Swapping japanese/generic electrolythics with solid-states

[Snayperskaya]

Re: Swapping japanese/generic electrolythics with solid-states

for the asus a8v, i think u missed a cap near the cmos battery right by the mosfet. i have this exact same board and the cap and mosfet there actually regulates the 5vsb supply for the board. i recommend replacing that as well so u have 16 kzg caps to replace not 15.

You're right! I've even marked it with a sharpie but forgot to red-circle it.

[Snayperskaya]

Re: Swapping japanese/generic electrolythics with solid-states

Got one more board for cap replacing - actually it just came back from a repair lab, asked how much just to have the caps changed would cost me and they told me the job would be ~US$250 (about twice I paid on the board) and I'd like to hear suggestions on which caps to choose since I'm going to buy them all and start swapping from the easiest boards first (this one have a lot of SMDs on the back pretty close to the caps terminals). I've circled the different caps in different colors for easier reading since the picture quality is pretty terrible.

TYAN S7012:

NCC KZG (red circle):
18x 470uF @ 16V

NCC KY (yellow circle):
4X 220UF @ 16V

SUNCON (blue circle):
5x 1000uF @ 6.3V

[ratdude747]

Re: Swapping japanese/generic electrolythics with solid-states

Got one more board for cap replacing - actually it just came back from a repair lab, asked how much just to have the caps changed would cost me and they told me the job would be ~US$250 (about twice I paid on the board) and I'd like to hear suggestions on which caps to choose since I'm going to buy them all and start swapping from the easiest boards first (this one have a lot of SMDs on the back pretty close to the caps terminals). I've circled the different caps in different colors for easier reading since the picture quality is pretty terrible.

TYAN S7012:

NCC KZG (red circle):
18x 470uF @ 16V

NCC KY (yellow circle):
4X 220UF @ 16V

SUNCON (blue circle):
5x 1000uF @ 6.3V

The KY's are probably fine... On my Tyan's I've recapped it's only been KZG's and KZJ's that I've replaced (other than an ancient dual slot 1 PII/PIII, that is).

I'd replace the KZG's with Nichicon HN (Post 2005 stock only!) or HE, personally. I've used the former on several Tyans with good results (the boards had other issues, but the caps weren't the culprit).

[Snayperskaya]

Re: Swapping japanese/generic electrolythics with solid-states

The KY's are probably fine... On my Tyan's I've recapped it's only been KZG's and KZJ's that I've replaced (other than an ancient dual slot 1 PII/PIII, that is).

I'd replace the KZG's with Nichicon HN (Post 2005 stock only!) or HE, personally. I've used the former on several Tyans with good results (the boards had other issues, but the caps weren't the culprit).

Thanks. I'm inclined to replace the KYs since I've seen one with a bulged "plug".

[momaka]

Re: Swapping japanese/generic electrolythics with solid-states

Thanks. I'm inclined to replace the KYs since I've seen one with a bulged "plug".

KY is a very reliable series. I don't suggest replacing every KY cap. Maybe only the one that has bulged on the bottom, and that's if it really does look bad. Though I am inclined to think perhaps there is/was something else wrong for that cap to bulge.

Also, the Suncon caps are actually Sanyo's. Depending on which series they are, you might not need to replace them. WF series tend to be fairly unreliable. WG is hit or miss (but somewhat okay). And WX is definitely okay. I'm not sure about their other series.

[Wester547]

Re: Swapping japanese/generic electrolythics with solid-states

KY is a very reliable series. I don't suggest replacing every KY cap. Maybe only the one that has bulged on the bottom, and that's if it really does look bad. Though I am inclined to think perhaps there is/was something else wrong for that cap to bulge.

Sometimes the rubber stand-off used for the bungs is crimped in such a way that it stands out “more” (I’ve seen such a rubber bung “look” on some Nichicon PWs before, yet they tested fine and haven’t leaked at all), but usually capacitors that pop their bottoms out (outgas) do so violently, and tend to show visible signs of electrolytic leakage.

Also, the Suncon caps are actually Sanyo's. Depending on which series they are, you might not need to replace them. WF series tend to be fairly unreliable. WG is hit or miss (but somewhat okay). And WX is definitely okay. I'm not sure about their other series.

WX I don’t know, I’ve had mixed experiences with them. It might depend on the datecode. I would consider the WX series (given their overall history) the least reliable of all equivalent Japanese series (HD, ZL, KZE, FF/FM, etc).

They (1500uF 6.3V 10mmx20mm case sizes) were used interchangeably with Nichicon HD (laminated cases) and Chemi-con KZE on old Intel 815 series (Socket 370) motherboards, and of those three series, as yet, I’ve only seen WX bulge and leak. Same goes for Asus dual Socket 370 motherboards (1500uF 6.3V WX and sleeved Nichicon HDs are used interchangeably on them, and I’ve only seen WX fail there).

Also, on many old Asus Socket 939 and LGA775 motherboards, for (what seems to be) RAM decoupling and filtering, they use 330uF 6.3V Nichicon HD, ventless Chemi-con KZE, Panasonic FF, LTEC LMB, and Suncon/Sanyo WX interchangeably, and I’ve only seen LMB and WX bulge to this day (the bulge is slight but unmistakable). I’ve also seen some fail on some rather elderly MSI GeForce 3 Ti200 video cards (470uF 16V WX). I too had mixed experiences with WG, at least those produced between 2001-2007, but I do consider them (overall) quite a bit better than the bad run of KZG and Nichicon HM (but definitely not as reliable as MBZ and FJ).

In regards to that Tyan board, if the capacitors say SUNCON on the sleeve, that probably means the board was produced after Sanyo sold their electrolytic division to their previous subsidiary (Suncon), which was back in 2009. So that board is at least 2009 vintage, if not newer. I don’t recall Suncon or KZGs with 2009 datecodes or later failing without a good reason, and KY are usually fine. It can’t hurt to replace the caps if the board isn’t working, but I’m starting to wonder if they’re actually the source of the problem.

[Agent24]

Re: Swapping japanese/generic electrolythics with solid-states

Even the best will fail eventually, with enough heat and time.

[Snayperskaya]

Re: Swapping japanese/generic electrolythics with solid-states

Updating the topic:

I've got a good desoldering tool now and I'm ready for changing the caps on the Asus (reattached picture for reference).

Marked caps: KZG (820uf @ 6.3V) (I've removed all of them already)
VRM near I/O are Panasonic FL (1000uF @ 16V).
VRM near CPU socket are Rubycon MBZ (1500uF @ 6.3V).

I've removed one of each for testing on my LCR and got the following:

MBZ: 1317 uF, 2% VLOSS, 0.06 ohms ESR
FL: 945 uF, 1.4% VLOSS, 0.07 ohms ESR

Should I replace the VRM caps too? (the readings seem pretty good for me)? If so, which models should I go for?

Thanks.

[Agent24]

Re: Swapping japanese/generic electrolythics with solid-states

Well, the datasheet for MBZ says a 1500uF 6.3v part in 10x12.5mm size should have an ESR of no more than 0.026 Ohm @ 20°C/100kHz.

Now, if your ESR test is at the same frequency and temperature, but you are measuring 0.06 Ohm, then the ESR is more than double what it should be in a worst case scenario. Additionally, as mockingbird pointed out, they are probably ~15 years old, so they may well be worn out.

What I would want to check, is if you measured their ESR straight after desoldering or not. If they were freshly desoldered and still hot, then their ESR should be lower than the datasheet specifications (ESR decreases with temperature increase). If they were hot and still measured 0.06 Ohm, then for sure they are not looking good at all.

While variations in test frequency, temperature, calibration of the ESR tester etc can skew the results, I think it's more likely that they're on the way out.

[Wester547]

Re: Swapping japanese/generic electrolythics with solid-states

I believe the ASUS A8V model is closer to fourteen years old, possibly not even that. I’d guess the board was made around the latter half of 2004 / possibly early 2005, and those MBZ appear to use newer black / white PET sleeves rather than the older blue / gold PVC sleeves. That said, those MBZ also appear to be 8mm rather than 10mm (8x20 and not 10x12.5), and the highest they should read in accordance to the datasheet is 0.019 Ohm at 20ºC/100KHz. So they are definitely out-of-spec (far above what they should be in the worst allowable case), unless your LCR meter isn’t calibrated correctly.

It sounds as though those MBZ and FL are starting to dry out based on the slightly high ESR readings. It may not hurt to recap the entire board for the sake of longevity.

[mariushm]

Re: Swapping japanese/generic electrolythics with solid-states

i wouldn't replace them.
the capacitors by back panel and agp and pci slots and bottom usb headers , they're all for filtering 5 rail locally, they just reused KZG series probably because they bought them in huge volume. they don't have to be as high spec as kzg series
replace with any 820uF -1000uF 6.3v or 10v or 16v very low esr seris..can use higher voltage rating , can go slightly higher in capacitance..

[Snayperskaya]

Re: Swapping japanese/generic electrolythics with solid-states

My tester is one of the transistor multitesters, just like this one:

https://ae01.alicdn.com/kf/HTB19WaqN...er-NPN-PNP.jpg

Not sure if it can handle precise readings of milliohms.

Also, I've tried searching for FL series datasheet, but no dice. Even on the local repository by a forum user I couldn't find it. Is it a low or a ultra low ESR series?

[Agent24]

Re: Swapping japanese/generic electrolythics with solid-states

Nobody ever found a datasheet for FL. It was likely only given to motherboard OEMs. I believe they would be ultra low ESR given their typical usage.

Your tester I think is accurate enough if calibrated etc, but out of the box from China, it's probably debatable.

[mariushm]

Re: Swapping japanese/generic electrolythics with solid-states

FL series is one of those "custom" / "on special order" / "on big customer request" series, and it's probably just a tweaked series to make the capacitors smaller diameter and taller (more suitable for VRMs and packed locations).
I would treat them as similar in specs to Panasonic FM series

see notes i put on the image:

[momaka]

Re: Swapping japanese/generic electrolythics with solid-states

Like agent24 mentioned, there hasn't ever been a datasheet posted for Panasonic FL. However, since Panasonic FL is often used interchangeably with Rubycon MCZ, Nichicon HN, and Chemicon KZG by most major motherboard manufacturers, we can assume (pretty safely) that FL is ultra-low ESR series and on par with Rubycon MCZ and Nichicon HN.

I also agree that you tester/meter may not be calibrated properly. I actually have one of those cheap testers as well (maybe not exactly the same model), but it always measures higher ESR between test pins 1 and 3. On the other hand, when I measure capacitors on pins 1 and 2, the ESR appears to be accurate. So just keep that in mind, as your tester may have some similar issue, if not incorrectly calibrated.

Last but not least, your motherboard appears to have 2-3 free (empty/unpopulated) cap spots on the CPU VRM low side (the side circled yellow in the above picture by mariushm). Instead of replacing any caps, you can also just add 3 more very low / ultra-low ESR caps in those spots, and that will help bring the total ESR of the circuit down and keep the mobo working for a long time. I don't suggest doing that when you have bad caps. But when you have good cap brands such as Rubycon MBZ and Panasonic FL and the cap themselves appear to have fairly reasonable ESR, you can do this "lazy recap" just for "peace of mind".

[Snayperskaya]

Re: Swapping japanese/generic electrolythics with solid-states

Nice, thanks everyone for all the input

I didn't know you could populate open caps slots on those motherboards. I have some other motherboards with incomplete VRM phases left completely unpopulated (caps, inductor and ICs). I assume you can't do that on those?

Quality caps here are pretty hard to come by. I've asked several sellers, but only one answered me with the following (I quoted for the original specs):

1000uF @ 16V: NCC KZG (yes, and they seem new/unused)
1500uF @ 6.3V: Panasonic FM/Nichicon HD

From what I've read from the 1500uF ones, they have pretty similar specs ripple and impedance wise (with each other, MBZs do better in both aspects). Would either fare well for replacement/adding more caps on the empty slots?

[stj]

Re: Swapping japanese/generic electrolythics with solid-states

FM is no good enough to replace FL.
FJ may be close to FL

[mariushm]

Re: Swapping japanese/generic electrolythics with solid-states

@Snayperskaya

If you look on your motherboard, you will see that he DC-DC converter (the VRM) that converts your 12v into the voltage your cpu is made out of a few almost identical "modules", which are called phases - there's a controller which sends pulses of energy to each of these phases and then the smaller voltage comes out from each of these phases and gets combined into the output capacitors to give the cpu a smooth voltage.

You can see in the picture below ... i may be wrong at some points but in large it's correct... you see circled in yellow the vrm controller which monitors current and voltage output and sends pulses to phases as required, then you see the three phases (like 3 cylinders in a car engine) that work together to convert power.
Each of these 3 main phases have a mosfet driver chip probably (the surface mount chip with 8 contacts), one or two input capacitors which act like buckets of energy, a high side and 1-2 low side mosfets (they use 2 to reduce losses and spread heat in a larger area), an inductor (that big coil of wire) and the output capacitor or capacitors.
Near the cpu socket, all the outputs are joined together.



On lots of motherboards you will see they often use 2 or 3 or even more capacitors for each phase, less on modern motherboards where they have now quality polymer capacitors.

There's two main reasons they used so many electrolytic capacitors. Basically, the idea is to reduce as much as possible the ESR keeping in mind that over time, capacitors age and their technical specifications can degrade.
When you install two capacitors in parallel, you get twice the capacitance but the ESR goes down in half, which is good.
So the motherboard designer can reduce ESR in two ways... either install more capacitors in parallel (within reason, because too much capacitance can damage things), or install capacitors with lower ESR.
The ESR is lower in direct proportion with the volume of the capacitor - the bigger the volume of the capacitor the lower the ESR. As a "side effect", typically with bigger volume you also get higher capacitance and higher voltage rating.
So the motherboard manufacturer tries to get the lowest ESR by using capacitors that are bigger - in this case, they went with 1500uF 6.3v capacitors, even though they probably only need around 3-4000uF in total, and even though the maximum output voltage of the vrm is maybe 2v, which would have allowed use of capacitors with lower voltage rating.

The problem is going bigger in volume is only possible up to a point, because the capacitors are so close to the socket you risk heatsinks and fans being blocked by the height of the capacitors. So motherboard manufacturers can only go so high and so big when it comes to capacitors and when they reach these limits they simply have to go with more smaller capacitors instead.

On your particular motherboard, the designers probably designed the VRM to be stable and work fine as long as the ESR is somewhere let's say below 20 mOhm (keeping in mind degradation over time, board has to last at least the warranty period).
By using only 3-4 MBZ series capacitors they probably figured something like this:

Hey, these MBZ caps are very low esr high quality capacitors so we'll be fine with only so many of these, no need to put 2 in parallel at the output of each phase to reduce the ESR further.

But, let's leave the capacitor locations on the motherboard, just in case we'll decide to use cheaper capacitors with higher ESR for this model in the future and then we'll have to install two in parallel to get the ESR below 20 mOhm. Or, maybe we'll want to reuse this VRM design on a "value" model of this board so instead of making the layout again, let's make the layout on this board and then we can just copy paste if needed.

So stj says that FM is not "good enough" to replace FL and that FJ would be better.... Yeah, if you want to be strict about, theoretically FM would not be "perfect" replacement.

But as I explained above, these boards are designed with some tolerance, with some accounting with degradation over time, so there is some percentage around those technical specifications where the motherboard will be perfectly happy with other capacitors. Things are NOT so super strict.

If the VRM wants less than 15 mOhm to be happy and the MBZ give it < 8 mOhm and the FM series will give it < 10 mOhm then who cares, both series are less than 15 mOhm
And if you install more capacitors, as they're in parallel on the motherboard, you're reducing the ESR even further, so you could get actually better performance with not quite as perfect capacitors.
If you go with polymer capacitors (recycled from other broken motherboards or whatever), those have technical specs even better than those original electrolytical capacitors.

So don't worry so much.

I'd suggest avoiding NCC KZG series,as it's a series that's problematic... for a long time, they tended to go bad even when components were not in use.
Panasonic FM is fine, Nichicon HD is a bit too low end, not suitable for VRMs. They can be used in those places I highlighted with orange, in those less important places, but anywhere else you must use better series like FM .

[momaka]

Re: Swapping japanese/generic electrolythics with solid-states

Nice write-up mariushm. I agree with everything that is said in the above post. The only part I have something to add to is this:

... in this case, they went with 1500uF 6.3v capacitors, even though they probably only need around 3-4000uF in total

Not always true.

The total capacitance needed often depends on the switching frequency of the buck circuit (VRM) as well as the voltage feedback loop design. Some VRMs may oscillate when the capacitance is too low, even if the switching frequency is fairly high (mainly due to the feedback loop).

So in general, it is not advisable to go much lower in capacitance. How low? Most VRMs tend to tolerate around 1/3 to 1/2 of the original capacitance... which is why polymer capacitor mods work on many motherboards. But for better voltage regulation and lower ripple output (especially under very high loads - i.e. if you plan to do overclocking with a very high-power CPU), it is recommended to stay closer to the original capacitance (i.e. 2/3 to 4/5). That said, you can also go higher in capacitance to some degree without issue. Typically, I find that up to 2x the capacitance is also acceptable (as some motherboards may have been designed with both more entry-level low-ESR caps, as mariushm mentioned, and fewer high-end ultra-low ESR caps.)

I didn't know you could populate open caps slots on those motherboards. I have some other motherboards with incomplete VRM phases left completely unpopulated (caps, inductor and ICs). I assume you can't do that on those?

You can populate empty [electrolytic] cap spots, but not ICs and inductors. Most motherboards that have ICs and inductors missing also have different SMD resistors/jumpers installed for their VRM controllers. So unless you have a complete board schematic and 100% know what you are doing, it is best not to populate missing ICs, MOSFETs, and inductors.

[Snayperskaya]

Re: Swapping japanese/generic electrolythics with solid-states

Thanks again for all the nice info. I really appreciate it.

What should I look into when choosing solids over electrolythics? Are they all (or mostly) high-ripple/low ESR? Is it safe to go for same voltage/capacitance as the original 'lythic or is there anything else I should look into the spec sheets?