ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

[MissingGhost]

Hi,
I purchased an ASUS A7N8X-E Deluxe 3½ years ago and have been using it 24/7 since then. The United Chemi-Con KZG capacitors are all dead on the secondary side of the CPU's VRM. They are all poped on both extremities. The primary side of the VRM also uses theses caps, but they seem fine.

What should I replace them with? I thought maybe I should replace them with the same kind (maybe it's just a bad batch), or is that a stupid choice? Is there some wide-spread problem with this kind of caps, or is this not likely to happen again?

Should I also replace the primary side of the VRM?
Also, where can I purchase them? I live in Canada.

The exact specifications are:
6.3V
3300uF
10mm pin to pin (or something like that)

[acstech]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

If there's any space between the caps, where you could put in 12.5mm, I'd put Panasonic FM's there. If not, you may need to order some Samxon's from Badcaps.net.

10mm pin to pin doesn't sound right. Are you sure you don't mean 10mm can diameter?

Anyway, here's a pic I found of the board:

http://www.firingsquad.com/hardware/.../images/02.JPG

It looks like there's room between some of 'em. I think it would be possible to put some 12.5mm Panasonic FM's in there. You might have to slightly offset some of them though.

I don't know if you can order from Digikey in Canada, but the numbers for Panasonic FM's in 3300uF 6.3v are P12345-ND for Digikey's number, and EEU-FM0J332 for the Panasonic number.

I don't see where it would be necessary to replace the input caps if there's no sign of damage. Just the output should be enough.

Good luck and post pics!

[MissingGhost]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

Yeah I meant 10mm can diameter, pin to pin is more like 5mm.
What's so good about the Panasonic FM? Some people told me not to put 12.5mm in a 10mm spot.

Yeah I think I can order from Digikey. I went to that site and also found "Panasonic NHG"s and "Nichicon VZ"s, are these bad capacitors?

On that picture you posted, the capacitors that are dead are the complete row of five large capacitors.
I think they may have died earlier than the other 3 because they are closer to the heat of the CPU. Maybe I'll replace them too for safety.

[PCBONEZ]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

NHG and VZ are not 'bad' but they are rated so far below what you need as to not even consider them.

If you like your CPU then look at ESR and Ripple ratings FIRST.
On a scale of 1-10 they are 10 important and capacitance is maybe 8.

>>> What's so good about the Panasonic FM. <<<<

Short of switching to polymer caps they are the only cap available at digikey with ESR and ripple close enough to KZG to use as a replacemnt.
Even then they are a slight downgrade from KZG.
KZG are comparable to Rubycon MBZ.
Very high grade but they don't like heat much and don't do so well in VRM circuits next to MOSFETs and the CPU.

3300uF 6.3v KZG 10x25mm - ESR=.012, Ripple=2800
ESR - less is better.
Ripple - more is better.

In 10mm sizes these are the same grade:
Rubycon MBZ, Sanyo WG, Samxon GD, Nichicon HM, Panasonic FL and FJ

These are one higher grade:
Rubycon MCZ, Samxon GC, Nichicon HN, Chemicon KZJ

On Nichicon HM&HN make sure the date code is 2005 or newer. (Long story.)

Digikey does carry Chemicon PS and PSA series polymer caps and they are a viable option.

.

[starfury1]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

I would replace both input and output regardless...since you seem to run it 24/7
(in fact Id probably recap the lot if I was going to cook it again 24/7)

3 1/2 years is like 30.000 hours run time
(being baked and worked in a brutal circuit)

And yeah panny FM seem to be more easily found

but that just my opinion.

Cheers

[MissingGhost]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

I think I will go with some United Chemi-Con PSA Series solid polymer capacitors.

Here are the specs:
680uF, 6.3V
7.0 mOhm ESR
5860 Rated ripple current (mArms/105c, 100kHz)
Manufacturer Part Number APSA6R3ELL681MJB5S
Digi-Key Part Number 565-3065-ND

So they seem nice, but the capacitance is quite lower than what I originally had. What are the dangers with this?

Also another option that I thought of is to use theses PSA for the secondary side, and replace the primary side with Panasonic FM. If you look at the picture of the board, it's all the caps on the secondary side that are dead because they are really close to the CPU and all the heat that eletrolytic capacitors hate. So maybe I could put the solid caps close to the CPU and keep high capacitance in the colder primary side. What do you guys think of this option?

[starfury1]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

Just on polly's it seems using say 820uf as opposed to 1500uf is quite out ok due the fact they have much lower ESR

Others will probably comment on the value you have selected, the above is a guide only (they know more them me on the ranges with them)

polly VRM output
Panny FM for input ...should be ok I think

but you might want to wait to see what others think

What your suggesting thought is the way it tends to be done polly VRM output and you right they are the one that get the most brutal treatment

I/p will probably be a wet electro

HTH

Cheers

[davmax]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

Just on polly's it seems using say 820uf as opposed to 1500uf is quite out ok due the fact they have much lower ESR

Others will probably comment on the value you have selected, the above is a guide only (they know more them me on the ranges with them)

polly VRM output
Panny FM for input ...should be ok I think

but you might want to wait to see what others think

What your suggesting thought is the way it tends to be done polly VRM output and you right they are the one that get the most brutal treatment

I/p will probably be a wet electro

HTH

Cheers

ESR is not the only VRM need. Bulk capacitance is also a factor. The VRM circuit monitors the voltage across the output C adjusting the PWM to maintain the voltage. The VRM has a response time that is dictated by the output C value and the series choke. A lower C will cause a transient load to pull down the output more quickly, the VRM will adjust the modulation to increase power output, this does take a finite time to compensate because the current through the output choke cannot rise instantly. Therefore there is increased risk of higher spike ripple due to a transient load change. Whether this is sufficient to be a problem depends whether this change takes transients outside of spec.

If you are running 24/7 keep temperature down at the caps. IF there are no heatsinks to cool the MOSFET's, get some to fit.

[davmax]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

One type of MOSFET cooler

http://www.acousticpc.com/thermalrig..._heatsink.html

The challenge with these coolers is fitting. Believe it or not a simple method is to get clear silcone adhesive and stick them on to the MOSFET's. The thermal resistance is still low enough to be effective and the silcone will tolerate the heat.

[PCBONEZ]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

Just an FYI: I looked at what some new boards that use poly in Vcore are using and I saw everything from 560 to 820 uF used. I noticed the ones with the smaller uF caps had more caps. The voltages were everything from 2.5v to 6.3v. (Mostly 4v.)

I think your choice is good but I don't have a solid 'feel' for how low you can take uF when switching to poly caps so I won't make a firm recommendation on that.
Here is a -weak- suggestion. (Just pointing out another option.)

- This ONLY applies to those caps on the CPU side of the VRM. !!!
- Those right next to the CPU and usually 6.3v.

If I were doing it (because I'm new to it and paranoid) I'd use the highest uF I could get.

If you check the voltage you will probably find it is less than 2 volts.
[Check the Vcore spec on all CPU's the board takes. Many people check with a meter and go by that but that only gives the voltage for the CPU currently installed. That can change later if you change CPU's. .. And you STILL should check across the caps with a meter just to make sure one or some aren't on a 5v rail vice Vcore.]

4v Poly caps would be fine in Vcore which means you have more replacement options.
2.5v are -probably- safe but I'm not that brave (yet).

Digikey has both PS and PSA in 4v 820uF.

Like I said.
Just pointing out more options.
.

[PCBONEZ]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

Not contradicting you davmax,,, you posted after I started writing..

[davmax]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

A few more comments about changing VRM cap values.
Lowering C
1.Will increase the the capacitive impedance Xc that is in series with ESR in direct proportion to the cap change. eg at 100Khz 1500uF Xc = 1.06 milliohm, a 680Uf Xc = 2.34 Milliohm. So as capacitance decreases the effect of ESR is steadily reduced.
2. Will cause the ripple voltage to increase, by halving the capacitance the voltage ripple will double and the ripple current will increase in proportion for small changes. As capacitance continues to decrease the ripple grows but the current starts to be increasingly limited by Xc.

It can be seen in 1. above that dropping from 1500uF to 680uf starts to have an impact on the effectiveness of a ultra low ESR cap.

[PCBONEZ]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

Okay so using the original cap at 100kHz:
3300uF KZG - ESR=12.0 mOhm
Xc = 0.48 mOhm
Xc and ESR in series = 12.48 mOhm

Proposed cap at 100kHz:
680uF PSA - ESR=7.0 mOhm
Xc = 2.34 mOhm
Xc and ESR in series = 9.34 mOhm

Looks like less 'resistance' to passing ripple to ground in this case and so ripple voltage should be reduced. - Or am I doing/thinking it wrong?
(Not being sarcastic, actually asking.)

.

[MissingGhost]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

One type of MOSFET cooler

http://www.acousticpc.com/thermalrig..._heatsink.html

The challenge with these coolers is fitting. Believe it or not a simple method is to get clear silcone adhesive and stick them on to the MOSFET's. The thermal resistance is still low enough to be effective and the silcone will tolerate the heat.

I don't think there is any problem with MOSFET cooling. If you look at the board's picture, the MOSFETs are rather large. Also I do not like MOSFET heatsinks. I think all of the cooling should be done at the PCB level, as the chip is designed with a large metal contact under it where heat can escape. Also the chips do not feel very hot when I touch them. However, I must say that my CPU (Athlon XP-M 2500+) is overclocked from 1.87GHz to 2.4GHz, and I run it 24/7 because of rosetta@home.
The problem with heat is that the VRM output capacitors are very close to the CPU in the exact direction where hot air exits the heatsink.

[MissingGhost]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

Just an FYI: I looked at what some new boards that use poly in Vcore are using and I saw everything from 560 to 820 uF used. I noticed the ones with the smaller uF caps had more caps. The voltages were everything from 2.5v to 6.3v. (Mostly 4v.)

I think your choice is good but I don't have a solid 'feel' for how low you can take uF when switching to poly caps so I won't make a firm recommendation on that.
Here is a -weak- suggestion. (Just pointing out another option.)

- This ONLY applies to those caps on the CPU side of the VRM. !!!
- Those right next to the CPU and usually 6.3v.

If I were doing it (because I'm new to it and paranoid) I'd use the highest uF I could get.

If you check the voltage you will probably find it is less than 2 volts.
[Check the Vcore spec on all CPU's the board takes. Many people check with a meter and go by that but that only gives the voltage for the CPU currently installed. That can change later if you change CPU's. .. And you STILL should check across the caps with a meter just to make sure one or some aren't on a 5v rail vice Vcore.]

4v Poly caps would be fine in Vcore which means you have more replacement options.
2.5v are -probably- safe but I'm not that brave (yet).

Digikey has both PS and PSA in 4v 820uF.

Like I said.
Just pointing out more options.
.

I did compare with different VRM designs too. As you said, newer boards with solid polymers on vCore have more caps. But also I noticed most of them have 3 or 4 phases. Probably this allows for a lower C value. My board is of an older design (VRM input on +5V!!) and has 2 phases. So I'll have to try to know if it works with PSA.

I did see the 4V caps too and they look interesting. I was just a bit scared, because as I understood the circuit there can sometime be spikes way higher than the target vCore. (or is this controlled by the inductor??) The stock vCore of the CPU is 1.65V, but I usually use 1.72V or a bit lower. No other CPU will be installed on this board, but anyway all compatible CPUs take about the same voltage. There is no secondary vCore, only 1.65v.

[davmax]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

Okay so using the original cap at 100kHz:
3300uF KZG - ESR=12.0 mOhm
Xc = 0.48 mOhm
Xc and ESR in series = 12.48 mOhm

Proposed cap at 100kHz:
680uF PSA - ESR=7.0 mOhm
Xc = 2.34 mOhm
Xc and ESR in series = 9.34 mOhm

Looks like less 'resistance' to passing ripple to ground in this case and so ripple voltage should be reduced. - Or am I doing/thinking it wrong?
(Not being sarcastic, actually asking.)

.

Yes you are correct there is still less resistance. Remember I was only creating awareness of what changes with cap value. In fact the case of the 680uf cap is even better than above(I over emphasised) because of the phase difference between Xc and ESR and in fact the actual final value computes to Z = 7.38milliohm so C has to drop more to affect ESR significantly. This brings us back to the fact that reducing C has other impacts on ripple as already described. Taking the ultimate limiting example of zero C and ESR remaining there would be a serious problem.

Unfortunately the text in this forum does not support super or sub script so the simple formula cannot be displayed correctly. Where you see a 2 it represents 2 superscript (value squared)

Z2 = Xc2 + ESR2

[davmax]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

I don't think there is any problem with MOSFET cooling. If you look at the board's picture, the MOSFETs are rather large. Also I do not like MOSFET heatsinks. I think all of the cooling should be done at the PCB level, as the chip is designed with a large metal contact under it where heat can escape. Also the chips do not feel very hot when I touch them. However, I must say that my CPU (Athlon XP-M 2500+) is overclocked from 1.87GHz to 2.4GHz, and I run it 24/7 because of rosetta@home.
The problem with heat is that the VRM output capacitors are very close to the CPU in the exact direction where hot air exits the heatsink.

If the MOSFETs run cool then no problem. Using an IR thermometer I have found many that do not run cool and have improved cap life by a temp reduction of up to 20C at the cap. The heatsink actually contributes to extracting heat from the motherboard.
Have you measured the temp at the top of the caps next to CPU?

[PCBONEZ]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

I did compare with different VRM designs too. As you said, newer boards with solid polymers on vCore have more caps. But also I noticed most of them have 3 or 4 phases.

I haven't looked into it and there may be other reasons but I think one of the the main reasons for using multiple phases is to spread the heat load out over more components.

My board is of an older design (VRM input on +5V!!)

In that case absolutely do not go below 6.3v on VRM input caps.
In fact I would choose 10v electrolytics but that's just me.
-
In suggesting 4v I was referring to VRM *output* caps -ONLY-.
Those ARE Vcore voltage.

I was just a bit scared, because as I understood the circuit there can sometime be spikes way higher than the target vCore.

Any spikes would be from PSU on VRM input side.
VRM output side IS Vcore and there should be no spikes there.
Is after input caps, inductor, MOSFETs.
Would have to be one hell'of'a spike to get through.

.

[acstech]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

I'll just throw this out here:

My theory is to replace the capacitors with good caps from an equivalent series with the same specs. For instance, if you are pulling out GP caps, replace them with GP caps! If you pull out "low impedance long life" caps, put back in "low impedance long life" caps. If you pull some ultra low esr caps, put those back in. If you pull 3300uF 6.3v caps, then put back in 3300uF 6.3v caps!

I know you wont be able to find replacement caps with the exact same ripple and ESR from a different manufacturer in an equivalent series, but they should be close enough.

I figure the more you try to change the circuit, the more can go wrong. The engineers more than likely knew what the hell they were doing. They probably figured in a fudge factor due to how capacitors age, as well as having pressure from bean counters. Usually this will cause things to tolerate somewhat more or less ESR, more or less capacitance, etc. But it's not ideal. We can look at the circuit, figure in the bean counter influence, and make an educated guess, but it's just that; a guess.

Throwing polymers in there will most work fine if chosen correctly, but I tend to error on the side of caution.

[starfury1]

Re: ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

ESR is not the only VRM need. Bulk capacitance is also a factor. The VRM circuit monitors the voltage across the output C adjusting the PWM to maintain the voltage. The VRM has a response time that is dictated by the output C value and the series choke. A lower C will cause a transient load to pull down the output more quickly, the VRM will adjust the modulation to increase power output, this does take a finite time to compensate because the current through the output choke cannot rise instantly. Therefore there is increased risk of higher spike ripple due to a transient load change. Whether this is sufficient to be a problem depends whether this change takes transients outside of spec.

Agreed
Thats why I left it and on the polly question as I am not sure how the interrelationship works with them and how far you could go in relation to the VRM circuit

True!
I should have thought about the capacitance a bit more
(what he stated, as opposed to what was originally there)
(it wasn't one of my better thought out replys...late here when I posted it)

Generally speaking;
Yes normally same with same is best idea.
(just make better quality if they arent that good)

I haven't fully read all the replys here
but yeah thats another point how does the phases used in VRM affect the situation.
in regard to the use of polly's
I didn't stop to think on that at all.

cheers