Capacitor self-resonant freq question

[Noel]

For 900MHz band operation, I have seen some ac-coupling capacitors in
the order of 100pF and others in the order of 10nF. If I use generic
NPO/X7R caps, which is better?

A typical 100pF NPO cap has self resonant freq around 1GHz while 10nF
is well below 100MHz. Since above the self resonant frequency, the
impedance goes up with frequency, looking like an inductor, I thought
it was better to use a 100pF cap. But then again, with the 10nF cap,
since the impedance is so much lower before it turns inductive, even
at 900MHz, its inductive reactance is still small so maybe it doesn't
really matter???

Can anyone out there tell me if I am right or wrong?

Noel

 

[Terry Given]

Noel wrote:

For 900MHz band operation, I have seen some ac-coupling capacitors in
the order of 100pF and others in the order of 10nF. If I use generic
NPO/X7R caps, which is better?

NPO. X7R is IIRC +/-15% over its temperature range, NPO is close to
zero. X7R also has a voltage coefficient of capacitance - as you
increase the voltage across the cap, its capacitance drops. This isnt
much, around 10% or so, but again NPO is almost perfect. X7R is also a
bit piezoelectric due to the materials used, whereas NPO aint.

So if you want a particular capacitance over a range of applied voltage,
temperature and thumping, NPO is the way to go.

If you only ever use NPO caps, that limits you to about 1nF or so. X7R
gets you up to about 1uF, but RF caps are almost always much smaller
than that.

A typical 100pF NPO cap has self resonant freq around 1GHz while 10nF
is well below 100MHz. Since above the self resonant frequency, the
impedance goes up with frequency, looking like an inductor, I thought
it was better to use a 100pF cap. But then again, with the 10nF cap,
since the impedance is so much lower before it turns inductive, even
at 900MHz, its inductive reactance is still small so maybe it doesn't
really matter???

if they are in the same package (eg 0603) then the inductance is roughly
constant, ie SRF varies with the square root of capacitance. To answer
your question, model each cap as R + L + C. use the same fairly low R
(1-10mOhm, can read off a datasheet maybe, or look at tdk website for
math models) and the same L. plot the magnitude of impedance vs
frequency for both caps, and look.

lets use 100pF and 10nF in the same package, with 0.01Ohms esr

100pF, 1GHz - L = 250pH (wow, thats low)
Z = 1.59 Ohms
ESR = 0.01 Ohms
Qu = 159

10nF 100MHz - L = 250pH
Z = 0.025 Ohms
ESR = 0.01 Ohms
Q = 2.5

so the 10nF is going to have a not-very-sharp dip from 0.025 ohms down
to 0.01 Ohms (resistive) at 100MHz. At 1GHz the impedance will be about
1.6 Ohms inductive, at 10GHz its 16 Ohms inductive and at 10MHz its
about 1.6 Ohms capacitive.

The 100pF cap is about 160 Ohms capacitive at 10MHz, 16 Ohms capacitive
at 100MHz and 16 Ohms inductive at 10GHz. At 1GHz it has a sharp
resonant peak down to 0.01 Ohms resistive.

So around the resonance of the 100pF capacitor the impedance is a *LOT*
lower than the 1.6 Ohms inductive of the 10nF cap. Obviously the actual
ESR is critical, as it directly controls the unloaded Q of the device.

Can anyone out there tell me if I am right or wrong?

Noel

always model your parts as RLC circuits and you wont be wrong.

cheers
Terry

 

[Tim Shoppa]

A typical 100pF NPO cap has self resonant freq around
1GHz while 10nF is well below 100MHz. Since above
the self resonant frequency, the
impedance goes up with frequency, looking like an
inductor, I thought
it was better to use a 100pF
cap. But then again, with the 10nF cap,
since the impedance is so much lower before it turns
inductive, even at 900MHz, its inductive reactance
is still small so maybe it doesn't really matter???

Order of magnitude reality check (emphasis on "magnitude"):

"perfect" 100pF capacitor at 900MHz: 1.8 ohms impedance

"perfect" 10nF capacitor at 900 MHz: 0.018 ohms impedance.

"Real" 10nF capacitor at 900 MHz: 80 ohms impedance (mostly inductive).

So the inductive reactance is not (in relative terms compared to
capacitive reactance) small at all, it's thousands of times bigger.

If the circuit impedance is 50 ohms to a few hundred ohms, the
impedance of the 10nF capacitor will have a measurable effect (gain
down a fraction of a dB to a few dB). Whether a fraction of a dB or a
few dB matter or not depends on what the circuit is there for.

It still decouples... which is probably the only reason it's there!

Tim.

 

[John Larkin]

On 10 Mar 2005 05:28:02 -0800, "Tim Shoppa" <shoppa@trailing-edge.com>
wrote:

A typical 100pF NPO cap has self resonant freq around
1GHz while 10nF is well below 100MHz. Since above
the self resonant frequency, the
impedance goes up with frequency, looking like an
inductor, I thought
it was better to use a 100pF
cap. But then again, with the 10nF cap,
since the impedance is so much lower before it turns
inductive, even at 900MHz, its inductive reactance
is still small so maybe it doesn't really matter???

Order of magnitude reality check (emphasis on "magnitude"):

"perfect" 100pF capacitor at 900MHz: 1.8 ohms impedance

"perfect" 10nF capacitor at 900 MHz: 0.018 ohms impedance.

"Real" 10nF capacitor at 900 MHz: 80 ohms impedance (mostly inductive).

Why does the 10 nf cap have a higher impedance than the 100 pf? If the
capacitors are the same types (say, both 0805's of similar
construction) the esl will be pretty much independent of capacitance,
and the bigger cap may well have lower esl (more layers.)

I routinely do multi-GHz dc blocks along microstrip transmission
lines, and TDR them to see how well they work. The only visible
difference between a 1 nF 0603 cap and a 2.2 uF 0603 is the improved
low-frequency coupling due to the obvious capacitance difference.

About the only time a lower-value cap will have a lower impedance is
when it hits series resonance.

Our observation is that, in general, bigger caps behave like bigger
caps.

John

 

[John Devereux]

John Larkin <jjSNIPlarkin@highTHISlandPLEASEtechnology.XXX> writes:

Why does the 10 nf cap have a higher impedance than the 100 pf? If the
capacitors are the same types (say, both 0805's of similar
construction) the esl will be pretty much independent of capacitance,
and the bigger cap may well have lower esl (more layers.)

I routinely do multi-GHz dc blocks along microstrip transmission
lines, and TDR them to see how well they work. The only visible
difference between a 1 nF 0603 cap and a 2.2 uF 0603 is the improved
low-frequency coupling due to the obvious capacitance difference.

About the only time a lower-value cap will have a lower impedance is
when it hits series resonance.

Our observation is that, in general, bigger caps behave like bigger
caps.

I found exactly this too. When I got a spectrum analyser with tracking
generator, the first thing I did was look at the impedance of various
chip caps.

Manufacturers datasheets often show power supply bypassing with
paralleled chip capacitors of say 100pF || 10nF || 1uF. What do you
think about this? Is there a reason not to just use say 3 x 1uF?

--

John Devereux

 

[John Woodgate]

I read in sci.electronics.design that John Devereux
<jdREMOVE@THISdevereux.me.uk> wrote (in <87y8cv8ewf.fsf@cordelia.devereu
x.me.uk&gt about 'Capacitor self-resonant freq question', on Thu, 10 Mar
2005:

Manufacturers datasheets often show power supply bypassing with
paralleled chip capacitors of say 100pF || 10nF || 1uF. What do you
think about this? Is there a reason not to just use say 3 x 1uF?

The Z/f characteristics depend critically on where, and how large, the
parasitic inductances are. Consider, for example, the difference between
the Z/f characteristics of three equal caps and three with very
different values, if they are measured via 30 mm long tracks.
--
Regards, John Woodgate, OOO - Own Opinions Only.
The good news is that nothing is compulsory.
The bad news is that everything is prohibited.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

 

[John Larkin]

On 10 Mar 2005 19:44:16 +0000, John Devereux
<jdREMOVE@THISdevereux.me.uk> wrote:

John Larkin <jjSNIPlarkin@highTHISlandPLEASEtechnology.XXX> writes:

Why does the 10 nf cap have a higher impedance than the 100 pf? If the
capacitors are the same types (say, both 0805's of similar
construction) the esl will be pretty much independent of capacitance,
and the bigger cap may well have lower esl (more layers.)

I routinely do multi-GHz dc blocks along microstrip transmission
lines, and TDR them to see how well they work. The only visible
difference between a 1 nF 0603 cap and a 2.2 uF 0603 is the improved
low-frequency coupling due to the obvious capacitance difference.

About the only time a lower-value cap will have a lower impedance is
when it hits series resonance.

Our observation is that, in general, bigger caps behave like bigger
caps.


I found exactly this too. When I got a spectrum analyser with tracking
generator, the first thing I did was look at the impedance of various
chip caps.

Manufacturers datasheets often show power supply bypassing with
paralleled chip capacitors of say 100pF || 10nF || 1uF. What do you
think about this? Is there a reason not to just use say 3 x 1uF?

IC manufacturers all think the world revolves around their part, so
they want three different caps on every power pin. The idea of
staggering the srf points of bypass caps is, in my opinion, silly. A
power plane or island adjacent to a ground plane, scattered with a few
0.33 uF or whatever surfmount caps, is good enough for most cases.

John