CR2032 vs 5.5v 1 Farad Super Capacitor?

[XOIIO]

Alright, here's something I was thinking about in my spare time, how effective would a 5.5v 1 Farad super capacitor be in comparison to a cr2032 battery, when being using for BIOS/CMOS, and real time clock backup applications? I have a few Panasonic ones laying around, and I was thinking of possibly adding one in to my alarm clock in place of the CR2032 it would usually use. For computers it is worthwhile to use the CR2032, but for something like an alarm clock that is not super crucial, I was thinking it may be a good candidate to replace the more expensive battery, especially since it would last longer than the CR2032, which from my experience die out anyways even if they have not been used during a power outage.

I did a quick search but I could not find a comparison anywhere online, I am going to try and find the discharge rate for these super capacitors, but I'd like to hear what other people think, or if anyone has done it.

I ordered the radial super capacitors, they would be pretty easy to swap out for the coin cell holder. They are fun to mess around with anyways, you can make some dubiously useful usb rechargeable LED flashlights with them that work for an actually decent time if you are working inside a  PC.

[BravoV]

Though not an expert, here my thought :

1. Discharge curve

Lithium battery has a very flat voltage at 3 volt (nominal) during it's life, and it will drop like a stone at the end of it's life. Capacitor has linear slope discharge which may cross the minimum voltage required to power your circuit even it has not fully discharged.


2. Self leakage

Ultra cap known to have a relatively high self discharge compared to Lithium cell, its like night & day, check the datasheet. Lithium cell self discharge is so low that usually will last for many years.

[XOIIO]

Though not an expert, here my thought :

1. Discharge curve

Lithium battery has a very flat voltage at 3 volt (nominal) during it's life, and it will drop like a stone at the end of it's life. Capacitor has linear slope discharge which may cross the minimum voltage required to power your circuit even it has not fully discharged.


2. Self leakage

Ultra cap known to have a relatively high self discharge compared to Lithium cell, its like night & day, check the datasheet. Lithium cell self discharge is so low that usually will last for many years.

Can't see the self leakage rate for my specific caps on the datasheet, but something I came across showed the voltage drop after 140 minutes or so at 2.5v, I'll try and find that again.

http://www.panasonic.com/industrial/components/pdf/ABC0000CE4.pdf

edit: here we are. http://www.robotroom.com/Capacitor-Self-Discharge-4.html

[mariush]

cr2032 have a big capacitance (200-250mAh) but they can only do 0.2-0.4mA continuously.. maybe 1-2mA pulses.
Supercapacitors are much better in that regard, but coin type supercaps have quite high resistance so they're not that much better than batteries and will discharge by themselves.

For clock backup, I guess it depends on the lowest voltage possible. A cr2032 will die at around 2v but a supercap can go to 0.. so if the clock handles as low as 1.2v-1.5v you could just put a 2.5-2.7v supercap and it will give you more than a day of backup for the time. Ex 0.22f 2.5v, 5x11mm, http://uk.farnell.com/cooper-bussmann/b0510-2r5224-r/cap-super-0-22f-2-5v-rad/dp/2148492

ps. if you want, i can do some discharge tests when i wake up (5-6 hours from now) but i only have 20-25f capacitors around.

[XOIIO]

cr2032 have a big capacitance (200-250mAh) but they can only do 0.2-0.4mA continuously.. maybe 1-2mA pulses.
Supercapacitors are much better in that regard, but coin type supercaps have quite high resistance so they're not that much better than batteries and will discharge by themselves.

For clock backup, I guess it depends on the lowest voltage possible. A cr2032 will die at around 2v but a supercap can go to 0.. so if the clock handles as low as 1.2v-1.5v you could just put a 2.5-2.7v supercap and it will give you more than a day of backup for the time. Ex 0.22f 2.5v, 5x11mm, http://uk.farnell.com/cooper-bussmann/b0510-2r5224-r/cap-super-0-22f-2-5v-rad/dp/2148492

ps. if you want, i can do some discharge tests when i wake up (5-6 hours from now) but i only have 20-25f capacitors around.

That would be neat. Tomorrow I am going to open up my clock, connect a variable power supply to the backup and see how low it goes before it drops out. I'm going to guess it can go to 1.5, or maybe even 1 volt before dying out, but it's just a stab in the dark. It probably can vary a lot between different kinds of clocks and the chips they actually use to keep the clock running. I wonder if they just keep the main bit running, or shift it to some lower power one for the backup period.

[Stonent]

Mjlorton (Martin) has a few videos on super caps and discharge rates on his Youtube channel.

[Psi]

yeah people say that super caps have a high self discharge and that they run flat in a week or two but its a myth, at least for the common 2.5V 2600F maxwell boostcaps.

I ran some tests on them before replacing my car battery and the leakage really is minute.
I cant remember the exact figures, but of the 9 caps i fully charged up the worst one lost like 0.05V over 2 weeks.

It would probably take 2 years for a maxwell 2.5V boostcap to drop ~1 volt form leakage. I can be pretty sure of that because i only used 6 of the 9 caps in my car. The other 3 have just been sitting around the lab since then and haven't been charged up again since the test.

I just went and checked them with the meter, they're at   1.99V, 1.56V and 0.83V
(shorted one of them out a few times 6 months back while demonstrating boostcaps to a friend, so that explains the 0.83V one)

I think the leakage slows down once the 2.5V cap gets down to ~2V.

[amyk]

I think the leakage slows down once the 2.5V cap gets down to ~2V.

It's just like discharging through a resistor, the curve is exponential.

[Rufus]

Alright, here's something I was thinking about in my spare time, how effective would a 5.5v 1 Farad super capacitor be in comparison to a cr2032 battery

The energy stored in a CR2032 battery is around 2350J, the energy stored in a 1F capacitor between say 4 and 2v (probably the range could expect an RTC to operate on) is about 6J.

So a 1F supercap is about 400 times less effective - depending how you define effective.

[mariush]

So a bit late but i guess better than never.

The test supercapacitor is a 25 uF 2.7v Cooper Bussman HV series capacitor: http://www.cooperindustries.com/content/dam/public/bussmann/Electronics/Resources/product-datasheets/Bus_Elx_DS_4376_HV_Series.pdf

HV1625-2R7256-R  : 25F 2.7v 0.027 ohm max esr @ 100hz,  45 uA leakage , 16x25

This particular supercapacitor stayed on my workbench unused for maybe a couple of weeks and the temperature was around 16-18c, because i left a window open by the workbench for some time (saying it in case it matters below).

Logging is done with Uni-T UT61E, data export made by UltraDMM. The UT61E does 2 readings per second and has a 10Mohm input impedance on DC voltage and 0.1% accuracy (+/- a few counts)

The supercapacitor was charged to about 2.684v using an adjustable linear regulator, then I simply disconnected it from the regulator and connected it to the multimeter leads.

Initially, I set up a laptop a the serial to usb adapter and i rushed to start the test and simply downloaded the default uni-t dmm software that was on the english version of the site.  It was probably an old version, because about 20-30 minutes after starting the test the software crashed. 

Thinking it may have been the screensaver or some other feature of the laptop, I recharged the supercapacitor and disabled power related stuff and left it running again only to find that it crashes again by itself.

Now what I did notice is that the supercapacitor was discharging up to around 2.21v when the software crashed.

So I installed UltraDMM, recharged the capacitor to 2.684v (or around that value) and again left it logging.
Software doesn't crash this time but what's interesting is that about 3h 20 minutes hours later, the supercapacitor still reported about 2.34v.

I don't know why the supercapacitor discharges much slower now, I can only assume the two semi-discharge/charge cycles and the room/workbench temperature increase (workbench and capacitor measures ~ 22.4c with an infrared temp sensor) contributed to this.

I'm attaching a first set of raw readings and a smaller ods datasheet with the pretty chart below:



I'll leave the laptop and multimeter logging for a few hours more, right now the supercap is at around 2.291v (at 20:42, last timestamp in attached file is 18:56 with a 2.348v reading) ...

offtopic libreoffice calc is painfully slow at doing such graphs with lots of values, takes 20-30s on an eight core cpu to plot 30k points... sigh.
 

[T3sl4co1l]

Notice you'll see dielectric absorption effects... it'll relax from nominal charge quite a bit, just as it will recover from a "complete" (0V short) discharge slowly.  I don't know offhand how much absorption those types have.

I recall reading an appnote, I think by Jim Williams at LT, which required extremely low leakage capacitors of high value (>10 uF), for AC coupling and low noise signal filtering.  The best are wet slug tantalum (hermetically sealed, mil spec, and around $100 each!).  After soaking to rated voltage for a week, then selecting for minimal leakage.

Electrolytics of course (tantalum and aluminum) have some electrochemistry going on, where leakage (true leakage, not absorption) drops over time when operated at in-circuit or nominal rated voltage.  AFAIK this shouldn't apply to super/ultra caps.

Tim

[Psi]

I think the leakage slows down once the 2.5V cap gets down to ~2V.

It's just like discharging through a resistor, the curve is exponential.

i got the impression some other chemical effect started to dominate which caused leakage to increase when the voltage got close to Vmax.

[mariush]

Ok, here's an update.

I stopped the logging at around 2.15v because the voltage drop was incredibly slow. I've attached the full log from ultra dmm and I've also attached two "cleaned" logs, with only date + voltage and the second one  "seconds from start" and voltage", so that you see a better representation of the voltage drop.
In these two files, I've also removed duplicate readings of the same voltage so that there's only about 12000 data points instead of over 100k .. ms office is stuck at 65k rows and libreoffice shits itself at more than 30k rows when doing charts.

There's about 6-8 minutes lost somewhere around 2.3v because ultradmm has its own quirks. As the logging was done on an old Centrino (single core 1.5ghz) I was afraid ultradmm may freeze due to lots of readings in the log so at around 50k readings in the meter log, I wanted to export and clear the log.. but udmm would interrupt the export because as it was taking 20-30s to save 100 MB of xml file to disk, the grid view changed so much it was aborting the export process. So i had to uncheck the auto update and then udmm froze for about 3 minutes. Eventually, I managed to do the export by pulling out the sensor from the ut61e so that udmm stopped receiving data.

Anyway... so here's the discharge graph :



X axis is seconds since start of test ... 47955s or 799.25 minutes or about 13 hours.

Raw logs, the filtered data etc can be downloaded from here (file size is over 1000KB so can't attach it)


Now I soldered a 15k resistor on the leads of the supercapacitor and I'm going to let it discharge for a few hours. Chose 15k because it's the same value used in cr2032 datasheets, for example see energizer cr2032, maxell cr2032, panasonic cr2032 lithium.

So far, 41 minutes in the test, the supercap discharged from 2.684v to 2.583v - It's going to be boring, the graph should be about the same.

[Rufus]

So far, 41 minutes in the test, the supercap discharged from 2.684v to 2.583v - It's going to be boring, the graph should be about the same.

But not nearly as boring as waiting the 1200 hours a CR2032 with 15k load would take to get to 2.4v.

Not that supercaps don't have valid application, but, detailed comparison with lithium batteries is pointless.

[mariush]

And the 15k resistor test is put to an end after a bit more than 12h.

Nothing unexpected.. supercap went from 2.682v to 2.18v without showing any signs of leaking or sensitivity to temperature (opened and left window open for about 30 minutes around the middle of the test cooling the room a few degrees C) or anything else.



So 12h to decrease by 0.5v.. but remember this is a 25f 2.7 supercapacitor with low esr (0.027 ohm). The 5.5v supercaps can have 10-40 ohm esr and and those cell type supercaps can have up to 200-300 ohm. Not sure how much this matters with the low power consumption of a time chip.
Anyway, even with a cheap 1f supercapacitor it should still give you a few hours of life, at least. And they recharge in seconds if you give them the current.

Raw readings, filtered data etc here.