ESR: How low is too low?

[Agent24]

I have a Delta PSU, a DPS-370AB A Rev 01F in fact.

It uses a mixture of Nichicon, UCC, LTec and wait for it... CapXon.
Delta! You should know better... and in an HP server too...

Guess which capacitor is bulging?


Due to the dimensions of the CapXon (GL Series: 0.024 ESR), the replacement comes down to Panasonic FC and Rubycon ZL. The ZL is cheaper since I don't have to buy in bulk. Imagine that. Saving money on GOOD parts!

The FC matches RCR and ESR closely: 0.022 Ohm
The ZL has much better RCR and ESR lower: 0.015 Ohm


I know if you lower ESR too much in PSUs you actually get more ripple. Considering that the Rubycon is probably half or more than half less ESR than the CapXon, is it OK to use? Or do you have to *seriously* decrease ESR to Solid Polymer levels to cause these kinds of problems?

[Wester547]

Re: ESR: How low is too low?

Which rail is the failed CapXon on? If it's on +5VSB then uzing ZL there would be no problem at all. +5VSB circuits actually benefit lots from very low ESR capacitors since it would be on the output of a flyback. And if it's a startup capacitor that went bad, same sentiments. The other rails on the secondary (except maybe -12V) might not be so generous. The FC would be a better option there since it's closer to the original.

[Agent24]

Re: ESR: How low is too low?

Sorry, forgot to mention that.

It's a 3300uF and it's on the 12v rail along with a 1000uF UCC KY that tested just fine for ESR and capacitance.
(The CapXon on the other hand has dropped to 560uF and is at 1.5 Ohms)

[mockingbird]

Re: ESR: How low is too low?

The 16V 1000uF UCC KY is 0.046 Ohms (Assuming it's the 10mm version)... In theory, you would be better off matching the ESR of the KY cap, in practice, it probably won't make an ounce of a difference. Use the most inexpensive cap you can get.

[Agent24]

Re: ESR: How low is too low?

In theory, you would be better off matching the ESR of the KY cap

Why?

[theOracle]

Re: ESR: How low is too low?

a bit lower ESR is fine and can even be an improvement, but going super low on a circuit that doesn't call for it can cause oscillation problems

on a 100uF cap, I would not go below 20 mOhm
on a 200uF cap, i would not go below 10 mOhm
and so on

going from 0.022 Ohm original spec to 0.015 Ohm should be fine

[mockingbird]

Re: ESR: How low is too low?

Why?

He said they're both on the 12V rail, so they're part of a pi filter. In theory, you're supposed to match the ESR of caps on the same pi filter.

[Agent24]

Re: ESR: How low is too low?

He said they're both on the 12V rail, so they're part of a pi filter. In theory, you're supposed to match the ESR of caps on the same pi filter.

But in parallel, the ESR seen by the filter will be different to that of the individual capacitors, right? Or is there something different which happens in this case?

If I make the new one the same ESR as the UCC KY (and hence, higher ESR than the original CapXon when it was new) - then the paralleled resistance will be higher than with the original capacitors and I assume, not a good thing.

This is just my guess, which is why I am asking

[Agent24]

Re: ESR: How low is too low?

Just had another look at the PSU.

Each rail after being rectified goes through a toroid and then on to the capacitors.
(In the case of the 12v, 5v and -12v rails they are on the same core)

Only the 3.3v, 5v and 5vSB rails actually have pi filters.

The 12v rail just has the UCC KY and CapXon GL capacitors in parallel, no coil between them. In combination with the toroid I guess this makes it an LC-filter?

(which would make the Pi filters actually LCLC filters?)


What effect does this have on ESR requirements? Lower ESR is better or even worse than with the pi filter?

[Wester547]

Re: ESR: How low is too low?

There is one last thing to consider, and that is the feedback loop. Even if the +12V rail forms a capacitor (C) filter in parallel instead of a PI filter, changing the ESR of the capacitors too much can mess up what the feedback loop is "tuned" to (on anything besides +5VSB), so to speak, so I don't recommend going lower than 0.015 ohms.

[mockingbird]

Re: ESR: How low is too low?

Good question, you're far more knowledgeable than I am on this matter.

[budm]

Re: ESR: How low is too low?

http://www.kemet.com/kemet/web/homepage/kfbk3.nsf/vaFeedbackFAQ/FABDB2AF5298C80A852576F80077B7B7/$file/2010-04%20CARTS%20-%20Can%20ESR%20Be%20Too%20Low.pdf


When you design the circuit, the ESR and Impedance parameter of the cap will be part of the consideration.

[momaka]

Re: ESR: How low is too low?

Since the CapXon was in parallel with the UCC KY, I'd say both the ZL and the FC will likely work fine.

If the PSU runs very hot, go for the FC since it uses non-aqueous electrolyte. You'll probably get longer life out of it. Not that you won't with the ZL. Both caps will likely outlast the useful life of the PSU.

[goop]

Re: ESR: How low is too low?

What about in graphics cards?

I'm looking for a replacement for a 1500uf 6.3V KZG, which I believe is rated 26mO, and about 1.5A for ripple current. HZ is less than half at 12mO. Would it be better to go with HM or HN at 18mO?

Am I correct in assuming that a higher ripple current rating can only help, regardless of the usage scenario?

[mariushm]

Re: ESR: How low is too low?

You can go with HN or HM, no problem at all. HZ would be a bit too low, but even then I don't think there's going to be a problem.

Keep in mind that KZG had 26 mohm esr when it was new, as the card was working and going through heat/cold cycles the esr was slowly going up as the capacitor was deteriorating. So the ESR is really not that critical and you could also probably go with capacitors that have higher ESR like Panasonic FR for example.. higher esr but more resilient to heat.

[goop]

Re: ESR: How low is too low?

The problem is I'm looking for 12.5mm height. The store doesn't have HN, only HZ (lower ESR) and HM (16mm). What might be the effect of a too low ESR?Hidden long term problems, or just doesn't work at all?

[Agent24]

Re: ESR: How low is too low?

As far as I know, in power supplies where the ESR matters as part of the design, a capacitor with ESR that is too low can cause higher levels of ripple than normal. This of course can be a problem for whatever the PSU is powering. I don't think the PSU itself will actually be damaged but I am not sure.

[goop]

Re: ESR: How low is too low?

budm attached above PDFs that talk about how too low an ESR can destabilize a PSU. I suppose this could apply to any circuit, but I don't know what the result may be in a graphics card.

[momaka]

Re: ESR: How low is too low?

What might be the effect of a too low ESR?Hidden long term problems, or just doesn't work at all?

On a video card, it won't matter at all, because DC-DC voltage regulation is usually done either with a buck-type regulator or linear regulator. Linear regulators don't really care about ESR and ripple current too much - as long as it's a functional cap (i.e. not gone high ESR), it will work. For buck type, you do want low ESR and going with the lowest possible doesn't have any adverse effects. In fact, going very low with the ESR usually allows you to use lower capacitance if needed. CPU VRMs are the same way (since a CPU VRM really is just a buck-type regulator). That's how some manufacturers swap the electrolytics for polymer capacitors.

So to answer your question more directly... no there won't be any adverse affect or long term problems with using caps with "too low" of an ESR.

But again, that's only for graphics cards and applies for the most part to motherboards as well. PSUs are a little different. There, the very low ESR can make the output filters resonate at some parasitic frequnecy, thus putting out more noise and ripple on the output of the PSU and also possibly throwing off the regulation feedback loop. So very low ESR in PSUs is experimental for the most part (with very few successful results).

[goop]

Re: ESR: How low is too low?

That's good to know. It looks like there isn't a large selection of small height wet electrolytics now. If lower ESR is fine there are more options, like polymers (aluminum) or SMTs. Nichicon NU in particular is not too expensive. And I think they may be more reliable, too?

What forms of DC-DC regulation would be sensitive to too low ESR?