many mobo's will benefit from polymer cap upgrades

[kc8adu]

i wound up with a bunch of polymer caps from a surplus outlet.
my experiments conclude that most if not all motherboards benefit from using polymers.
noise/ripple on vcore of this imperial gl is half what it showed with ncc kzg.
next up is a vp6.
note these are 2.5v caps so only useful on vcore and vdimm.

[willawake]

Re: many mobo's will benefit from polymer cap upgrades

hehe you probably doubled the value of that imperial

[kc8adu]

Re: many mobo's will benefit from polymer cap upgrades

hehe you probably doubled the value of that imperial

i bet the original cost of the caps exceed the oem cost of the board.
like a car who's value doubles when you fill it up.

[Super Nade]

Re: many mobo's will benefit from polymer cap upgrades

Actually man, I bought a whole bunch of OSCON SEPC 4V 560uF recently. Planning on recapping all MB VRM's I have with it.

I'm trying a little experiment-->
http://www.ocforums.com/showthread.php?t=536054

Suggestions are most welcome!

[kc8adu]

Re: many mobo's will benefit from polymer cap upgrades

cool!
i got a big mess of those too.
nice thing is they are the short version needed for 775 boards.
on the next boards i do i will record before and after readings.
looks like a friends bp6 is the next test subject.
i got a ton of the 820@2.5 in fujitsu and ncc.

[Super Nade]

Re: many mobo's will benefit from polymer cap upgrades

I want to replace the 6.3V Fujitsu's on my ASUS P5K, but I wonder if that would make any difference? So you reckon these big MB makers are avoiding Sanyo because of the price?

[kc8adu]

Re: many mobo's will benefit from polymer cap upgrades

what value are the fujitsu's?
i would say if the ones you intend to install have the same value but lower esr there may be improvement esp if you overclock.
i have been using the 820@2.5v on vrm output(vcore)
i tried 560@4v on vdimm to give a safety margin.
so far this board has been folding 2 weeks with no problem.
the box it came from would bsod several times a day.thats why my neighbor tossed this board.
he found out too late i repair them after he bought a asus board and bent over for his billy gates reaming for an xp license.

[gonzo0815]

Re: many mobo's will benefit from polymer cap upgrades

I have done a recap too with Oscon SEPC series. But the problem with those polymers is probably to maintain capacitance.

My first and only polymer recap job was an Asrock K7S8XE+, but the poor Tbred died shortly after the mod.
But ok, it was oced right from the start up to the limit and may be further (but always prime, memtest86+ etc. stable) and was running several years w/o a problem.

I had to reduce the capacitance of the four KZG 3300uF caps by using four 2700uF polymers.

I know, that may be it isn`t to smart to tweak a value board, but it was just a project

Now, for the Intel P4 boards, there are already smaller value caps used, may be due to a higher switching speed and the very fast response time Intel specified for it.

[Super Nade]

Re: many mobo's will benefit from polymer cap upgrades

Fujitsu's have the same value i.e 560uF and all spots are not populated. I'm replacing them with SEPC 560uF, 4V. ESR is about the same i.e around 7mOhm.

[kc8adu]

Re: many mobo's will benefit from polymer cap upgrades

i would replace them and stuff all the unpopulated spots too.

[Super Nade]

Re: many mobo's will benefit from polymer cap upgrades

Yep, that is what I'm gonna do. I'll be replacing all Fujitsu 560uF's across the board with SAnyo. So, do you think I'll see more stable voltages?

[kc8adu]

Re: many mobo's will benefit from polymer cap upgrades

maybe.
but make sure you only change vrm output side caps.
you will be safe to replace those on vdimm too but do a bit of testing to ensure you dont install these on vrm input which are +5 or +12 as lifespan vill be very short(pardon the pun)
these dont like overvoltage and often turn to very low value resistors when they go.

[gdement]

Re: many mobo's will benefit from polymer cap upgrades

I have done a recap too with Oscon SEPC series. But the problem with those polymers is probably to maintain capacitance.

My first and only polymer recap job was an Asrock K7S8XE+, but the poor Tbred died shortly after the mod.
But ok, it was oced right from the start up to the limit and may be further (but always prime, memtest86+ etc. stable) and was running several years w/o a problem.

I had to reduce the capacitance of the four KZG 3300uF caps by using four 2700uF polymers.

I know, that may be it isn`t to smart to tweak a value board, but it was just a project

Now, for the Intel P4 boards, there are already smaller value caps used, may be due to a higher switching speed and the very fast response time Intel specified for it.

If the ESR is reduced, then you don't need as much bulk capacitance.

===

I somewhat recently recapped a Socket-370 board. It originally came with 4x 1000uF electrolytics on Vcore. I replaced them with 4x 560uF Oscons, and it passed every stress test perfectly with the most stressful CPU I could find to plug into it (Celeron 1100MHz). It's probably more stable now than it ever was when new. Intel and AMD both have some docs which show the relationship between ESR and required capacitance, which is what convinced me to try doing this.

[gonzo0815]

Re: many mobo's will benefit from polymer cap upgrades

Yes, that is right, but the control loop has to be fast enough to respond faster, if the bulk capacitance is lowered (or i am wrong in this case??)

If the VRM was designed with an few 1000uF capacitors, i think it would be no problem to use an 560uf to 820uf polymer capacitor.
But how about a VRM with a few 2200Uf to 3300uF at the output?
I am not sure, if the control loop would handle a drop to 1500uF or even lower w/o changing the feedback filter.

[Super Nade]

Re: many mobo's will benefit from polymer cap upgrades

AFAIK, the output side caps have nothing to do with the feedback loop. They only serve to create a low impedance path for AC. More important here is the whopping 6200mA ripple current rating of the OSCONs. If you have a small number of phases, inductor ripple current is the biggest problem.

In fact OSCONs are specifically recommended against being used in any kind of feedback/integrator loop. The tech docs say so.

It is tackled in two ways, on the CPU package or the socket, you see a whole bunch of MLCC's which have very good frequency response across the board, have zero temperature de-rating and have a very low ESR. You may also see a bunch of Tantalums thrown in there to accommodate low frequency cutoffs. All this is ofcourse after the output side is filtered.

[MixMasta]

Re: many mobo's will benefit from polymer cap upgrades

If the ESR is reduced, then you don't need as much bulk capacitance.... Intel and AMD both have some docs which show the relationship between ESR and required capacitance, which is what convinced me to try doing this.

Hey could you post a link to these docs. I am interested in reading them. Thanks!

[Super Nade]

Re: many mobo's will benefit from polymer cap upgrades

I've got some stuff in there for AMD.
http://www.ocforums.com/showthread.php?t=536760

Look at VRM 11.0/11.1/8.XX/9.XX Tech docs from Intel's website.

[gdement]

Re: many mobo's will benefit from polymer cap upgrades

Hey could you post a link to these docs. I am interested in reading them. Thanks!

Here's the AMD document which I was looking at more recently:


I think this is the Intel doc I read before, or at least it has similar info. It was a few months ago and I can't find it on my computer anymore:

[gonzo0815]

Re: many mobo's will benefit from polymer cap upgrades

From my understanding, the bulk capacitance is required for transient response. Sure the transient voltage drop depends on the ESR, but if the control loop is not fast enough, the caps are drained before the switcher controller can adapt to it.
Furthermore, the inductor sets the minimum time, how fast the current could be delivered.

I refer to the L6917D app note:


I have read a lot of other more general notes about VRM design over th time from ST, but i forgot the document numbers.
Basically at the bottom of page 25 there is the control lope mentioned, and at page 15 the inductor slope time and capacitor choice and minimal bulk capacitance formula.

Transient response is from my understanding critical, as it is occurring frequently when powering a fast CPU.

Many board manufacturer (including Asrock) had been disabled bus disconnect feature of Athlon Xp in the BIOS, to reduce transient stress from the VRM.

I have enabled it on my Asrock K7s8XE+, and i experienced Vcore fluctuations as high as 0,2v.

[gdement]

Re: many mobo's will benefit from polymer cap upgrades

From my understanding, the bulk capacitance is required for transient response. Sure the transient voltage drop depends on the ESR, but if the control loop is not fast enough, the caps are drained before the switcher controller can adapt to it.
Furthermore, the inductor sets the minimum time, how fast the current could be delivered.

I refer to the L6917D app note:

Thanks for that link, I've saved it and will be interested to look at those formulas more carefully sometime.

It seems to me that the response time is already factored into the capacitance formula:

        deltaI * deltaT
Cb>=    --------------------------
      (deltaV - (deltaI * ESR))

deltaT is defined as the "voltage regulator response time" in the AMD document (pdf pg 17).
deltaI is the max current transient
deltaV is working voltage * tolerance

If deltaT remains constant, and the only change is in the ESR, then required bulk capacitance should be reduced, if I'm understanding this formula correctly.
It's a simplification though, the Intel doc points out that the formula is assuming that the voltage regulator has no response during deltaT, then suddenly responds 100%. But that should only make it more conservative than it has to be.