Polymers to replace wet electros

[davmax]

Considering caps at the VRM output.

Using Polymers to replace Low ESR electrolytics on a motherboard.

Polys are small, have very low ESR at lower capacitance and can operate at higher frequencies (up to several Mhz). On an existing board the design is fixed around non polymer and the higher freqency characteristic is of no advantage.

A number of people are using the polys based on the fact that the ESR is low
eg a 820uF has an ESR of 7-8 milliohms compared with
3300uF 6.3V = 9
2200uF 6.3V = 14
1000uF 6.3V = 18

Consider replacing a 3300uF with an 820uF poly the capacitance has dropped by 4:1 increasing ripple and transients to >3 times. With only a small difference in ESR between the two products the output ripple will be far worse with the 820uF.

Replacing a 1000uF with an 820uF the cap ratio is now 1.22:1 so ripple and transients will increase marginally, however there is a drop of ESR of about 2.5:1 therefore ripple current will increase by 2.5 making sure that less ripple reaches the CPU.

Polys are the best for new higher frequency VRM designs.

[bgavin]

Re: Polymers to replace wet electros

Davmax,

I have a number of boards that are hyrbid VRM: polymer + electrolytic. All the Intel boards in my possession are like this. My new Gigabyte GA-P35-DS3 is entirely polymer.

Nearly all the non-polymer are KZE and KZG, which I replace. I did this last night at a client site for a D845GBV board because the 6.3/3300 KZE were bulging. I don't have 16v/2200 in stock, so I left those alone.

The KZE have been the failure point on all the Intel boards I've worked so far.

[Fizzycapola]

Re: Polymers to replace wet electros

Voltage uF Height Dia. Pitch Ripple mA ESR Ω Mftr. code
100Khz 100-300Khz
100uF 10.5 10 5 3210 30 20SA100M

Permissable ripple looks very high for 100uF capacitor, better than a 2200uF Rubycon ZL. Despite not being apparent on paper, solid capacitors may be better in high temperature environments than liquid capacitors despite what spec sheets say.

I think I read somewhere
For smoothing, a single 10uF Oscon out-performs upto 3 1000uF electrolytics.

Only drawback of Solid type caps is to do with rush current which can cause damage when rapeatedly switching devices on/off , apparently at a certain point it becomes beneficial to use a resistor to eliminate this issue or a thermistor type thing.

[davmax]

Re: Polymers to replace wet electros

I think I read somewhere
For smoothing, a single 10uF Oscon out-performs upto 3 1000uF electrolytics.

This is an extreme statement that is totally misleading. Smoothing has everything to do with capacitance eg no capacitance = max ripple, larger capacitance = less ripple. ESR is related to bypassing ripple current away from the processor. Both the cap and CPU are connected in parallel. A 1.4V 60W CPU is a resistance of 32.6milliohm. For a capacitive filter to effectively to bypass 99% of the ripple the output caps must have a combined ESR of 0.32 milliohms, the caps take most of the ripple away. This is theoretical example 99% is not normally achievable.
In the quest for low ESR specs it would seem many are forgetting the role of capacitance.

[davmax]

Re: Polymers to replace wet electros

Davmax,

I have a number of boards that are hyrbid VRM: polymer + electrolytic. All the Intel boards in my possession are like this. My new Gigabyte GA-P35-DS3 is entirely polymer.

Nearly all the non-polymer are KZE and KZG, which I replace. I did this last night at a client site for a D845GBV board because the 6.3/3300 KZE were bulging. I don't have 16v/2200 in stock, so I left those alone.

The KZE have been the failure point on all the Intel boards I've worked so far.

What you are saying is not clear. Hybrid in what way? Wet at the VRM input and polymer at the output?

Yes my Gigabyte board is all polymer. It is designed that way.

I hope you got my point about care in replacing wet caps in older designs.

[bgavin]

Re: Polymers to replace wet electros

Yes indeed I got the point. I am quite happy to replace KZE with MCZ because it works for me.

By hybrid, I mean there exists mixed poly + lytics on the vrm side of the socket. This is the side between the I/O shield and the socket proper. I don't know VRM well enough to know if this constitues In vs Out sides.

I don't replace any of the polys... just the KZx caps. I have equivalent replacements in stock, so 6.3/3300 is replaced by MCZ 6.3/3300, etc.

[davmax]

Re: Polymers to replace wet electros

Looking at polymer caps available it is clear that the product range is targeted at specific computer sections.
2.5V for VRM output Athlon XP and Pentium 4 typically 1.5V OC < 2.5V
4.0V for 3.3volt circuits
6.3V for 5 volt circuits
16V for VRM input

So for replacement of wet caps

Replace 1000u 6.3V with 820u 6.3V usually these caps are scattered around the main board.
3300u 6.3V (VRM out) 2700u 2.5V ( next to processor.)
16V wet caps really have no equivalent polymer for older system designs.

[davmax]

Re: Polymers to replace wet electros

It would seem that polymers are for the enthusiast. Their cost seems to be about 3 times higher than the usual wet caps.