What happens when you seriously increase capacitance?

[PCBONEZ]

Re: What happens when you seriously increase capacitance?

Many 9's
Show me how the power to an AGP slot is checked before the BIOS code runs.
- Hint. It's not.
It's checked indirectly WHEN the BIOS code runs by checking for the video card's PROM.
No video -> post error -> no BOOT.

[You also completely disregard pre-ATX systems which believe it or not some people still use especially in parts of the world where people HAVE to replace their own caps.]

~~~

-
"Scientific Method" works like this.
- Create a theory.
- Try to prove it WRONG.
- It's assumed to be correct until ONE example is found where it's wrong.
-
I just gave you one example where your theory is wrong.
(Two with pre-ATX systems.)
-
You gave me an example where my theory is wrong.
-
Neither theory is always true.
-
I knew what I said wasn't ~always~ true which is exactly why I moderate such statements with "possible that" or "could happen".
-
You can call that vague if you want.
I call it realizing that not everyone else in here doesn't the same kind of equipment on their workbench as I do.

I use generalized ideas for a valid reaon.
There are far FAR too many ways to design a motherboard and far too many motherboards out there for me to test or investigate them all. (And as some manufacturers 'cheat on' the design guidelines that would be the only way to know.)
-
As soon as I (or anyone else) says something is ~always true~ about how a motherboard is designed an example will be found where it's not true.

-
One example is power to memory slots.
Some recent Intel design guidelines for motherboard power distribution specify that power to the memory slots be provided by +5vsb through a DC-DC converter to 3.3v.
-
I would say 'some' motherboards take memory power of +5vsb and you'd call that being vague. - But there is no way I could look at every motherboard (supposedly) designed to that spec to verify compliance, I didn't look at every design guide for systems with +5vsb available to see where memory power comes from, and I didn't look at AMD design guides. - Have/would you even try?

-
Another example that comes up in here a lot is what grade (ripple rating/ESR) of caps should be used in VRM's for xx CPU type of motherboard.
-
There is NO 'always correct' answer to that question because caps used in parallel raises the Ripple Current rating and lowers the ESR rating for the overall circuit.
-
With a given CPU type and in a given position in the VRM, one mobo design may use 1 or 2 high grade caps another design may use 3, 4, 5 lower grade caps in parallel.
-
Along that line, the grade or quanity of caps you see on a mobo doesn't *necessarily* mean a 'better' motherboard as a lot of people seem to think.
- MCZ's in there may be just barely enough if there's not enough of them.
- 15-20 caps in a VRM might mean el-cheap'o if they are just barely low ESR caps.

BTW Many 9's:
I'm not angry and I appreciate the argument.
You are making me look at things I haven't looked at in a while and that's a good thing.
Also other people reading might benefit from our bantering.

.

[PCBONEZ]

Re: What happens when you seriously increase capacitance?

starfury1

That's kind of my point.

Unless you are 'experimenting' (read 'playing'),,,,
and WILLING to write off the motherboard as a total loss,,,,
the smart thing to do is stay as close to the original specs as possible.

.

[davmax]

Re: What happens when you seriously increase capacitance?

The topic originally posed the question “What if capacitance is greatly increased around RAM slots?”

This has been considerably covered. The essence to the answer is “Is there any benefit?” Very often there is little benefit in increasing capacitance if the circuit is working well.

Typically a RAM slot cap does not have a heavy task compared with application to a VRM output filter. If a VRM example is taken as the worst case then conclusions can be drawn for RAM slot use. Take the example of 1000u being replaced with 4700u and used in a VRM filter with ripple frequency of 100Khz

Using the Samxon Data for GC range

1. 1000u Xc = 1.6mohm

ESR

6.3V = 20mohm

16V = 14mohm

2. 4700u Xc = 0.34mohm

ESR

6.3V = 9mohm

10V = 8mohm

What are the gains seen in the figures above for this VRM application?

• Increasing the capacitance gives marked improvement in ESR reducing heat loss in the capacitor, increasing can size and improving the output ripple.
• The output bulk capacitance has increased to reduce voltage sag on longer term heavy current draw downs.
• Using a higher voltage rating cap reduces ESR, in general reduced at a faster rate than with increasing capacitance value.

What are other implications?

• A major increase in output capacitance in the VRM will slow the closed loop response time.
• Xc in both cases is relatively insignificant.
• There will be a longer in rush current time. Usually no problem in this type of example.
• High voltage caps cost more and are larger, they may not fit. Additional they have a larger surface area to lose heat (the ambient to case temperature rise will be reduced). As cap diameter increases the heat path from the core to ambient in increased raising core temperature. This is not usually a problem in miniature caps used in this type of application.
• Running caps at below their rated voltage is quoted by Sprague as no issue today (it was in the early days of less pure aluminium). In hand with the new technology this has dramatically reduced the need to be concerned about forming a cap. The initial forming leakage current is not high and soon reduces.
• The great cap killers are core temperature (very close to case temperature), exceeding the rated voltage and reverse voltage above 1 volt.

To conclude.



Hopefully the above covers most aspects that relate to the original question. Using high voltage rating caps does cost more and they may not fit, a resulting lower operating temp will extend life, this benefit can only be gauged on the actual operating environment temperature because this is the largest determinant of cap operating temperature (life), the can size efficiency and internal dissipation benefits of a larger capacitor can then be far less significant. A prospective case for not changing to a larger cap would be where the operating case temp is not higher than 50C. A cap rated at 2000hrs @105C would have a lifetime 90,000hrs @ 50C (10.3 years 24/7).



1000u 6.3V caps are widely used for RAM slot filtering and little trouble is experienced with this value for the task required. Problems that have been experienced are more to do with low quality caps.

[PCBONEZ]

Re: What happens when you seriously increase capacitance?

Many 9's

Your PSU examples ARE the wrong kind of cap.
You specified on 5 volts Va.
The high volt caps on 5v in a PSU are used for feedback and have nothing to do with reducing ripple (same ripple on both sides = zero ripple felt) and so they don't degrade from it.
They are those little tiny ones with no vents and are not the kind of cap the discussion (or this site) are about.
.

[PCBONEZ]

Re: What happens when you seriously increase capacitance?

2000hrs @105C
Is an endurance rating.
It has nothing to do with the operating lifetime.

[PCBONEZ]

Re: What happens when you seriously increase capacitance?

Sprague contradicts themselves:

They give a table for the % of rated voltage used in their calculation of estimated operating life.

% of rated voltage ---- conversion factor
100% --- 2
75% --- 1.4
50% --- 1.0
25% --- 0.8
10% --- 0.6

Noted the effect of lowering applied voltages is less than in the Cornell Dubilier documentation I drew on earlier. That info may be older and the lesser effect due to improved aluminum purity since. (Or maybe Sprague uses better aluminum??)

I agree a larger can will have a longer life due to a lower core temperature.

But if you answer a rhetorical question about the effect of changing one or two parameters and you allow all the other variables to change also, then the answer is of no value.

That would be like answering "will a more powerful engine make a car go faster" with "No, because the car is bigger".

~~~

As we are off on considering the odd-ball variables:
-
Another consideration would be that to go up to 25v you may have to use a lower grade cap.
-
Few of the higher grades of (good brand) low ESR caps even HAVE caps above 16v.
-
The following series do not have caps over 16v:
Ruby: MBZ, MCZ,
Samxon: GA, GC, GD
Nichicon: HZ, HN, HM,
UCC/NCC: KZJ, KZG,
Sanyo WG has some 25v but only at 1000uF and below.
Panasonic FJ has some 25v but only at 1000uF and below.

I would think that if 25v caps on a motherboard was a good idea or even -SLIGHTLY- beneficial the premium grades would at least have them.

.

[davmax]

Re: What happens when you seriously increase capacitance?

PCBonez.
The rating 2000hrs@ 105C is what all temperature lifetime calcs are based. An example was given for life @ 50C. Check temp lifetime calc methodology, tables and the formula. Your statement that this has nothing to do with operating lifetime is totally incorrect, temp lifetime calculation is well understood and has been thoroughly covered in this forum.

See lifetime calculator: https://www.badcaps.net/forum/showth...citor+lifetime

Elna link:

[davmax]

Re: What happens when you seriously increase capacitance?

9999999 has made a valid input in that lifetime, which is very temperature dependent, can be improved by using a higher voltage capacitor. His aim is longer life rather than voltage safety margin.

By increasing can size and lowering ESR his way the operating temperature can be lowered by 5 - 10C.

This means a small cap operating at 65C can be replaced by a larger cap operating at as low as 55C. A 10C gain means doubling the cap lifetime.

[999999999]

Re: What happens when you seriously increase capacitance?

Many 9's
Show me how the power to an AGP slot is checked before the BIOS code runs.
- Hint. It's not.

There is no need, a higher voltage potential exists and it takes less time in uS, ms at most. If that potental did't exist the PSU would shut off.

It's checked indirectly WHEN the BIOS code runs by checking for the video card's PROM.
No video -> post error -> no BOOT.

Having a failed POST because there's no viable prom has nothing to do with the scenario you posed.

[You also completely disregard pre-ATX systems which believe it or not some people still use especially in parts of the world where people HAVE to replace their own caps.

Please quit making up nonsense. AT also has power good signal, there is no reason it would be an exception.

-
"Scientific Method" works like this.
- Create a theory.
- Try to prove it WRONG.
- It's assumed to be correct until ONE example is found where it's wrong.

Hahaha, nice try but no. You have not tested your theory and I have. End of story?

[999999999]

Re: What happens when you seriously increase capacitance?

Many 9's

Your PSU examples ARE the wrong kind of cap.

No, but why do you keep writing this? SAME model and family of cap, just higher rated capacitance and voltage is the fair comparison.

You specified on 5 volts Va.
The high volt caps on 5v in a PSU are used for feedback and have nothing to do with reducing ripple (same ripple on both sides = zero ripple felt) and so they don't degrade from it.

Show us a schematic of any AT or ATX PSU where this cap is that is "used for feedback". You are either using the wrong terms, or just misunderstanding a circuit or schematic.

A cap in parallel to such a circuit would reduce ripple. A cap in series would be a high-pass filter. There is no high pass filter needed in the feedback loop and I already wrote that there "might" be one for ripple reduction, but usually there is not, usually the IC comparitor or controller chip may have such a decoupling cap for it's own supply (vcc) and that's it, the rail being monitored with this feedback loop usually only has the caps on the main supply output stage in the LC filter(s).

They are those little tiny ones with no vents and are not the kind of cap the discussion (or this site) are about.
.

I never implied there were, whatever you might be talking about. Why don't you clarify what you ARE talking about by indicating them by component # on a schematic or at least firing up photoshop/etc and circling them in a picture?

[999999999]

Re: What happens when you seriously increase capacitance?

Sprague contradicts themselves:

They give a table for the % of rated voltage used in their calculation of estimated operating life.

% of rated voltage ---- conversion factor
100% --- 2
75% --- 1.4
50% --- 1.0
25% --- 0.8
10% --- 0.6

What is this supposed to mean? It makes no sense without more info. In idealistic terms, maximum life from a cap is achieved by a working voltage around 70-80% of it's rated voltage. Of course, this includes a bit of ripple (that's why the cap is in the circuit).

However, there is a context when they write "life", in that it means time until no longer adhereing to the specs. If a cap rated for 50V were ran at 5V, eventually the oxide layer is too thin from doing that to be able to meet the specs running at 50V anymore, so it's "life" is up, expired for running at 50V (until it's reformed, or suffers a loss while reforming), BUT, it's still viable at the working voltage because the oxide layer is correspondant to that working voltage. The main drawback in this scenario is a loss of some capacitance, but not enough to matter for our purposes, and further we were also discussing caps of higher capacitance so this very minor effect is completely countered in some cases.

But if you answer a rhetorical question about the effect of changing one or two parameters and you allow all the other variables to change also, then the answer is of no value.

You are still overlooking that other factors besides idealism are what determine which cap values are chosen. Budget and availale space are two of the primary factors on these dense and (relatively) low cost parts. Today you can get a 600W PSU for $100. Some industrial 50W PSU cost over $200. Part of the cost difference is the lower volume, but it's also a matter of raising the BOM ceiling.

That would be like answering "will a more powerful engine make a car go faster" with "No, because the car is bigger".

Lots of cars have more than one engine option. There's gearing and microchip speed limiting too but this tangent is off topic.

As we are off on considering the odd-ball variables:
-
Another consideration would be that to go up to 25v you may have to use a lower grade cap.

Depends on the specifics. Yes this may be true in some cases, and when it is true, I recommend staying with the better series of cap. I was only referring to doing it when staying with the same family/series/etc of cap, or equivalent family/series from a different manufacturer.

Few of the higher grades of (good brand) low ESR caps even HAVE caps above 16v.
-
The following series do not have caps over 16v:
Ruby: MBZ, MCZ,
Samxon: GA, GC, GD
Nichicon: HZ, HN, HM,
UCC/NCC: KZJ, KZG,
Sanyo WG has some 25v but only at 1000uF and below.
Panasonic FJ has some 25v but only at 1000uF and below.

THere are plenty of caps that prove viable for many years that aren't in your short list. Caps from most or all the above brands. Even so, you wrote they aren't above 16V, but that means you do have the option of using a 16V cap at 5V working voltage, which is below your supposed 1/3 working/rated idea.

I would think that if 25v caps on a motherboard was a good idea or even -SLIGHTLY- beneficial the premium grades would at least have them.

As already written, the variables for us are access to parts, and the available space, and build budget. On a "premium grade" board, if it really is this in build, not just market price, you shouldn't have any caps failing except the rare defect specimens. If you do have them failing, it's time to figure out why, or if you can't, at least address the core parameters of caps that would make one more robust than another. As for what options are available in premium grades, it's a matter of perceived market demand, plus you ignore something crucial - you don't NEED a supposed "Premium grade" cap if you aren't trying to shrink it down to smallest size possible. A supposed "lesser" grade cap (still very low impedance/ESR type of course) from any of the above companies will have superior specs for our purposes except a tiny bit more loss, by simply moving up to larger size.

[PCBONEZ]

Re: What happens when you seriously increase capacitance?

Many 9's

As I said the little chart contains the same factor (Mv) used in the previous post.
You'd have to read back if you don't remember that far.
Basically
Operating Life = (the factor in the chart) x (the other factors)
Using relational math (if that's the right word) you can see by what % the factor will affect operating life.

~~~~~~~~

Manufacturing cost concerns and what may end up on a board as new really isn't part of the discussion.
- I didn't read that part in depth but on a once through I don't disagree with you on that point.

~~~~~~~~

Note: ~~ By grade I am referring to comparing Ripple and ESR ratings. (So I don't have to explain that too.)

I'm surprised I have to explain this though:

I never said the other caps are bad. - I said they are lower grade.

The point was to look at what is found on modern motherboards where higher grade caps are required.

An early P3 board, a cheap P3 board, or a P2 or earlier board could certainly use your old favorites and they'd be appropriate there.

But I'm only considering grades that are appropriate in the VRM of a P4, a P3/P4 server board. (Modern boards.)

You might find these (premium grade caps) on older high quality boards in the VRM but you won't likely lower grades in the VRM of a modern board. (You 'could' if they used a bazillion in parallel but even cheap modern boards use these grade because using a bazillion is more expensive. They just use these grades from crap brands.)

~~~~~~~~
The objective was:

- Look at boards where they want the highest quality possible and see if they use 25v caps.

Answer: No
- They don't even make 25v in the premium grade caps.

Conclusion was:
If they are going to have a series of caps that is intended for use on high end boards and 25v had any advantages at all they would at least make them. - And they don't.

~~~~~~~~

I'm surprised you don't have a better feel for which series are higher grades than others.

It would take me umpteen pages to explain how I got this but the grades go like this:
I'm only going to give numbers for one can size but the same grade sequence follows through the other can sizes.
(The can size is where Ripple and ESR are consistent withing a given series.)
Numbers are Ripple/ESR.

~~~
2880/.009 -- - HZ, GA
~~~
2220-2350/.012 -- - KZJ, GC, HN, MCZ,
~~~
1870/.016 -- - FJ, FL
~~~
1870/.018-.019 - - HM, WG, MBZ, GD
~~~
1870/.021 -- - KZG
~~~
1500-1700/.020-.033 -- - ZLG, GE, FM, HV, ZLH, KZM, KZH
~~~
1250/.041 -- - ZL, HD, GK, KZE, WX,
~~~
1250/.054 -- - ZT (high temp)
~~~
1050/.069 -- - KY, HE, YXG, YXH
~~~
<1000/.069-.085 -- - FC, PW, LXZ, PM (Some others fit here but I didn't care this low.)

And before I get accused of making rules again.
-
This is only a quick reference guide to tell you which series's are best to look in when replacing caps so you don't have to open up every flippin' spec sheet on the planet to find what you need.
-
You still have to look at the specific cap you are thinking about because for a given volt/uF cap one mfr may not use the same can size as another and can size directly affects Ripple and ESR ratings.

That's what I use for me and I'm not going to spend 2 days explaining where it came from.
- If you don't like it then don't use it.

~~~

Back to those PSU caps. - Pard'ner. You are the confused one.
- Or you haven't even looked at a schematic.

You don't remove ripple by dumping it through an IC.

The high volts caps you pulled as example are connected at one end to the output of the VR(s) and on the other to the feedback pin on the IC controller chip through a resister. The is no connection to ground and there is no connection to Vin of the VR.

- They are not filters.

~~~

If you read back you'll see what I said was:

>>>> The working voltage (circuit voltage) needs to be no less than (approximately) 1/3 of the rated voltage of the cap. <<<<

Note: "approximately"
The 1/3 is just a rule of thumb. (Ball park.)
It's not a hard limit or an actual spec. - I never said it was.

~~~

Why do I need to clarify to you what kind of capacitors I'm talking about when it's the subject of the site?

Clarification should only be needed if I'm talking about some other kind of cap.

~~~

[PCBONEZ]

Re: What happens when you seriously increase capacitance?

davmax

Your elna link is redirecting me to some site that wants to sell me an Elna sewing machine.

I'm looking for the right one.

.

[davmax]

Re: What happens when you seriously increase capacitance?

Sorry the link was corrupted. Try this one:
https://cdn.badcaps-static.com/pdfs/...0dc91bca69.pdf

[PCBONEZ]

Re: What happens when you seriously increase capacitance?

davmax

I see here and other places where people look at that data sheet and they see the xxxx hrs @ 105C and think it's the estimated lifetime of the cap.

If you can take that number and determine the lifetime of a cap with your equation that's cool (very!) but I'll have to look at that later.

~~~

For those other folks that don't realize what that number is:

They take a cap and abuse it...
- put it at 105C
- apply max rated voltage
- apply max rated ripple
- (think there's a specific frequency, don't remember.)
- hold it there for xxxx hours

Then:
They return it to 20C for 24 hours and test it.
- The capacitance has to be within 25% of the rating.
- The dissipation factor has to be not more than 200% of the rating.
- The leakage current can be no more than the rating.

The xxxx hours it is abused and still passes the last test at becomes the "xxxx hrs @ 105C" you see in the data tables.

~

It's not lifetime as such.

It's more how long you can abuse it without breaking it.

..

[davmax]

Re: What happens when you seriously increase capacitance?

I am very sorry to inform you but your statements do not relate to what most of us and the manufacturers state about lifetime. Where the major determinant of life is operating core temperature (case is within 2C of core).

Your input regarding the life impact due to lowered operating voltage is interesting, however the evidence is contradictory and manufacturers do not included data or reference to this parameter in their published data. In general the evidence both practically and in terms of data presented, makes this a doubtful academic area of argument to follow.

[PCBONEZ]

Re: What happens when you seriously increase capacitance?

That number in the data sheets is not Lifetime.

It's Endurance.

.

[davmax]

Re: What happens when you seriously increase capacitance?

Your explanation of the 105C rating as simply endurance and abuse is a fanciful and an unfortunate misunderstanding of manufacturers data which is placed there to allow engineers and others to calculate lifetime.
Read and research more carefully and for your sake be very careful what statements you make. Endurance and load life are interchangable terms.

An example from a data sheet is attached.

[davmax]

Re: What happens when you seriously increase capacitance?

A caution for those who would like to adopt the higher voltage/larger can size method of increasing cap lifetime. Be careful with the leads, the lead spacing increases from the original component. Make sure that no stress is placed on the lead sealing by forcefully squeezing in the leads. Any gas leak created by such action will remove any life gain improvement and may cause premature failure.

[999999999]

Re: What happens when you seriously increase capacitance?

Many 9's

As I said the little chart contains the same factor (Mv) used in the previous post.
You'd have to read back if you don't remember that far.

And as I wrote, the chart is meaningless without more contextual info.

Basically
Operating Life = (the factor in the chart) x (the other factors)
Using relational math (if that's the right word) you can see by what % the factor will affect operating life.

Please supply all info, your attempt at summary leaves out all vital paramaters.

Manufacturing cost concerns and what may end up on a board as new really isn't part of the discussion.

It is exact what the discussion is about, you are arbitrarily assuming the parts that are chosen based on cost, are somehow ideal in functionality. They aren't. Getting the part to work acceptibly at lowest cost is objective #1 for these low cost consumer parts, but we see (having to do the repair at all), that a mistake was made, that it wasn't the right part choice to achieve an acceptible lifespan. To be blunt, we are correcting their mistakes.

I'm surprised I have to explain this though:

I never said the other caps are bad. - I said they are lower grade.

The point was to look at what is found on modern motherboards where higher grade caps are required.

False. Supposed "higher grade" caps are only required if as I wrote previously, you are needing or trying to put the smallest part possible in the location. Sometimes it is manditory, there just isn't enough space otherwise, but this not what we were discussing, we were discussion when the space isn't a limitation and the higher spec'd voltage part WILL fit. There's nothing "higher grade" about using small cap with better specs for the size, versus a larger cap that achieves the same specs. People thing higher grade means lower ESR, but using the larger cap also lowers ESR. YOu have no point about "grade" of caps if you don't point out a specific functional parameter that matters, where the supposed "higher grade" cap still betters a larger, still very low ESR grade from same manufacturer. I think you never look at spec sheets.

An early P3 board, a cheap P3 board, or a P2 or earlier board could certainly use your old favorites and they'd be appropriate there.

But I'm only considering grades that are appropriate in the VRM of a P4, a P3/P4 server board. (Modern boards.)

YOu're not considering anything, you're randomly making a guess. The FACT is, in all parameters we can consult the datasheet and see whether the larger cap meets the specs of the ones you claim as superior. They are superior in size, that's it... the moment you aren't limited in size, the supposed superiority vanishes in many cases. The topic was about what a suitable functionality cap would be and if one of higher spec would work acceptibly, and it will, but you only cited a vague concept without testing it.

You might find these (premium grade caps) on older high quality boards in the VRM but you won't likely lower grades in the VRM of a modern board. (You 'could' if they used a bazillion in parallel but even cheap modern boards use these grade because using a bazillion is more expensive. They just use these grades from crap brands.)

You keep writing "premium grade" but you are again wrong. If what they chose was ideal for the circuit, we would have no need to replace them. We are replacing what was inappropriate and failed.

- Look at boards where they want the highest quality possible and see if they use 25v caps.

I already covered this, but your vanity prevented learning.

Answer: No
- They don't even make 25v in the premium grade caps.

Also already coverd, but yes there are some premium grade caps like Pansonic FM in 25V range.

Want the bottom line? What you claim is a problem, I do fine. So do others. YOu are the one who through misunderstanding, can't achieve the success others do by consulting datasheets for any potential replacement and qualifying it for the job instead of just clinging to some idealized list and not recognizing that the whole point was what parameters were gained with larger size. You again ignore that 5V working voltage using a 16V caps is also below the threshold of your arbitrary proposed rule, and yet is proven to work fine.

please just do the work instead of repeating hearsay about an overgeneralized concept.

Conclusion was:
If they are going to have a series of caps that is intended for use on high end boards and 25v had any advantages at all they would at least make them. - And they don't.

Wrong, the conclusion is as mentioned already, that they are building to a budget, that it would cost more for these caps and they take up more space. You keep trying to refer to what an original design used, but if the original design was ok, we wouldn't be replacing caps!

I'm surprised you don't have a better feel for which series are higher grades than others.

I'm surprised you keep repeating things which you have no experience with. You are just recanting something you tried to memorize without "knowing" or "doing". YOur supposed proofs are just a house of cards.

I'll repeat it again - I don't just write it, I do it.