Announcement

**Harvey** · 02-15-2008, 05:48 PM

Re: PC-Chips(ECS) 810lmr/810lm

Ugh, PC Chips.

- Other than that, it really depends how hard your going to be pushing the VRM. I've retro-fitted an aOpen board that had 8x2200uF with 8x1800uF as it was all I had handy; and it worked well enough with a slower CPU (1.5Ghz) but became unstable when using a CPU that was at the top end of the boards capability (2.8Ghz).

I'd risk trying one by replacing the 2200's in the VRM with 1800's and see if it can cope - a good thrash test (or even installing Windows) will soon show if its stable or not.

**Bookworm** · 02-15-2008, 06:09 PM

Re: PC-Chips(ECS) 810lmr/810lm

Duron 750's.

**PCBONEZ** · 02-15-2008, 11:27 PM

Re: PC-Chips(ECS) 810lmr/810lm

What are the ESR and ripple ratings of the old caps vs the ratings on the new caps.
That matters a lot more than nailing the uF dead on.

.

**Bookworm** · 02-15-2008, 11:52 PM

Re: PC-Chips(ECS) 810lmr/810lm

Hard to say. The new ones I can use are either 10v,1500uf chemicon KY, 6.3v, 1500uf JXA Rubycon (low ESR, high ripple), and 16v,1800uF MCZ (Rubycon). If I could get away with it, I'd prefer to use the JXA's, I have more of them.

The originals were(are) G-Luxon LZ series.

I don't have any charts that tell me what the ESR and ripple are on the various caps.

**Bookworm** · 02-16-2008, 12:15 AM

Re: PC-Chips(ECS) 810lmr/810lm

Looks like the Max ripple is 1220 on the JXA 1500uF.

Max ripple on the KY would be 1650.

**Bookworm** · 02-16-2008, 12:21 AM

Re: PC-Chips(ECS) 810lmr/810lm

The lowest grade of 6.3v, 2200uf "Good" caps is about 1440 ripple.

Okay - the _lowest_ is about 960.

**PCBONEZ** · 02-16-2008, 07:54 AM

Re: PC-Chips(ECS) 810lmr/810lm

G-Luxon LZ series specs
9 - 6.3v, 2200uf (10mmx25mm=1220/.051)(10mmx30mm=1310/.048)
3 - 6.3v, 1500uf (10mmx16mm=1070/.055)(10mmx20mm=1250/.044) ..
4 - 10v, 1000uF (8mmx20mm=1080/.065)

The best idea would be to buy the right caps for the application.
Changing values is never a sure thing.
If you are going to use on-hand caps and lower uF that much you need to reduce ESR significantly. (Maybe 1/2) - Of those you listed I would use the MCZ which is overkill in the safe direction, JXA isn't low enough ESR and KY is iffy.
-
.

**Bookworm** · 02-16-2008, 10:56 AM

Re: PC-Chips(ECS) 810lmr/810lm

If the 2200uF G-Luxon are 1220 ripple, that means that the JXA is identical (in ripple, but not capacitance)

I have to wonder, if you can double capacitance (I've seen people do it a lot here), why is it that it can't be lowered much.

If I was going to wait to buy, I have _five thousand_ in transit. I just don't know when they'll show up, and I wanted to fix two boards well enough so that they'd work with Duron 750's for a year or so - by Sunday.

**PCBONEZ** · 02-17-2008, 07:08 AM

Re: PC-Chips(ECS) 810lmr/810lm

This is the kind of thing I have trouble explaining without being able to draw pictures. LOL
-
It's not really good to go too far up in uF either.
- Yeah, I know people do. Anyone knowledgeable and sensible doesn't go up more than one value unless it's a project/experimental (hobby) board anyway. And that's fine.
-
In over-clocker forums with those clueless wonders offering all their advice I see people going up in uF thinking the improvement is from the uF when in reality they lowered ESR in the process (because they used better caps) and that's where the improvement actually came from.
- If they go up in uF and don't lower the ESR there is a chance of overheating and damaging the capacitors inside the CPU. (Explained below.)
-
The capacitors are only one part of the filter and only one of the filters.
The uF value is chosen to filter best at a specific design frequency.
(It's a bell curve. The design frequency is the highest point.)
If you change the uF then you change frequency the filter is optimal at.
(Moves the curve to right or left.)
-
People forget that these filters do the relatively low frequencies and the the higher frequencies are taken care of by small caps either built into the CPU or in/near the CPU socket (depending on the CPU type). These two sets of filters work TOGETHER.
[Two *overlapping* bell curves side by side.]
Same-same in non-CPU circuits. Memory, chipset, add-in cards all have these small high-freq filter caps.
-
Making the lytic caps smaller in uF shifts them to be better at higher frequencies and that takes a bit of work away from the high-freq filters.
[Bell curves become closer together.]
Going the other way (higher uF) makes the high-freq filter caps work harder.
[Bell curves become further apart and there is less sharing of the load.]
-
Changing the uF affects 'what frequency' is filtered best.
Changing the ESR changes 'how easy' it is for the cap to filter it.
[Lowering ESR shifts the bell curve UP.]
Changing the ripple changes 'how much' the cap can filter without overheating.
- All three of these things work together.
- It's not just the one frequency that is affected. They filter nearby frequencies almost as good as the optimal and the effectiveness lessens as the frequencies get further away from that value. (The bell curve.)
-
-
Using bogus numbers for explaination::::
Lets say the optimal is 200kHz, with 3300uF, ESR=.040.
Effectiveness at 200kHz is peak = 100%.
-
Change uF to 2200 but leave the ESR as is.
Now 100% effective occurs at (say) 300kHz.
At 200kHz it is only 75% effective now.
[Curve moved left. Curve at 200KHz point is lower.]
-
So, we lower the ESR which makes it 'easier' for the cap to filter (at all frequencies) and hopefully that returns the effectiveness at 200kHz to back near where it was before.
[Bell curve move up. Value at 200kHz is (ideally) at least as much as before.]

That same thing happens if you raise capacitance.
[Move bell curve right instead but it's still lower at 200kHz just as moving left.]
You again use a lower ESR to improve performance of the whole frequency spectrum and hope that gets the filtering at the design freq into spec.
[Moved bell curve up again to restore value at 200kHz.]
-
If you raise uF and do not lower ESR there is a risk that the larger uF lytic (being less effective at higher freq end of the curve than it used to be) will pass more work on to those caps in or near the CPU.
[Reducing the amount of overlap in the two bell curves. Where they shared the load before they no longer share.]
-
Those aren't caps you'd really want to work hard or to overheat. You can't fix caps inside your CPU and those little SMD buggers under the ziff socket are all but impossible to work with.
-
-
It suffices to say when you go changing uF values there is a lot more going on than one might think.
-
It's best to stick to at/near the original uF if possible.
- But it isn't always possible so it's good to know what to do when you can't.

And if it's just a hobby board, HEY, experimenting is FUN!
.

**Bookworm** · 02-17-2008, 11:55 AM

Re: PC-Chips(ECS) 810lmr/810lm

Eh - it's a 750 mhz, so it won't be pulling much juice, compared to say, a 2200+ (or even a 1400).

I recapped with the 1500 JXA's, and ran it overnight doing a hard drive test. Mosfets were staying cold, so were the caps.

I'm going to call it a 'good enough for this job' - which boils down to 'if it lasts a year, it's longer than I need it to' - since I'm donating it to a reservation, along with five identical boards that have good caps sets, (six machines, total), and it'll be used by people who don't have one, probably for schoolwork.

Interestingly enough, all the G-Luxon caps I tested, other than the obviously bulged one, were registering around 2200-2300uF (once I pulled them out. In circuit, they say OL). The bad one was registering 1300uF

And the board was still running fine. I hadn't had reports of a crash on it at all. It was replaced due to upgrades, not reboots.

My guess is that they overdid the caps because it was 1) cheap, and 2) a way to make up for cheap caps.

**kc8adu** · 02-17-2008, 01:21 PM

Re: PC-Chips(ECS) 810lmr/810lm

that capacitance meter is worthless to check mobo caps.
they can read in tolerance but by nearly open esr wise

**Bookworm** · 02-17-2008, 07:41 PM

Re: PC-Chips(ECS) 810lmr/810lm

"That capacitance meter".

Which one? The one that I've never mentioned the name of, so you can't actually make a judgement call?

It's a Fluke 112 DMM with capacitance reading ability.

**PCBONEZ** · 02-17-2008, 07:49 PM

Re: PC-Chips(ECS) 810lmr/810lm

How is the CPU temp doing?
You changed that too.

**Bookworm** · 02-17-2008, 11:28 PM

Re: PC-Chips(ECS) 810lmr/810lm

CPU temp was running at ... 97F, as I recall. Left the original 750 on the board. I did replace the fan, because the original was copper, but didn't want to cool really well. I just put a new fan on, new thermal compound, (after cleaning off the old crap with paint thinner/mineral spirits)

I ran it overnight, then checked the temp again, and it was staying under 100F

**PCBONEZ** · 02-18-2008, 12:33 AM

Re: PC-Chips(ECS) 810lmr/810lm

Cool.

**kc8adu** · 02-18-2008, 12:59 AM

Re: PC-Chips(ECS) 810lmr/810lm

Originally posted by Bookworm

"That capacitance meter".

Which one? The one that I've never mentioned the name of, so you can't actually make a judgement call?

It's a Fluke 112 DMM with capacitance reading ability.

still useless for testing mobo caps.unless it has an esr function.which iirc even fluke never put on a dmm.

**PCBONEZ** · 02-18-2008, 02:21 AM

Re: PC-Chips(ECS) 810lmr/810lm

Meter:
Capacitance meter doesn't help much when ESR matters.
The ESR can be hundreds of times what it supposed to be and capacitance still be in spec.
.

**Bookworm** · 02-18-2008, 12:03 PM

Re: PC-Chips(ECS) 810lmr/810lm

Eh - I replaced them. I just wanted to see the capacitance.

I just thought it was funny that the one that visibly bulged was half the capacitance it should be, the others were normal.

Again, this machine was in constant operation with no issues until it was replaced with a newer box. I just didn't want to give anyone a machine with obviously bulged caps. (well, maybe in the power supply... *whistle*)

Announcement

PC-Chips(ECS) 810lmr/810lm

PC-Chips(ECS) 810lmr/810lm

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment