sorry, I was wrong - Murata... Have only been able to find at Mouser. 0805 package.

 

Murata 

BLM21A102S

      Mouser: BLM21AH102SN1D

 

Yes, cheap is nice. But, there will be a time, not very long in the future, where my products will have to be CE approved. Ferrite beads will be a crucial  part of that and I will have to have consistency and repeatability. For that, I will not be able to use EBay or offshore-nobrand parts.

 

Jim

Jim Wagner Oregon Research Electronics, Consulting Div. Tangent, OR, USA http://www.orelectronics.net

Last Edited: Fri. Jul 22, 2016 - 09:08 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The gotcha with ferrite beads is the current rating. They become less effective as the current increases. Push too much current through them and the ferrite saturates and it becomes just a piece of wire. So it pays to read the specs carefully.

Jim, it is my understanding that the emissions requirements for FCC and CE are similar. Where they differ is in susceptance. Ferrite beads are good for stopping rf getting in and getting out.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Those BLM21A102S are 6.3 cents in quantity 100 at Mouser. Pretty reasonable assurance there.

The largest known prime number: 282589933-1

It's easy to stop breaking the 10th commandment! Break the 8th instead. 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0
  1. I'm not sure anyone would pick out an Arduino board as an example of conservative and perfect hardware design.  (that said, older versions had more caps...)
  2. Arduino Micro has two caps for 3 power pins, and seems to have additional caps for VUSB, UCAP, and AVCC1.
  3. These are small boards with (mostly) on-board voltage regulators, so there are usually additional caps near the regulator that probably still qualify as "near" the chip when all factors are taken into account.

 

As for Ferrite beads; aren't those usually specified by their resistance at ~100MHz, for use mostly in RF noise suppression?   For AVCC, don't you want an LC circuit whose resistance is "high" at the sort of frequencies that the ADC is actually measuring (ie MUCH lower frequencies, implying a much higher inductance?)  I guess a bead couldn't hurt, but it doesn't seem like it would help a lot, either.

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Kartman -

 

Through the number of FCC/CE certifications I've had a hand in, I think that your comments are pretty much correct. FCC basically says "You shall emit no more than some limit and you have to accept what ever interference is around". CE says "you shall emit no more than some limit and you must accept local fields up to some limit". In all cases, the limit is frequency dependent. CE is also very specific about what "accept" means: no undesired or faulty behavior. For example, I worked on a remotely controlled (by wire) pan/tilt/zoom video camera mount. At certain  interference frequencies of certain field strengths, the thing would start panning, with no control possible. That was a failure and had to be fixed for CE. FCC didn't care - as long as you can put up with it, and it does not burst into flames, its OK. 

 

Yes, ferrite beads protect both ways, PROVIDED there is sufficient bypassing on both sides. It is all just a voltage divider. The (hopefully) large series Z (the bead) with a (hopefully) small shunt Z (bypass cap). This is one of the reasons why high frequency performance of bypass caps is so important. If the cap has a low self resonance, it can be relatively high reactance at some high frequency and the beneficial voltage divider is shot.

 

westfw - logically, your argument seems correct, but it does not quite square with my experience. To be fair, most of my concern has been stuff leaking out and that is mostly at higher frequencies. But, if you look at the actual supply noise, it tends to be a negative going spike at every clock edge. Given that the MCU clock runs a lot faster than the ADC clock in an AVR, I think there is plenty of opportunity for those spikes to cause ADC mayhem. Because the spikes are so narrow (roughly same as gate rise time), there is lots of high frequency energy in them and the beads seem to do a pretty good job. That is not to say that something else might do better, maybe even a lot better, but for my applications, they have seemed to be more than good enough.

 

Jim

Jim Wagner Oregon Research Electronics, Consulting Div. Tangent, OR, USA http://www.orelectronics.net