Estimating PC pad capacitance?

[John Larkin]

John Larkin a écrit :

I wasn't speaking of measurement instrument capabilities, I can easily
get down to the fF, but rather meant: how do you measure it without
introducing perturbations, i.e. field modification with your probe,
wire, whatever...?

At this pad size I guess you can't neglect this anymore.

The true test would be a check between a field solver and a careful
measurement, but we also need to accurately measure Er.

Not hard. Measure Er on a big hunk of copperclad. Now cut the test pad
and hook up the AADE meter. Ground one side and use a needle probe on
the other. Place the probe close to the pad, zero the c-meter, then
touch the pad. I've done this on small pads, and it seems legit to me.

As long as the c-meter reads about the same fairly far from the pad as
it does just before contact, which it usually does, this should be OK.

John

 

[John Larkin]

We used an embedded pad on our HV circuits (those that needed it) to
achieve a feed forward effect on our feedback loop. It was in
parallel with our HV feedback resistor voltage divider network.

It gave us about 15 pF for a quarter inch wide pad at about just
under an inch length. One "plate" was embedded, and the other was top
side. Of course where in the embedding makes a huge difference for,
as we all know, plate proximity distance directly affects capacitance
achieved.

My ROT for FR-4 is 15 pf per square inch, copper on both sides of an
0.062 thick board. If thinner, scale as 1/t.

I've measured the capacitance temperature coefficient of some random
FR-4 samples and generally get numbers like +900 ppm/degC, pretty
damned bad. Dielectric absorption is nasty, too.

John

 

[joseph2k]

My ROT for FR-4 is 15 pf per square inch, copper on both sides of an
0.062 thick board. If thinner, scale as 1/t.

I've measured the capacitance temperature coefficient of some random
FR-4 samples and generally get numbers like +900 ppm/degC, pretty
damned bad. Dielectric absorption is nasty, too.

John

John you are way, way over it's head. From the content if it's posts it is
the lesser scion of somebody that actually can and does do what it claims.

 

[joseph2k]

Not hard. Measure Er on a big hunk of copperclad. Now cut the test pad
and hook up the AADE meter. Ground one side and use a needle probe on
the other. Place the probe close to the pad, zero the c-meter, then
touch the pad. I've done this on small pads, and it seems legit to me.

As long as the c-meter reads about the same fairly far from the pad as
it does just before contact, which it usually does, this should be OK.

John

With pads of 0.062 on a side i would only treat that as an approximation.
At that size the fringing from the probes themselves is a significant error
source.

 

[Tom Bruhns]

With pads of 0.062 on a side i would only treat that as an approximation.
At that size the fringing from the probes themselves is a significant error
source.

I see a couple things here...

First is that, as I understand it, John is first zeroing out the
capacitance from the probe to the ground plane...or at least to some
combination of the plane and the pad. I would think that would
mitigate the effect of the capacitance from the probe itself in the
final measurement configuration. To make and even more accurate
measurement, we could put a coaxial shield around the probe, but then
the question becomes, what potential should that shield have? If it's
grounded (the same potential as the ground plane), then the pad has
additional capacitance to it; if it's the same potential as the probe,
then it will disturb the field around the pad when measuring the pad's
capacitance. It seems to me that zeroing the measurement with the
probe close to the pad is a reasonable compromise.

Second, what I really had in mind when I posted the question in the
first place is the capacitance not just of the pad but of the pad and
the part that's soldered to it. Of course that gets much more
complicated than just a planar pad against a ground plane. It's
unrealistic to think I'll end up with a perfect model of the parasitic
effects in my circuit, but I'd at least like to get an idea of what to
do to compensate for them. I do know that they are not insignificant,
and that I will need to compensate. Unfortunately, the pad
capacitance to ground is not the only parasitic effect involved.

Cheers,
Tom

 

[John Larkin]

John you are way, way over it's head. From the content if it's posts it is
the lesser scion of somebody that actually can and does do what it claims.

Still, the gross tc of FR4 may be interesting, to others if not to
him.

And I'll politely engage anyone, MiniPrick or PhilA or anyone, on
technical subjects as long as they stay under control.

I recall one time when Phil said something that was true, and I backed
him up. He was shocked that I would agree with him merely because his
statement was true. He would not have done the same for me. That
situation, of course, couldn't happen with AlwaysWrong.

John

 

[MassiveProng]

I've measured the capacitance temperature coefficient of some random
FR-4 samples and generally get numbers like +900 ppm/degC, pretty
damned bad. Dielectric absorption is nasty, too.

Works fine for a feed forward on an HV feedback loop.

 

[MassiveProng]

John you are way, way over it's head. From the content if it's posts it is
the lesser scion of somebody that actually can and does do what it claims.

You are a pile of shit. You don't even qualify as an it, dipshit.

 

[MassiveProng]

Still, the gross tc of FR4 may be interesting, to others if not to
him.

Took the bait, I see.

So it's ok for others to be an asshole, just not me, eh?

 

[MassiveProng]

recall one time when Phil said something that was true, and I backed
him up. He was shocked that I would agree with him merely because his
statement was true. He would not have done the same for me. That
situation, of course, couldn't happen with AlwaysWrong.

Examine this thread, and find yourself to be incorrect.

 

[John Larkin]

Took the bait, I see.

So it's ok for others to be an asshole, just not me, eh?

It's not OK, at least to me, that anyone be ill-mannered. That you are
rude and crude to *everybody*, and beligerent about knowing things you
obviously don't, is just laughable. And we do laugh.

Hell, you don't even have a name.

John

 

[John Larkin]

Works fine for a feed forward on an HV feedback loop.

Yeah, a 5 or 10% error there wouldn't affect loop dynamics a lot. It
can/does really trash the tc of some LC oscillators. We cut the ground
plane out from under some parts of our circuits to reduce the FR4
capacitance for this reason.

When Tektronix started making fast (like, 300 MHz) scopes on FR4
boards, they also discovered transient "hook" effects from dispersion,
the fact that in FR4 the Er changes with frequency.

At 900 PPM, you could use a bit of board as a temperature sensor!

John

 

[John Larkin]

I see a couple things here...

First is that, as I understand it, John is first zeroing out the
capacitance from the probe to the ground plane...or at least to some
combination of the plane and the pad. I would think that would
mitigate the effect of the capacitance from the probe itself in the
final measurement configuration. To make and even more accurate
measurement, we could put a coaxial shield around the probe, but then
the question becomes, what potential should that shield have? If it's
grounded (the same potential as the ground plane), then the pad has
additional capacitance to it; if it's the same potential as the probe,
then it will disturb the field around the pad when measuring the pad's
capacitance. It seems to me that zeroing the measurement with the
probe close to the pad is a reasonable compromise.

Second, what I really had in mind when I posted the question in the
first place is the capacitance not just of the pad but of the pad and
the part that's soldered to it. Of course that gets much more
complicated than just a planar pad against a ground plane. It's
unrealistic to think I'll end up with a perfect model of the parasitic
effects in my circuit, but I'd at least like to get an idea of what to
do to compensate for them. I do know that they are not insignificant,
and that I will need to compensate. Unfortunately, the pad
capacitance to ground is not the only parasitic effect involved.

Cheers,
Tom

Regular, cheap surface-mount resistors make good terminations at least
to 5 GHz. They tend to be a tad inductive, so two 100 ohm resistors,
in a V or T splay at the end of a 50-ohm trace, is even better.

You can also AC couple a signal well into the GHz. Make a small gap in
a 50 ohm trace and bridge it with 2 or 3 0603 or 0402 caps, regular
cheap ceramics.

Tiny surface-mount parts can do things that used to be done only with
distributed (transmission-line based) circuits. It's a lot easier to
tune a lumped circuit.

John

 

[MassiveProng]

Yeah, a 5 or 10% error there wouldn't affect loop dynamics a lot. It
can/does really trash the tc of some LC oscillators. We cut the ground
plane out from under some parts of our circuits to reduce the FR4
capacitance for this reason.

When Tektronix started making fast (like, 300 MHz) scopes on FR4
boards, they also discovered transient "hook" effects from dispersion,
the fact that in FR4 the Er changes with frequency.

At 900 PPM, you could use a bit of board as a temperature sensor!

What about G10?

 

[MassiveProng]

Isn't that about the same chemistry, without the flame retardant
maybe? I wouldn't think it would be much different. FR4 isn't very
consistant to start with.

I don't know that I have any G10 to try. I think you've said it's
better for HV applications, but we don't do a lot of that lately.

I don't recall saying anything about it.

Some hits for FR4 and G10

http://en.wikipedia.org/wiki/FR4

http://www.thegundcompany.com/DataP...essure Laminates/NEMA G10 G11 Glass Epoxy.pdf

 

[John Larkin]

What about G10?

Isn't that about the same chemistry, without the flame retardant
maybe? I wouldn't think it would be much different. FR4 isn't very
consistant to start with.

I don't know that I have any G10 to try. I think you've said it's
better for HV applications, but we don't do a lot of that lately.

John

 

[Marra]

I have almost never had capacitance problems between tracks on my
projects in 25 years.

I have seen problems with ribbon cables and crosstalk.

If I was having problems with a circuit I would be looking elsewhere
like logic races, decoupling or software not coping with glitches.

www.ckp-railways.talktalk.net/pcbcad21.htm

 

[Helmut Sennewald]

Anyone have any favorite reasonably accurate applets or formulas or
the like for calculating the capacitance of a small rectangular
surface against an extensive ground plane? That is, for example, the
capacitance of the pad for a surface-mount part, where there's a
ground plane about 1.5mm behind the pad, through FR4 PC board
material, er about 4.5.

Hello Tom,

I successfully tried with the concept of fringe-capacitance as used
in SPICE for Mosfets.

C = C0 + Cfringe

C0 = e0*er*Ap/d

Cfringe = Cf' * Pp

Ap = area of pad
d = distance of pad to the plane
er = 4.5 for FR4
e0 = 8.85e-12A/m
Cf' = fringe_capacitance/m
Pp = perimeter of pad

From my calculations: Cf' = 24pF/m, er = 4.5

You can get this number indirectly from your transmission line calculator.
Choose Microstrip:
http://www.rogerscorporation.com/mwu/mwi_java/Mwij_vp.html

Calculate the usual C' (capacitance/meter) for a trace with the
pad-width with your impedance calculator.
..
Calculate capacitance of a trace with the length of your pad
C'' = C'*pad_length

Calculate pad capacitance C0
C0 = e0*er*Ap/d

Fringe capacitance per meter
Cf' = (C''-C0)/pad_length/2 (2 because of two sides)

Total capacitance of pad:

C = C0 + Cf

C = C0 + Cf'*Pp

Cpad = C0 + 24pF/m * Pp

Pp is the perimeter of your pad in meter.

The good news is that it's roughly 24pF/m with low sensitivity
regarding the ratio of pad dimensions to the plane distance.
This means you can use this constant for your multilayer board too.

Best regards,
Helmut