Decoupling caps

[Paul Burridge]

I recently bought 4,000 1uF SMD caps from e-bay. I can't recall why
exactly, but they were cheap, anyway and you'd kick yourself if you
needed to build a circuit that called for 3,500 1uF caps and you
hadn't bought them!
Now, seriously, one often sees values for RF bypass caps quoted
typically at 0.1 or .01uF or similar value depending on whatever. I
don't see any reason why those values shouldn't be regarded as
*minimum* for the purpose of decoupling and I should be able to use my
neat little 1uFs for any RF decoupling circuit cap. It can't cause any
problems, can it? I assume they only bothered to specify .1 or .01 in
the days when physical size had to be considered. What's the view of
the panel?

 

[Clarence]

I recently bought 4,000 1uF SMD caps from e-bay. I can't recall why
exactly, but they were cheap, anyway and you'd kick yourself if you
needed to build a circuit that called for 3,500 1uF caps and you
hadn't bought them!
Now, seriously, one often sees values for RF bypass caps quoted
typically at 0.1 or .01uF or similar value depending on whatever. I
don't see any reason why those values shouldn't be regarded as
*minimum* for the purpose of decoupling and I should be able to use my
neat little 1uFs for any RF decoupling circuit cap. It can't cause any
problems, can it? I assume they only bothered to specify .1 or .01 in
the days when physical size had to be considered. What's the view of
the panel?

You didn't say if these are Ceramic, or Tantalum. But assuming Ceramic they
will work fine up to many MHZ. Probably not good above something like 100 MHZ

 

[James Meyer]

Now, seriously, one often sees values for RF bypass caps quoted
typically at 0.1 or .01uF or similar value depending on whatever. I
don't see any reason why those values shouldn't be regarded as
*minimum* for the purpose of decoupling and I should be able to use my
neat little 1uFs for any RF decoupling circuit cap.

Obtain the full specs for your caps and use those specs in your favorite
spice simulator to look at the equivalent impedance for the caps for the
frequency of interest. You may find that the caps are excellent RF bypass
capacitors or you may find that they look like inductors in series with
resistors.

There are good reasons why .1 and .01 uF caps are often used as bypass
caps instead of 1.0 uF.

Jim

 

[John Fields]

I recently bought 4,000 1uF SMD caps from e-bay. I can't recall why
exactly, but they were cheap, anyway and you'd kick yourself if you
needed to build a circuit that called for 3,500 1uF caps and you
hadn't bought them!
---

Now, seriously, one often sees values for RF bypass caps quoted
typically at 0.1 or .01uF or similar value depending on whatever. I
don't see any reason why those values shouldn't be regarded as
*minimum* for the purpose of decoupling and I should be able to use my
neat little 1uFs for any RF decoupling circuit cap.

---
Hmmm... sed, right?^)

The reason is because once you start to get into high frequencies the
inductance of the cap (leads, end terminations, _winding_) starts to
matter and, generally, the higher the capacitance for a given family
of caps, the higher the inductance of the cap. At low frequencies it
doesn't matter much, but once you leave the slop of audio behind and
things start to get important, you'll eventually get to a point where
instead of soldering in bypass caps what you'll really be doing is
soldering in peaking coils, where some really interesting things start
to happen when you hit the self resonant frequency of the cap. Look
it up. Especially, think about what an aluminum electrolytic with
_coiled_ plates...
---

It can't cause any problems, can it?

---
Right. Like you really _need_ an 0.1, but since you can't get one to
fit you'll settle for an 0.01 instead?

That's kind of like if you were starving to death and someone handed
you a shrimp instead of a lobster.
---

 

[John Larkin]

There are good reasons why .1 and .01 uF caps are often used as bypass
caps instead of 1.0 uF.

Why?

John

 

[Spehro Pefhany]

Why?

John

They are 0.001 cent cheaper and work as well.

Best regards,
Spehro Pefhany

 

[Guy Macon]

The reason is because once you start to get into high frequencies the
inductance of the cap (leads, end terminations, _winding_) starts to
matter and, generally, the higher the capacitance for a given family
of caps, the higher the inductance of the cap. At low frequencies it
doesn't matter much, but once you leave the slop of audio behind and
things start to get important, you'll eventually get to a point where
instead of soldering in bypass caps what you'll really be doing is
soldering in peaking coils, where some really interesting things start
to happen when you hit the self resonant frequency of the cap. Look
it up. Especially, think about what an aluminum electrolytic with
_coiled_ plates...

Simple solution: just put ten of those 1uF caps in series, and you
have the 0.1uF cap that everyone likes!
(GUY runs away as everyone throws rotten fruit and rock at him...)

 

[John Larkin]

They are 0.001 cent cheaper and work as well.

Oh. Right.

John

 

[Jeff]

Why?

Because, due to series inductance, etc, a 0.1uf cap is likely good for a few
MHz, a 0.01uf cap good for a few tens of MHz, etc. Case size (1206, 0805,
0603, etc) and type (SMD TH) have a huge play in this.

 

[Spehro Pefhany]

Because, due to series inductance, etc, a 0.1uf cap is likely good for a few
MHz, a 0.01uf cap good for a few tens of MHz, etc. Case size (1206, 0805,
0603, etc) and type (SMD TH) have a huge play in this.

What is the difference in series inductance between a 0603 1uF part
and a 0603 0.1uF ceramic part?

Best regards,
Spehro Pefhany

 

[Boris Mohar]

What is the difference in series inductance between a 0603 1uF part
and a 0603 0.1uF ceramic part?

Best regards,
Spehro Pefhany

They don't say but they say many other things for your inference

http://www.avx.com/docs/techinfo/parasitc.pdf



Regards,

Boris Mohar

Got Knock? - see:
Viatrack Printed Circuit Designs http://www3.sympatico.ca/borism/

 

[John Larkin]

Because, due to series inductance, etc, a 0.1uf cap is likely good for a few
MHz, a 0.01uf cap good for a few tens of MHz, etc.

A little more precisely, a 0.1 uF cap is likely worthless below a few
MHz, a 0.01uF cap is worthless below a few tens of MHz, etc.

For a MLC of a given case size, series L is pretty much independent of
capacitance. So the issue becomes, would you rather have, as a bypass
or coupling cap, a 1 nF cap in series with 0.7 nH, or a 1 uF cap in
series with 0.7 nH? At very high frequencies, they both reduce to 0.7
nH.

A 0.33uF 0603 surfmount cap is a very good coupling cap/DC block in a
50-ohm system, to 10 GHz maybe.

John

 

[Jeff]

What is the difference in series inductance between a 0603 1uF part
and a 0603 0.1uF ceramic part?

http://rocky.digikey.com/WebLib/Panasonic/Web data/ECJ Series.pdf

Take a look at page 15, Only the impedance values are given, and it's for
0805 caps. Note that the ceramic material also has a significant effect on
impedance. Add in a little trace inductance, and the graphs will shift to
the left, likely enough to get close to my ballpark #'s.

Best regards,
Spehro Pefhany

http://www.speff.com

 

[John Larkin]

http://rocky.digikey.com/WebLib/Panasonic/Web data/ECJ Series.pdf

Take a look at page 15, Only the impedance values are given, and it's for
0805 caps. Note that the ceramic material also has a significant effect on
impedance. Add in a little trace inductance, and the graphs will shift to
the left, likely enough to get close to my ballpark #'s.

At high frequencies, the impedances converge. You *can* use the SRF
notch to advantage in narrowband applications.

John

 

[John Larkin]

They don't say but they say many other things for your inference

http://www.avx.com/docs/techinfo/parasitc.pdf

Nice. Fig 5 is great.

John

 

[Carl Ijames]

They don't say but they say many other things for your inference

Nice. Fig 5 is great.

Doesn't Fig 5 say that the 1 uF's are better (or no worse) for bypassing
everywhere except between maybe 30 and 80 MHz where a .001 uF has maybe
4x lower impedance but the 1 uF is still under .5 ohm? I guess you
could parallel a .001 and 1, but it looks to me like the 1 uF by itself
would be just fine.

 

[Spehro Pefhany]

They don't say but they say many other things for your inference

http://www.avx.com/docs/techinfo/parasitc.pdf

Figures 4 and 5 on on page 3 give a pretty good idea (the latter for
1206, but whatever).

Best regards,
Spehro Pefhany

 

[PaulCsouls]

You didn't say if these are Ceramic, or Tantalum. But assuming Ceramic they
will work fine up to many MHZ. Probably not good above something like 100 MHZ

Yes your probably right for decoupling applications but ceramic caps
tend to get worse as they get bigger in value. The ones upto 1000pF
are usually the NPOs, then you hit the X7Rs upto 0.1uF, as you
approach 1uF and above you'll need to compomise down to Z5Us.
Basically because you need funkier dielectrics for bigger capacitances
in small packages. In decoupling it probably doesn't matter but if
your designing a filter it helps to keep the caps small.

Paul

 

[Ray Anderson]

.
For a MLC of a given case size, series L is pretty much independent of
capacitance. So the issue becomes, would you rather have, as a bypass
or coupling cap, a 1 nF cap in series with 0.7 nH, or a 1 uF cap in
series with 0.7 nH? At very high frequencies, they both reduce to 0.7
nH.

A 0.33uF 0603 surfmount cap is a very good coupling cap/DC block in a
50-ohm system, to 10 GHz maybe.

John

For the most part, the partial inductance contributed by the ceramic MSD
decap is fairly independent of actual capacitance value for a given case
size. (i.e. .01uF, .1 uF and 1uF in a 0805 package have about the same
inductance which is mainly determined by the height dimension). The
actual mounted inductance is what really matters and is composed of the
partial inductances contributed by the capacitor, by the escape
geometries and vias, and the spreading inductance of the planes which is
a function of the stackup thickness.

If you have poor mounting technique then the partial inductance
contributed by the capacitor itself can almost get lost in the noise.
With proper attention to detail one can achieve a mounting inductance of
300-400 pH. At this point the capacitor inductance is a significant part
of the total loop inductance which, with the capacitor's nominal
capacitance, will determine the parts SRF in use.

Reverse geometry capacitors (0306 as opposed to the more common 0603 and
0204 etc.) have considerably less mounted inductance than their more
common counterparts for a couple of reasons. 1.) Less intrinsic
inductance due to internal construction 2.) Lower mounting inductance
due to the smaller current current loop encountered in the mounting process.

Someone quoted an Agilent app-note near the beginning of this thread
that addressed the measurement of capacitor inductance. Not to cast
aspersions on Agilent, but the methodology and fixture they supply for
the impedance meter as discussed in the app note does not produce very
accurate or repeatable numbers (particularly ESR). There are much more
accurate ways of doing the measurements by utilizing a VNA to do S21
measurements.

-Ray

-Ray

 

[Clarence]

Yes your probably right for decoupling applications but ceramic caps
tend to get worse as they get bigger in value. The ones up to 1000pF
are usually the NPOs, then you hit the X7Rs upto 0.1uF, as you
approach 1uF and above you'll need to compromise down to Z5Us.
Basically because you need funkier dielectrics for bigger capacitances
in small packages. In decoupling it probably doesn't matter but if
your designing a filter it helps to keep the caps small.

Paul

There are many considerations for using a cap in a filter. I prefer film types
at low frequencies, and Mica for RF. But there are always exceptions.