ASUS A7N8X-E Deluxe leaking United Chemi-Con KZG

[starfury1]

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

MG
what is the VRM chip used, knowing the chip may shed a bit more light on how far you can go

here a couple of links

This chip is history but the pdf is fairly though so worth a read

ADP3188

The guy below designed one for a thesis so its gets a bit involved, the links are Doc files

Sean Kelly

I d say stick with the PDF that should give you enough background on them VRM design

the simple short answer is to replace with some better suited wet electro's but there is always room for improvement.
(esp I think in a 24/7 application)

As I said I don't know much on the polly's except some are using them, one point I may missed is some MB's do have extra unused positions.
(like my 2 Gigabyte MB's)

Unfortunately you don't seem to have that available if needed.

thanks too, you guys for posting the finer details of what needs to be considered with a mod like this.

I gotta chew over the posts a bit more

Cheers

[MissingGhost]

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

It's an L6917BD from ST Microelectronics.
Datasheet

[starfury1]

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

Thanks
I'll have a read but I dont design these things
(I have an understanding of the basics at best)
So I really wont be of much use at all on the finer points.
(if of any real use at all)

But the pdf might help others with addressing the finer points

cheers

[starfury1]

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

just on my post it was a somewhat off the top of my head reply and it probably should have been 820 1000uf (not 1500uf)
davmax has posted why, from his analysis polly's probably aren't the best idea in an older VRM circuit else were.

you can get away with uppping the values in a VRM a little and it wont affect circuit behavior too much (but effect it, it will)

reading the 2 pdfs above seems caps selection can get a little on the complicated side

normally safest option is "same with same" perhaps a tad more capacitance wise may be of long term benefit.
(due to aging and going out of tolerance or possibly over clocking)
(talking wet electros here)

on polly's I just dont know enough to have any real valuable input on.

Cheers

[MissingGhost]

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

I'm going to start working on it tomorrow I wish. Here are some pics.

The board as I took it out of the case, after dusting.


Faulty KZGs on the VRM output.


Notice that the second cap leaked a lot from the bottom.
There is a large bubble of this thing that got dry on the board. (How do you remove that?) You can see that the input caps look perfect.



New capacitors. They are large! I also ordered some polymers to try them. I don't care if the board fries or anything. If both work, overclocking results will tell what's the best. Capacitors on the left are KZG from another board to show the size difference.

[bgavin]

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

Damn those KZx... Every dying board I get these days, has these things installed.

They are also present on the brand new Intel D875PBZ boards I just acquired... time to recap those also.

[acstech]

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

Those Panasonic FM's? I think they'll work good, if they'll fit. How come you got 4700's instead of 3300's? Just that urge to experiment?

[MissingGhost]

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

The board ready for work.

Close-up of bad caps.

The five output caps after removal.

It looks like this.


I must say it was hard! I thought it would be easier. After doing that I don't have very high hopes that it will work. I will try it later, but I expect massive shorts.
Also it really needs a lot of heat. I worked all the time at 450C because that was the max on my soldering iron, but I would have used 500+C if I could.

[willawake]

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

its a fine job
yuk those kzg blowing the bottom bung out like some real cheap shit
you went kinda low on the vcore out replacing 3300 with 820 polymer. still i think 5 820 is pretty ok for that use. how is the overclocking?

[acstech]

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

It looks fine from above. Any pics of the other side of the board so we can see the solder joints?

If you want to test for shorts, a simple multimeter across the caps usually works unless it's solder bridged to some data line or something.

[gonzo0815]

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

I know this two phase VRM desing from my old Asrock K7s8XE+. It had the same VRM controller too.
I recapped it with 2700uF 2,5v Oscon SEPC polymer capacitors.
From my experience, this VRM needs the bulk capacitance, but ok on the Asrock there where only 4 caps, Asus was generous with 5 o/p capacitors.

My Asrock developed higher voltage drops after the recap.
I have to admit, that the poor old Athlon xp 1500 was pushed to it`s limit right from the start (2000mhz, 200mhz x 10), and has reached his end of life after some 3+ years.
Haven`t got any usable skt. A cpu since that to test it further. May be the VRM needs some change in the control loop or may be a higher frequency.

Any way, nice experiment, i am very interested what your results will be.

[davmax]

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

MissingGhost appears to be another focused on the ESR of poly's at the same time reducing the output filter capacitance by 4:1. Capacitance does have a real role. By decreasing C by 4 the noise and ripple voltage will increase by about the same amount. Has the ESR decreased by 4:1 to compensate. It may work but we need understand the risk of reducing bulk capacitance.

The relevant formula is V = IT/C Where V equals voltage change on C, I is current in or out of C and T is the time the current flows. It can be seen by this formula that as C decreases the amplitude of V increases in proportion. V can represent ripple or load current transient generated voltage.

This voltage increase will cause a proportionaly higher ripple current through the capacitance and if the ESR has not dropped by the same amount the resultant ripple at the CPU will be higher than the original design intent.

[davmax]

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

Those five caps should have been 2700uF 2.5V Poly's to ensure same and probably better performance.

http://www.pcbulb.com/index.php?main...93679bdf67e9f7

[kc8adu]

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

yes thats a bit to far to push those 820's
you can use the 820 to replace up to a 1500 safely.
this proven in testing here.
a compromise that will improve performance is to leave polys in the last 3 spots closest to the cpu and put 3300@6.3 mcz in the rest.
thats an older 2 phase low speed vrm that needs more bulk cap.

[gonzo0815]

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

Following the guideline from the data sheet, i think the missing capacitance may not be the real problem, but it affects the voltage loop bandwidth.
There are two formulas, which are affected by the capacitance and the ESR, which determine the voltage loop bandwidth.

Following the "CPU Power Supply: 5 to 12VIN; 1.7VOUT; 45A" example, which is probably quite similar to the used circuit on the board i have recalculate the two components affected:

R8 which is 6.2k with a total of 11000uF installed capacitance and c2 with 15nF (both values are from the circuit digram).
Adjusted to the polymer capacitors with 820uf 7mR ESR and a total capacitance of 4100uF the values for an 20khz voltage loop should be:

R8 6.99K and C2 ~7nF.

The minimum needed capacitance could be derived from another equation, stated at page 15 in the manual.
I haven't calculated this one, as i could not find the max. duty cycle required by this equation.
If the total capacitance would not be sufficient, the voltage drop during a transient would simply to high (which would easily be detected by a oscilloscope) or even with a on board voltage monitoring. Worst case would be a no boot thing or simply instabilities.

A lower ESR will on the other side reduce the voltage drop caused by a transient, but sure the control loop and the inductor should be fast enough to prevent a discharge of the capacitors below the acceptable transient voltage drop

Any way, even considering that my examples are not derived from the real world circuit from the board, i think that the only negative side effect would be a altered control loop, which could theoretically easily adjusted.
But may be, the side effects are not that dramatically.
Any way, as clearly as it is, reducing the capacitance that much is an experiment, and i think the author of this topic is simply aware to this.

Generally, the polymer capacitors have very low ESL, which will reduce high frequency ringing & ripple.
May be, this is not necessary at the cpu VRM, as a bunch of MLCC`s in the cpu socket will address this issue already.

But for other VRM`s like chipset, RAM etc. the lower ESL of polymers will may be benefit a stable operation, as at those locations, there are not that many MLCC`s present (from the total capacitance derived from MLCC).

[starfury1]

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

you can use the 820 to replace up to a 1500 safely.
this proven in testing here.

if you have posted that before KC8 that is probably why those numbers stuck in my mind.

BTW for some reason (probably at my end I can't view the photos MG)

So if you dont mind in future can you please post photos on the forum,
rather then link them.

Willawakes post on how to do it HERE do read it all thought

As I said I dont know enough of how the VRM works to really comment
I would have thought that jump was a bit on the dramatic side but it does sound like you know what your on about just not familiar with VRM design.
(so I would have thought the 2700uf would have been a more close to the mark)

I do know they seem to operate over a wide capacitance range

probably why they keep chugging alone with crap GSC's with popped tops
(thats not to say they are doing it at optimal performance thought)
(not sure about newer designed ones thought either)

Anyway I'll leave it to you guys that understand its operation better then I.

Cheers

[davmax]

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

Gonzo. Yes VRM response will change by reducing capacitance. This response is used to adjust the pulse width of the energy pulses at 100Khz to maintain the DC output voltage. However we must consider the steady load condition (no response time required) to understand ripple. Under this condition PS ripple will exist relating to the load, the inductance and output C. If now the capacitance only is reduced by 4:1 for this same condition the ripple will increase by 4:1. (The controller does not compensate or react to ripple frequency). Ripple is generated by V(ripple) = IT/C (earlier post) A positive ripple excursion whilst C is being pumped and negative excursion when energy comes from the cap.

As stated before ESR has to drop dramatically to compensate ie 1:4.

[gonzo0815]

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

I understand this, but from my point of view, the high capacitance was never needed for ripple reduction.
I assume that 3300uf capacitors where only used to avoid transient voltage drop and to reach an low enough ESR value.

I simply can`t see any statement in the data sheet, which specifies a min. capacitance because of ripple voltage.
The only equation which addresses min. capacitance requirement is the one on page 15.
And from my point of view, this one is dealing with the response time, as the max. duty cycle is required.

May be meeting the transient requirements in such an circuit will be overkill to the steady load condition and thus the ripple current is that low, that it is not worth to consider.

If i got you right, then the only way to decrease capacitance and maintaining ripple would be to increase the switching speed either by using more phases or a higher switching frequency.

[davmax]

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

Gonzo I agree that the 3300u caps may have been chosen to get ESR low enough. Unfortunately we do not know that for sure or have worked through the maths or tested the circuit to see what C has enough safety margin for CPU. So safest bet is to restore to as close as possible the original design/component characteristics. For example if polys are used get close to C value and there may be benefit in a lower ESR.

A frequency change upward may reduce PS ripple but may not affect any transient load spikes in the same way. Higher frequency of course brings other issues depending on how you go.

With multi phase VRMs. Some may be truly multiphase but I suspect that in terms of cost and simplicity many may simply be switching MOSFETs on at the same time thus getting load sharing via each choke inductance. I have not checked to see what variants there are.