Electrolytic caps

[JSF]

When using electrolytic caps in a power supply is it any advantage in
using a cap that is twice or three times the output voltage, like a 35 volt
cap on a 13 volt supply.
Also any ideas on electrolytic caps behavior when the caps are tested
at -55 C, any changes?

 

[Spehro Pefhany]

When using electrolytic caps in a power supply is it any advantage in
using a cap that is twice or three times the output voltage, like a 35 volt
cap on a 13 volt supply.

They'll tend to last longer. I don't believe there's much advantage
over about double, though.

Also any ideas on electrolytic caps behavior when the caps are tested
at -55 C, any changes?

Assuming you're talking about aluminum electrolytics, capacitance
drops and ESR increases by a fairly large factor. Limits may not be
specified as low as -55°C (maybe at -40° or -25°C), even if they are
rated for operation that low. Check the data sheets carefully if
that's a real consideration for you.


Best regards,
Spehro Pefhany

 

[Paul E. Schoen]

When using electrolytic caps in a power supply is it any advantage in
using a cap that is twice or three times the output voltage, like a 35 volt
cap on a 13 volt supply.
Also any ideas on electrolytic caps behavior when the caps are tested
at -55 C, any changes?

I have heard that capacitors may become leaky if not used reasonably close
to their rated voltage, because the voltage helps maintain the aluminum
oxide dielectric. I have used 35 VDC capacitors for 5 VDC regulator outputs
without problem, but usually I specify 10 or 16 volt. I don't think it is as
much a problem with tantalum.

Paul E. Schoen
www.pstech-inc.com

 

[Terry Given]

They'll tend to last longer. I don't believe there's much advantage
over about double, though.




Assuming you're talking about aluminum electrolytics, capacitance
drops and ESR increases by a fairly large factor. Limits may not be
specified as low as -55°C (maybe at -40° or -25°C), even if they are
rated for operation that low. Check the data sheets carefully if
that's a real consideration for you.


Best regards,
Spehro Pefhany

Hi Spehro,

did you read Michael Gaspari's paper in thenov/dec 2005 IEEE industry
apps? very good.

Cheers
Terry

 

[Spehro Pefhany]

Hi Spehro,

did you read Michael Gaspari's paper in thenov/dec 2005 IEEE industry
apps? very good.

Cheers
Terry

No, is this the right publication?

http://ieeexplore.ieee.org/iel5/2943/32594/01524724.pdf?isnumber=32594&arnumber=1524724

 

[Terry Given]

No, is this the right publication?

http://ieeexplore.ieee.org/iel5/2943/32594/01524724.pdf?isnumber=32594&arnumber=1524724

thats the professional comic; I meant the journal

The guts of his model is this:

Michael Gaspari wrote a great paper in IEE trans. Industry applications,
vol.41 no.6 nov/dec 2005, pp1430-1435.

his cap model is:

---[Ro]---[R1]---+----[R2]----+----[C1]----
| |
+----[C2]----+

R0 = resistance of foil, tabs & terminals
R1 = resistance of electrolyte
R2 = dielectric loss resistance
C1 = terminal capacitance
C2 = dielectric loss capacitance


R2 and C2 give a large variation in ESR with frequency. typically the
effect of R2,C2 peters out above 10kHz, so you can take the ESR at
100kHz as the combined value of R0 and R1. this can be seen from the
ripple current multiplier tables a decent cap data sheet has.

R0+R1 = ESR @ 100kHz

R2 = ESR @ 100Hz - (Ro + R1)

and pick C2 to get the right values for ESRs in the 100H - 10kHz range

the reason the ripple current varies with temperature is the loss in the
cap is kept constant (for a given lifetime) so lower ESR means more
current. you can thus translate a ripple-current multiplier table (eg
see LXZ cap datasheet) into an ESR multiplier.

ESR_multiplier = 1/(ripple_multiplier)2

the LXZ table for 220uF - 560uF caps is:

120Hz 1kHz 10kHz 100kHz
0.5 0.85 0.94 1.00

so the ESR multiplier is:

4 1.4 1.13 1


the ESR of these caps have at 100Hz is 4x th 100kHz value....

ESR variation with temperature is due to the increased conductivity of
the electrolyte.

R1(T) = R1o*exp[(To-Tcore)/E]

R1o = ESR at temperature To

E = temperature sensitivity factor

R1 is usually 5x Ro, and you can calculate E from the two ESR
measurements (-10C, 20C) for a given cap family.



Cheers
Terry

 

[Spehro Pefhany]

On Sun, 19 Feb 2006 23:12:18 +1300, the renowned Terry Given

<deliberately TOP posted so that Terry's synopsis gets repeated below
for archiving>

Excellent summary, Terry. I'll save this, it will likely come in
handy.

There are some huge differences (5:1) between different capacitors in
their very-low-temperature ESR behavior. Eg. 3* to 15* room
temperature ESR.

Thanks!

The guts of his model is this:

Michael Gaspari wrote a great paper in IEE trans. Industry applications,
vol.41 no.6 nov/dec 2005, pp1430-1435.

his cap model is:

---[Ro]---[R1]---+----[R2]----+----[C1]----
| |
+----[C2]----+

R0 = resistance of foil, tabs & terminals
R1 = resistance of electrolyte
R2 = dielectric loss resistance
C1 = terminal capacitance
C2 = dielectric loss capacitance


R2 and C2 give a large variation in ESR with frequency. typically the
effect of R2,C2 peters out above 10kHz, so you can take the ESR at
100kHz as the combined value of R0 and R1. this can be seen from the
ripple current multiplier tables a decent cap data sheet has.

R0+R1 = ESR @ 100kHz

R2 = ESR @ 100Hz - (Ro + R1)

and pick C2 to get the right values for ESRs in the 100H - 10kHz range

the reason the ripple current varies with temperature is the loss in the
cap is kept constant (for a given lifetime) so lower ESR means more
current. you can thus translate a ripple-current multiplier table (eg
see LXZ cap datasheet) into an ESR multiplier.

ESR_multiplier = 1/(ripple_multiplier)2

the LXZ table for 220uF - 560uF caps is:

120Hz 1kHz 10kHz 100kHz
0.5 0.85 0.94 1.00

so the ESR multiplier is:

4 1.4 1.13 1


the ESR of these caps have at 100Hz is 4x th 100kHz value....

ESR variation with temperature is due to the increased conductivity of
the electrolyte.

R1(T) = R1o*exp[(To-Tcore)/E]

R1o = ESR at temperature To

E = temperature sensitivity factor

R1 is usually 5x Ro, and you can calculate E from the two ESR
measurements (-10C, 20C) for a given cap family.



Cheers
Terry

Best regards,
Spehro Pefhany