noise smoothing capacitors

[CTho9305]

when you're replacing them, how closely must the values match the originals? I knocked 4 off my slotA tbird pcb. I still haven't verified that this is the cause for my crashes (heat and need of reformat are still possibilities), but if it turns out to be the only probably cause, how close do I have to go? I would think that if its bigger, its fine regardless of how much too big, right?

 

[DeeK]

Theoretically, any size larger than the original should work. What the caps do is they hold charge which is supplied to the system as current if the voltage dips. When charge moves off of a capacitor, the voltage on it decreases. Larger capacitors lose less voltage when current moves off of them, keeping their output voltage more stable.

 

[CTho9305]

so would the replacement have to be an smd capacitor then? or do the big cylinders "react" too slowly?

 

[ERJ]

I realize this is not a "Very Technical" comment but:

You knocked 4 capacitors off your motherboard and it still works?

ERJ

 

[dkozloski]

The problem comes with higher clock speeds. All discrete electronic components exhibit small amounts of other electrical characteristics. Inductors display capacitance, capacitors can act to a very small degree as inductors or resistors. At low speeds these effects can be ignored but at high speeds they become significant. If this were not the case any dummy could design electronic equipment and expect it to work. Learning how to get around the problems I've described is what separates the men from the boys. The point I am making is that you have the best chance by duplicating the original parts the best you can. For a source you might examine the components on the old stuff in your junk box like unused video or soundcards and salvage what you need. Good Luck!

 

[Dulanic]

For use on a MB you want to use a cap that for sure is the same Voltage rating, the ferad rating should be as close as possible but I guess it could be higher if needed. Caps being used for cleaning a signal work like this...basically what they do is sit on the line and the signal passes by them, that signal can vary slightly (noise), the cap will smooth out the signal.... think of it like this... your car battery... While the voltage going in can vary depending on the RPM of the car since that rotates the alternator at different speeds, the voltage going out wont vary... thats almost exactly what a cap does to smooth out noise. So you could put a bigger battery in (ferad) but you couldnt put a 6V battery in there.

EDIT: Btw check www.digikey.com for the parts you need, theyll have them. I get parts from them all the time for work. Theyll charge a $5 surcharge on any order under $25 tho.

 

[CTho9305]

dkozloski - i have yet to find out their values.

ERJ - i knocked them off the PCB of my SLOT A thunderbird. slot cpus have a pcb of their own.

 

[Dulanic]

<< dkozloski - i have yet to find out their values.

ERJ - i knocked them off the PCB of my SLOT A thunderbird. slot cpus have a pcb of their own. >>



That makes sense, caps used for noise reduction dont complete a signal path so if you knocked them off the circuit would still work, but you would have alot more noise, which could easily cause crashes... you could get a noise spike which makes the CPU think the signal is a 0 instead of a 1 and that can EASILY cause crashes.

 

[CTho9305]

right. which is why i'm hoping that its ok and a reformat will solve it. unfortunately, i doubt that is the case. i figure the caps had to be for noise - any other purpose i can think of would result in an instantly dead CPU

 

[Bleep]

There are some things posted that are not quite true relating to this matter. A capacitor will under no circumstances pass direct current only alternating current, the capacity is related to its ability to either reject or pass alternating current. In a smoothing type circuit that is where AC is converted to DC there will always be a AC component still active so to speak on the direct current. A capacitor is then put in parallel with the DC current the capacitor will short out the AC component thus preventing it from remaining on the DC current. The value of the capacitor determins what frequencies are shorted out and not allowed to remain on the DC. If the AC component is of high frequency it takes a very small capacitance to remove the AC component, conversely if the AC component is of low frequency such as 60 Hz the capacity must be large, several hundred Microfarads, if the frequency is high 50mhz or so the capacity will be small maybe even in the picofarads area. So to answer your question you should get a capacitor as close to the value as the board or circuit was designed for. As to the relating it to a car battery, if you look at the construction of a car battery other than the chemical involvement it is a large capacity capicator and as such filters out all low frequency AC components that may be on the DC that comes from the alternating scource.At the same time I might mention the battery in a car does NOT regulate the voltage this is done externally from the battery with a voltage regulator circuit. 1 other thing I might mention a diode in conjunction with a capacitor used in a steering mode allows you to use the same wire to transport ac and dc over the same wire and allow the DC to go 1 way by not allowing it to pass through the Cap. and to allow the AC to go down another path because of the blocking of the AC by the diode. this is a subject that books are made of and the reasoning for all of this is not subject to a dissertation on a small basis. I hope this helps a little in understanding what a capacitor does in some circuits.
Bleep

 

[dkozloski]

The larger a capacitor, the more inductance it can exhibit. Electrolytics are rolled up foil of various types separated with an electrolyte. The rolled up foil becomes like a wound coil. Rather than filtering out noise you might be building an oscillator with too large a selection. A diode can show all the characteristics of a capacitor as well. This element can be optimized to the point that the capacitance of a varactor diode can be varied by varying the voltage across the diode and thus tuning a circuit. Many, many radios, TV's, cell phones and the like are tuned this way. Stick to the original components to the extent possible.

 

[Demon-Xanth]

With filter caps generally bigger is better. However too big and you get excessive surge current and/or rounding off of the signal. Make sure you get the correct voltage and at least get the capacitance within +1000%/-90%

I know that looks like a huge range, but a 33pF and 1uF cap can look alot alike

 

[CTho9305]

Bleep - I dont think the capacitors are being used in the way you describe... if they were setting timing, the CPU would probably not work at all. I think they are just in parallel with some DC circuit. if there is a drop in the current, the capacitor can supply current for a short period of time. if there is a voltage spike, the capacitor should &quot;absorb&quot; some of it

 

[UnixFreak]

As a general rule +/- 15% is optimal, with more being better. Voltage has to be matched or exceeded, though. Believe it or not, too much capacitance will affect audio, but in your case, its not much to be worried about.

 

[Bleep]

CTho9305
I should not even reply to the statement you made but in no way is a capacitor a voltage regulator or will act in such a manner in fact with ac applied to a capacitor it will charge to Peak voltage of the AC. in a DC circuit there is NO difference beetween the voltage applied and the voltage charge on the capacitor. This will be my last word on this subject.
Bleep

 

[CTho9305]

replied in wrong thread, and I can't delete my post

 

[DeeK]

<< CTho9305
I should not even reply to the statement you made but in no way is a capacitor a voltage regulator or will act in such a manner in fact with ac applied to a capacitor it will charge to Peak voltage of the AC. in a DC circuit there is NO difference beetween the voltage applied and the voltage charge on the capacitor. This will be my last word on this subject.
Bleep >>



We're not talking about AC. As near as I can tell, we're talking about one-way capacitors that store charge for when tranisitions in the IC cause current spikes. Without a filtering cap, a current spike will cause a momentary very high current draw from the power supply. With the power supply being so far away, the current must travel a long distance with a significant amount of resistance. So when the spike occurs, the current and resistance will cause the voltage at the IC to drop. With a filtering cap, the current is drawn from nearby storage along a path with much less resistance. With a significantly large enough capacitor, the loss of voltage across it from the charge removed becomes negligible, and the noise spikes on the IC supply are removed.

 

[pm]

They are bypass caps used to reduce localized step current induced voltage spikes. They are connected between VDD and VSS (power and ground) to minimize the effect of step current changes caused by the clock generation, distribution (on the rising edge of the core clock signal a huge change in current occurs). In fact it's a little known fact that a sizeable portion of most CPU die contains bypass cap.

CTho9305: I would call AMD and find out the exact size, value and manufacturer of the capacitors in question. They may even be willing to replace the CPU, mail you replacement caps for free, or direct you to their distributor. I would not try to guess them since that will probably just waste your time and money and could nuke the CPU completely. As others have pointed out the value and the composition of the capacitors is carefully chosen - which is why a lot of small caps are used rather than one big one. Parasitic effects are very prominent at the frequencies modern CPU run at.

Bleep: No offense intended but I think you need to chill out a little bit. As Anand said, &quot;leave your egos at the door.&quot;

Patrick Mahoney
IPF Microprocessor Design
Intel Corp.

 

[CTho9305]

pm - please enable pm'ing



<< bypass caps used to reduce localized step current induced voltage spikes >>

- huh? can you explain that in simpler terms? while you're at it, although I think I understand the rest of that paragraph, a paraphrasing might help.

so between clock cycles there is a huge change in power usage, and the cpu is actually drawing alterantely huge amounts of power followed by very little, hundreds of millions of times per second?

 

[pm]

<< pm - please enable pm'ing >>

pm is now pm-able. Sorry about that - I forgot to re-enable PM's when I returned to the forums. In any case, you (or anyone else) are welcome to email me too.


<< - huh? can you explain that in simpler terms? while you're at it, although I think I understand the rest of that paragraph, a paraphrasing might help... So between clock cycles there is a huge change in power usage, and the cpu is actually drawing alterantely huge amounts of power followed by very little, hundreds of millions of times per second? >>

Your explanation is spot on. That's exactly what happens. On the rising edge of the core clock signal, a huge number of transistors all fire roughly simultaneously as they charge and discharges wires and transistors throughout the chip. Just switching the clock itself can account for 30% or more of the total power an entire CPU expends.

So, what you have is that millions (or billions nowadays) of time per second, the current load of the CPU changes dramatically. So without the caps you would see a localized (ie. on the chip, or even just on parts of the chip) voltage variations in time. To explain in simpler terms, capacitors are charge collectors that act to oppose changes in voltage by sourcing or discharging current. So if you increase voltage, the capacitor will act to oppose that change by requiring current to charge the capacitor, and visa-versa for discharging. The amount of &quot;opposing capability&quot; that a capacitor can have is known as it's capacitance (C). In equation form, this would be &quot;i = dV/dt * C&quot; or current is equal to the change in voltage in time multiplied by the capacitance. In effect, by storing charge capacitors can act like (very small, very poor) DC power supplies.

 

[WarCon]

Not precisely on-topic, but curiosity has me. Are there &quot;bypass&quot; caps before the voltage regulators also? If not, what would the effect of adding several do? Also are there bleed resistors in the circuit, so the caps don't hold the charge when off. And would a large capacitor disturb the output from the power supply?

The reason I ask, is that there has been speculation and even some reasonable tests to show that power supplies can effect overclocking capability. Would a reasonably sized cap on the +5 and +12 possibly help the situation without being forced to purchase an $80 + supply?

 

[CTho9305]

thats a good question - I would think yes, but you'd need big ones (one farad maybe!). I would think it would help though. you could even get some sort of molex-passthrough for it... it might help when, say, a cdrom drive spins up.

 

[jamarno]

In practice, much larger capacitance values, even 2-3 times the originals, work fine for filters, although larger capacitors have higher series resistance, which can make the filtering worse or cause the capacitors to run hotter and wear out sooner. However, on most circuit boards there isn't enough room for bigger capacitors than the originals, so check the dimensions of any replacements. For switching power supplies (any regulator on a motherboard, most of a monitor's high voltage circuitry, the low voltage part of any PC supply), make sure they're rated for 105 Celcius and for high frequency operation or they may wear out in less than a year. Also look at the lifespan rating, which ranges from about 1,000-7,000 hours @ 105C. Digi-key and B&amp;D Enterprises have large selections.

Filter capacitors are used ahead of the voltage regulators to maintain minimum DC voltage because the voltage comes in as AC (sometimes converted from DC) that's fed through diodes to convert it to pulsating DC, and the capacitors then turn this into more steady DC.

Filter capacitors are also needed to prevent excessively large voltage spikes because chips draw large amounts of power in brief bursts each time they toggle signals high and low. Without adequate filtering, these spikes can cause erratic operation and even damage the chips.

 

[CTho9305]

Hrmmm... would the programs that are being run matter in instability caused by missing caps? I have been up for 6.5 hours now flawlessly - i'm not running genome@home (my normal cpu-heater), but I ran drempels (very cool animated desktop) which keeps CPU usage pegged. Would the software matter?