Maker Pro
Maker Pro

Microwave dc blocking caps

T

Tom Derham

Jan 1, 1970
0
I have seen it written many times that for dc blocking (e.g. series cap on
the input to a microwave amplifier), the cap is chosen as large as possible
such that the impedance is a fraction of an ohm at the operating frequency,
but the series resonant frequency must still be well above the top operating
frequency.

However, in practice I can't find a value to meet both these requirements.
e.g. at 2.4GHz, an "ideal" cap at 100pF would give fraction-of-an-ohm
impedance, but all real caps of this size have their resonant frequency well
below this frequency and so are inductive in the range of operation...
although the (negative) impedance is still fairly small at this range (maybe
a few ohms).

So how is a blocking cap chosen? The ESR of a good microwave cap (e.g. AVX
AccuP) might be only 0.2 Ohms, but this is swamped by the reactive impedance
everywhere but at the resonant frequency!

Lastly, if the application is only for a fixed frequency, is it fair game to
spec a capacitor for use *at* its resonant frequency? I can see this is not
wanted in filters because it could lead to instability, but is it ok for dc
blocking purposes?

Thanks

Tom
 
P

Paul Burridge

Jan 1, 1970
0
I have seen it written many times that for dc blocking (e.g. series cap on
the input to a microwave amplifier), the cap is chosen as large as possible
such that the impedance is a fraction of an ohm at the operating frequency,
but the series resonant frequency must still be well above the top operating
frequency.

I don't know where you get this figure of "a fraction of an ohm" from.
That sounds idealistic in the extreme. So long as the impedance is
chosen to be very small compared to the load impedance, there
shouldn't be a problem; i.e., a ratio of 100:1 or something of that
order.
 
J

John Larkin

Jan 1, 1970
0
I have seen it written many times that for dc blocking (e.g. series cap on
the input to a microwave amplifier), the cap is chosen as large as possible
such that the impedance is a fraction of an ohm at the operating frequency,
but the series resonant frequency must still be well above the top operating
frequency.

However, in practice I can't find a value to meet both these requirements.
e.g. at 2.4GHz, an "ideal" cap at 100pF would give fraction-of-an-ohm
impedance, but all real caps of this size have their resonant frequency well
below this frequency and so are inductive in the range of operation...
although the (negative) impedance is still fairly small at this range (maybe
a few ohms).

I can't see why the series resonant frequency should be above (or
below, for that matter) the operating fequency; what matters is that
the cap have a low impedance. For a coupling cap, it doesn't matter if
the reactive part is inductive or capacitive. If adding more pf's
doesn't add series inductance, more pf's don't hurt, regardless of
what happens to the theoretical srf of the cap.
So how is a blocking cap chosen? The ESR of a good microwave cap (e.g. AVX
AccuP) might be only 0.2 Ohms, but this is swamped by the reactive impedance
everywhere but at the resonant frequency!

An 0805 cap has maybe 700 nh inductance, and an 0603 is less. At very
high frequencies, one can parallel several small (0603, 0402) caps
across the width of a microstrip trace and have damned-near perfect
coupling... essentially invisible on a 20 GHz TDR. A single cap, if
it's smaller than trace width, can suffer from the current crowding
from the wide trace into the skinny cap, which looks like extra
inductance. The cap makers sometimes cheat and use a high-k substrate
to force the trace width to match the cap size when they measure cap
s-params.
Lastly, if the application is only for a fixed frequency, is it fair game to
spec a capacitor for use *at* its resonant frequency? I can see this is not
wanted in filters because it could lead to instability, but is it ok for dc
blocking purposes?

This is sometimes done, but it's the *in-circuit* impedance of the cap
that matters, and that depends on geometry. Datasheet s-params may be
meaningless unless you can duplicate the physical situation.

1 nh is 15 ohms at your frequency. A couple of 0603's in parallel
should get you down to a few ohms. For narrowband RF, this impedance
can just be lumped into the overall circuit tuning strategy. Most RF
parts and circuits are sloppy enough that a few ohms of reactance
likely won't matter.


John
 
T

Tim Wescott

Jan 1, 1970
0
Tom said:
I have seen it written many times that for dc blocking (e.g. series cap on
the input to a microwave amplifier), the cap is chosen as large as possible
such that the impedance is a fraction of an ohm at the operating frequency,
but the series resonant frequency must still be well above the top operating
frequency.

However, in practice I can't find a value to meet both these requirements.
e.g. at 2.4GHz, an "ideal" cap at 100pF would give fraction-of-an-ohm
impedance, but all real caps of this size have their resonant frequency well
below this frequency and so are inductive in the range of operation...
although the (negative) impedance is still fairly small at this range (maybe
a few ohms).

So how is a blocking cap chosen? The ESR of a good microwave cap (e.g. AVX
AccuP) might be only 0.2 Ohms, but this is swamped by the reactive impedance
everywhere but at the resonant frequency!

Lastly, if the application is only for a fixed frequency, is it fair game to
spec a capacitor for use *at* its resonant frequency? I can see this is not
wanted in filters because it could lead to instability, but is it ok for dc
blocking purposes?

Thanks

Tom
I have seen the idea of using a blocking cap at it's series resonant
frequency advocated, I think in the ARRL's "UHF Experamenter's Manual".
I've never designed a microwave circuit myself, so take it for what
it's worth...
 
I

Ian Buckner

Jan 1, 1970
0
John Larkin said:
An 0805 cap has maybe 700 nh inductance, and an 0603 is less. At very
high frequencies, one can parallel several small (0603, 0402) caps
across the width of a microstrip trace and have damned-near perfect
coupling... essentially invisible on a 20 GHz TDR. A single cap, if
it's smaller than trace width, can suffer from the current crowding
from the wide trace into the skinny cap, which looks like extra
inductance. The cap makers sometimes cheat and use a high-k substrate
to force the trace width to match the cap size when they measure cap
s-params.


This is sometimes done, but it's the *in-circuit* impedance of the cap
that matters, and that depends on geometry. Datasheet s-params may be
meaningless unless you can duplicate the physical situation.

1 nh is 15 ohms at your frequency. A couple of 0603's in parallel
should get you down to a few ohms. For narrowband RF, this impedance
can just be lumped into the overall circuit tuning strategy. Most RF
parts and circuits are sloppy enough that a few ohms of reactance
likely won't matter.


John
I think John meant 700pH inductance rather than 700nH!

From the AVX site at: http://www.avx.com/docs/catalogs/hi-q.pdf
there is a graph of typical SRF for the U series. A 10pF 0805 is shown
as having a SRF of around 2.4GHz, which translates to a series
inductance
of around 440pH. As John says, if you need lower put capacitors in
parallel, or lump it into the overall circuit.

Regards
Ian

(For lots of fun, try ac coupling over 100kHz to 6GHz!)
 
T

Tom Derham

Jan 1, 1970
0
Ian Buckner said:
I think John meant 700pH inductance rather than 700nH!

From the AVX site at: http://www.avx.com/docs/catalogs/hi-q.pdf
there is a graph of typical SRF for the U series. A 10pF 0805 is shown
as having a SRF of around 2.4GHz, which translates to a series
inductance
of around 440pH. As John says, if you need lower put capacitors in
parallel, or lump it into the overall circuit.

(For lots of fun, try ac coupling over 100kHz to 6GHz!)

Excellent stuff guys, thank you.
I feel rather more confident now about operating close to the resonant
region for dc blocking. As for the wideband 100kHz-6GHz, think I'll leave
that to the more experienced!
-t
 
J

John Larkin

Jan 1, 1970
0
I think John meant 700pH inductance rather than 700nH!
Right!


From the AVX site at: http://www.avx.com/docs/catalogs/hi-q.pdf
there is a graph of typical SRF for the U series. A 10pF 0805 is shown
as having a SRF of around 2.4GHz, which translates to a series
inductance
of around 440pH. As John says, if you need lower put capacitors in
parallel, or lump it into the overall circuit.

Regards
Ian

(For lots of fun, try ac coupling over 100kHz to 6GHz!)

For a reasonably wide trace, like 100 mils or so, you can make a
z-slice in the trace

------------------------| |-----------------
-----------| |
| ------------
-----------| |------------------------------


(darn, that's hard to draw)

and bridge the little end gaps with small caps, and put a row of
bigger caps across the big horizontal center cut; toss in a tantalum
in the middle. That seems to work pretty well.

Working coaxial, one can machine clever overlapping telescope-like
things with a big cap inside.

John
 
S

Steve Kavanagh

Jan 1, 1970
0
Lastly, if the application is only for a fixed frequency, is it fair game to
spec a capacitor for use *at* its resonant frequency? I can see this is not
wanted in filters because it could lead to instability, but is it ok for dc
blocking purposes?
For fixed frequency narrow band circuits I often do this for blocking
and bypassing (use a capacitor with its series resonant frequency
equal to the frequency of interest). You will probably end up in the
general vicinity of 10 pF at 2.4 GHz, depending on type and size of
capacitor you choose. Seems to work OK for VHF and up. It's any
parallel resonant frequency you need to stay away from (capacitor
impedance -> infinity).

Steve
 
G

gwhite

Jan 1, 1970
0
Tom said:
I have seen it written many times that for dc blocking (e.g. series cap on
the input to a microwave amplifier), the cap is chosen as large as possible
such that the impedance is a fraction of an ohm at the operating frequency,
Fine.

but the series resonant frequency must still be well above the top operating
frequency.
Strange.

However, in practice I can't find a value to meet both these requirements.

If it is self-resonant, it likely will be a "fraction of an ohm," and it will be resistive at SRF by
definition.
e.g. at 2.4GHz, an "ideal" cap at 100pF would give fraction-of-an-ohm
impedance, but all real caps of this size have their resonant frequency well
below this frequency and so are inductive in the range of operation...
although the (negative) impedance is still fairly small at this range (maybe
a few ohms).

For an 0402, once they pass SRF, they get up to about 0.1 nH, maybe up to 0.25nH, but that is about
it (inductance is related to the physical dimension). (That is almost without regard for pF
value.) For example, a 47 pF and 220 pF cap will "look" pretty much identical to signal energy at
2.4 GHz.
So how is a blocking cap chosen? The ESR of a good microwave cap (e.g. AVX
AccuP) might be only 0.2 Ohms, but this is swamped by the reactive impedance
everywhere but at the resonant frequency!

If you want wideband coupling, the low frequencies are what you'll mostly care about.
Lastly, if the application is only for a fixed frequency, is it fair game to
spec a capacitor for use *at* its resonant frequency?

I do it all the time. But series self-resonance is not all that sharp when the charateristic
impedance is up around 50 Ohms or so. So don't get stressed about it.
I can see this is not
wanted in filters because it could lead to instability,...

Using at SRF doesn't work in filters by definition. It is a very low ohm resistor at the SRF. Or
we could say it is a filter in and of itself, but a poor one.
...but is it ok for dc
blocking purposes?

Yes. I've tested 0402's for SRF at 2.4 GHz. It is somewhere in the neighborhood of 15 to 22 pF,
IIRC.
 
P

Paul Burridge

Jan 1, 1970
0
If it is self-resonant, it likely will be a "fraction of an ohm," and it will be resistive at SRF by
definition.


For an 0402, once they pass SRF, they get up to about 0.1 nH, maybe up to 0.25nH, but that is about
it (inductance is related to the physical dimension). (That is almost without regard for pF
value.) For example, a 47 pF and 220 pF cap will "look" pretty much identical to signal energy at
2.4 GHz.


If you want wideband coupling, the low frequencies are what you'll mostly care about.


I do it all the time. But series self-resonance is not all that sharp when the charateristic
impedance is up around 50 Ohms or so. So don't get stressed about it.

This business of choosing a cap for its SRF is new to me. Is this a
microwave-specific technique? In 'ordinary' RF., I've not come across
it and the only thing that appears to matter is getting the cap's
impedance at the frequency of interest in relation to the source and
load impedances as low as reasonly practicable. Not that SRF is
generally a consideration at HF, of course.
 
G

gwhite

Jan 1, 1970
0
This business of choosing a cap for its SRF is new to me. Is this a
microwave-specific technique?

I don't know. I don't keep score. I suppose I personally do it more often at microwave.
In 'ordinary' RF., I've not come across
it and the only thing that appears to matter is getting the cap's
impedance at the frequency of interest in relation to the source and
load impedances as low as reasonly practicable.

That's right. But it is "easy" or "reasonable" to get that Z low by just operating at the SRF. By
definition it is the lowest Z possible.
Not that SRF is
generally a consideration at HF, of course.

Watch yourself.
"What is now proved was once only imagin'd." - William Blake, 1793.

Grau, teurer Freund, 1st alle Theorie
Und grün des Lebens goldner Baum.

(All theory, dear friend, is dull gray,
And only the golden tree of life is green.)
GOETHE
 
P

Paul Burridge

Jan 1, 1970
0
That's right. But it is "easy" or "reasonable" to get that Z low by just operating at the SRF. By
definition it is the lowest Z possible.

Thanks, I'll bear it in mind....
Watch yourself.

Well I mean if you're using today's subminature components. Granted it
wasn't always the case 30 or more years ago.
 
Top