Bypass capacitors... I don't get it (I'm a twit)

Thread Starter

MagicMatt

Joined Sep 30, 2013
117
I tried to read the link in the sticky, but the page has been dead for the last 24 hours:
http://www.interfacebus.com/Design_Capacitors.html

I understand I need them. I don't fully understand quite how they do what they do, but I understand that the chips use power, which makes the power supply ripple and wibble, and the caps counteract that. I think.

I don't really understand the logic of using a 22uF electrolytic and 0.1uF mica on each chip, but I'm prepared to just "go with it".

However, is there a problem using larger value electrolytics? I ask as I've already got quite a few 16V and 25V at various values between 100uF and 1000uF, which if I could use would mean I don't have to buy any more. :)

Also, should I just "stock up" on 0.1uF mica caps to use all the time next to any chips? (I don't have any mica caps in my box) or again could I just go for higher value as a lazy "catch all" - ie 1uF mica?
 

wayneh

Joined Sep 9, 2010
17,496
Yes, I have a stock of 0.1µF ceramic capacitors for bypassing every single IC that I use. I recommend you do the same.
The larger electrolytic is good for low frequency, "DC" smoothing, whereas the small ceramic can send high frequency noise to ground, bypassing the IC. You'll sometimes even see 3 bypass caps on an IC.
 

crutschow

Joined Mar 14, 2008
34,285
You generally only need one high value electrolytic on the power supply to smooth the lower frequency fluctuations. The value is not critical and usually any value above about 10μF is adequate.
But each chip needs a 0.1μF ceramic (not necessarily mica) cap across the power pins to ground for the high frequency transients. Larger capacitor values tend to be less efficient at this since they have more parasitic inductance (and I doubt you will find a 1μF mica capacitor).
 

alfacliff

Joined Dec 13, 2013
2,458
too many high value electrolytic s can cause problems of surge currents when power is applied. having small and medium caps makes the supply rails lower impedance to all frequencies. large electrolytic caps have significant inductance that make them unsuitable for higher frequencies. the smaller ones in parallel with them bypass the higher frequencies.
 

Lestraveled

Joined May 19, 2014
1,946
The longer a wire is, the more inductance it has. The more inductance, the lower the voltage will drop when a chip demands current (instantaneous). De-coupling caps cancel out the wire inductance and act as short term energy reservoirs. De-coupling caps are your friends.
 

Thread Starter

MagicMatt

Joined Sep 30, 2013
117
nsaspook, I think you're confusing being lazy with simply finding it difficult to understand. Useful PDF though, I will print it and read it through a couple of times, and see if it can make the water a little less like mud.

I have ordered 100x 0.1uF ceramic and 25x 0.33uF ceramic. Thanks for your help! :)
 

WBahn

Joined Mar 31, 2012
29,979
The default strategy I used on the test boards I did for our ASICs, which were mixed-signal and sometimes pretty sensitive to noise, was to decade-tier the bypass caps. I generally used every other decade. I used 0.1uF ceramic at every chip and also to stitch the power planes together, I used 10uF tantalum distributed around the board, generally one every two to four square inches, as bulk charge storage, and I used a 100uF to 470uF aluminum at the power entry point of the board. That generally worked very well. On particularly sensitive or high speed (keeping in mind that "high speed" for us back then were external signal frequencies up to a couple hundred megahertz) I would bypass chips that connected directly to ours (as well as ours) with 0.01uF caps (in addition to the 0.1uF) and, for one design, actually had to add 0.001uF caps as well (and I did that on a hunch to just rule it out and it actually made quite an improvement). I also used ferrite chokes in several designs to further keep digital noise out of the analog sections, though I never really got a good feel for just how well they did or did not help.
 

Thread Starter

MagicMatt

Joined Sep 30, 2013
117
My frequencies are somewhat lower... the only high speed clock is the internal PIC at 16MHz... the PWM is 250Hz and the I2C just under 100kHz. :)
 

nsaspook

Joined Aug 27, 2009
13,086
nsaspook, I think you're confusing being lazy with simply finding it difficult to understand.
I know but usually when something is difficult to understand it's not because of complexity, it's because you're missing some basic fact that links the information into your current knowledge so you can't see the pattern.
 

Thread Starter

MagicMatt

Joined Sep 30, 2013
117
True... but lazy would be if I just said "Ah f*** it, I can't work it out, I'll just slap anything in there and hope it works"... ;)
 

k7elp60

Joined Nov 4, 2008
562
MagicMatt,
I have read the entire post and I did not see this reason for the .1uF capacitors other than the high frequencys. Going back to capacitor theory in that capacitive reactance= 1/(6.28xfC). In reality the XC decreases as the frequency increases. If you are using a 0.1uF capacitor at 16Mhz that is connected to the DC supply of the circuit and ground, the 16Mhz will take the capacitor to ground because the XC is very close to 10 ohms. If this the way I look at it.
 

WBahn

Joined Mar 31, 2012
29,979
The idea is to bypass (hence the name) high noise frequency signals from the near-chip supply rails from power to ground instead of letting those signals get into the IC (or whatever circuit is being bypassed). So want a low impedance at the noise frequency.

What your analysis isn't taking into account -- and which is why you decade tier bypass banks -- is the effect of parasitics. Above a certain frequency, the parasitic inductance of a physical capacitor begins to dominate and the device, as a whole, looks more like an inductor than a capacitor. The point at which this occurs is generally known as the self-resonant frequency. In general, the higher the capacitance, the lower the self-resonant frequency.
 

MrChips

Joined Oct 2, 2009
30,716
If you want to cover a wide range of frequencies you have to stagger the capacitance values.

You may even see 1nF, 10nF, 100nF and 1000nF all on the same bypass bank of capacitors.

100nF and 10μF together is a common combination.
 
Top