Series caps for decoupling

[Patrick Timlin]

Here is something I am thinking about that I am not sure of the
answer. Normally when you decouple something you use a single cap, but
what if you needed a value you didn't have but instead had twice that
value and series connected them to get the desired value?
Theoretically you get the value you need, but in real life, does a
series connection of two caps still give you decoupling?

The real life data... for a product we have a real time clock running
off a 32768 crystal, pretty standard stuff, but we have an issue with
the clock gaining time. This is for a Motorola PowerPC based design
and Motorola acknowledges certain problems with thier chip that can
cause this. One suggestion is to add a 10pF cap off the XTAL pin to
ground in addition to the normal crystal circuit. So I find I don't
have 10pF caps around, but do have 18pF caps, so I figured two of
these in series gives me 9pF, close enough. These are also 0603
ceramic chip caps and found a way to mount them so that I don't need
to add any additional wires (the two caps nicely bridge across the
point I need them and a ground point) so no added inductance due to
wires.

Anyway, I am letting it run to see if it still gain time, but I
started thinking if the series of two caps is really valid for
decoupling like this. I will try just a single 18pF if this doesn't
work, but thought I would throw it out for discussion to see if anyone
might know for sure about this situation. Like I said I never really
gave it much thought since I never had a need to do this in the past,
but now that I am thinking about it, I am wondering if this is one of
those "real life does not match theory" kind of things where my "9"pF
won't look like a single 9pF cap in this sort of situation. Thoughts?

PT

 

[Ban]

Here is something I am thinking about that I am not sure of the
answer. Normally when you decouple something you use a single cap, but
what if you needed a value you didn't have but instead had twice that
value and series connected them to get the desired value?
Theoretically you get the value you need, but in real life, does a
series connection of two caps still give you decoupling?

The real life data... for a product we have a real time clock running
off a 32768 crystal, pretty standard stuff, but we have an issue with
the clock gaining time. This is for a Motorola PowerPC based design
and Motorola acknowledges certain problems with thier chip that can
cause this. One suggestion is to add a 10pF cap off the XTAL pin to
ground in addition to the normal crystal circuit. So I find I don't
have 10pF caps around, but do have 18pF caps, so I figured two of
these in series gives me 9pF, close enough. These are also 0603
ceramic chip caps and found a way to mount them so that I don't need
to add any additional wires (the two caps nicely bridge across the
point I need them and a ground point) so no added inductance due to
wires.

Anyway, I am letting it run to see if it still gain time, but I
started thinking if the series of two caps is really valid for
decoupling like this. I will try just a single 18pF if this doesn't
work, but thought I would throw it out for discussion to see if anyone
might know for sure about this situation. Like I said I never really
gave it much thought since I never had a need to do this in the past,
but now that I am thinking about it, I am wondering if this is one of
those "real life does not match theory" kind of things where my "9"pF
won't look like a single 9pF cap in this sort of situation. Thoughts?

PT

Patrick,
what you descibe is not called "decoupling" but loading. It will work at
this low freq.(32KHz) just as well as 1 cap of half the capacity.
When it really comes to decoupling the performance of 2 caps in series will
be inferiour and shouldn't be employed, because the resulting parasitic
inductivities are adding up (2L in series), so the resonance point and Q
will be lower.

 

[John Woodgate]

I read in sci.electronics.design that Patrick Timlin
sting.google.com>) about 'Series caps for decoupling', on Fri, 23 Apr
2004:

Here is something I am thinking about that I am not sure of the
answer.

What really worries me is the confusion between decoupling and tuning,
not to mention the elementary nature of the question.

 

[Fred Bartoli]

Patrick,
what you descibe is not called "decoupling" but loading. It will work at
this low freq.(32KHz) just as well as 1 cap of half the capacity.
When it really comes to decoupling the performance of 2 caps in series will
be inferiour and shouldn't be employed, because the resulting parasitic
inductivities are adding up (2L in series), so the resonance point and Q
will be lower.

Not quite Ban :

Lnew = 2L but Cnew = C/2 so the resonant frequency stay the same.
Also Q = L w0 / (Rs_eq) = 2 L w0 / (2 Rs_eq)
This for the cap alone, obviously.

Thanks,
Fred.

 

[Ban]

Not quite Ban :

Lnew = 2L but Cnew = C/2 so the resonant frequency stay the same.
Also Q = L w0 / (Rs_eq) = 2 L w0 / (2 Rs_eq)
This for the cap alone, obviously.

Thanks,
Fred.

What you are referring to is the first case of relativly low frequencies.
For decoupling capacitors this is not longer applicable. The self-resonance
point is when the capacity has disappeared and the component starts behaving
purely inductiv. This is already lower for a single higher capacity cap, so
2 in series will not cancel this phenomenon. Have a look at the datasheets
of capacitors. The Q refers to the ESR of the capacitor, which might be
lower for a single one of higher capacity, but here again the two add up and
will give a higher ESR value than one small cap. The most important
parameter for decoupling is the parasitic inductance, which will limit the
di/dt capability and thus the effectivness.

 

[Fred Bartoli]

I was just answering to your statement about the lower resonant frequency
and I stand my case.

What you are referring to is the first case of relativly low frequencies.

This has nothing to do with frequency.

For decoupling capacitors this is not longer applicable. The self-resonance
point is when the capacity has disappeared and the component starts behaving
purely inductiv.

Sure. Put in series to identical resonant devices and you'll get the same
behaviour at the same frequency, but with twice the impedances.

This is already lower for a single higher capacity cap, so
2 in series will not cancel this phenomenon.

Self resonance is lower for a higher value single cap, because you increase
the value. Putting 2 in series halves the value.

Have a look at the datasheets
of capacitors. The Q refers to the ESR of the capacitor, which might be
lower for a single one of higher capacity, but here again the two add up and
will give a higher ESR value than one small cap. The most important
parameter for decoupling is the parasitic inductance, which will limit the
di/dt capability and thus the effectivness.

I'm pretty well aware of that : I've designed a mixed (linear and pwm)
supply for wafer testing that had to withstand 100A/ns (yes no mistake)
di/dt while providing under 100mV regulation for a 100A step, right at the
wafer supply planes. Do the maths and you'll be scared. Plus there was the
probing tips resistance that ate 90% of the regulation margin...


But the OP wanted to reduce his cap value by a 2 factor, so the point is not
to make an equivalent with a string of higher caps.

Another matter is of course that he wasn't looking for bypass caps.

Thanks,
Fred.

 

[Paul Burridge]

I read in sci.electronics.design that Patrick Timlin
sting.google.com>) about 'Series caps for decoupling', on Fri, 23 Apr
2004:

What really worries me is the confusion between decoupling and tuning,
not to mention the elementary nature of the question.

I have some ceramic disk caps in the 1pF - 10pF range: 1, 2, 3p3, 5,
7, 8 & 10pF to be specific. They all check out as marked on a bridge.
They have the black tips, too. Would I be right in thinking that these
will be low K types? I can't imagine why on earth any manufacturer
would make high-K in these low values...

 

[John Popelish]

I have some ceramic disk caps in the 1pF - 10pF range: 1, 2, 3p3, 5,
7, 8 & 10pF to be specific. They all check out as marked on a bridge.
They have the black tips, too. Would I be right in thinking that these
will be low K types? I can't imagine why on earth any manufacturer
would make high-K in these low values...

I think the black tip indicates NP0 (COG) low Q types.

 

[Active8]

I think the black tip indicates NP0 (COG) low Q types.

I thought Paul's black tip caps were covered some time ago. Hey
Paul. Does this apparent lapse of memory have anything to do with
you taking up smoking again?

 

[Paul Burridge]

I thought Paul's black tip caps were covered some time ago. Hey
Paul. Does this apparent lapse of memory have anything to do with
you taking up smoking again?

I haven't actually got around to it, yet, Mike so I'll have to plead
senility.
Those particular caps weren't covered in the discussion - at least I
wasn't referring to the very low value ones in the question I
originally asked. It just seems daft to make a poor quality hi-K cap
specifically for the low values shown, don't you think? What would be
the application for such a device?

 

[Active8]

I haven't actually got around to it, yet, Mike so I'll have to plead
senility.

Well I figured if you had Alzheimers, you might want to cut things
short. Better yet, party your ass off before you get to the point
where you don't know WTF is going on.

Those particular caps weren't covered in the discussion - at least I
wasn't referring to the very low value ones in the question I
originally asked. It just seems daft to make a poor quality hi-K cap
specifically for the low values shown, don't you think? What would be
the application for such a device?

High power UHF tube apps? You don't really need high-Q for a LP
harmonic filter.

 

[Paul Burridge]

Well I figured if you had Alzheimers, you might want to cut things
short. Better yet, party your ass off before you get to the point
where you don't know WTF is going on.

Thanks for the advice, Mike. I'll try to remember it. ;-)

High power UHF tube apps? You don't really need high-Q for a LP
harmonic filter.

Nope. They're all 50V max.