Low ESR capacitor lifetime improvements?

[Asuka]

I was looking for replacement capacitors for a network switch, and noticed that compared to the old low-ESR Nichicon motherboard capacitors in my switch, from 2007, which have a lifetime of 2000 hours at 105 degrees, the newer ones seem to have a lifetime of 8000 hours at 105 degrees. Did the lifetime of low ESR capacitors increase drastically recently?

[David Hess]

Some series of capacitors have a higher lifetime rating than others.  They are of course usually more expensive or may compromise other specifications.

[Asuka]

The original capacitors are Nichicon KZG series, and the 10V 1500 ?F ones are all leaking:


According to the datasheet for the current model,

Endurance with ripple current : 2,000 hours at 105C

I have been looking for a replacement and found that whereas obsolete HM series and HZ series Nichicon also promise 2,000 hours, the in-production Rubycon ZLH series

After applying rated voltage with rated ripple current for specified time at 105?,
the capacitors shall meet the following requirements.
...
?D? 8 8000

Similarly the Panasonic FR series gives 6,000 hours.

The obsolete versions of the capacitors can be purchased more cheaply but it seems that up to date ones have much longer lifetimes. Did I misunderstand something?

[David Hess]

The obsolete versions of the capacitors can be purchased more cheaply but it seems that up to date ones have much longer lifetimes. Did I misunderstand something?

Check that the replacements have similar ripple current ratings.

[Asuka]

The obsolete versions of the capacitors can be purchased more cheaply but it seems that up to date ones have much longer lifetimes. Did I misunderstand something?

Check that the replacements have similar ripple current ratings.

The rated ripple current of the Panasonic FR is 1560, the ripple current rating for the ZLH series is 1500, and the existing capacitors have a rating of 1540. Here is a screenshot of the spreadsheet I made with the other relevant-looking characteristics. What I was surprised about was the lifetime increase for similar-looking capacitors.

[mariush]

If I remember correctly, the KZG capacitors have a water based electrolyte in them and they were designed to have very low ESR. It's actually one of the bad series from United Chemi Con, at least during a period of time of a few years they had something bad with this series and a handful of others, I think the story is they used too much of some chemical in the composition.  The most recent batches of this series should be fine.

The FR series uses a different electrolyte that's not water based, so that's why it has a much higher lifetime. The compromise is slightly higher ESR and in general they're slightly taller or thicker (more volume) compared to KZG or other ultra low esr capacitors (nichicon hz,hn for example)

That switch doesn't need so low esr capacitors, they probably used those kzg because they were cheap, they're mass produced for motherboards so they probably got a good price buying them in volume.

You should be fine using panasonic FR series, or united chemi con KZE and even KY, it should be good enough for that switch.

You could also do a polymer capacitor mod , for example replace the 1000uF 16v with 1000uF or even 820uF 16v polymer capacitors and the 1500uF 6.3/10v ones with 1500uF or 1200uF polymer capacitors. It seems like the 16v capacitors are in parallel (for lower esr or maybe they couldn't fit a single taller capacitor) so lowering to 820uF shouldn't be a big deal.

The lifetime rating is different on polymer capacitors, don't compare their life with the electrolytic capacitor life...  with electrolytic capacitors you sort of double their expected life with every 10 decrease Celsius so a 2000h @ 105c will last 4000h@95c ambient temperature, 8000h@85c, 16k@75c, 32k@65c ambient and so on  ... but with polymer capacitors the formula is kinda like this  Lifetime = original lifetime x 10 ^ [ Tmax - Tamb)/20]  so for example a 2000h @ 105c polymer at 65c ambient will last 2000h x 10 ^ [(105-65)/20] = 2000 x 10^2 = 200'000h , so about 6.25x longer life compared to an electrolytic.

[Asuka]

If I remember correctly, the KZG capacitors have a water based electrolyte in them and they were designed to have very low ESR. It's actually one of the bad series from United Chemi Con, at least during a period of time of a few years they had something bad with this series and a handful of others, I think the story is they used too much of some chemical in the composition.  The most recent batches of this series should be fine.

This switch is a Corega SW24GTR, manufactured in Taiwan in October 2006, which seems to be the era of the "bad capacitors". I bought it second hand a few weeks ago. It currently works OK, but when I opened it to see if it was full of dust or something, I found these bad-looking capacitors.

The FR series uses a different electrolyte that's not water based, so that's why it has a much higher lifetime. The compromise is slightly higher ESR and in general they're slightly taller or thicker (more volume) compared to KZG or other ultra low esr capacitors (nichicon hz,hn for example)

I tried to get the similar dimensions as you can see from the spreadsheet I posted. In one place the capacitors are quite close to each other, and obviously the same through hole size would be helpful.

That switch doesn't need so low esr capacitors, they probably used those kzg because they were cheap, they're mass produced for motherboards so they probably got a good price buying them in volume.

I've done some searching about why low ESR capacitors are necessary. For example, https://electronics.stackexchange.com/questions/109385/when-should-i-use-a-low-esr-capacitor
says

The requirement for "low-ESR" capacitors normally arises in output filters of switching power supplies

That doesn't seem to be what the 1500 uF capacitors are doing, they are smoothing the power supply for the big chips I think.

You should be fine using panasonic FR series, or united chemi con KZE and even KY, it should be good enough for that switch.

You could also do a polymer capacitor mod , for example replace the 1000uF 16v with 1000uF or even 820uF 16v polymer capacitors and the 1500uF 6.3/10v ones with 1500uF or 1200uF polymer capacitors. It seems like the 16v capacitors are in parallel (for lower esr or maybe they couldn't fit a single taller capacitor) so lowering to 820uF shouldn't be a big deal.

The 16V/1000uF capacitors are not looking dead yet but I thought I should replace them since they are the same brand as the failing ones (6.3V/1500uF).

The lifetime rating is different on polymer capacitors, don't compare their life with the electrolytic capacitor life...  with electrolytic capacitors you sort of double their expected life with every 10 decrease Celsius so a 2000h @ 105c will last 4000h@95c ambient temperature, 8000h@85c, 16k@75c, 32k@65c ambient and so on  ... but with polymer capacitors the formula is kinda like this  Lifetime = original lifetime x 10 ^ [ Tmax - Tamb)/20]  so for example a 2000h @ 105c polymer at 65c ambient will last 2000h x 10 ^ [(105-65)/20] = 2000 x 10^2 = 200'000h , so about 6.25x longer life compared to an electrolytic.

I think I will replace them with the polymer capacitors if that's an option. Clearly I have a lot to learn about capacitors.

[David Hess]

What I was surprised about was the lifetime increase for similar-looking capacitors.

I am not surprised at all.  Even for capacitors with a similar temperature rating, ESR, and ripple current rating, operating life can be very dependent on construction and materials.  A 50% increase in price can yield a part with 5 times the operating life.  Operating life can also be increased by derating the ripple current.  This could be done by using a higher voltage part assuming that it will fit.

I think I will replace them with the polymer capacitors if that's an option. Clearly I have a lot to learn about capacitors.

The printed circuit board photographs show that these are the input and output capacitors for a switching power supply.  You can probably get away with replacing the input capacitors with polymer capacitors however stability of the regulator may rely on the ESR of the output capacitors so using much lower ESR polymer capacitors on the output could result in oscillation of the regulator's feedback loop.

[mariush]

That switch doesn't need so low esr capacitors, they probably used those kzg because they were cheap, they're mass produced for motherboards so they probably got a good price buying them in volume.

I've done some searching about why low ESR capacitors are necessary. For example, https://electronics.stackexchange.com/questions/109385/when-should-i-use-a-low-esr-capacitor
says

The requirement for "low-ESR" capacitors normally arises in output filters of switching power supplies

That doesn't seem to be what the 1500 uF capacitors are doing, they are smoothing the power supply for the big chips I think.

You should be fine using panasonic FR series, or united chemi con KZE and even KY, it should be good enough for that switch.

You could also do a polymer capacitor mod , for example replace the 1000uF 16v with 1000uF or even 820uF 16v polymer capacitors and the 1500uF 6.3/10v ones with 1500uF or 1200uF polymer capacitors. It seems like the 16v capacitors are in parallel (for lower esr or maybe they couldn't fit a single taller capacitor) so lowering to 820uF shouldn't be a big deal.

I think I will replace them with the polymer capacitors if that's an option. Clearly I have a lot to learn about capacitors.

What I mean is that the circuit requires Low ESR capacitors, since it's a switching power supply, but it doesn't necessarily need a capacitor series considered in the "ultra low esr" sub-category of the "low esr" category of electrolytic capacitors.

If the KZG series has 10 mOhm ESR for that 1500uF 6.3v capacitor, you can use UCC KZE with let's say 12 mOhm or KY with 15 mOhm and it's not going to be the end of the world, there's quite a lot of margin designed in these circuits. Keep in mind also that those capacitors were going slowly bad over time, with their ESR values and peak current ripple capabilities changing and the switch still worked until eventually the capacitors just got bad enough to cause visible problems.

This KZG series was designed for VRM circuits, the DC-DC converters which take 12v from the computer power supply and produce 0.8v .. 1.3v for the processor at up to 100-120A of current (spread across multiple power phases and multiple capacitors) and these DC-DC converters operate at high frequencies, up to 300-500kHz

The power supplies in your switch are switching power supplies, but they're not at the same level, they're way more simple and run at much lower frequencies, typically 60kHz to 200 kHz, and they basically convert 9v..12v to the voltages those chips need, which would be 2.5v , 3.6v, something like that. Basically, the power needed for the ICs.

In your particular case,  one of the power supplies in the switch uses a switching controller made by Richtek, it's that chip RT9214 which runs at 300kHz .. You can see the datasheet here: http://www.richtek.com/assets/product_file/RT9214/DS9214-15.pdf

You can practically compare the schematic in the datasheet with how the parts are arranged to the right and bottom side of the chip, you have the inductor (the wire around the ferrite), the two mosfets and the two capacitors to the right of them installed in parallel to reduce ESR and have enough capacity (due to height)


The KZG series was not one of those in the 2002-2004 times where Nichicon screwed up their HM/HN/HZ capacitors and a lot of Dell computers failed. It's also not one of those made with incomplete electrolyte formulas that were stolen. For a long time, UCC made this series but the electrolyte inside I guess was unstable, they aimed for too low ESR, and A LOT of them went bad even on shelves, not installed in computers, never used.  If you see a board that's before let's say 2016 and it has KZG series capacitors, you can suspect them.

[tautech]

Clearly I have a lot to learn about capacitors.

Not so much.

It's a bit simpler than what you think.
EVERY manufacturer walks a tightrope of: service life vs cost !

Components have been selected to give an acceptable MTBF in whatever application they are used in.
Sometimes lesser quality products are used BUT well inside their specifications.

For this repair, select a good brand with better specification for Ripple, ESR, temp and hrs and be quite confident they will give excellent service.