LN2 capacitor testing

[TiN]

Lot of extreme people know about adding extra bulk capacitors for VGA's power convertors, and such. There are great guides online, like Shamino's one (http://kingpincooling.com/forum/showthread.php?t=185). But maybe not all understand why to add them, and which capacitors are good, which are bad for these works.

I'll describe this a bit more in this thread.

There are a lot of capacitor sizes and types manufactured today, and not expiriensed person can easily confuse,what cap to choose.

http://www.xdevs.com/kb/caps.jpg

To make right decision even books was written by experts, but we only stop here on one exact ability of different capacitors to work on extreme low temperatures. Not warmer than 77K of boiling LN2. This is practical lesson, to get idea what happen to caps on frozen hardware.

So I choose cap caps to test

http://www.xdevs.com/kb/gcaps.jpg

From left to right.

Rubycon MBZ 1000µF/6.3V ---- Nichicon 470µF/6.3V SMD ---- Sanyo OS-CON SVP 1200µF/4V SMD ---- Fujitsu 820µF/2.5V

Nichicon KZG 1500µF/6.3V ---- Nichicon KZE 1000µF/6.3V ---- Sanyo OS-CON SP 510µF/4V ---- Fujitsu 820µF/4V

Tantalum electrolyte caps from EPCOS.

All these caps are low-esr, very often can be found on switching DC-DC convertors, and widely used as good filter/charge store.

Those with caps with shiny metal cans not covered by plastic are solid polymer type, plus violet OS-CON and yellow Fuj too polymer. Most of cases you can see if cap polymer or usual liquid electrolyte, by presents of cuts on top. If top of cap can have cross or star cut - it's 90% liquid. Cuts are there to break hermetic can more easy if electrolyte boil in can due excessive load/voltage or temperature. This prevent cap from destructive explosion. Solid caps don't boil so their can top is flat usually.

Polymer caps better than liquid, have more stable characteristics, and usually much lower ESR, which mean that they can filter ripples better, and feed more current.

There is good basic presentation from Sanyo about their OS-CON's (http://www.xdevs.com/pdf/E_oscon_basic.pdf). - 1.1MB.

Pity that OS-CON's now used only in few computer parts (mainly only server hw now), coz greenpeace and such guys found their waste as harmful for environment. :headshake:

So lets show begin.
I used few caps for test.

Test is simple. Hook up Fluke 87V DMM in capacity mode to cap and read value on ambient temp, and with cap submerged to LN.

http://www.xdevs.com/oc/captest/test1.jpg

Chineese generic axial cap 470u x 16V

Nominal:

Ambient temp photo (http://www.xdevs.com/oc/captest/test1a.jpg)

Capacity drop fast when cooling started

Light subzero (http://www.xdevs.com/oc/captest/test1b1.jpg)

During cooldown it got malfunction, maybe shorten even. That's because liquid electrolyte in cap frozen.

Short occur deeper subzero (http://www.xdevs.com/oc/captest/test1c.jpg)

After further cooling, it recovered but almost with no capacity. 15nF is hundreds times less than nominal 535uF measured on ambient. So on LN temps it's just non-existant as capacitor

Cap lost all capacity in LN2 (http://www.xdevs.com/oc/captest/test1d.jpg)

Chineese LowESR 470u 16V

All similar to previous. With same shorted issue.

Ambient temp photo (http://www.xdevs.com/oc/captest/test2a.jpg)
Same shorts (http://www.xdevs.com/oc/captest/test2b.jpg)
Lost all capacity in LN2 (http://www.xdevs.com/oc/captest/test2c.jpg)

Sanyo LowESR generic 470u 6.3V

439uF ambient

Ambient temp photo (http://www.xdevs.com/oc/captest/test5a.jpg)

With frozing almost instantly got to few nF. Same no any use under LN.

No use for LN2 (http://www.xdevs.com/oc/captest/test5b.jpg)

Sanyo OS-CON SP 510u 4V

Ambient 514uF

Ambient temp photo (http://www.xdevs.com/oc/captest/test3a.jpg)

Frozen...406uF

Dropped to 80% cap, and no worse. Usable! (http://www.xdevs.com/oc/captest/test3b.jpg)

Holded it for a min in boiling LN, but no less drop. Perfect use for us!

Tantalum smd cap 220u 10V

Ambient 228uF

Ambient temp photo (http://www.xdevs.com/oc/captest/test4a.jpg)

Frozen 210uF, almost no change

Looks good for LN temps (http://www.xdevs.com/oc/captest/test4b.jpg)

Ceramic SMD MLCC 10uF

No pics coz i holded it by probes and nobody put shutter :D
No matter still, it don't change it's cap.

Of course such "testing" don't show all effects present with extreme cold, but can show a bit more to us, even if we don't know ESR,ESL and real work changes.

Maybe later, if people would interest, i'll make special testing curcuit and PCB to test caps with subzero under load.

So now we have the answer, that putting on just any capacitor for modding is NO way to go. If use generic caps, you can ever damage power curcuitry badly.
That's why only good solid caps are best for extreme benching, they loose capacity and bit but doing well in general. Tantalum and ceramic small SMD caps are best, but they aren't much big capacity, so need to use very big amount of them to get sufficient filtering/power storage. (Don't look on Volterra, coz it switching on near 1 to 2.5MHz , this make ceramic enough for nominal use and moderate oc).

Updates, error reports, suggestions are always welcome.

[FatAlbert]

so much helpful information here ;)

I think your Frankenstein G80 will be enough good do show how the capacitors work in very low temp.

[Shammy]

interesting info tin thanks!
what about Nichicon 470µF/6.3V?

[k|ngp|n]

Good stuff TiN :up: Thanx for your contribution and time!

[TiN]

Will test more caps later then :) No ln2 left now.

Forgot to add later tests too.

Sanyo OS-CON SEPC axial 560uF 4V

495uF ambient.

Ambient (http://www.xdevs.com/images/ln2_cel/capsepc1.jpg)

451uF in LN2, ok for extreme

Frozen SEPC (http://www.xdevs.com/images/ln2_cel/capsepc2.jpg)

Yellow Fujitsu polymer (with cuts on top,btw), 820uF 4V axial

830uF ambient (no pic)

731uF in after 1 min LN2, ok for extreme

Frozen Fujitsu axial (http://www.xdevs.com/images/ln2_cel/panas2.jpg)

[Expat GriZ]

Thanks for this!! Great info & testing. :up: Now to find a shop in Belgium.....:headshake:

[steponz]

Thanks Tin, Very helpful...

steponz

[carpo93]

very nice

[chispy]

Thank you TiN , very informative and useful info and explanation. Appreciated :up:.


Regards: Angelo.

[Putittogether]

Thanks TiN. I learned something:clap:

[Brian y.]

Very very nice info.

Appreciate it:up:

[Dualist]

Nice work, even nicer info :up:

[Gatman]

Excellent info, very good to know. :up:

[Shammy]

cna you test at -60C?
i think the coldest any cap will go is -60C maybe, prolly -20C at most times, -190C on a cap will prolly never happen

[TiN]

It will be bit hard, but will test for sure later.
Maybe assy some simple ATE unit to get graph "capacity/ESR related to cap can temperature".

[Shammy]

no , use kingpin f1 on the cap and control temp :P

[Raja@ASUS]

cna you test at -60C?
i think the coldest any cap will go is -60C maybe, prolly -20C at most times, -190C on a cap will prolly never happen

Yeah that's about right I think, I think even when you run a board for a few hours right down at -180 the caps probably don't get much past -60.

Polypropylene's are another option but they are expensive (very low ESR too). I think a lot of MLCC's in parallel is good enough, plus with the Volterra solution swithching over 1MHz the need for absolute 'bulk' capacitance is reduced. Of course, Gulftown's current draw and rail slew rate requirements at 6GHz are probably somewhat unique, so I guess it does not hurt to have a lot of reserve if the PWM can handle it without oscillation.

I wonder how a 'super' cap (double layer types with very high capacitance) on the high side of the rail would perform in very high current scenrios. Not sure how the initial charge would interact with PSU's at first switch on though.


Very Nice study though Tin. Unique stuff.

[BenchZowner]

Yeah TiN is my type of guy ( woah, that sounded kinda gay :D ).
He wants to be prepared to be ready.
Testing lots of stuff before going LN2 with the hardware.
Might take ages, but when the session takes place, it's nearly 100% that it will go well ;)

[Jor3lBR]

Superb thread TiN thank you!

[Hyperhorn]

I have to admit that I am not experienced with advanced voltmodding, but I would like to thank you for sharing your knowledge. Much appreciated, TiN! :clap:

[T_M]

Testing in DICE would be fairly straight forward