Capacitor Mods 101

[Holst]

Ive used those Nichicon 4700's in the past on modded boards with some success Charlie.

One point to note is that haing way too much capacitance can be a bad thing, especially on boot up.
As the voltage regulation on the board has to chage the caps up this can put an unusually high load on the voltage regulation.
I did mod an epox board that would trip the OCP uppon power up as I added so much capacitance. Had to disable to OCP to get it to power up. Even then the board OC worse than before as I had overdone the cap mod.

[Revv23]

Ive used those Nichicon 4700's in the past on modded boards with some success Charlie.

One point to note is that haing way too much capacitance can be a bad thing, especially on boot up.
As the voltage regulation on the board has to chage the caps up this can put an unusually high load on the voltage regulation.
I did mod an epox board that would trip the OCP uppon power up as I added so much capacitance. Had to disable to OCP to get it to power up. Even then the board OC worse than before as I had overdone the cap mod.

Good info...

BTW i've used DIGIKEY.com for pots and such before and i was astounded by the quality/speed of thier service... priced very fair as well. website is hard to browse though.

[SPL15]

Ive used those Nichicon 4700's in the past on modded boards with some success Charlie.

One point to note is that haing way too much capacitance can be a bad thing, especially on boot up.
As the voltage regulation on the board has to chage the caps up this can put an unusually high load on the voltage regulation.
I did mod an epox board that would trip the OCP uppon power up as I added so much capacitance. Had to disable to OCP to get it to power up. Even then the board OC worse than before as I had overdone the cap mod.

Yea Most people dont realize a cap becomes another load on the power supply and regulation circuits. When you power up, theres going to be a massive current surge to charge those caps. A cap is not a battery!!

a CAP filters noise and ripple by acting as a low impedance for AC current. The more capacitance you use, the more stress it puts on Diodes and rectification circuits because of a shorter charge time which skyrockets peak currents in rectification diodes.

I dont know how many times I gotta say it, but a cap is for noise/ripple filtering only. If your using a cap for energy reserve purposes, your wasting time and money, and could be hurting stability of your motherboard regulation circuits which will ultimately lead to a worse OC.

QUALITY OVER QUANTITY!!!!!!!!!!

[Hell-Fire]

Yea Most people dont realize a cap becomes another load on the power supply and regulation circuits. When you power up, theres going to be a massive current surge to charge those caps. A cap is not a battery!!

a CAP filters noise and ripple by acting as a low impedance for AC current. The more capacitance you use, the more stress it puts on Diodes and rectification circuits because of a shorter charge time which skyrockets peak currents in rectification diodes.

I dont know how many times I gotta say it, but a cap is for noise/ripple filtering only. If your using a cap for energy reserve purposes, your wasting time and money, and could be hurting stability of your motherboard regulation circuits which will ultimately lead to a worse OC.

QUALITY OVER QUANTITY!!!!!!!!!!

I would hope people know most of this by now...atleast for anyone that is considering doing any type of cap mod to their hardware. Then again, alot of people juse follow guides or try mods without knowing what is actually going on with the circuit loop components without bothering to learn about what is really happening. But it is important that most users here look at these types of things as abstractly as possible and leave the rest to more skilled people.

Regardless, a big part of this is whether the cap is in series of parallel with the diodes/rectifiers.

At the instant the supply is fired up those caps respond like a short...if a series connect is present for diodes...the current will be pretty high. Its the opposite for the parallel connect...all current will flow through that cap as the diode will respond like a short.

Thats only right after hitting the power switch of course.

With the larger capacitance, it is going to charge slower which will increase the current through any series component..diodes included. Again, the opposite is true for the parallel connect. And for petes sake, dont anyone connect caps in series unless you know what you are doing.

I really hope that anyone adding caps to a circuit is not doing so for the purpose of powering another device. Well, atleast not your average computer user that is. There are many applications where the cap is used to store charge for power usages.....but the big use is for the filtering of AC signal in DC applications. Thus a poor responding psu can be the result of faulty, poor quality capacitors that just arent up to snuff. You get enough spikes and say buh bye to stability. We saw this when chilly1 and FUGGER hooked up my PCP&C 510 Dlx to an O-Scope. Saw some nice 60mV+ spikes.

This is the kind of discussion that prob belongs in its own specific forum as most of what SPL has to say is well above the average modder. Maybe this wouuld be a good idea, then truly technical information is available by simply checking that forum.

[STEvil]

Sounds like an idea to me, you had mentioned trying to clean this place up a bit once before

[Hell-Fire]

If I can only survive this week STE...lol.

I have some Guides in the works, as well as a review on a product. But getting killed right now...and thats no joke. 4 exams, 2 lab reports, 1 quiz, and 2 homework assignments....talk about hell to be me.

[LostInSpace278]

How would an average person, like me, measure the line noise on a PSU? Is there a special tool or do I need to call an electrician? (Which happens to be a friend, no charge)

I am really curious to how much noise is along the rails on my PSU's.


On another note, I have heard that on an NF7-S, cap modding has increased stabilty some. Is this true?

BTW, Thankyou for answering my questions about volt modding my NF7-S. All where successful!!!! (Vdimm, Vdd, and Vtt) Also modded my 9600 XT I am having so much fun with a soldering iron, it should be illegal. My wife won't even let me turn on her puter any more


Any one have pics of the vcore mod to this mobo? Just a pic of the area with what pins to connect. Not the pencil one either , I want to be able to more than 1.8~1.85 volts to it. I don't have the skills to add a 4th phase yet. I know I read it another post somewhere, Just wasn't explained very well.

[Hell-Fire]

Doesnt the NF7 have bios options above 2v for Vcore??

To be able to read the "noise" in your psu rails you will need to have access to an oscilloscope bud.

[LostInSpace278]

It has over 2v in the bios, but it fluctuates to much under load. I had read on another post about an IC chip, were by you connect 2 legs via a VR and it will control the fluctualtion a little better. And if your brave enough, you can add a 4th phase, for the best results.

EDIT: I believe its the one where you solder a 5k VR to pin 7 and pin 10 on HIP6301. Does this sound right?

[Hell-Fire]

Not sure if thats right or not honestly. I am in the middle of writing a 6-8 hour lab report and just checking the forums sporatically. Dont have time to check the data sheets.

Adding a resistor between 2 pins sounds like the standard droop mod to me. Would also help to consider adding possibly a choke near the 12v aux connector, and a couple of caps in parallel with the stock ones. Might help to filter out any additional noise.

[STEvil]

yup, choke and caps would help the board I think.

Got a 600w powerstream on mine and max I can get is 2650mhz 2.03v 2700mhz will run around in windows fine but once I put a load on it it crashes in about 5-10 minutes. Voltage doesnt help either... this is air cooled, too.

[LostInSpace278]

Well I took apart a PSU a week or so ago and found some low ESR caps on it. The mobo has 4 16v caps. Only 3 16v caps in the PSU. I do have enough for the 6.3 caps. Would doing these help as well or just stick with doing the 12v caps? I see the choke your talking about, the tiniest on the board? Its a tight squeeze in there, will it matter if it touches anything, like the cap or the plastic shroud for the 12v Aux line?

I can do about the same as you STEvil, but only need 1.95 vcore.(on air as well) This is with a Antec True 430. I am just trying to cure the fluctation problem. Maybe later on If I go water chiller, DI, or direct die, I may need more, but right now I am good.

[STEvil]

voltage isnt the important number (other than it should be higher than what the max the cap will have going through it), and you have to keep track of what each capacitor you are replacing or adding to does. Some circuits wont be able to handle the extra capacitance so well due to lower requirements vs what a CPU would need. Most of the 6.3v caps will be for vdimm or the northbridge or agp (just did an 8rda+ today, replaced every 6.3v cap on the board because they were all buldging and leaking).

Wish I had more room under my board so I could add another choke and a couple more caps..

[LostInSpace278]

So I can add the choke under the board? I just solder it to the existing choke?

I have read that on the capacitors, I need to keep the legs as short as possible, is this correct? how do I check the voltage comming from the capacitor? Can I take out my video card, flip the mobo over and turn it on? I haven't even bothered to put the mobo back in the case, I figured as many things as I want to do to it, it would be a waste of time!

[SPL15]

By putting another choke in parrellel with the existing inductor, your going to be greatly reducing the inductive component of the PI filter (CLC filter). This will lower it's noise filtering capability and by adding another core in parrelel to the existing one, your going to be increasing saturation current of the total inductance which may prove to also be highly detrimental to noise/ripple filtering in addition to the lower inductance value of the chokes in parrelel which will hurt noise filtering. Know what your doing before you do it.

Also, VOLTAGE IS NOT COMING FROM THE CAPACITOR!!!!!!!!!!!!!!!!!!!!!! The power supply produces pulses of current which are FILTERED by caps!!! Caps are not batteries!!! They hold a static charge, they dont hold enough energy to power a high current draw for any meaningful amount of time. If your trying to relying on caps to make up for a crappy power supply, there's a simple solution to it and I hope you guys are bright enough to figure it out. Caps on motherboards should ONLY be used for noise filtering. Again, know what your doing before you do it.

If you hook up a scope to your computer, KNOW HOW TO USE IT!! depending on your grounding situation, you can do anything from getting false results to destroying the scope and computer. I highly recommend isolating the scope's ground connector. Also, your ripple measurements with a scope will be WAY off if you stick the ground on one point and put the probe somehwhere a foot away. Just cause ground is supposed to be the same throughout the system, doesnt mean it is. Again, know what your doing before you do it.

[STEvil]

How do chokes respond to current load, SPL15? Oppainter has had a choke unsolder itself due to heat on his x800 series card and I am wondering if part of the problem on the nF7 is the choke by the P4 connector cant handle the current? I also have a dead 9700 card and one working one and have considered placing the chokes from the dead card in parallel with the ones on the good card to see if I can increase the voltage regulation circuitry on the card somewhat. Adding caps only had minimal gains..

Glad to have you here btw, I only know enough to be dangerous

[Revv23]

im pretty sure replacing a small choke with a nice big looking one couldnt hurt to much...

[SPL15]

im pretty sure replacing a small choke with a nice big looking one couldnt hurt to much...

How do you know this??? Do you know what it does? Why its doing it? and how much it is doing it? Statements like these spread bad information and bad information is virally rampant in consumer electronics especially where enthusiasts are involved. I urge you to make statements that are true based on real knowledge NOT because it seams like it would work because it is bigger which obviously means it is better.

As for STEvil's question, I wont go into inductor physics because it would take an hour for me to upload the massive post needed to throughly explain how they work. Basicaly an inductor acts as a low impedance to low frequencies and a gradually rising impedance to frequencies higher in frequency. Inductors do this because they have a magnetic field that must be opposed/chagned in order to change the current through the coil. Inductors reduce high frequency voltage by -6dB per octave above the crossover frequency. The crossover frequency is determined by the frequency value where the load impedance matches the inductor impedance. This is a very very very basic decription of the function of an inductor.

A low pass filter (Which an inductor is) has a higher impedance to fast high frequency votages which is good for noise/ripple filtering but can be bad for cases where a fast transient response is needed like in a power supply.

In a power supply filtering application, the choke is just about always ran in saturation mode which in very simple terms decreases it's impedance to fast current pulses, but maintains it's ripple/noise filtering abilities decently. If a choke is in saturation, it's ferrite core cannot be magnetized anymore than it is. Air-core chokes are almost immune to saturation because they have no ferrite core to be magnetized.

A choke gets hot because of the coils of wire dissipating power.... Duh. The coils dissipate power because of the natural resistance of copper (DCr ~ Direct Current Resistance) and the reactive impedance (Xl) of the inductance of the coils when DC current with AC ripple flows through them. Let's only factor in the Direct current resistance and say one of the output filtering inductors on a three phase motherboard has a direct current resistance of .02 ohms (Which is probably unrealistically low). Power = Current Squared times Resistance (P=I^2*R) so lets say your running a highly overclocked prescott which will draw around 90 amps from your motherboard power supply which divided between 3 phases equally (Which is also very unlikely) will equal 30 amps per phase. So this give us 30 amps squared times .02 ohms which will equal 18 watts being dissipated by the inductor in a perfect world during perfect conditions. 18 watts is a lot of power when there is little surface area to dissipate the heat. In the real world the heat dissipation will be much higher.

What I've done to my motherboards was rewind the inductors with a SLIGHTLY lower inductance and SLIGHTLY larger cores with multiple strands of enamal coat wire that equal a much larger cable diameter to lower DCr. BUT if your having inductors unsolder themselves from the board, then your either vastly exceeding power design limits of the board, or the board was designed for cheap manufacturing costs or both.

[LostInSpace278]

Well you have done a nice job on your board. As you know what you are doing.

We or I was looking for a way to control the fluctuation on the vcore, in a simple yet effecient way. Please feel free to school me in the art of choke rewinding. A few links to the products I would need, would be most helpful.

[STEvil]

Makes sense as to why chokes sometimes "squeal" then.. can squealing be an indication you are beginning to run too much current through it maybe?

LostInSpace - vdroop mod takes care of vcore fluctuations much of the time, although it might drop your max vcore somewhat (selecting 2.3v would result in 2.1v max for instance). On the DFI Infinity and LanParty-B this was done by soldering to a VR (10K max) or a fixed resistor between 2K and 5K from the output side of one of the chokes supplying vcore to the vsense pin for vcore... I believe there is a guide for the nF7(-s) in the mod forum also..

http://www.xtremesystems.org/forums/...773#post633773
http://www.xtremesystems.org/forums/...=nf7%2A+vdroop

those two links have mention of them, dont know how valid the first link is and I havent read the second one yet...

Thanks for the schooling SPL15, looking forward to futher info

[SPL15]

I think the question should be is the fluctuating VCore that the BIOS, MBM, etc report an accurate reading? When I viewed my VCore on my fully calibrated/certified tektronixs scope, it didnt look nearly as bad as MBM makes it look. My min max range was 1.4-1.44 with this prescott with the voltage set at auto in the BIOS and had an average Peak to Peak ripple of 11 mV. RMS ripple was calculated by the scope to be only 5.4 mV!!!! considering how much current is drawn by the CPU and the noisy environment, this is very respectable by any standard. MBM5 however during the same time frame showed my VCore min/Max as 1.36 to 1.48 volts.... I will believe my $20,000 DPO scope that does 20 GigaSamples per second over free monitoring software anyday.

You have to remember the chip on the motherboard that is getting this information is being subjected to all the EMI/RFI and ground noise the computer is producing. This isnt a good thing for accurate and precise measurement. I would use this mainly as a reference and not an absolute measurement for basing decisions and effectiveness of mods.

Squealing Inductors are caused by coils resonating from the lower frequency/longer duration current pulses from the switching mosfet. When current demand is high, the PWM IC gives the switching mosfet lower frequency,high duration voltages on it's "gate" lead. Since we now know the filtering on the mosfet's output is a low pass filter, we should be able to figure out that the switching lower frequency will have a lower impedance to the filter AND a longer charge duration due to the lower frequency and longer pulse duration thus enabling voltage to stay constant which will mean more current will flow when the resistance of the load is dropped.

Most things run in constant voltage mode, this means output impedance of the supply power source is 0 ohms (In a perfect world). When load resistance drops, current rises and voltage stays 100% the same. We dont live in a perfect world though, so the power source's output impedance is not 0 ohms. there are parasitic resistances I've talked about before and these resistances are in series with the load creating a voltage divider. As the load resistance drops, the ratio of parasitic resistance to load resistance drops which simply means that the voltage across the parasitic resistance grows and the voltage across the load drops (Like Vcore).

Now seeing as how the inductors are in series with the Vcore, and the inductors are running VERY hot, there has to be a voltage drop across these inductors in order for them to be getting hot. Here is one of the main parasitic resistances of the Vcore. We all know that noise and ripple is bad for stability, but we also know that voltage drops can have the same effects. What do you do??

People who know basics would think the magical capacitor will solve all their problems, because they are taught a cap is a energy storage device, which it is, but a very very small energy storage device that stores a static charge. They are not batteries! can you power up a 100 watt light bulb (which takes less power than a CPU) for any appreciable amount of time by shocking it with your finger? NO!!! I'm not saying dont do cap mods, but know what they are doing. At the frequencies and current the power supply is working at, the length of the leads on the caps (mm count) will make or break the effectiveness of these caps, and 99% of the mods I've seen have lengths that render the cap worthless. On amplifiers I've designed and built, a few millimeters of lead length on filter capacitors made a distortion difference of 0.3% at 20 KHz which is massive. Computer power supplies/regulation circuits works at 20Khz and above. The purpose of caps on motherboards is for FILTERING purposes ONLY. It makes sense to spend time making it quality filtering NOT QUANTITY.

So the inductor is the main source for VCore drop due to parasitic resistance. what can you do about it? Well I'm fairly certain that Asus and Abit dont have an inductor making factory so they have to be buying these parts from someone else. There are many catalogs with pages of inductors to experiment with. Digikey, Mouser, MCM, Allied, etc. have pages of inductors that would work. If you have the gonads to roll your own. you can take the coils from your motherboard or a dead one and cut all the wires off, but remember how many turns there were. You then need get small diameter like 22 Gauge enamal coat wire and use several strands about 6 or so to wrap around the core the same or ONE turn less than what was on there before. your basicly making a larger gauge wire from mulitple smaller gauge wires. MAKE SURE your windings are nice and neat!!! it matter !! also different winding patterns have different effects on parameters. you may want to do some searches on google about it, cause I'm not going to go into inductor physics in a computer forum.

I've taken this thread way off topic so I'll quite now before I get booted.

[SPL15]

Makes sense as to why chokes sometimes "squeal" then.. can squealing be an indication you are beginning to run too much current through it maybe?

LostInSpace - vdroop mod takes care of vcore fluctuations much of the time, although it might drop your max vcore somewhat (selecting 2.3v would result in 2.1v max for instance). On the DFI Infinity and LanParty-B this was done by soldering to a VR (10K max) or a fixed resistor between 2K and 5K from the output side of one of the chokes supplying vcore to the vsense pin for vcore... I believe there is a guide for the nF7(-s) in the mod forum also..

http://www.xtremesystems.org/forums/...773#post633773
http://www.xtremesystems.org/forums/...=nf7%2A+vdroop

those two links have mention of them, dont know how valid the first link is and I havent read the second one yet...

Thanks for the schooling SPL15, looking forward to futher info

If you need more than 1.8 volts to the CPU, you will need to change the surface mount tantalum caps in the CPU socket if it has them. They are rated at 2 volts or 1.8 If I remember correctly. Tantalum caps do not like out of spec voltages, ripple currents, and high heat. They also like to flare up when they are abused so it would be wise to change them. Also tantalum caps are the highest quality cap money can buy, but they dont like abuse and will degrade very quickly if not ran within tolerances.

[STEvil]

Very good info and definately not off topic

Tantalum caps not liking above ~2.0v would certainly answer for why some boards run great for long times at or above ~2.1v then deterriorate.

[SPL15]

Very good info and definately not off topic

Tantalum caps not liking above ~2.0v would certainly answer for why some boards run great for long times at or above ~2.1v then deterriorate.

Just wanted to clear up that those caps I'm talking about are RATED for 1.8 volts max and that is why they do not handle higher voltages, it is not because tantalum caps do not like over 2 volts as there are tantalum caps rated for 50 volts, although they are large and EXPENSIVE!!!. Tantalum caps like to short out and burn up when they fail. This is usually precluded by a skyrocketing ESR which basicly makes its high frequency filtering capabilities worthless.

Tantalum caps are the best money can buy quality wise, but they are very unforgiving of abuse.

Any cap no matter what type is supposed to be ran within 20-80% of its rated voltage. This is especially true for Electrolytic caps because the Dielectric will not form/maintain correctly if there is not at least 20% of the rated voltage present during normal usage.

On caps you see a tolerance rating, most people know that this is for capacitance value tolerance, but few people realize that this also applies to WVDC (Working voltage Direct Current), Ripple Current, ESR, Dielectric Absorption, and all other specs of a cap. Running caps at the edge of voltage ratings and Ripple Current ratings will always mean bad news sometime down the road. A high ESR reading signals a failing cap usually caused by excessive ripple current... But I doubt any of you have a Capacitor ESR meter which kinda leaves ya in the dark.

[LostInSpace278]

I think the caps you are refering too are like these

I did a google search and there are many types of tantalum caps, most of the varieties you see used on mobos. I am just making sure I am following correctly.