I know a digital a$$ kicking is coming for this post...
sunblock
Posts: 55
I've been doing so much reading on bypass, decoupling, analog and digital grounds, separating them, don't separate them, etc. Please help me with a simple programmer-type logic train of the differences when and where to use what. I'm making sure I have 0.1uf capacitors between the VDD and VSS pins on everything I own except the dog. But I have replies to posts also saying 33uf caps from analog source to analog ground. Then some example suggest 10uF caps from analog source pins to analog ground. Do I need 33uF or 10uFs from source to digital grounds, like the 0.1uF bypasses?
If someone needs an example to help me understand, on page 40 of the Nuts and Volts of Basic STAMP (Volume 1) is a connection/schematic diagram. Several questions I have: 1) the 10uF tantalum capacitor I assume is connected to a analog ground? 2) does that mean that the GND pin is connected to a digital ground? 3) If so, do you split it down the middle (sounds stupid I know)? 3) Shouldn't there also be a 0.1uF cap between the VCC and GND as a bypass? Or don't yu use those if the 10uF tantalum is used.
Be kind folks, please, not everyone is as smart as the general audience in these forums.
If someone needs an example to help me understand, on page 40 of the Nuts and Volts of Basic STAMP (Volume 1) is a connection/schematic diagram. Several questions I have: 1) the 10uF tantalum capacitor I assume is connected to a analog ground? 2) does that mean that the GND pin is connected to a digital ground? 3) If so, do you split it down the middle (sounds stupid I know)? 3) Shouldn't there also be a 0.1uF cap between the VCC and GND as a bypass? Or don't yu use those if the 10uF tantalum is used.
Be kind folks, please, not everyone is as smart as the general audience in these forums.
Comments
The rule with capacitors is that the 0.1 uF ceramic caps need to be distributed around the board near their respective ICs in order to keep short-term transients at bay. The reason for ceramic is that they have the lowest equivalent series resistance (ESR) of the various capacitor families, so are the most effective short-term transient clamps.
Tantalums are next up the ladder for ESR and are typically used as filter capacitors. These filter out noise from off-board sources such as EMI. They are also spec'd as the output caps for some voltage regulators that require not only high bulk capacitance but also relatively low ESR. They are placed very near teh outputs of the regulators they filter and do not substitute for the distributed capacitance provided by the ceramic bypass caps spread about the board like a multitude of small struts holding Vdd and Vss a fixed distance apart. In addition to their regulator output filtering, my personal rule is that every board I design must have a tantalum cap or equivalent high-value ceramic cap (typically 10 uF) on every incoming supply line, regardless of how close it is to its power source.
Up the ESR scale from tantalums are aluminum electrolytics. These are typically used for providing high bulk filtering for loads with high transient current requirements, such as the RC servo motors used on, say, the BOE Bot. They are sometimes used as filter caps on the inputs or outputs of those regulators that remain stable with their higher ESR values. Again, they are not a substitute for the distributed ceramic bypass caps.
In mixed digital/analog systems, or in systems that switch high currents, the analog or load ground needs to be kept separate from the digital or control ground except at the common point where the mixed signals come together. This is done to prevent digital current spikes from affecting the accuracy of the analog readings or to keep high load transients from corrupting the control signals by sending current surges through the control gournding. It is often the case, in these situations, that the analog or load side have it's own filter and bypass caps, which further isolates each side of the system from noise emanating from the other side.
-Phil
Generally one should not use tantalum capacitors in timing circuits such as in RC oscillators.
The problem is tantalum capacitors have a weird characteristic where they exhibit tiny shorts.
In the normal filter capacitor usage these tiny shorts are blown out by the relatively stiff current
from the power supply and regulator, No fowl.
However, most timing circuits are in relatively high impedance circuits, 10K to 1M resistance or so.
These circuits don't have the Moxy to clear these tiny shorts and the timing circuit quits running.
Tantalums have pretty small leakage current specs deceptively suggesting it's acceptable to use
very high valued timing resisters.
This shorting thing is apparently unrelated to the leakage current spec.
I got really bit by this one time in a commercial application.
If high value timing capacitors are required stick to electrolytic types even though they
generally have higher leakage currents.
Duane J
-Phil
* That's as close as I'll come to an a$$-kicking, unless you ask how big the input cap has to be.
it took me 3 months to build a prop board to my satisfaction becuase i had to find all this info phil posted all over the net! as a matter of fact duanes post is yet another rule of thumb that ive never heard. i will say always go x7r grade cermic over tant when possible. tants can also lead to saftey issues when charged backwards, or if a user accidently shorts the leads of the tant by by setting om a pile of jumper wires and stuff while prototyping a circuit
Next to a analog or digital device of course, between the VCC and VSS pins.
But, this is only really easy to do on; push boards, protoboards, deadbug or experimental circuits, just push the caps in, and/or solder as necessary.
But then, if you are actually planning to build a complex PCB's (i.e. more than a few devices), it is a lot of work to make sure the right cap (as per the schematic) is next to its associated device. All of the bypass caps are connected in parallel to the same supply and ground rails (i.e. VCC and VSS), therefore on the initial PCB layout they are connected to the same "rats nest" and not really associated with a particular device. The task is even more difficult, because bypass caps are typically all the same value.
Some designers just scatter the bypass caps around the board next to a likely devices. Other (better) designers take the time to place the correct bypass caps near its associated devise, as per the schematic - this can be a lot of work!
But, does it matter? - it is only a reference symbol printed on PCB in silkscreen. To me; "Yes, it does matter". And therefore, I make sure the correct bypass cap is next to its associated device as per the schematic.
I use the following technique to make the PCB layout of bypass caps and associated devices easy. On the schematic, I insert a "Zero Ohm" resistor in the supply line (VCC) that feeds each bypass cap. Now regardless of PCB layout software that is used, the initial "rats nest" will suggest close layout of the correct bypass cap with its associated device. From PCB layout perspective, the Bypass caps are NOT attached to the VCC rail. The associated "Zero Ohm" series resistor is easily located and placed next to its bypass cap (because it is in series and not parallel on the schematic). Its a simple trick that just makes PCB layout much-much easier!
Also, for most low current devices, I use a 0201 "Zero Ohm" device footprint. Instead of an actual device, a simple solder bridge is all that is needed to make the connection. And, this technique makes trouble shooting a new designs all that much easier. A troubling device can be removed from the circuit with a little solder wick.
This technique; may, or may not, be useful for Production Products, but for the Hobbyist it works great!
BTW, I also put in 0201 "Zero Ohm" in all I2C lines, next to each I2C device, for easy initial testing. A bad I2C device can easily take the entire bus down - which is not fun to troubleshoot.
Eldon - WA0UWH
-
The rest of the field is just providing 'rules of thumb' that may or many not be old wive's tales.
Back in 2004, I locked on to tantanalum capacitors as being the apex of excellence and bought quite a few for board building. But the truth is that they are rather complex little beasts and can create trouble in some situations. So my collection of tantanalum capacitors just sit around while I use electrolitics and ceramics -- like everyone else.
I suppose that if I actually bought a good oscilloscope, I'd use up those tantanalum capacitors, but that is an expensive way to go.
As an alternative, I have created my own 'rule of thumb'. First do what the voltage regulator's PDF document tells you (actually read the thing, don't just look at the schematics suggested) and always add by-pass caps to the power leads of every logic chip. If the chip has more than one V+ and GND, do all of them.
And only, then - after I power up and use the device and find problems -- will I consider more and bigger or better capacitors.
BTW, the smart PCB designers provide space to attach extra capacitors if they are needed. But they then actually assemble a minimal configuration and only add fixes after they discover problems. And when they do discover problems, they use an oscilloscope to resolve them.
here is a little blurd and pic that shows scope shots of some digital chips with and without the .1uf caps
http://hackaday.com/2011/10/25/do-you-know-why-youre-supposed-to-use-decoupling-capacitors/
@loopy i dont like tants much but i had bought some 22uf (doesnt seem any better than a 22uf electro) to put on the output side of my 5v regulators when i build boards to OC the prop. I figured with all the jazz about using a 33uf tant on the power pins of the crystal side when overclocking a puvtting a tant on the 5v rails output should be the better choice. after reading what you just wrote about tants, im curious to hear your opinion
If you must read about audio hi-fi electronics, Elliot Sound Products is just one of a very small minority of websites that offer a sane view of the art and science of audio amp building.
sound.westhost.com
But even he at times will ask for capacitors that are hard for me to come by. I just use what I can get and if it sounds okay, I forget about it.
My opinion about tantanalum?
"The Art of Electronics" by Horowitz, et. al. claims to have seen tantanalum capacitors in parallel with ceramics causing oscillations on the power bus. We can't have that, can we? These devices are suppose to remove oscillations, not make them.
Besides, there is a lot of other confusing nonsense - like I have read that one should use a 1uf tant where you would normally uses a .1uf ceramic. Why have a system of values that is supposed to apply to everything in a clear scientific way and then tell people that they have to ignore the system for one kind of capacitor. Farads relate directly to time and that fact should not change with a different product.
Avoiding electrolytics is an excellent idea if one can do so. The problem with all electrolytic capacitors since they were first invented is that they are less durable than other capacitors - they leak, the bulge, and the chemistry in them will change over time and heat.
http://nwavguy.blogspot.com/?m=1
as far as the tant thing ill look over the art of electronics, what chapter? the whole thing is sephiea recomends a 33 tant which would be in parallel with a .1uf at the least depending on the power supply.
There is /one/ point about audio caps that I just learned recently: Ceramic caps should not be used in audio filters because they have a slight piezoelectric effect, and the capacitance will increase/decrease with voltage, causing intermodulation distortion in your filter. Film caps don't have that issue, which is why they're frequently used in speaker crossovers.
Never mind the digital a$$ kicking.
What you should bear in mind is that when dealing with decoupling capacitors and such we are not in the digital world. This is real life analog electromagnetism. Those high speed rising/falling edges of the zeros and ones you are moving around your circuit around are basically high frequency analog signals. In that world the inductance of your connections, the capacitance between connections can have a large effect. The audio analog guys cannot help you as they normally have no idea about radio frequency behaviour. The software guys cannot help because, well, they are software guys.
Potentially every part of your circuit can affect every other part due to capacitive or magnetic coupling. The physical layout can change how things behave. For this reason it seems the art of the decoupling capacitor is "black magic".
Add to that the switching of large currents, like motors, through your circuit traces, which have resistance, and you see how odd voltages and hence logic levels can end up in the wrong place.
All that is not very helpfull I know, but it's a big topic and all I am saying basically is that you are excused from being befudled.
The rules of thumb offered in this thread will help in most cases so go with that even if they are not totaly clear.
For audio filter design stable capacitance value, linearity and low losses are important - for a high-Q filter stage polystyrene are the best IIRC.
When I used to pull apart old 1970's computers to salvage parts, I noted that they had one 0.1uF cap next to each chip, and they had a 10uF tantalum on each row of chips (maybe 10-20 chips per row). I tend to copy that.
Exceptions - I think the prop needs a 10uF near its supply pins for the PLL (as well as needing two 0.1uF caps as it has dual supply pins). And I've also tended to put a 10uF tant near the supply pins of an SD card, as these can draw several hundred milliamps sometimes.
So on many boards, by the time there are two 10uF tantalums on the board anyway and the total chip count is usually under 10, there probably is no need for more 10uF tantalums.
Would never think of being anything but kind to one with questions.
I'll add my $0.02 with a few links for your reading enjoyment, in addition to the info previously given by other Forum members:
http://www.murata.com/products/emicon_fun/2011/04/cap_en14.html read the pages before and after for good insight.
http://www.electronics-tutorials.ws/capacitor/cap_1.html lotsa good, basic info here.
http://www.radio-electronics.com/info/data/capacitor/capacitor_types.php another good overview of the topic and technology.
I don't think they exist. A quick scan down futurlec's list and I think 0.1uF is the highest value. Ceramics are more in the picofarad range.
Multilayer ceramics (monolithics) go up to 2.2uF but they are rather large at that value.
-Phil
I sure hope they do! I just designed a board with 10uF 603 ceramics... Mouser
I think you units are wrong... It's probably .75 dollars
I stand corrected. Hmm - me thinks I must get out of the futurlec website more and wander over to Mouser every now and then
Yikes!
Do you have a link? The closest that I could find is XR7 6.8uF through hole for $0.12@25K (link, DK part #445-8304-ND). With the data from that page, here is the graph of the price vs quantity:
From this, it appears that 10M units would be roughly $0.10 each.
You are correct on both accounts, but having to sort out ESR is a whole additional bally=wick.
The 1uF tantalum will work for a .1uf ceramic according to popular belief.
And ceramics can actually have 'mircrophonic effects', where they pick up audio noise bia their piezoelectric qualities. I am digging through "The Art of Electronics" right now. Page 22 has a comparison chart and tantalum rate poor for accuracy and poor for temperature stability - in addition to that mention somewhere else that they can oscillate when in parallel with ceramics.
I will keep digging. Ceramics rate POOR in both the same categories.
If you want good capacitors, just use any plastic - Polyester (Mylar), Polystyrene, Polycarbonate, Polypropylene, or Teflon.
Polyester(Mylar) ranges from 0.001uf to 50uf. That pretty much covers the usual computer applications other than driving crystals.
In fact, Page 331 and Page 600 both recommend using 0.1 - 0.01uf ceramics at logic chips (varies between one per chip or less) and a few 10uf tantalum caps spread throughout the board to provide a reserve of engery. NOTHING so far about a ceramic and tantalum causing an oscillation when used together.
FYI, I have checked Chapter 1 in detail and I have looked at all the indexed entries for by-pass capacitors.
I suspect I read this idea of incompatiblity somewhere else and it is just plain wrong.
Paul Horowitz, Winfield Hill ... A classic situation is the use of a pair of bypass capacitors (one tantalum, one disc ceramic), often recommended to improved bypassing. The pair can form a lovely paracitic tuned circuit somewhere in the HF to VHF region (tens to hundreds of megahertz), with self-oscillations!
Silly of me to just try to flip through the book when I have Google.
Not sure what 10Mhz to 500Mhz would do to the average project, maybe a beat frequency on the crystal oscillator.
@loopy, so the book advises agains paralleling ceramics and tants, then goes ahead and does it anyway? i wish someone with a suitable scope could decouple a chip with a ceramic/ceramic ceramic/electro ceramic/tant set up and post some pictures so we can clear this up
But even then, he changes from one ceramic per logic chip to one per two logic chips to maybe even one to four.
In other words, it is not an easy book to read. And these days it is getting more out-dated as it doesn't discuss NiMH batteries or power MOSfets. Time marches on. I can only hope and pray that the next tome on electronics will be more tightly organized. But it is just getting harder and harder to have good facts and clear sailing.