Will a capacitor's value remain constant with frequency?

[DragonPetter]

Will real practical capacitors have a constant capacitance across the entire frequency range, or do physical limitations change the capacitance at high frequencies?

I see in capacitor datasheets that the impedance of a ceramic capacitor changes with frequency, but I see this as the ESR and ESL of the part playing a factor.

 

[Floid]

If you get to a high enough frequency, the ESL will start to exceed the capacitance and the part starts acting as an inductor instead of a capacitor. Depending on the ESL (which depends on the material and package of the cap), this usually will happen in the 10MHz - upper 100s of Mhz range (800+).

 

[sophiecentaur]

The main reason why your average capacitor appears to change its capacitance with frequency is do to the inevitable Inductance that is there because of its construction. A typical capacitor has a lead and is probably spiral wound. The series inductance introduces an inductive reactance in series. This reactance adds (vectorially) to the Capacitive reactance to produce a lower one so it appears that the capacitor has a higher value than it measures at low frequency.
Eventually, the L and C will resonate (XL + XC =0) and, above this resonant frequency, the C 'becomes' an L!
It's a real problem to make a high value Capacitor that will operate exactly as you expected at UHF and above.

 

[DragonPetter]

Thanks for the explanations. I knew that the impedance is changing and you help make that clear. I guess the other part to the question is if there is sort of a non-linear effect that the actual materials can have that makes them lose capacitance as frequency is increased?
A similar effect that I'm trying to compare to is that in electrolytic capacitors, as the voltage is increased very close to its maximum voltage rating, its effective capacitance is actually decreased because of the dielectric properties. So in this case, its some property of the capacitor aside from its capacitance, inductance, resistance parasitics that is affecting its effective capacitance.

 

[es1]

This app note could help
http://www.n4iqt.com/BillRiley/multi/esr-and-bypass-caps.pdf

 

[yungman]

All the above are true. Just want to add one more. When you said "all" frequency, this include very high frequency. As the dimension of the capacitor larger than 1/20 of the wave length, you enter into a different world of traveling wave and all the lump approximation are out the window. You'll start seeing potential pattern even on the capacitor plate.

 

[Antiphon]

I'll add a real ball buster.

If you use low voltage ceramic caps near their rated voltage they can loose more than half their capacitance. A 1uF ceramic cap rated for 10 volts is way less than 1uF if charged to 9 volts. The same cap if rated for 50 volts will still be 1uF. Nobody ever seems to talk about this.

I never use any ceramic caps at more than 50% of their rated voltage.

 

[yungman]

Holy, I did not know this.

 

[Antiphon]

Holy, I did not know this.

Glad I could help Yungman. I know you've been designing small low voltage mixed signal circuits. That's just where you will run into it.

"For example, the capacitors may undergo a change of capacitance (Y5V ceramic capacitors will loose 80% of the initial capacitance under rated input voltage). Also, the ESR of input capacitors will depend on the rise time of the waveform."

From Page two: http://cds.linear.com/docs/Application%20Note/an88f.pdf

Edit: here's the definitive reference:
http://www.enpirion.com/bcf35454-46a3-4835-8ded-2ebf9fb53c27/resource-center-product-literature-delivery.htm

 

[yungman]

Thanks for the info. I just never thought of this.

 

[Kholdstare]

Thanks for the explanations. I knew that the impedance is changing and you help make that clear. I guess the other part to the question is if there is sort of a non-linear effect that the actual materials can have that makes them lose capacitance as frequency is increased?
A similar effect that I'm trying to compare to is that in electrolytic capacitors, as the voltage is increased very close to its maximum voltage rating, its effective capacitance is actually decreased because of the dielectric properties. So in this case, its some property of the capacitor aside from its capacitance, inductance, resistance parasitics that is affecting its effective capacitance.

http://www.ami.ac.uk/courses/topics/0184_dp/index.html

http://en.wikipedia.org/wiki/Dielectric#Dielectric_relaxation

 

[sophiecentaur]

There are clearly two factors affecting the value of effective capacitance. One is linear (effects of parasitic inductance) and the other is non-linear (voltage dependent Capacitance). Non-linear effects could give you some very strange results if you are unlucky, generating cross and intermodulation products at embarrassing frequencies. Once they've been generated they are virtually impossible to eliminate. But, to be fair, it is the linear effects due to the construction that are more common.

 

[Mike_In_Plano]

Typical effects:

Voltage dependence - High dielectric ceramic caps such as Y5U have less capacitance with increasing voltage.

Soakage - Capacitors are subject to having little "nooks and crannies" of capacitance that slowly become acclimated to a change in voltage. Thus, you can start with zero volts, apply 1, come back, and it will have say, .9 volts. High dielectric ceramics and electrolytics are the worst offenders, while some polypropylene film show very little of the effect. Note that this effect can be slooooow.

Frequency dependence of electrolytics - Electrolytics typically show less ESR and an improved looking capacitance as you go up into the 100 kHz range.

Equivalent inductance - As you go up in frequency, every cap will eventually suffer from equivalent series inductance. Thus, it stops appearing to be a cap, becomes a resistor, and then becomes an inductor as the frequency goes up. Chip capacitors can serve with very little inductance, hence many ceramics are well suited to high frequency use. The shorted and wider these caps are, the less inducatnce they exhibit, so these types are preferred for bypassing ultra fast processors.

Mike